PDA

View Full Version : LW8 benchmark



sandman300
05-07-2004, 05:04 PM
I thought it would be a good idea to start a thread to see comparitively real results of lightwave 8 on different hardware.

I run Win XP pro on a Athalon64 3200+ w/ 1gig FSB and 512 MB RAM. The scene is from the LW content CD2 called Radiosity_BOX.
On my computer it rendered at 51m 3s.

kmscottmoore
05-07-2004, 05:13 PM
I got 23 minutes 27 seconds on my PC.

3D Boxx
Dual 2.8 Ghz Pentium 4/Xeon
1GB PC 2100 RAM
Windows XP Pro

I haven't had time to render it on my Mac yet. If anyone has one, I would honestly be interested in how the G5 does on this one. Beamtracer? are you out there?

--Scott

WizCraker
05-07-2004, 05:45 PM
And with FPrime it takes what? 5 Minutes.

kmscottmoore
05-07-2004, 05:54 PM
Actually, FPrime gives you a preview very fast, but to get any kind of decent anti-aliasing takes forever. I gave up after about 15 minutes when everything was still grainy. Worley is aware of this and they say they are working on it.

I have found that I mostly use FPrime during set-up, but I have been favoring LW's render engine for finals.

Hopefully FPrime 1.5 or 2.0 will improve the anti-aliasing and then I will be able to use it for final renders. (Don't forget the other SDK problems as well.)

theo
05-07-2004, 06:09 PM
Computer built In-house
Asus PC-DL Board
Dual 2.8 Xeons
2000 MB Ram
WinXP

------------------------------------------------------------

bloontz
05-07-2004, 06:09 PM
34m 1s

dual 2ghz xeon
2gb pc2100 ram
8 threads

gregsduncan
05-07-2004, 06:14 PM
I know it makes sense but having the dual processor just about halves the render time. Or it is having double the memory (comapred to sandman) which halved the time? Or is it because sandman was using AMD?

meanlebh
05-07-2004, 06:23 PM
most of it probably is the dual processors, but LW is also optimized for intel, and not for AMD...I am assuming that even a dual AMD would not quite be able to match the intels....I wish that I had my LW 8....I would also be interested in seeing how the G5 does compared to the intels....but alas, it got sent to the wrong address, so I can't run the test as of yet....if anyone out there reading this is running a 2 ghz dp G5, and has the time, please post your results....:D

theo
05-07-2004, 06:42 PM
One thing that a lot of guys overlook is the motherboard which is why we built the system in-house. The Motherboard is crucial if you want to max out your Xeon's potential. Problem is, a lot of boards that say they can handle Xeons are not nearly as capable when it comes down to the benchmarks.

The best board in the world for Xeon's is the Asus PC-DL, which is what I am using. Tom's hardware has a review on this board proving this.

If Scott doesn't have an Asus he is probably using the second best board as his numbers look great as well.

I am just not convinced that RAM is the grail of rendering- it comes down to raw CPU power and a motherboard with an open pipeline. RAM is important of course but more so in the production side.

hrgiger
05-07-2004, 07:13 PM
I'm guessing that FPrime is going to do much better.

Here it is at 1m 10 s.

And just to make it fair, I'm tying one of Fprimes hands behind it's back (gave it two bounces instead of Lightwave's one).:D

hrgiger
05-07-2004, 07:15 PM
Here it is at 4 minutes.

hrgiger
05-07-2004, 07:17 PM
Oops.

hrgiger
05-07-2004, 07:19 PM
Sorry, I think Fprime rendering in the background messed up the file attachement. I paused it. (Another benefit of Fprime) Let's try this again.

hrgiger
05-07-2004, 07:25 PM
And just under 10 minutes...

kmscottmoore
05-07-2004, 07:26 PM
Here is FPrime at about 7 minutes or 23 of its quality passes.

kmscottmoore
05-07-2004, 07:27 PM
A little over 20 minutes/64 quality passes.

theo
05-07-2004, 07:30 PM
HR you crazy guy- we're just benching LW's render not trying to see what is the fastest render on the market

I would use FPrime as well with this particular image- heck with LW's render.

kmscottmoore
05-07-2004, 07:31 PM
This is Lightwave's render.

FPrime doesn't have a built-in timer, so all times are estimates (I'm not patient enough to sit there and watch it)

Notice the shadows in the foreground are much more smoothed in the LW render than in FP. Also, FP doesn't compute the glow effect around the ceiling light.

Oddly enough, the reflection of the ceiling in the largest ball is smoother in FP than the one in LW.

MH9
05-07-2004, 07:33 PM
Dual AMD Opterons Model 242; 1.6ghz
1gb 333mhz ECC Registered SDRAM by Transcend
Tyan Tiger K8W Motherborad

34 minutes.

Interestingly, for all the AMD to Intel comparison ppl, this is the same time mark as Bloontz' dual xeon system, running w/ 2 Xeon 2ghz processors, 2gb ram of ram that's one step slower than mine.

Wheeeeeeeee. :)

L&R,
S9:Micah

hrgiger
05-07-2004, 07:37 PM
And finally, here it is at 15 minutes. I figure another 5-7 minutes and the noise in the shadows will be about gone. So if we could agree on that, I figure I've got you beat by at least 20 minutes if not almost a half hour. (Based on your first render of 51 minutes)

If that would be the case, if you were rendering a 30fps movie of an hour and a half....you just saved your self over 9 years of rendering (If you were foolish enough to render an hour and a half movie on a single machine!).

Hour and a half movie at 30 fps = 162.000 frames.

Time saved per frame (30 min) * 162.000 frames = 4,860,000 min

Divided by 60 gives you 81,000 hours

Divided by 24 gives you 3,375 days

Divided by 365 gives you 9.25 years.

kmscottmoore
05-07-2004, 07:37 PM
This is FPrime at 100 passes (about 32 minutes) We are now way past LW's render time.

hrgiger
05-07-2004, 07:39 PM
Ok kmscottmoore, I was just having some fun with numbers. I'll have to let Fprime go a little more and see if I come up with results like yours.
Also, are you using FPrime preview or are you actually doing an FPrime render. I recall there being a quality differnece between the two. Mine is a render.

kmscottmoore
05-07-2004, 07:40 PM
This is the most frustrating thing about FPrime:
this is what you get after 1 pass (about 20 seconds.) That's right I said 20 seconds.

You are so close, and yet so far. It is painful to sit and wait for the grain to disappear after you see so much in so little time.

theo
05-07-2004, 07:41 PM
If you have an LCD monitor (I am using dual Viewsonic 19's) scroll the image really fast up and down while looking at the ceiling area in the image- the LCD refresh causes a weird, visually hallucinogenic sensation.

theo
05-07-2004, 07:46 PM
Hold on a sec... you can't look directly at the ceiling just use your periphery and let it hold in the area between the ceiling and the walls. The LCD refresh and the obtuse angle of the walls meeting the ceiling creates a cool sort of animated expansion effect.

Don't look too long though you might get sick. Visual Seizures coming up.

hrgiger
05-07-2004, 07:48 PM
kmscottmoore,

My bad, I thought that was your render in the first part of the thread at 51 minutes. Here's Fprime at 23 minutes which is how long your pc render took. Although, like I said, mine is two bounces.

kmscottmoore
05-07-2004, 07:52 PM
And no glow!!

kmscottmoore
05-07-2004, 07:56 PM
hrgiger,

In my experiments 1 vs. 2 bounces doesn't seem to effect FPrime much.
I guess I'm just using the FP previewer, I usually open it when I start working, then zoom it up to full size and let it crank away.

**My wife is yelling at me for playing on the computer on Friday night.**

hrgiger
05-07-2004, 08:03 PM
Actually, I think it's the area light that really is slowing down Fprime. Just out of curiosity, I'm trying it without the area light and just upping the radiosity values.

The glo can easily be faked.

MH9
05-07-2004, 08:09 PM
At 10min and 55 passes, the FPrime render has reached a totally smooth image quality. Perhaps it reached this quality sooner, but it definitly makes it there in 10min, using only 1One of my two 1.6ghz Opterons.

Very nice.

Yet again... wheeeeeeeeeeeee.

L&R,
S9:Micah

kmscottmoore
05-07-2004, 08:34 PM
MH9,

What do you consider "totally smooth image quality"?

If you look at my post above, at 100 passes in FPrime, the shadows under the balls still have some grain in them, as opposed to the LW render.

policarpo
05-07-2004, 08:38 PM
here's my 9m18s tweak on the subject on a 1ghz G4 PowerBook with 768mb of ram.

i need to tweak it some more to improve quality.

:D

MH9
05-07-2004, 08:41 PM
Hey Kmscott,
By "totally smooth", i mean everything, shadows included.

I think it's important to note that i'm using the actual "FPrime Render" button rather than the "FPrime" realtimepreview which you were using, if i'm not mistaken.

The FPrime realtime preview renderer not only has splotchy shadows, but also has the gradient banding on the walls, as can be seen in your post.

I recommend using the "FPrime Render" button. It's much quicker in reaching quality.

L&R,
S9:Micah

MH9
05-07-2004, 08:42 PM
Polycarpo:
Looks like you were using FPrime, is that correct?

L&R,
S9:Micah

GeorgeDittmar
05-07-2004, 08:45 PM
i dont wanna know how long it will take on my machine. man i wish i had a dual machine

MH9
05-07-2004, 08:46 PM
And the quality of the "FPrime Render" button's quality is smoother than HRGiger's 23 minute version. You simply couldn't complain with the image. So, let me get it up here. I had not posted the image because i forgot that this forum doesn't have the same image size limit as CG Talk.

Here it is:

MH9
05-07-2004, 08:47 PM
The banding in this image is caused by the JPEG compression. The actual rendered image is totally clean.

L&R,
S9:Micah

kmscottmoore
05-07-2004, 09:22 PM
Micah,

You don't have radiosity turned on. It must be turned on separately for FPrime in the master plug-ins panel. The settings you want (to match the Lightwave scene) are:

Monte Carlo
1 Bounce
140% intensity

Once you turn that on, you'll get quite a different result. Compare your posted render to mine. Notice how yours is much darker and has no color bleeding?

hrgiger
05-07-2004, 09:49 PM
Yeah, I was about to say that. It renders pretty durn fast when you don't have radiosity turned on in Fprime.

Architook
05-07-2004, 09:52 PM
The biggest quality difference seems to be fprime's lack of the Shading Noise Reduction blur that LW uses. That really helps hide LW's noise a lot! Fprime needs this option too but it's not supported (yet?).

But you can see fprime's basic quality is higher than LightWave by looking at the REFLECTION of the grainy shadow in the flat side of the mirror cube. Notice that it's noticably grainier in LW than it is in fprime.

Please add SNR to fprime! ;)

SLAYER
05-07-2004, 10:28 PM
Here is mine.

Win XP
AMD Athlon 64 3400+
1GB RAM (PC-4000)

meanlebh
05-07-2004, 10:41 PM
all of these f-prime numbers are great...but I thought this post was to compare LW's renderer....is no one out there using a G5??? I would love to see if it is even in the same ballpark as the top of the line xeon processors even without direct LW optimization for it.....

TripD
05-07-2004, 10:47 PM
44 min 2 sec

P4 2.8 gig, Gigabyte 8knxp mobo

1 gig corsair twin ram

looks just like Sandman300's first post.

I gotta get me a dualee for this kinda thing. Time to start saving.

sandman300
05-07-2004, 11:45 PM
Do you think it might be a RAM issue that is making it take as long as it did (I only have 512MB)?

Sometime this next week Im going to try and get this rendered on machine with the Win64 beta OS, see if there is a difference. Ill post as soon as I can.

FPrime looks nice but I dont see any reason to want it, especially with all Ive seen with grainyness and not being able to handle certain effects (glow). Someone should start a thred showing what it does with partical effects, and other volumetrics.

hrgiger
05-08-2004, 04:50 AM
FPrime doesn't do anything with volumetrics. Yet. It's a limitation caused by the SDK of Lightwave which reportedly is being worked on between Worley and Newtek. Hopefully we should see limitations removed in subsequent releases of both softwares.
However, if you haven't experienced the joys of Fprime, you just have to appreciate what it can do. When you start up Fprime, it shows you a preview of your scene almost immediately.Look earlier in this thread where kmscottmoore shows an FPrime render after only 20 seconds. Even with radiosity. Even with multiple bounces you'll get results like that.
Also, how long have we wanted to pause renderings with Lightwave and pick it back up later? This is exactly what Fprime can do.
In a production environment you can render out a low quality animation with FPrime in a fraction of the time it would take in Lightwave. Then you can let it refine itself until it's a finished quality animation.

BTW, it's probably not a RAM that causes your slowdown. The only reason RAM would be causing a drastic slowdown is if your computer was accessing virtual RAM which in the scene above, your computer shouldn't be doing unless you're running other things in the background that's eating up RAM.

gregsduncan
05-08-2004, 05:27 AM
Hi hrgiger / sandman

I have FPrime and agree with what you say - setting up lights, textures etc are so grrreat now - you don't have to do F9s to see what a small light rotation / move has made. And yes, with radiosity you can see an instant (albeit low quality) preview almost immediately. With Lightwave you have to wait, as you have seen here with this example, up to 52 minutes for a single frame. Now if you don't like what that frame looks like, you have to make (normally very small) changes and wait another 52 minutes to see if it's better. Not with FPrime. You can see an almost instant preview.

On your other point, I would say that with Lightwave you can stop rendering and pick it up later if you are exporting TGAs and giving Lightwave a frame range to render. HOWEVER, your second sentence in that para was very good for production purposes. I have just done a very basic, very short animation which took LW about 3 hours to complete. I need to tweak the motion on it and then let it bake another 3 hours! Now FPrime creates an initial "low res" image for each frame in an animation and then goes back over those images and refines them. If I use FPrime to render my animation, I can see a "low res" version first where it only takes about 10 seconds per frame to render (total 20 minutes). If I don't like it I've only used 20 minutes. BUT if I DO like it, FPrime can just take these low res images and put them through another pass to refine them.

FPrime is really a great tool.

----------------------

Anyway, what do people think of FPrime as a final renderer? Does it seem (forgetting it can't do vols / shaders) that it is quicker as a renderer? How do you know the number of passes it has done is as good as what LW would do?

gregsduncan
05-08-2004, 05:31 AM
Hey hrgiger

Why not make your signature with hyperlinks so ppl can just click to go to your tuts?

Copy and paste this to your sig, but take out all the **s.

[**URL=http://giger3d.com]http://giger3d.com[**/URL]

My tutorials:

[**URL=http://giger3d.com/Tutorial_Menu.htm]http://giger3d.com/Tutorial_Menu.htm[**/URL]

Aegis
05-08-2004, 06:39 AM
23m 19s on my self-built dual 3.06Ghz Xeon - not too shabby :)

hrgiger
05-08-2004, 08:01 AM
Done. Thanks Greg.

kmscottmoore
05-08-2004, 09:38 AM
Just for fun, I loaded the scene in 7 last night and let it render. To my suprise, 7 actually rendered a few seconds faster than 8. (23m 27s) for 8 vs. (23m 12s) for 7

Now, keep in mind, I have rendered the same scene on the same machine two or three times in a row, and the time has varied by a much as 5-6 seconds (on a long render) so you can't put too much emphasis on a few seconds difference. But, at least with this scene, there's no dramatic render difference between 7 and 8.

Sandman, your problem is not RAM.(although for saving larger files, more wouldn't hurt) All of the dual processor machines show basically half of your render time. The 64 bit OS won't help Lightwave, you need a 64 bit Lightwave.

Also notice that Aegis' reported render time is nearly identical to mine (dual 3.06 vs. dual 2.8) According to clock speed, he should be about 10% faster than me. You need a much bigger jump in clock speed to make the price/performance tradeoff worth it.

hrgiger
05-08-2004, 10:14 AM
I did all the benchmarks in the LW 7 folder to compare LW 7 and LW 8. LW 7 was faster on all of them if I recall, sometimes by as much as 10-20 seconds. To make sure I was getting an honest render, I rebooted before each one. I'm not sure who at Newtek added faster rendering as a LW 8 feature but they were surely smoking crack that day. Not that it matters to me. I use FPrime most of the time now. Hopefully the sdk problem will be resolved and then I can use it all the time.

kmscottmoore
05-08-2004, 10:58 AM
Ok, this one's from my Mac. 44m 36s

Dual G4 1.25 Ghz
1.25 GB RAM

Pretty respectable compared to the single processor systems with higher clock speed, but obviously no match for the Dual Xeons.

I don't have a G5 :(
C'mon guys, where are the rest of the Mac render times?

TSpyrison
05-08-2004, 11:03 AM
Dell Inspiron XPS notebook computer
3.4 Ghz, 1 gig Ram

And I can take it wherever I go :)

NigelH
05-08-2004, 11:17 AM
OK, Ok, here you go:

Dual 2Ghz G5 2.5Gb Ram
LW[8]
25m 47sec

As for FPrime, also keep in mind that it is not multithreaded (as yet), so it runs on only one processor.

Aegis
05-08-2004, 12:07 PM
23m 2s on LightWave 7.5c (dual 3.06Ghz)

http://www.agmorgan.pwp.blueyonder.co.uk/Radiosity_BOX.png

NigelH
05-08-2004, 12:34 PM
Another thing to consider when rendering with FPrime, is that it seems more akin to 'enhanced' rendering with the LW renderer. That is to say, that the entire image is refined with each pass as opposed to just the edge antialiasing. I'm guessing this is why the reflections in the balls are smoother than in thre LW renders. Since the scene is set to 'Low' antialiasing (as opposed to 'enhanced low'), only the edges are refined on subsequent passes. maybe setting the scene to 'enhanced low' AA and 1 thread in LW would be a fairer test of pure rendering power between the 2 engines.

Of course, hobbling the LW renderer is admittedly pointless. I suppose my point is simply to look at where FPrime is starting from and to expect great things from it in the future.

gregsduncan
05-08-2004, 06:31 PM
Pls could you give any thoughts on this:

For the price of a Dual Xeon ..... ??? .... computer you could buy even 4 cheaper computer but with a fairly high spec in terms of processor.

I'm not talking about the modelling and scene setups, but purely for rendering, what would you rather have :

a) One Dual Xeon computer
b) 4 cheaper, say single 2.8/3.0GHz 512MB RAM machines as a small render farm.

sandman300
05-08-2004, 08:32 PM
gregsduncan and friends of Fprime

I dont mean to be rude but could you start threds for your topics. This one has a topic. thankyou.

theo
05-08-2004, 10:32 PM
Yeah- poor Sandman's gettin' sandbagged here. Leave FPrime for another thread guys. It gets a little thick in a thread every time the words "LW render" are mentioned- all the FPrimers come out of the woodwork as if we don't already get the fact the FPrime is slightly faster at some things.

Original1
05-09-2004, 02:06 AM
Originally posted by sandman300
gregsduncan and friends of Fprime

I dont mean to be rude but could you start threds for your topics. This one has a topic. thankyou.

Interested in G5 benchmarks in particular.

gregsduncan
05-09-2004, 03:47 AM
Sorry Mr Sandman. I would post a comparison but don't have 8 yet :( Should have in next couple of days. Will post then.

cagey5
05-09-2004, 04:21 AM
2 days? Glad to see you've not lost your sense of optimism then, despite all the delays. ;)

MH9
05-09-2004, 12:48 PM
Kmscott:

O_o

[smacks fore head]

I'll be back with the FPrime set up with radiosity, haha.

L&R,
S9:Micah

SLAYER
05-09-2004, 08:03 PM
WTF?

I have posted here my test of this scene which was 46min, 44sec with 1GB Ram.

Today I added another 512MB card of the exact same RAM so I now have 1.5GB.

After this, This scene renders at 47min, 17sec. I was expecting the render time to be slightly faster, but certainly not longer.

Any ideas on why this may be?

Thanks

wacom
05-09-2004, 09:31 PM
Hey for all you FPrime Benchmark testers- don't forget to triple that LW logo- it's an NGon cancer child- and we know what that can do to good'ol FPrime...

OK...go into modeler- the walls and the logo are one! SUPER NGON CANCER! YIKES!

Oh NO! I just realized that all the objects are on the same layer too! CRAP!

This must be a FPrime killer on purpose...

wacom
05-09-2004, 11:55 PM
Here is what I got in Fprime after about 21 min on a p4 1.8 w/ 768MB RDRAM while surfing on the web. I re-worked the geometry of the scene to be "FPrime" able. Shading noise reduction would help sooo much.

wacom
05-10-2004, 12:01 AM
Here is the re-worked file in case anybody cares to take a look at it/use. Every thing has its own layer, there are no N-Gons etc.
I have a feeling that if the LW logo was an image map instead of a 318 polygon object it would help...

PS- Sorry to those who were wanting just a LW render test...

silverlw
05-10-2004, 07:59 AM
Stop wining and download NEAT IMAGE (http://www.neatimage.com/) hehe
No seriously even if neatimage is a damn fine noisereduction tool it should still be implemented in Fprime and im sure Worley is working on it already. Here is a fine example of what to expect with noisereduction. Vray compared to Fprime (http://vbulletin.newtek.com/showthread.php?threadid=22337&highlight=fprime)

TSpyrison
05-10-2004, 08:59 AM
Here is my work machine, a 2.6 Ghz P4 (running at 2.8 Ghz) and 1 gig of memory

I like my notebook better :)

silverlw
05-10-2004, 10:20 AM
Fprime 60 seconds refienemnt 2 and same image noisereduced
http://hem.bredband.net/b223277/[email protected] http://hem.bredband.net/b223277/[email protected]

wacom
05-10-2004, 11:18 AM
Originally posted by silverlw
Stop wining and download NEAT IMAGE (http://www.neatimage.com/) hehe
No seriously even if neatimage is a damn fine noisereduction tool it should still be implemented in Fprime and im sure Worley is working on it already. Here is a fine example of what to expect with noisereduction. Vray compared to Fprime (http://vbulletin.newtek.com/showthread.php?threadid=22337&highlight=fprime)

Yeah- I tried using PSP to reduce the noise after four min. and it looked real good. Most programs these days let you run scripts too so that you can just automate the process if you have an animation. I just thought that If I posted a "noise reduced" FPrime shot a buch of people would say it's "cheating". I'm with you though, and when FPrime has noise reduction I'm sure it will be a better implementation of it than the native LW version. I hope worley also makes it adjustable...

Did you render using a modified version of the BOX file? The origional is loaded with FPrime killers... If you didn't please try out the moded version I made and see if there is a speed change for you...at least in the Preview window...

gregsduncan
05-14-2004, 03:49 PM
Got 8 today so as I said, I'm posting my results.

Slowest computer so far :( Only a laptop tho - not my desktop.

Sony Vaio, P4 2.8GHz, 512MB, NVIDIA GeForce FX Go 5600.

Time to upgrade I think :)

gregsduncan
05-14-2004, 03:55 PM
To go O.T. again, here is the same in FPrime after ONLY 90 SECONDS, but with noise reduction applies in PSP. Total time taken 93 SECONDS!!!

Can y'all see now the benefit of FPrime??????? :p

I know it's not like the final image but you sure get a good idea of how it looks in very quick time....

sandman300
05-14-2004, 05:00 PM
for the price, Id rather wait the time it takes for Lightwave own render engine to cook. I think $400 would be better spent on hardware upgrades. Perhapse when whorley makes an fprime that can do everything that LW's stock renderer can do (and more, quicker), I might think about it.

wacom
05-14-2004, 05:06 PM
Originally posted by sandman300
for the price, Id rather wait the time it takes for Lightwave own render engine to cook. I think $400 would be better spent on hardware upgrades. Perhapse when whorley makes an fprime that can do everything that LW's stock renderer can do (and more, quicker), I might think about it.

Well if you use netrendering you may have a point, but that's not including the part of FPrime that makes it sooo much easier to set up most scenes and do texture work. Maybe I'm just experienced enough to work blindly though...

I really hope Worley will make a demo. That will show people the other 50% of it that they don't get to see in the final renders.

Lewis
05-14-2004, 05:44 PM
Hi !

Interesting render times guys.

I'm just typing to explain to wacom why scene is made that way it is (not too friendly to FPRIME). When i was making that scene FPRIMe wasn't even announced so i couldn't know that he would prefer triangles and separate layers and all that stuff. That's just current limitation of FPRIME (which is sure great but have it's pros and cons) but AFAIK FRPIME isn't topic of this subject ;).

BTW scene was finsihed on 31st May 2003.

amorano
05-15-2004, 08:41 PM
Is there any reason you guys are not using chris blanos's site

www.blanos.com

for benchmark results? Maybe if someone asked him to also support Fprime now it would be a nice addition.

Elmar Moelzer
05-16-2004, 01:31 AM
Hey!
Well, if it is allowed to tweak the scene a bit...
Here is mine:
http://www.mediastudio-graz.com/images/rad_test1.jpg
Rendertimes in LW: 25 minutes 53 seconds (and I had all sorts of sh*t open in the BG, was browsing the web, checking emails, etc).
This is with full AA, not Adaptive sampling, but I reduced the amount of samples quite a bit in return (2x6 rays/evaluation) and I used an Area- light- quality of 3 instead of 4.
System: P4 2.4 Ghz
1024 MB Ram
The quality is a bit lower than the one of the original scene, but not much and it is wayyyy faster...
Not that bad for the good old LW IMHO...
Thing is, that people always do their test- renderings without AA (or Adaptive Sampling) forgetting, that you get 5 times the amount of samples with Low AA and Adaptive Sampling Off (9 times the samples with medium AA, etc).
Here is another rendering that uses the noise- reduction of Digital Fusion instead of LWs. Note, that here the noise seen in the reflections is reduced as well.
http://www.mediastudio-graz.com/images/rad_test_DF.jpg
CU
Elmar

silverlw
05-16-2004, 06:23 PM
Does digital fusion have noisereduction? How does it work? gaussianblur?

kmscottmoore
05-16-2004, 07:47 PM
Sandman,

I'm sorry your thread got hi-jacked. I feel responsible because I was responding to Wizcraker's over-enthusiastic statement about the abilities of F-Prime. I just wanted to show realistically what could be expected. I do think that the comparision of the same machine in LW7 and LW8 was a valid and worthwhile comparison. To make it up to you, I complied all the results and sorted them, fastest to slowest.

Policarpo, wacom, silver lw, and Elmar Moelzer, your results were not included because you modified the scene in one way or another. It is only a valid comparison if we all use the same scene.

amorano, Mr. Blanos has not updated his site to include the new scenes that came with version 8 of Lightwave. Currently, it is not possible to post these results there.

Ladies and Gentlemen, the results are:

theo
Dual 2.8 Ghz Xeon
LW 8
22m 36s

Aegis
Dual 3.06 Ghz
LW 7
23m 2s

Aegis
Dual 3.06 Ghz Xeon
LW 8
23m 19s

KM Scott Moore
Dual 2.8 Ghz Xeon
LW 7
23m 12s

KM Scott Moore
Dual 2.8 Ghz Xeon
LW 8
23m 27s

Nigel H
Dual 2 Ghz G5
LW 8
25m 47s

bloontz
Dual 2Ghz Xeon
LW 8 (?)
34m 1s

MH9
Dual 1.6 Ghz Opteron
LW 8
34m ?s

TSpyrison
3.4 Ghz P4 (notebook)
LW 8
36m 46s

Trip D
P4 2.8 Ghz
LW8
44 m 2s

KM Scott Moore
Dual 1.25 Ghz G4
LW 7
44m 36s

Tspyrison
2.6 Ghz P4 (overclocked to 2.8 Ghz)
LW 8
46m 8s

Slayer
Athlon 64 3400+
LW 8
46m 44s

Sandman
Athlon 64 3200+
LW 8
51m 3s

Gregsduncan
2.8 Ghz P4 (Vaio notebook)
LW 8
63m 40s


So far, it looks as if theo is King of the Hill.

These results show that there is more to consider than clock speed alone when purchasing a system (or building your own.); however, when it comes to rendering, dual processors are definitley the way to go. I tried to note whenever someone was using a notebook, because generally the front side bus on a notebook is slower than a desktop.

Also, if I were buying a new system today, I think I would skip the 64 bit processors. There just isn't enough software support right now to justify the expense. Maybe in another year. . . ?
Of course, you übergeeks who have to have the latest and greatest will buy them anyway, just to be the first kid on the block to have it.

lots
05-18-2004, 03:57 AM
Just something i'd like to note..

While LW has been "Optimized" for Intel chips, it was more specifically optimized for the SSE2 instruction set. Thus on a Pentium 4, when compared to any Athlon XP, the Pentium 4 would always win. However the Athlon 64/Opteron are different. They also include the SSE2 instruction set, and will feature SSE3 in the next core release (Prescott, the new Pentium4 already supports it) Thus the Athlon64/Opteron are in much better positions to compete in SSE2 optimized programs like LW. Granted, from this thread, the Pentium 4/Xeon's main advantage comes from its greater clock speed when compared to the AMD chips. Which, as said above my post, the best thing in rendering is number of cpus and how much MHz they have.

mattclary
05-18-2004, 05:26 AM
Given both chips have SSE2 instructions, the next deciding factor seems to come down to raw cycles per second.

Lightwolf
05-18-2004, 06:04 AM
Well, Opterons perform SSE2 as fast as the P4 _per cycle_, so the current P4s outperform the Opterons in that respect...
On the other hand, the Opterons have a much stronger FPU (as had the Athlons), so you'll see them scream in non-SSE2 optimized apps (or non optimized parts of apps, like for example volumetrics in LW).
Cheers,
Mike

caesar
05-18-2004, 08:54 AM
What kind of things in LW uses FPU besides volumetrics? HDR? HV? Particle collisions? Cloth FX?

NigelH
05-18-2004, 10:04 AM
I feel I need to point out (though I can't claim to be unbiased), that the Dual 2ghz G5 did very well in comparison to the Dual 2.8 and 3 Ghz Xeons - based on clock speed alone (averaging about 3-4 minutes slower, but 8 minutes faster than the dual 2 Ghz Xeon). Particularly considering that there in NO significant optimization for the G5 processor.

OK, I know I paid $3000 for it, but it was worth every dime to me :-).

Chris
05-18-2004, 10:30 AM
I haven't had a chance to go through the new benchmark scenes for LW8.

What additional scenes would you like to see up on my site - I can have them in place in a few minutes.

- Chris

Lightwolf
05-18-2004, 10:55 AM
Originally posted by caesar
What kind of things in LW uses FPU besides volumetrics? HDR? HV? Particle collisions? Cloth FX?
Just about everything, more or less. I mean, heck we're talking 3D here, and there is tons of floating point math involved....
So: Yes, yes, yes, yes, yes... IK, bones, shading, viewport redraws... :)
NigelH: Nowadays GHz can really only be used to compare between processors of the same brand and family, everything else is so dependant on th architecture that the clocking is just about meaningless (for CPU performance). The SSE2 example mentioned above is one of the few exceptions here...

Cheers,
Mike

caesar
05-18-2004, 01:04 PM
okey dokey

NigelH
05-18-2004, 01:38 PM
Originally posted by Lightwolf

NigelH: Nowadays GHz can really only be used to compare between processors of the same brand and family, everything else is so dependant on th architecture that the clocking is just about meaningless (for CPU performance).

My point exactly. :)

Lewis
05-18-2004, 03:32 PM
Originally posted by NigelH
I feel I need to point out (though I can't claim to be unbiased), that the Dual 2ghz G5 did very well in comparison to the Dual 2.8 and 3 Ghz Xeons - based on clock speed alone (averaging about 3-4 minutes slower, but 8 minutes faster than the dual 2 Ghz Xeon). Particularly considering that there in NO significant optimization for the G5 processor.

OK, I know I paid $3000 for it, but it was worth every dime to me :-).

Generaly I agree in this MHz - speed not importan ratio but i must say that LW do HAVE optimizations for AMD, Intel and MAC. LW includes MMX, 3Dnow+, SSE and from Lw 7.5b it includes SSE2 and AltiVec for MAC so it's not true that it's optimzed for P4 only :).

Mac is a very nice thing but comparing price and performance it doesn't have chance against Intel/AMD machines :(.

NigelH
05-18-2004, 03:57 PM
Originally posted by Lewis
Generaly I agree in this MHz - speed not importan ratio but i must say that LW do HAVE optimizations for AMD, Intel and MAC. LW includes MMX, 3Dnow+, SSE and from Lw 7.5b it includes SSE2 and AltiVec for MAC so it's not true that it's optimzed for P4 only :).

Mac is a very nice thing but comparing price and performance it doesn't have chance against Intel/AMD machines :(.

Since The G5 is manufactured by IBM, I don''t think it benefits anything from Ativec at all. The G4 does, since it was manufactured by Motorola. At least that's my understanding. Please correct me if I'm mistaken.

As far as price/performance, I've already said all I'm going to say.

CaptainKirk
05-18-2004, 04:09 PM
IBM itself uses AMD Opterons in it's dual workstations like Intellistation A Pro.

They do not use G5 chips despite the fact that they make them, because AMD chips are faster. It's just common sense ( something Mac people generally lack ), G5 is not really competitive with the best Intel or AMD have to offer, and they look even worse when you consider the price ( you can have a dual Opteron system , 2.4 ghz each, for less than a dual 2ghz G5 system , and Opterons were faster across the board even when they were running at only 1.8 or 2ghz where Macs are still stuck a year after announcing they'd have a 3Ghz machine in a year ).

My questions for Steve Jobas are again:

Where are these 3ghz dualies which you announced last year?

How is it possible that, as you say, with G5 you not only Caught up with PCs, but surpassed them, when all along you were telling everybody G4s were faster than the fastest PCs.

How exactly do you catch up and surpass when according to you, you are always ahead.

Do you ever stop lying?? Everybody does it, but I have never seen anybody quite like you Steve.

It's crap like this and total disregard for common sense + their constant blaming of everybody else for their problems ( New Tek, this is not optimized, that is not optimized, etc.......) by most Mac users that just gives you no choice but hate them with a passion.

kmscottmoore
05-18-2004, 04:16 PM
Lightwolf's statement about comparing clock speeds in different architectures is true, but it has always been true (not just with the latest multiple Ghz machines.) It also is impractical to compare the clock speeds of generations of the same family of processors. If anyone remembers, the first generation of Pentium 4's were actually SLOWER than Pentium III's running at the same clock speed.

To defend Nigel, the G5 actually did quite well. There were several other improvements, beside clock speed, in the G5 over the G4 machines. The G4 really got a bad rap because Motorola wasn't able to keep up with Intel in terms of clock speed. But, one of the more significant problems had to do with the front side bus. Basically, there were a lot of wasted clock cycles because the processors weren't getting data fed into them fast enough.

Lewis' statement about when you compare price/performance of a Mac and a PC is horribly misinformed. Notice that I'm the only one who posted BOTH PC and Mac rendertimes? I have also commented in this thread:
http://vbulletin.newtek.com/showthread.php?s=&threadid=21304

I actually paid a little more for my PC than Nigel paid for his G5, and I didn't get a dvd burner or any of the excellent apps that come with every new Mac. (Luckily I had most of that stuff with my G4)

If I were starting out from scratch as an independent artist, and I was buying only one machine I would still go with the Mac as my first choice.

kmscottmoore
05-18-2004, 04:21 PM
Nigel,

The first Power PC chip was actually co-designed by IBM and Motorola. The G5s still have the Altivec. Altivec is the brand name for a sub-processor, that for lack of a better term, pre-crunches some of the data before it gets to the main processor.

The problem was that very few software companies made thier products take advantage of the scheme.

kmscottmoore
05-18-2004, 04:26 PM
Captain Kirk,

Actually, the G5 is the same as IBM's Power4 processor, which they use in their high-end servers.

As for your statement that they don't use the G5 because AMD is faster: most IBM desktops and Laptops use Intel chips, not AMD.

Like I said in my previous discussion on the topic, sit down and seriously use a Mac for at least a year before you go off ranting and raving about what they can and can't do, and what bone-heads Mac users are.

wacom
05-18-2004, 04:28 PM
I know this is all good info, but if the FPrime posts were OT then this stuff is way OT. Lets keep to the render time testing...no? We don't need another Chip Vs. this chip debate war waged here. If the render time says it's slower, it's slower.

CaptainKirk
05-18-2004, 05:27 PM
Actually, the G5 is the same as IBM's Power4 processor, which they use in their high-end servers.

Not even close.

Celeron is closer to a Xeon than a G5 chip is to the real deal Power 4.

It's just based on it, G5 is a severely watered down version. Dual G5s would cost 3 times as much if those were the real Power 4 processors.

NigelH
05-18-2004, 05:52 PM
Originally posted by CaptainKirk
It's crap like this and total disregard for common sense + their constant blaming of everybody else for their problems ( New Tek, this is not optimized, that is not optimized, etc.......) by most Mac users that just gives you no choice but hate them with a passion.

Issues? :p

Lightwolf
05-19-2004, 02:41 AM
Originally posted by wacom
I know this is all good info, but if the FPrime posts were OT then this stuff is way OT. Lets keep to the render time testing...no? We don't need another Chip Vs. this chip debate war waged here. If the render time says it's slower, it's slower.
Hi wacom...
the thing is, CPU architectures actually come into play when you compare rendertimes. Just look at x86, AMD vs. intel benchmarks, where either chip design has faster render times depending on the scene.
So, if you do loads of radiosity and raytracing, go for intel, if you do more FX stuff (volumetrics) go for AMD.
Cheers,
Mike

Lightwolf
05-19-2004, 02:45 AM
Originally posted by CaptainKirk
...
It's crap like this and total disregard for common sense + their constant blaming of everybody else for their problems
...

It is crap like this that totally degrades decent discussions, gets people banned and is generally responsible for a hostile atmosphere in these forums.
I think a bit of courtesy hasn't hurt anyone yet.
Cheers,
Mike

caesar
05-19-2004, 12:03 PM
Originally posted by Lightwolf
It is crap like this that totally degrades decent discussions, gets people banned and is generally responsible for a hostile atmosphere in these forums.
I think a bit of courtesy hasn't hurt anyone yet.
Cheers,
Mike

Captain Kirk´s posts are not so polite...neither wise....
IBM has Intel, Power4,5 and AMD machines...guess what?! The #3 in supercomputer are made of "weaks watered down version of Power4" G5 systems out -of -shelves, each processor reaching ONLY 8 Gigaflop (the same top performance delivered by Itanium 2 and Nec´s EarthSim processors) at less than 10% of their price...c´mon !

Lewis
05-19-2004, 02:43 PM
Originally posted by kmscottmoore

Lewis' statement about when you compare price/performance of a Mac and a PC is horribly misinformed. Notice that I'm the only one who posted BOTH PC and Mac rendertimes?

I actually paid a little more for my PC than Nigel paid for his G5, and I didn't get a dvd burner or any of the excellent apps that come with every new Mac. (Luckily I had most of that stuff with my G4)

If I were starting out from scratch as an independent artist, and I was buying only one machine I would still go with the Mac as my first choice.

Hi kmscottmoore !

You can stay with Mac it's your right and i generaly like MAC very much (I'm old Motorola (amiga) User so i know how good is to work on stable MC 68040 machine :). BUT (there is always but) you really can't convince me that you can buy cheaper MAc than PC in same speed. I myself built hundred of PC based Machines (worked in PC service) and i never buyed premade branded machine (except Notebooks). I always pick my MB, RAm, GFx, HDD ... and all components by myself so i have much broader choice. You simply can't do that with MAC (what have some Pros but Cons also) so price is always higher. You can buy P4 MBO (single) for less than 90$ or you can buy best single MB for 200 US$. It's just personal choice and if you know how to pick components both machines can be very stable no matter how big differnece is (More expensibe MB does have more gadgets but maybe you don't want them so you buy cheaper version).

Also Single HT Prescott on 3.4 Ghz on good MB can compete with Dual Macs so i don't see any price/performance advantage on Mac. The Fact that you payed more for your current PC it's just "wrong" choice and I'm sure i can always build cheaper PC in same performace range than MAc since i loooked for Mac to be my second Machine but i ended with another PC and saved cash for brand new iiyama 19" monitor.

Don't get me wrong . I like MAc and i don't search for any PC-MAC war but facts are there.

Sorry for BIG OT guys :).

kmmscottmoore - I'm glad you included Mac benchmark and it would be good for Mr. Blanos to iclude that scene on WEB when it's ready :).

kmscottmoore
05-19-2004, 04:21 PM
Originally posted by Lewis
you really can't convince me that you can buy cheaper MAc than PC in same speed.

Attached is the invoice for my PC. I paid $500 more than Nigel did for his G5.

The render times:

KM Scott Moore
Dual 2.8 Ghz Xeon
LW 8
23m 27s

Nigel H
Dual 2 Ghz G5
LW 8
25m 47s

The PC was faster, but not end of the world faster. The configurations are different, but, there are some things that Nigel got with the G5 that I didn't get:
faster RAM (although less of it), DVD burner, better software bundle etc.


Originally posted by Lewis

Also Single HT Prescott on 3.4 Ghz on good MB can compete with Dual Macs so i don't see any price/performance advantage on Mac.


No one posted a single processor render time that was anywhere close to the G5. Particularly not the AMD processors that Captain Kirk seems so fond of.


Originally posted by Lewis
The Fact that you payed more for your current PC it's just "wrong" choice and I'm sure i can always build cheaper PC in same performace range than MAc

I'm sure that Eric at Boxx will be glad to know that I made the "wrong" choice, and that you will be putting him out of business with your Croatian built PCs.

kmscottmoore
05-19-2004, 04:23 PM
Poor Sandman, all he wanted was a benchmark thread. :(

CaptainKirk
05-19-2004, 05:27 PM
Attached is the invoice for my PC. I paid $500 more than Nigel did for his G5.


Why don't you double the RAM in G5 and stick that Quadro V. Card in a Mac as well, add another hard drive to match the Boxx. See how much more Mac is then.

You are comparing a machine with twice as much RAM and a $600 video card to a Mac with the basic $125 v.card.

Similarly configured G5 will be over $4000, so there is no way Macs are a reasonable man's alternative.

And , yes you did pay too much. To be fair, I don't know when you bought it, but a dual 2.8 Xeon machine like yours, today with a good video card and 1MB of RAM can be had for under $2200.
And I'm talking about a better case, power supply, Raptor 10 000 rpm drive, etc......... better components than those you'll find in a Mac, Dell, or that Boxx ( except maybe a v.card, because that is overkill, a $150 NVidia 5700 will do just as well )

kmscottmoore
05-19-2004, 05:36 PM
CaptainKirk,

why is it that on a benchmark thread, you haven't posted a single render result?

I checked your other posts, all you do is talk about what COULD be built and then you ***** about Macs.

In fact, you have never posted a single screenshot, render, technical tip or anything worthwhile.

15 year old kids with cracked copies of Lightwave should stay off these boards.

CaptainKirk
05-19-2004, 05:50 PM
How is any of that relevant to the fact that you paid too much and that Macs are slower while costing more?

Stay on the subject. And I'm probably older than you with a legitimate copy of LW8 and DFX+


each processor reaching ONLY 8 Gigaflop (the same top performance delivered by Itanium 2 and Nec´s EarthSim processors) at less than 10% of their price...c´mon !

Those gigaflop numbers Apple uses are theoretical numbers.

It's like me saying that in theory I could melt a LW8 CD , drink it, then stick a Pentium 4 up my ***, and then render complex scenes by simply blinking my eyes.

Obviously my statement is just as ridiculous as Apple's performance claims.

kmscottmoore
05-19-2004, 06:00 PM
So where is YOUR render time?

sandman300
05-19-2004, 06:12 PM
'sOK kmscottmoore, the thread got me the info that I was looking for. the comparison between hardware was a good comparison as a real world benchmark. I often wonder when you see benchmarks on websites or in articles if the computers that were tested were real peoples computers with comparible hardrives and other parts.

As far as my PC's preformance I was a little disapointed by comparison, I thought "dam if only I had known before I bought the hardware" but then I looked to see what I would have gotten (a dual opteron) and I would likely have to spend around $3000 more.

As far as optimization, I found out today that the Athalon64's are optimized for SSE2 so that is good.

As far as optimizing for the 64bit OS, the beta version still has a lot of bugs in it. If NT is going to work with it (this is my own speculation) they likely wont start any development untill the official release of the system.

All in all Im happy with my Athalon64.

Perhapse someone might know if there are any bios tweeks or anything else that you can do to an existing system (aside from upgrading) that will improve performance?

as for the rest, take what you want and ignore the rest.

wacom
05-20-2004, 01:08 AM
Originally posted by Lightwolf
Hi wacom...
the thing is, CPU architectures actually come into play when you compare rendertimes. Just look at x86, AMD vs. intel benchmarks, where either chip design has faster render times depending on the scene.
So, if you do loads of radiosity and raytracing, go for intel, if you do more FX stuff (volumetrics) go for AMD.
Cheers,
Mike

I know...I know...

I don't really have a problem with it...but I know some people take it personally and I just didn't want the posting of render times to stop due to the thread degrading into a Chip Vs. Chip debate. Oh well...here we go again...

caesar
05-20-2004, 10:38 AM
Originally posted by CaptainKirk


Those gigaflop numbers Apple uses are theoretical numbers.
...
Obviously my statement is just as ridiculous as Apple's performance claims.

Wow, Apple is good at "talking", theirs "theoretical numbers" put the VirviniaTech Cluster in #3 - http://www.top500.org/lists/2003/11/3/!
And those "ridiculous performance claims" can be viewed here - http://computing.vt.edu/research_computing/terascale/pressrelease.html

And the 8 GFlop claim can be viewed here from the the Virginia´s Computer Eng. Chief Dr. Srinidhi Varadarajan -The theoretical limit of a dual-processor Power Mac G5 is 16 gigaflops, or 8 gigaflops per processor. He was getting 95 percent efficiency on a single processor, which was amazing for the short amount of time he had. using BLAS benchmark http://www.apple.com/education/science/profiles/vatech/optimization.html

Please CaptainKirk talk with facts, show the numbers and we´ll read, analyse and possibly agree with you. Your personal opinion can´t change anything about what processor is better for what.

Anyway, this thread has good info ( from the people who collaborate) for LW performance in a lot of processors, Ill think twice when upgrading!

CaptainKirk
05-20-2004, 11:52 AM
or 8 gigaflops per processor.

I do talk with facts, as your own links confirm.

8 Gigaflops is just THEORETICAL performance of the chip.


Besides than you gotta be insane to believe anything you read on Apple.com

The facts are G5 is slow compared to AMD or Intel chips.

There was just yesterday another article finally testing Dual G5 against dual Opterons and Xeons in videoediting environment. In every single test ( mostly various aftereffects and video editing related stuff ) G5 was between 20% and 100% slower. We know it's slower in photo editing and we know it's slower in ALL 3D application performance. And they weren't even using the latest Opterons which are another 20% faster.

So why exactly would you want to buy it?

kmscottmoore
05-20-2004, 12:32 PM
CaptainKirk STILL hasn't posted a render time, and he is still bitching about Macs.

I'm setting my phaser on IGNORE.

Dickigeeza
05-20-2004, 01:06 PM
My system is based on dual 3 gig xeons, with 2 gigs of ram. 8 threads. asus pc delux mobo.

This 1st render is on version 7.5c.

Dickigeeza
05-20-2004, 01:08 PM
This is the same machine, all settings the same but now version 8.

fully 10 seconds longer.

I wonder why that is.

Rich

electropulse
05-20-2004, 01:23 PM
3.0 P4 w/1 gig crucial pc-3200 ram

electropulse
05-20-2004, 01:25 PM
Originally posted by electropulse
3.0 P4 w/1 gig crucial pc-3200 ram

Dickigeeza
05-20-2004, 01:26 PM
The system was a custom job from these guys - who I can throughly recommend.

www.armari.co.uk

Sorry for this plug, I have no commercial tie up whatso ever.

Rgds Rich

Meshbuilder
05-20-2004, 02:09 PM
Often people buy their computers from cheap PC-store. But then they take the risk of having troubles with the computer later because the store have put together the computer with cheapest components. Sometimes it works great and sometimes it doesn’t.
You can have the same problems when you build your own PC and don’t know much about computers.

That’s why “many” people buy their computers from big companies like DELL, HP, Compaq etc. And what does a fast workstation cost from say DELL and BOXX?

I have really tried to make the computers as similar with graphic card, ram, DVD, Harddrive ect.


Dell Precision Workstation 650
Dual Intel Xeon 2,8 GHz
Nvidia Quadro NVS 280, 64 MB dualview
(I have this “cheap” graphic card at work and it’s not faster than Ati 9800 Pro)
1 GB Ram
DVD / CD Burner
120 GB HD
Keyboard & Mouse
56k Modem
PC Speakers
Wireless network
Windows XP
$ 3,870


Ok let’s buy a similar machine from the super workstation company BOXX.

3D BOXX 8106
Dual Intel Xeon 2,8 GHz
Ati Radeon 9600XT
(I don’t know if this is faster or slower than Ati 9800)
I GB Ram
DVD / CD Burner
120 GB SATA
Keyboard & Mouse
Windows XP Pro
$ 4,334


Ok now let’s see what the super expensive PowerMac from Apple own store cost?

PowerMac G5
Dual IBM G5 2,0 GHz
Ari Radeon 9800 Pro
I GB Ram
DVD / CD Burner
160 GB SATA
Keyboard & Mouse
FireWire
Wireless network
Mac OS X
$ 3,598


The Mac isn’t that expensive when you compare it to workstations from this companies.
And none of the other PC run Mac OS X ;)

I know many of you can build a great working "cheap" PC, but not all people can do that.

wacom
05-20-2004, 03:15 PM
OK here's my 2cents.

Render times are render times. If you want really fast render times then invest in buying several computers, or just send stuff out to respower or some other network render farm. It costs hardly anything to send your stuff to a render farm in the end.

The cost of a PC can be very cheap and it can be very fast as long as you do your research. Going to the main Dell site to get pricing is only for fools. Do you know how over priced simple things are like a DVD burner or larger HD is from sites like this? Try going to their preowned systems section and get a system. Many are dual proc systems...

http://outlet.us.dell.com/Dispatcher?target=InventoryPage&action=filter&lob=PREC&unique=1085086897671&sessionID=Atcw1aO8!-2079325356!1687085411!1085086896749&tgtSeg=I

Go somewhere else andget more RAM, a DVD burner, and tell me it's not cheaper than what some of your are posting. They even come with PRO 3D video cards.

If you have a Mac, you could still make a render farm for fairly cheap that used bare bone AMD based systems. All you need is a cheap, small harddrive, 512MB or more of cheap RAM, a CPU and a ten dollar network card. Add a few $30 Dlink hubs and routers and some cable and for 2000-3000 bucks you've got a nice little render farm of five to eight systems running at around 2+ Ghz each. Hell, you could even buy a bunch of cheap laptops and network them to save on space and still get a nice setup going for as much as these new fancy pants systems are costing you. Get off the hotrod mentality and get into the affordable commuter mode of thinking. Lightwave has free render nodes because it's so slow at most rendering they almost expect you to take advantage of it.

I have a Dell 8200 dimension P4 1.8 with 768 MB of RDRAM and my system still smoked many others because of FPrime. I payed 279 bucks for FPrime (and had a little help from neat image)- and it looks like I'd need to buy two dual systems to get the same resaults for this test (which is hard on FPrime BTW). I was going to buy a new system, but I decided to go with FPrime because I'd rather support what I and others really need: faster rendering for Lightwave REGARDLESS of what processor we have been using. That, my friends, is an investment that makes sense for now AND the future.

So you can go and waste money on clock cycles or look at a way to get more out of them. Just my 2 cents.

Lewis
05-20-2004, 05:31 PM
Originally posted by kmscottmoore

I'm sure that Eric at Boxx will be glad to know that I made the "wrong" choice, and that you will be putting him out of business with your Croatian built PCs.

Jeez man yoou aren't reading my posts or what ? Where did i said that BOXX isn't good ? I just want to say that you coulg get better price if you build comp on your own. And is this "Croatian PC's" supposed to be Insult ??? What's that iwth MAC guys ? I have few friends and they are so thick headed with Mac that it's unbelivable and they all also have PCs at home :). My first Mac was Apple IIc 15-16 years ago so i'm not unfamiliar with them either.

Liek i said i built hundred of PCs form scratch and i can build you cheaper PC than MAc at same speed at any time. If you don't belive me show me MAc price and components and i'll built you same or better Pc config for less money. If you can't build system on your own and need brand name config that's not my problem but that's not proving that PC isn't cheaper. It's not my problem that you can't buy "no-name" MAC :).

Bye

mattclary
05-20-2004, 05:36 PM
Right on, Lewis!

CaptainKirk
05-20-2004, 05:57 PM
Often people buy their computers from cheap PC-store. But then they take the risk of having troubles with the computer later because the store have put together the computer with cheapest components. Sometimes it works great and sometimes it doesn’t.

You don't have to put it all together yourself.

BoXX makes good PCs, but they are overpriced. You can buy a PC which will perform pretty much exactly the same as that dual 2.8 ghz Xeon for over a $1000 less.

I know because I'm considering one right now ( either Xeons or Opterons ). At GamePC.com I can configure a dual 2.8 gig xeon machine and still be under $2500, despite using 1GB of memory, Lian-Li aluminum case with a top of the line PS, 10 000 RPM Western Digital Raptor hard drive, and so on. I can save more too, because I don't need to buy another OS ( saving $200 by not buying Windows XP again ), I don't need 2-3 huge drives because I already have hard drives up the wazoo, all I need in the new machine is a fast 10 000 rpm raptor drive. I could get it for close to $2000 which would be at least $1500 less than a new G5. And like Lewis says, it's not our fault Apple gives you no choice in these matters.

You don't need a $600 video card or an overpriced motherboard used by Boxx.

Point is , this is all at least a $1000 less than a dual 2gig G5 machine ( with same amount of memory ), and with better components which I chose, then they put it together for me. So, it's not like anybody has to do it themselves.

Fact is, G5 is an overpriced, slower machine, often unsuitable for audio work ( check for problems with PS noise ). And will drive a lot of people crazy with it's whining, crackling, popping noise coming out of power supply every time you click something on you desktop. Simply a slower, badly designed machine.

Beamtracer
05-20-2004, 06:16 PM
Originally posted by CaptainKirk
Fact is, G5 is an overpriced, slower machine, often unsuitable for audio work ( check for problems with PS noise ). And will drive a lot of people crazy with it's whining I know what drives me crazy with whining, and it's not my G5 workstation! :rolleyes:

The Apple G5 is the premiere machine for professional audio work. The most widely used software for film & tv audio mixing is Protools. Most professional facilities choose to use it on the Mac rather than Windows.

For music sequencing, the Apple G5 is also the preferred machine. More professional musicians sequence their music on Mac than Windows.

wacom
05-20-2004, 06:44 PM
Here is my render on my C64 after 24hrs. This ROCKS. I bought a polish C64 at a thrift store for 15 bucks, with a tape drive, and boy am I loveing it. More pro audio and graphics people use C64s to fill up their basements and closets than anyother system! I tried to build my own, but it was going to cost $25! No DVD burner though...

Can someone post some render times? Please be sure to include your machine and any changes you've made to the scene. If you can render it in on a Texas Instraments Speak and Spell then go ahead- I don't care what you use, more power to you.

This AMD Vs. Intel Vs. IBM Vs. Apple etc etc. debate needs to be moved to the general discussion forum if you can't keep it about this benchmark scene! I don't care if the TI-82 is the best calculator for graphing cos and sin unless it's rendering this LW scene.

kmscottmoore
05-20-2004, 07:06 PM
Damn, Wacom! That is funny!

Actually,the C64 was the first computer I ever used. Learning BASIC my first year of high school. That seems so long ago now.

kmscottmoore
05-20-2004, 07:25 PM
Lewis,

I'm sorry I insulted you.
Unfortunately, in my mind:
Croatian Built = Yugo
I don't think you guys will ever live that one down.

I'm trying to be reasonable here and not start a flame war. Now, as calmly as I can:

The whole point of this thread to begin with, was to benchmark machines with the new scene (which I believe you may be the author?)

However, you clearly stated that a 3.4 Ghz Prescott was the equal of the Dual 2 Ghz G5. No one has posted a render result that substantiates that claim.

Also, the price/perfomance issue was brought up. I showed, as an example, the price and render time for Nigel's G5 and my Boxx. I think that the two are close enough to negate that the G5 is not competitive.

Personally, I don't want to build computers. My clients don't pay me to tinker with logic boards and soldering irons. For tax purposes, with my business, it makes more sense to buy a complete system from an established company. (Not to mention the three year warranty if something goes wrong.) The IRS would not take kindly to me trying to deduct $2000 for something that someone put together in his garage.

If you build your own PCs, good for you.

kmscottmoore
05-20-2004, 07:36 PM
In the midst of all this, we have a new leader!

Dickigeeza
Dual 3 Ghz Xeon
LW 7
21m 3s

Dickigeeza
Dual 3 Ghz Xeon
LW 8
21m 12s

theo
Dual 2.8 Ghz Xeon
LW 8
22m 36s

Aegis
Dual 3.06 Ghz
LW 7
23m 2s

Aegis
Dual 3.06 Ghz Xeon
LW 8
23m 19s

KM Scott Moore
Dual 2.8 Ghz Xeon
LW 7
23m 12s

KM Scott Moore
Dual 2.8 Ghz Xeon
LW 8
23m 27s

Nigel H
Dual 2 Ghz G5
LW 8
25m 47s

bloontz
Dual 2Ghz Xeon
LW 8 (?)
34m 1s

MH9
Dual 1.6 Ghz Opteron
LW 8
34m ?s

TSpyrison
3.4 Ghz P4 (notebook)
LW 8
36m 46s

electropulse
3.0 Ghz P4
LW 8
40m 49s

Trip D
P4 2.8 Ghz
LW8
44 m 2s

KM Scott Moore
Dual 1.25 Ghz G4
LW 7
44m 36s

Tspyrison
2.6 Ghz P4 (overclocked to 2.8 Ghz)
LW 8
46m 8s

Slayer
Athlon 64 3400+
LW 8
46m 44s

Sandman
Athlon 64 3200+
LW 8
51m 3s

Gregsduncan
2.8 Ghz P4 (Vaio notebook)
LW 8
63m 40s

kmscottmoore
05-20-2004, 07:45 PM
Originally posted by Dickigeeza
This is the same machine, all settings the same but now version 8.

fully 10 seconds longer.

I wonder why that is.

Rich

http://www.newtek.com/products/lightwave/product/8/features.php#SpeedandWorkflowOptimizations

Speed and Workflow Optimizations:

Faster Rendering
Faster Inverse Kinematics
Faster Dynamics Calculations
Faster OpenGL Performance
New, workflow oriented, menu layout
Improved viewport manipulation

Dickigeeza, so far everyone seems to be saying that 8 renders a little slower than 7. I think I'll start a new thread about it and see if we can get a response from NewTek (if they don't boot me off this chatboard)

Beamtracer
05-20-2004, 08:54 PM
Originally posted by wacom
More pro audio and graphics people use C64s to fill up their basements and closets than anyother system! Hehehehe!!! I think you're right, Wacom. I must admit that I have a Commodore 64 sitting in my closet. True!!!

I don't have any argument with people like Lewis who like to tinker around and build their own computers.

My gripe is just with the anti-Mac hysteria that comes from Captain Kirk / Panini (previously banned from the Lightwave forums). Comments like "the G5 is unsuitable for audio work" are just silly.

theo
05-20-2004, 09:35 PM
Man... from benchmarking to fossil-hunting.

Anyone with a Commodore 64 sitting in their closet must instantly and without delay be given a large red letter C that is to be sewn onto their garments until they confess to the sin of commodore-clinging and renounce their evil ways.:D

CaptainKirk
05-21-2004, 02:07 AM
#1 Yugo was made by Serbians not Croatians

#2 New benchmarks show that 3.0 ghz dual xeons are already almost 25% faster than the fastest dual G5 ( that is pretty significant considering they can be had for a lot less than the same slower G5.

#3 this is how you buy the fastest PC for less money and better components than BoXX ( Boxx is OK, but lack of options and prices are a bit on the ridiculous side )

go to:

http://www.gamepc.com/shop/systemfamily.asp?family=gpgx2

Pick either 2.8 or 3.0 dual xeon machine. $2340 or $2700, and they do come with DVD burners.

Great case, and quiet too, not like Apple or Boxx.

Now, I don't need to pay for windows again, I can use what I already have saving me $200, I don't need the default card, I can get NVidia 5700 and stick it in there ( game PC will do it for you if you don't want to touch it ), saving another $150. I would switch HD option to a faster WD Raptor drive. I don't need another DVD burner either , and so on.

Tweaking a few options here and there for under $2000 I can have a dual 2.8 with a better case, 1GB of memory and faster HDrives than that Boxx , or I can have the dual 3 ghz machine for around $2200-$2300 This is between $1200 and 1800 less than 1GB Ram G5 or a Boxx, Boxx will render the same , G5 20+% slower than this $2000 PC.

It doesn't take a rocket scientist to figure this out.
That is how you do it and you don't need to do absolutely anything yourself. Just pick components and they will assemble it for you. If they do not have what you want, you can buy it at NewEgg, ship to them, and they will add it .

Forget about Dell and other cheap crap. These are real machines for a decent price. Even if unlike me you do want to pay for another copy of windows again, and include DVD burner, and so on, you'll still pay hundreds less than what you'd pay for a slower Mac.

I've done business with GamePC many times and they are as good or better than anybody. Much better than Apple and their crap support and quality control. ( yes, I too have Mac friends with plenty of problems and wasted days on phone with Apple )

Dickigeeza
05-21-2004, 02:27 AM
[i]Originally posted by kmscottmoore

Dickigeeza, so far everyone seems to be saying that 8 renders a little slower than 7. I think I'll start a new thread about it and see if we can get a response from NewTek (if they don't boot me off this chatboard) [/B]

Cheers KMscottmoore,

A new thread on render time comparisons between version 8 and earlier releases could make interesting reading.

From my experience NewTek are not monsters like at some other software co's I have dealt with. I'm sure they wont boot you off, but hopefully will explain the decrepecies in render times between 8 earlier versions.

Rich

caesar
05-21-2004, 07:41 AM
Originally posted by wacom
Get off the hotrod mentality and get into the affordable commuter mode of thinking. Lightwave has free render nodes because it's so slow at most rendering they almost expect you to take advantage of it.




Very smart thinking

theo
05-21-2004, 12:13 PM
Hey CaptainKirk, I wouldn't knock Dell to badly. As a backup unit to my main monster box I have a Dell dual Xeon that has performed admirably for the last couple of years. We build our own machines now so I'll never go back to Dell but I don't have any complaints either.

hairy_llama
05-21-2004, 03:28 PM
hehe, I got seven 2.8ghz 1gig ram p4 for $3500, that is 19.6 ghz, 7 gig ram.. :)
Main computer was only like $1200(3 ghz p4 2gig ram, 150gb hdd)... so for a grand total of $4700(and $300 for butterfly net render, which you don't HAVE to have) I have 22.6ghz and 9GB ram. Why would anyone buy a BOXX or G5?

wacom
05-21-2004, 04:03 PM
Originally posted by hairy_llama
Why would anyone buy a BOXX or G5?

For the same reason some people want a maned mission to mars: because it's cool.


BTW I bet you've never payed a heating bill.;)

hairy_llama
05-21-2004, 06:25 PM
For the same reason some people want a maned mission to mars: because it's cool.

Haha, now it makes sense :)


BTW I bet you've never payed a heating bill.

You've got that right!

Lewis
05-22-2004, 06:29 PM
Originally posted by kmscottmoore
Lewis,

I'm sorry I insulted you.
Unfortunately, in my mind:
Croatian Built = Yugo
I don't think you guys will ever live that one down.



OK no problem i'm not insulted , even more now when i see you need a small history lession :))

Here is one small for you :

Croatia is NEVER built Yugo(s). Croatia WAS part of Yugoslavia 14 years ago and then we went i war from 91-95 (Serbia and Yugoslavian Army attacked Croatia and later Bosniaadn Hercegovina). Yugo was made by Serbia (city named Kragujevac) and they started to making them AGAIN :)). And belive it or not Yugo wasn't too popular car in Croatia for last years ('coz they were made by our enemy) and Now (in peace for last 9 years) Serbia is resumed making of Yugos :). Car is still "ugly" and it's almost not changed a bit , except few small facelifts and they now have new engine (1.4 injection).

See you were in wrong all the time :D.

caesar
05-24-2004, 06:15 AM
What is this Yugo car all about :confused:

wacom
05-24-2004, 10:02 AM
Originally posted by caesar
What is this Yugo car all about :confused:

They were talking about how Yugo sucks at rendering.

Sure it's a cheap form of transportation, but man does it suck for renders. On the plus side it's much easier to kill yourself with a Yugo then with LW8. Since there is no real manual the box has very little force when dropped on your head. Now LW 7, that might...

OK, but who makes the best ride'n lawnmower that's what I want to know!

Forget rendertimes!

Ztreem
05-29-2004, 12:07 PM
It took 48 min and 8 sec on my computer.
Dual AMD 1,6 GHz AthlonMP
1 Gb RAM
GeForce 4600Ti
80 Gb HDD
:mad:

foghat
05-29-2004, 07:25 PM
Ok here are my results using LW 8 on PC and MAC platforms

P4 3.0e 1gb DDR400

48min 47 sec

and .....

G5 1.6 512 DDR400

1hr 1min 50sec

Time to start saving for dual Xeons.......

Ztreem
05-30-2004, 06:06 AM
Here is the original radiosity render.
Rendertime 48min 8sec.
:(

Ztreem
05-30-2004, 06:08 AM
Here is a modified Fake radiosity render
Rendertime 38sec.:D
All renders in LW8.

Stewdent101
05-30-2004, 08:11 AM
46min 39sec

AMD 64 3400
1gig ram
120 gig sata h/drive

dwburman
05-30-2004, 01:49 PM
@Ztreem: NICE :D

You need to add some noise to the fake picture:

wacom
05-30-2004, 04:15 PM
Originally posted by Ztreem
Here is a modified Fake radiosity render
Rendertime 38sec.:D
All renders in LW8.

What did you do to fake it? Just position a bunch of lights and use raytracing only? Do tell.

Wireframex
05-30-2004, 04:35 PM
Abit IC7
P4 OC 3,45 Ghz
2Go RAM
QUADRO XGL 980

Time : 39 mn 31s

Phil

foghat
05-30-2004, 04:58 PM
Ztreem

Please do share your settings!!!

:D :D :D

Ztreem
05-31-2004, 03:20 AM
It's quite simple. It's just a bunch of lights. it takes some time to setup, but not too long. If you have fprime it should be a breeze.

The secret is that you use negative light sources to darken certain areas. :cool:

nigebabe
06-06-2004, 07:27 AM
Err Sandman how do u have a 1gig FSB on your 3200+ mate? Even overclocked you cant hit 1gig fsb on an Athlon 64.

Exception
06-07-2004, 10:30 AM
This benchmark is useless as Newtek said that they forgot to implement the optimized code for LW8, and that there will be a patch for it when LW8 gets released officially, adding 5 to 10% to the render speed depending on the processor.

gregsduncan
06-07-2004, 10:33 AM
Newtek said that they forgot to implement the optimized code for LW8, and that there will be a patch for it when LW8 gets released officially, adding 5 to 10% to the render speed depending on the processor

Is that code for "we've been working on 8 some more and when we release the next version (8a) there will be some speed improvements" ???

Any enhancements are welcome whatever the case.

Is there a rough release date for full distribution of 8?

phil lawson
06-07-2004, 11:52 AM
Late June/start of July if I remember. :)

Exception
06-07-2004, 12:13 PM
Gregsduncan:

No its: "code that we had already but forgot to compile into the pre-release, we're very sorry".

caesar
06-07-2004, 12:37 PM
"code that we had already but forgot to compile into the pre-release, we're very sorry".

LOL

sandman300
06-07-2004, 12:49 PM
quote nigebabe: "Err Sandman how do u have a 1gig FSB on your 3200+ mate? Even overclocked you cant hit 1gig fsb on an Athlon 64."

1gig FSB is what I paid for. 1 gig FSB is what it came with. no overclocking involved. perhapse you were thinking of the XP version?

Aegis
06-07-2004, 04:24 PM
Is there a rough release date for full distribution of 8?

Siggraph would be my guess...

jeremyhardin
06-09-2004, 02:04 PM
26m 58s
LW 8
Dual 2GHz G5
1GB RAM

jacross
07-29-2004, 01:26 AM
Although we still haven't seen any benchmarks from the fastest dual AMD Opterons (The 250's), I figured I'd throw my hat in with my new Dual Optie setup @ 29m 44s.



2x AMD Opteron 244
2GB PC2700 CL2.5 (2x 1GB sticks)


I really can't wait until Newtek releases the rendering speed enhancements.

sandman300
09-19-2005, 02:04 PM
It's been a while since I posted to this thred. I got a job at a computer shop, and I've had the oppertunity to test some of the systems.

Below is the results for a dual AMD 244 1.8Ghz with 1 GIG of RAM
:thumbsup:


As soon as LW64 is released I'll be testing that. :D

Third quarter is almost over... hint, hint Newtek.

Wireframex
09-19-2005, 03:37 PM
http://img382.imageshack.us/img382/6896/bench3jp.jpg

Dual Xeon 3.4 Ghz EMT64 - 4 Go RAM - windows XP

Waiting for XP64 results too :jester:

Phil

krimpr
09-19-2005, 05:35 PM
Two Opteron 275's, 2Gb ram. (ran outta money.)

Liber777
09-20-2005, 01:27 PM
Dual EM64T Xeon 3.2GHz, 2GB RAM, Win Server x64 Enterprise

Lightwave 8.0 = 21 min, 26 sec

Lightwave 8.3 = 19 min, 44 sec (32-bit of course)

LW is running in discovery mode because this is a test machine, but I doubt that would affect the numbers. Can't wait to test 64-bit LW 8.5!

~Stivan

silverlw
09-20-2005, 03:00 PM
Dual Opterons pffftt! Feel the pow'a of my mighty athlon 2400+ :D

prospector
09-20-2005, 04:57 PM
Kray is THAT good??

or

is it your photoshop skills ;D

krimpr
09-20-2005, 04:58 PM
OOOOOOOOooo.. That Kray render looks pretty good, and the time looks even better.

Tiger
09-20-2005, 09:02 PM
My darling...mmm...F-Prime 2 :hey:
My photoshop skills :rock: :D

silverlw
09-20-2005, 09:22 PM
haha the war is on :hey:
Doesnt your image look a bit pale with one bounce?
I used 24 bounces and optimized quality and rendertime a bit.

prospector
09-20-2005, 11:09 PM
Boxfight!!!!!!!!!!! :D

Medatron
09-21-2005, 12:51 AM
Just for the laugh factor, who can post the SLOWEST render time.

For me, I cajoled 7.5 to run on an old pentium 233. I aborted the render after 4 days and it still wasn't 1/2 done.

Captain Obvious
09-21-2005, 02:08 AM
Just you wait. I'll get LW[whatever runs on Mac OS 9] running on my PowerMac 6100 and render the scene. 60 megahertz of PowerPC goodness. ;)

Tiger, what kind of hardware do you have?

Exception
09-21-2005, 07:55 AM
Silverlw: whats up with the difference in reflection between your first and second kray post?

I havnt tried Kray now for a while... are there presets yet? I can remember fighting with it for a while. It was okay on simple scenes but when there were tons of polygons involved, it was impossible for me to find the right settings. Every test render turned out with weird colors, grainy or otherwise nutty, with high render times... I suggested presets or some sort of wizard or auto-detection for the settings back then... don't know if they did anything with those suggestions.

Captain Obvious
09-21-2005, 08:14 AM
Ehh, not to rain on everyone's Fprime/Kray parade, but Lightwave's renderer isn't really that slow. You just need to tweak it a little. ;)

This was rendered on a single 1.6GHz G5, hardly top-of-the-line hardware. I was using the computer in the meanwhile. Had Lightwave been allowed to monopolize the CPU, it would've been even faster. Oh, it was two bounces as well, instead of one.

Captain Obvious
09-21-2005, 09:05 AM
This took 108 seconds. Yes, that's right, 108. What, Lightwave slow? :D

Yes, it is a little noisy, but it's not so bad some noise reduction wouldn't fix it. Shading noise reduction was obviously turned off, and I removed the luminosity from the coloured walls. Not too shabby for less than two minutes, wouldn't you say?

silverlw
09-21-2005, 09:07 AM
Exception: By accident i had smoothing on the flat surfaces of the cubes resulting in weird shading.

There is still no presets and the more i use Kray i realize it will be hard to do.
The values change quite dramatically depending on what you want to do.
Examples:Light's far away or close,strong or faint,lot's of photons or not sufficient with photons,irradiance size, Arealights size and quality, blurred reflections, High contrast lightning or low, The size of the objects involved and so on....
There is simple not any settings that will be sufficient to a majority of renders or a certain type of renders. For rather quick testrenders i recomend changing the defaults to this: FGrays 100-300, Sample density 0.3, Oversample 25% Always use prerender 50% since it clears up noise quite alot.

I always monitor this thread wich i recomend about Kray. Kray thread (http://vbulletin.newtek.com/showthread.php?t=36461) I happily helps those with questions there.

silverlw
09-21-2005, 10:40 AM
I feel childish but here is my 86 seconds version :neener:
I dont think i can push the scene much further now without getting it really noisy and splotchy.
I inklude the Kray scene so someone with Opterons/Xeons could try it and post the rendertimes.

Here i latest Kray Kray 1.511 (http://hem.bredband.net/b223277/Kray/kraytracing_demo1_511.zip)

my computer specs: Amd athlon 2400+ 266mhz sdram 640mbyte.

viewnathan
09-22-2005, 12:02 PM
I got a mac g5 dual 2 gig here...like to try but where is the file?
viewnathan

viewnathan
09-22-2005, 12:11 PM
mac G5 dual 2 gig- 43.4 secs but ther is a plugin missing...
probably only for pc..

Captain Obvious
09-22-2005, 12:24 PM
Kray is currently Windows only, yes. But the original scene is supposed to be in Lightwave/Content/Benchmark or something such.

sandman300
09-22-2005, 06:36 PM
It's strange. I started this thred as a comparison of hardware under the context of lightwave performance. It seemed to work for a while and then several people began comparing F-Prime results, which was useless in this context. Others modified the scene, making their results meaningless. Now Krey enters the thred. By all accounts it sounds like a great program, prehaps a little pricy but then again the cost would be justified by the results. Unfortunately, this thred is not the "how much faster is Krey than Lightwave's native renderer" thred. I'm not trying to flame anyone. But it would be nice if your going to post results of the rendered scene, do it as is without any 3rd party software. I only ask this because I'm genuinely interested in finding out how Lightwave performs on different hardware.

Thankyou, again I'm not trying to flame or whatever else mean people do. :lightwave