View Full Version : X800 vs. LW8...will it cut the mustard or just cut the cheese?

05-22-2004, 08:36 PM
The board is in and installed:D

After removing every NVidia driver I could find (even went thru regedit), I installed the full package.

After a reboot, I set my desktop, grabbed a ciggy and a coca cola and began to giv'r hell.

Now this isn't a report like the 'others' you see when looking at reports for a video card.
We don't need to see all the steenki'n hype, Just the LW results.


THAT is the question.

The FIRST thing I noticed was the WEIGHT of the card.....HOLY TONNAGE Batman.
It's like 3 times heavier than my Ti4200.

You WILL be using the AGP cardlock;)

It uses an extra plug in for power and a cable comes with it, along with many many other cables.

But on to LW.

With my ciggy half gone and the coke going flat, I clicked on Modeler icon.

Cool...it worked so far:) Modeler is open.

Grabbed the center frame to resize windows....smooth as a babys butt, no hesitation, no multiple window frames showing when dragged around.


Set texture res to 1024.
UNCHECKED show cage and show point handles (as I never use them).

Now I started to load models...

NOTICE >>>>>>
The FPS are estimates but are very close (+or- 2 at worst as having been into video for 12 years, I can get mighty close at framerates, but you can take it with whatever skeptacism you want, but I do believe I am mighty close)

The Gorilla.....10,846 polys with 6,790 in Sub'D mode
Set perspective window to TextureWireframe

This model spun around and zoomed in and out in REALTIME

Centurion.....24,784 polys with 19,947 in Sub'D mode
Again in texture wireframe mode

This spun and zoomed at 27 FPS, some very slight hesitation.

All screen changes were instant (IE single frame and back to quad frames...windows)

Satellite....8488 polys with 8440 in Sub'D mode.

Everything in realtime

Shuttle Endeavour......174,311 polys with none in Sub'D mode

In smooth shade, Textured, TexturedWire, spin and zoom was at 20 FPS
Hesitation, but workable
Nothing like before when ya moved a model, ya let go of mouse and waithed for model to catch up.

In wireframeshade on down to wireframe, speed jumped to about 27 again. Very slight hesitation.
As I most always work in wireframe shade, this was great.

Now the BIG KAHUNA Alpha 43 model.....220,799 polys with 1848 in Sub'D mode.

Texturewire, texture, smoothshade...about 20 FPS
Wireframeshade and lower..27 FPS...almost REALTIME !!!!

As I sais, I work mostly in wireframeshade, this was beautiful.

There were no MT frames (ie: no model showing), no artifacting with dissapearing points

Now the BAD news in modeler. With these large models (shuttle and Alpha 43) screen change was 10 to 11 seconds.
Going from quad mode to single or back, top left screen updated immedietly, last 3 screens took 10 to 11 seconds.

Screen resize for the same models (grabbing the intersection of the quad view) and moving about 1 inch any direction causes a 3 second delay. Not intolerable but there nevertheless.

So what does it look like
Somewhere between the gorilla 6790 Sub'D polys (realtime movement)and the Centurion 19,947 Sub'D polys (27 FPS movement) there is a slowdown.
So if your models are around 8000 Sud'D size, you'll like it
If your models run in the 100,000+ poly without sub'D, you'll like

So far it's all mustard:)

I'm doing Layout now so.....

Layout update tomorrow

05-23-2004, 12:13 AM
Thanks for the review.
I find with my FX5950 ultra, I get quite a performance hit in layout
when the surfases on an object is transparent, relative to the complexity of the object of course, how dose your card handle transparent surfases.
LW7.5c OGL performance increase is a lot more notisable than LW8, after changing my card, I hope NT, and Im sure they will, give the OGL in LW8 boost. Or evern better, DirectX support!!!.

05-23-2004, 08:25 AM

Upon Opening Layout I set bounding box to 3,000,000
(my usual default)
Set Textures to 1024X1024..Again, my usual

Loaded the massive Moonbase.lws

File size:
2,615,229 polys
186.6 megs object memory
16 images

It loaded to default shaded solid

First I hit play

On my Ti4200, doing this I waited a few min and when still nothing happened I clicked it off.
Now 5 seconds after I hit play it jumped to frame 59 then to frame 0
it recycled every 5 sec

Not good, but at least I SAW movement;)

BUT.... If I just grab slider and manually move it, the delay is 1 sec for slider to move, then 1.6 sec later the scene moves.

Once ya click on skider the cursor dissapears so no way of knowing how many frames ya move it till it changes 1 sec later.

So upon moving with the keyboard 1 frame at a time, I get the same results, 1 sec for key and 1.6 sec for the scene.

I then went to options panel and UNCHECKED EVERYTHING

moving slider with mouse or keyboard took 1 sec again, but scene updated in 2/3 the time 1 sec

So now...move slider scene updates......2 sec

Went to Textured solid mode

Everything stayed same

Upon going from single window to quad took 11 sec for all windows to refresh
going from camera window to perspective took 3 sec
going from default position in perspective to zoom out to see whole scene was a 4 sec delay

I then went to object properties and brought all subpatch levels to 0 display and 3 render

Filesize went to 467,097 polys
68.4 megs object memory.
Hit play

2 FPS!!
Houston, we have viewable movement!!
Now scene is at least workable with everything showing
If ship on ground is dissolved out and you only are working on flightpath of flying one
you get a 50% increase in speed 3FPS
And there is a total time lapse of 1 sec when moving slider manually.

I would now consider this a 'workable' scene in all it's glory
and not bounding boxes.
But a full speed check still requires a preview to be made.

Changing the one in flight to bounding box gives me a 4 FPS working scene
Never liked bounding boxes tho:)

Onto other scenes!!

After playing with 2.5+ million polys these should be quite easy?


Tire bounce scene
17,245 polys
4 FPS in play mode
Almost smooth when moving slider manually
All subpatches set to 0 and it jumps up to 7 FPS


LW_Promo scene

66 FPS
Window changes and resizing....instant

Chug scene
60 FPS
Window changes and resizing....instant

Firewalk scene

20 FPS
Windows change instant
Quad view play drops to 10FPS (all still default as loaded) So it is textured in all windows

In conclusion

All scenes here were done with the framerate meter applied

The base machine is;

P4 2.8 GHz
on a P4S533 mobo
1 gig of PC3200 ram
X800 Video card

There are NO other programs on the machine besides LW and the VT2.
So no resources are being used other than Windows 2000 SP4 and LW.

The UPside
everything is greatly speeded up
There are no blank windows or OGL abnormalities.

The DOWNside
The problem that originally sent me on this quest is STILL THERE!!!!!
The spreadsheet STILL crashes LW when clearing scene
I am now convinced that after trying an NVIDIA card and 5 different drivers (and even soft quadroing it) and this card and the only 2 sets of drivers out for it, that Spreadsheet is at fault.
7.5 works flawlessly...8 crashes
No-one can tell me otherwise

Coming from the Ti4200 to the X800, and ALWAYS buying NVidia and going to ATI, I am going to REALLY enjoy this card.
Soon there will be ways listed to crank the ramspeed up (haven't seen any yet). Maby jump it to the X800XT (haven't seen that yet either). And everythin will move along a little better.


when changing window layouts or zooming on object in either modeler or layout
the holdup is where?
The video card redraw ...or...in LW trying to reconfigure the point placement in the new windows?
Which one handles that?

In layout...when moving slider or hitting play
does the slowdown occur in the video card or LW ?
Which one handles the point and object placement from frame to frame?

I know rendering is pretty much handled by the CPU so I didn't do those tests, but which part handles the actual point and object placement when working with a scene?


If you'd like a comparison to your card.....give me a scene (from LW8)or send one of your own, or send what you need checked in modeler.

Post your results and I will post mine (use the FPS meter in layout so everyone is on the same page tho).

I will give an honest (LW'ver to LW'ver) and not sales hype, answer. As best I can.

Well, my ciggy is burned to the filter and my coke is MT.
So I bid you a fond farewell;)

It cut the mustard and NOT the cheese :D

05-25-2004, 09:26 AM
An interesting read!
I bought a new pc a few weeks ago, (went from 1333ghz, 256mb ram to 3ghz 1gb ram) and was also quite pleased with the overall speed increase (not talking about the obvious speed in rendering), even though the gpu stayed the same.

If possible, do you have the fps values of your 'old' gpu, so we can see the difference between the two?

05-25-2004, 10:30 AM
didn't do before I switched cards, cuz I just remembered the bad stuff from old card to see what changed.
Overall the X800 was much faster.
But will switch the old card back in and do a comparison.

05-25-2004, 10:57 AM
Hm somehow i can't believe that any video card can speed LW like you described...although i really wish there is!

Or do they have some secret "only ATI will cut the mustard now" feature??

would really like to have a direct comparison...
is there something like FRAPS that can measure your framerate in modeler? I'd really like to have measured results.

When i read "174,311 polys with none in Sub'D mode (...) at 20 FPS" i can only dream of that because i'm perhaps getting 2fps, although apps like 3ds max give me ~10fps (measured with the 3dsmax.ini entry to display FPS)

05-25-2004, 11:26 AM
Thanks for your detailed analysis, prospector.

05-25-2004, 11:27 AM
To my surprise FRAPS works with Modeler!!
However it does not display fractional frame rates, so i am really getting between 2 fps on a my chapel with 108k polys, flat shaded :(

version 1.9D is free

05-25-2004, 06:11 PM
Got it and will test.

05-25-2004, 06:54 PM
Dunno if it really does work within modeler.

with the shuttle and Frap I get 11 FPS
In wireframe to flat I drop to 4 ???

Wire up to Flat ALWAYS go faster.

When I spin the model in perspective there is no noticable hesitation. Just as smooth as a babys butt.


Lemme try the others.

But at least other readers can now get a comparison reading from thier compys with the program, tho I really don't trust the numbers

AVIs produced at those frame rates produce very noticable jumps, and I see none of that in modeler except when I get to 20 (my estimate) and then it looks like a AVI film.

But hey this proggy gets a base measurment that if wrong will be just as wrong for everyone at the same rate.:)
So a comparison can be made.

05-25-2004, 06:55 PM
Dunno if it really does work within modeler.

with the shuttle and Frap I get 11 FPS
In wireframe to flat I drop to 4 ???

Wire up to Flat ALWAYS go faster.

When I spin the model in perspective there is no noticable hesitation. Just as smooth as a babys butt.


Lemme try the others.

But at least other readers can now get a comparison reading from thier compys with the program, tho I really don't trust the numbers

AVIs produced at those frame rates produce very noticable jumps, and I see none of that in modeler except when I get to 20 (my estimate) and then it looks like a AVI film.

But hey this proggy gets a base measurment that if wrong will be just as wrong for everyone at the same rate.:)
So a comparison can be made.

Using ver 2.1.1

05-25-2004, 06:56 PM
oops...double post:(

05-25-2004, 08:44 PM
Thanks for the post prospector. I am very interested in this card (well the X800XT version). I am holding off a new PC purchase till I can decide which gfx card I want to get. Have you tried this card with a dual monitor setup? I am most interested if that is working properly with the ATI card. I am still a bit hesitant with the ATI line cards because of some of the posts I have read of them having drawing errors in LW (I think the 9700 or 9800 cards). I kind of want to get a Nvidia 6800 based but they seem to have been delayed.

05-25-2004, 10:13 PM
all my compys run dual monitors.
everything seems to be working on both monitors on the compy with the x800.

05-25-2004, 11:07 PM
OK here are the numbers

I STILL go with the above results,and when you see the layout results you'll understand why.


Alpha 43....4

And here's why I don't trust those

LW_Promo......Frap..76_______LW Frame rate counter...62
Chug...............Frap..42_______LW Frame rate counter...59
Firewalk..........Frap..35_______LW Frame rate counter...9
Tire Bounce....Frap...17_______LW Frame rate counter...4

Frap was all over the place (higher and lower than LW) but none were close

LW Promo +14
Chug -17
Firewalk +26
TireBounce +13

So you can take the numbers for what they are.

As in my top reports, if there was no,slight, or much hesitation, then that's what I saw.

Just tried it in VT but won't do video from there.
Tried to see what it says for what I know to be 30FPS video.

If it seems to anyone that I might be 'pushing' this board, I am not, could care less, as no-one had it to compare what I had to, I thought I would.

05-26-2004, 01:52 AM
Ok, seems like Fraps does not work that well with multi-viewport apps...at least on your machine.
Hadn't really time yesterday to give it a real testrun...

Thanks for trying it, now i only wish i could afford a X800 Pro... :(

05-26-2004, 07:40 AM
An I like hardware a lot heheheh..BUT lets remember X800, like any Gforce card and their drivers are not projected to 3d apps! , they´re made for OGL and DX games = real-time gtaphics. ATI Fire and Nvidia´s Quadro are made for pro/workstation use (with a "pro" price too), handling high poly scenes/models better than the desktop ones, with drivers specific configs for each major app, including LW.
Often we see people "adapting" workstation drivers to desktop cards - of course, its a hack to gain the driver performance and not having to pay the tubes in a similar gpu/vpu pro card

05-26-2004, 09:18 AM
Yea and what exactly does LW have to benefit from pro-cards?
special driver support? negative...
special feature support? negative...

Only thing you could hope is that the drivers are more geared towards higher polycount than high fillrates, but all tests of consumer vs. pro-cards in LW i read came to the conclusion that LW is just totally CPU limited and there's no noticeable difference that would justify the price-plus.

But I'd be perfectly happy if LW would just queeze out consumer cards as much as possible.
I hear Newtek saying that the CPU is that important because of all the subpatch and bones etc stuff, ok, but then why is it slower than other apps when i just put a static objects in Layout and rotate the perspective view?
Sure my GeForce2 GTS really is not the fastest anymore, but seeing other apps giving 3times or 5 times the framerates on the same cr**py card doesn't motivate me to invest money in hardware that doesn't get fully used...

05-26-2004, 03:11 PM
It use to be "pro' cards had 64 megs ram, then 128, then 256, now 512 or 387 (forget which) and the game cards had but half of the 'pro' version.

The ram was a faster ram on 'pro' cards
But now a days there is little difference in cards of the same family

100 to 200 dollar cards of any kind 'pro' or 'game' pretty much are close to the same results

same for 200 to 400 and 400 to 600 ones.
So sure, a 600 'pro' card will do better than a 200 'game' card.

When all is said and done the X800 cost me 437 with tax and shipping, so I could have gotten a 400 dollar 'pro' card, would I have gained anything?
Not by any figures I saw.
An 800 dollar card would have given me more, and the 1500 'pro' card would have too.

Doing tests with this and other cards, it really comes down to the CPU.
The firewalker test proves that, very little polys, but slow because of the CPU trying to keep up with the particle system,
HVs are slower, IK is slower, hard and soft dynamics are slower.

If I got a dual xeon and run the same tests then I mosdt likely would see a speed increase across the board.

05-26-2004, 04:32 PM
Actually you wouldn't see much of a speed increase with dual CPUs in Layout. It seems that LW only uses 1 CPU in its Display rendering. So for working in Layout you would be better off getting 1 REALLY fast CPU then 2 medium or fast CPUs. (Same it seems like for MD, Particles, HardFX...) Basically Layout IS NOT Multi-Threaded.

Also if you've seen Messiah (which also does SubD's) it's pretty fast, and has been much faster then LW all the way back to version 5.6, so that tells me that LW can get faster (mostly by relying on the GPU more.)

I hope to see support for OGLSL in the LW SDK along with direct support in Layout/Modeler itself (i.e. you can write a OGLSL that Layout would take into account in its' display render.) By using Vertex deformers & shaders a LOT less CPU power would be used for IK and HyperVoxels. The Pixel Shaders don't seem to help out as much in this paticular area. Rather the Pixel Shaders would be used for Lens Flair, Specular Highlights, Refractive Transparency...

05-26-2004, 10:26 PM
Yup second CPU doesn't get touched.
I already thought about getting a single Athlon 64 for modeling and animation work and use the dual 1800+ as renderslave...

05-27-2004, 06:50 AM
Then maby that should be something that Newtek should work on.
Getting modeler and layout multithreaded and not just the renderer.

05-27-2004, 10:05 AM
Here's the deal and difference with video cards.

Typically you find you have 2 types of cards, game cards and workstation cards. Radeon x800 (game) FireGL X2 (workstation).

Typically, when using the workstation card you will see a large increase in visual performance. Even though LW's opengl is one of the worst in industry, you will still notice it improved with a workstation card.

Why? a few reasons:
1) Workstation cards run at a slower clockrate for stability and quality. Game cards typically are allowed

2) Workstation cards are optimized for OpenGL. Game cards will support it, but not to the degree they do with DirectX.

Too LW isn't developed for the platforms it supports. Otherwise, Apple users could enjoy Quartz and PC users enjoy DirectX.

05-27-2004, 11:12 AM
Originally posted by Lynx3d
Yup second CPU doesn't get touched.
I already thought about getting a single Athlon 64 for modeling and animation work and use the dual 1800+ as renderslave...

If you have FPrime then that would be the way to go (seeing as that isn't Multi-threaded yet either.) However I would go with an Opteron instead. AMD will be coming out with Dual cores that will work in MBs that support the 104w supply Socket940 Standard (ask your MB provider for info)... Well that and a BIOS update.

And for the Record I'm thinking about the same thing.

05-27-2004, 11:47 AM
I don't think a single dual-core Opteron will be more cost effective than a dual-Opteron system...IMHO quite the oppsite, at least for some time after launch.
And then, it's still not giving me any performance advantage over a single Athlon64 for modeling and animating. Don't really need another renderslave...

But i think i am going to wait until Win XP 64bit is out and PCI-E is a common feature anyway. Doesn't really make sense buying top of the line hardware right now. PCI-E is more or less ready, and with it lots of new hardware (not only video cards). I'm not going to buy an AGP Radeon X800, not because of performance but because my coming system will most likely not support it, and my budget is rather tight since i'm not making any money with LW (yet...you never know).

And i don't have FPrime because i don't want even more software that defeats the sense of SMP...
Sure it runs as seperate thread AFAIK, but right after i stop adjusting parameters LW sits completely idle anyway.

05-27-2004, 02:23 PM
It´ll be released soon, with a plugin for Maya. Its a software that makes your 3D app use the quadro´s gpu (not compatible with gforce) like a processor. We´ll just need to see someone writing a plugin for LW.

09-15-2004, 01:27 AM
hi i just got the X800 pro and i am really overwhelmed with the
results i mainy use LW for modeling with silo and zbrush to fill the workflow

still renders are my main concers..although animation (done in 2d mostly) comes in some tiimes.

i dont quite get all these "profile" settings for the ATI card using the Redline V2.0
most of them are game concerned but ..what are the optimal settings
for LW (espcecially Layout with Fprime which some times i find slight sluggish)
i am sure i tweaked something and changed a setting so...all help will be appreciated..

09-25-2004, 07:02 PM
Cool, I just did a search for ATI cards and I just happened to come accross this thread. I'm picking up the grand-daddy of cards in a couple of weeks (Insanely backordered) MSI ATI X800 PCI-X (http://www.overclockercafe.com/Reviews/VGA/HIS_Excalibur_X800XT/index.htm)
This is a great thread. I was very hesitant of buying ATI because of people having issues with it, but I still really want a PCI-X version of the nVidia 6800 Ultra (Asus says by the end of October).

09-25-2004, 07:57 PM
And people still confuse PCI-E (also written PCIe) with PCI-X... :D
PCI-X does NOT stand for PCI-Express...

09-25-2004, 08:39 PM
And people still confuse PCI-E (also written PCIe) with PCI-X... :D
PCI-X does NOT stand for PCI-Express...Dudn't matter ;), a fast video-card by any other name is still sweet...

PCI-X just fit's a heck of a lot better in my mind that PCI-E. I usually write it PCI-xp ;)