PDA

View Full Version : can you testrender this scene for me? (on fast machine)



erikals
01-25-2010, 05:52 AM
i'm basically just curious on how this frame number 485 renders, how fast.

it's an old scene Peter Profetto (PeterPro) shared at CGtalk waay back.
http://forums.cgsociety.org/showpost.php?p=1026854&postcount=58

on my QuadCore 2.6GHz it rendered in 4minutes 13seconds.
would you happen to have an i7 or Xeon? could you test render it for me?

http://www.newtek.com/forums/attachment.php?attachmentid=81408&stc=1&d=1264423918

BigHache
01-25-2010, 09:14 AM
Just rendered on a dual Quad-Core Intel Xeon 2.8 GHz Mac Pro, 3m14s.

Matt
01-25-2010, 09:48 AM
Dell Precision 490 Intel Quad Core Xeon (X5355) 2.66Ghz (Work Machine) Windows XP Pro SP3, 32Bit, LightWave 9.6

Rendered frame 485 in 5m 38s

Which is pretty crap! I wanna see what my box at home can do it in! Better be less!

OnlineRender
01-25-2010, 10:42 AM
Dell Precision 490 Intel Quad Core Xeon (X5355) 2.66Ghz (Work Machine) Windows XP Pro SP3, 32Bit, LightWave 9.6

Rendered frame 485 in 5m 38s

Which is pretty crap! I wanna see what my box at home can do it in! Better be less!

lol ,ohh its a race to see who has the fastest machine .

ivanze
01-25-2010, 10:54 AM
LW 9.6 Win 7 i7 860 2.8Ghz = 3m 28s

I can't believe this machine is only 14s slower than a dual Quad-Core Intel Xeon 2.8 GHz Mac Pro.

Twisted_Pixel
01-25-2010, 11:28 AM
Oh er 7min43s here. Dread to think what it would have been on my old system.

Matt
01-25-2010, 11:47 AM
Just tried this on my home machine.

Custom build, Intel Quad Core 2 Extreme (QX6700) 2.66Ghz Windows 7 Pro, 64Bit, LightWave 9.6

Rendered frame 485 in 6m 51s

Right, time to upgrade I think, not happy with that speed at all.

btomer
01-25-2010, 12:07 PM
Rendered on a i7 920 oc to 3.9Ghz = 2m 28s

OnlineRender
01-25-2010, 12:10 PM
remember its not just your hardware , its your whole setup .

i would suggest amd over intel everytime

Shiny_Mike
01-25-2010, 01:17 PM
I'm slow, 5m28sec on a stock AMD 955 (with web surfing in the background). But hey, not bad for a $160 CPU.

edit: knocked it down to 4m30s with no background surfing and using 32bit 9.6 instead of 64

erikals
01-25-2010, 05:30 PM
wow, all these test. thanks all, i appreciate it http://erikalstad.com/backup/anims.php_files/smile.gif

it's interesting how the Xeon's don't keep up.
but this is a bit special scene though, being Hypervoxels...

that overclocked i7 ~4GHz sure beat everything else. 2m 28s http://erikalstad.com/backup/anims.php_files/eek.gif
any way to have dual i7 cpus ? http://erikalstad.com/backup/anims.php_files/question.gif

erikals
01-25-2010, 05:48 PM
...Custom build, Intel Quad Core 2 Extreme (QX6700) 2.66Ghz Windows 7 Pro, 64Bit, LightWave 9.6
Rendered frame 485 in 6m 51s...

this is strange,...

yours - Quad Core 2 Extreme 2.66Ghz - Win7, 64Bit, LW96-? > 6m 51s
mine - Quad Core 2 -------- 2.83Ghz - Vista, 64Bit, LW96-32 > 4m 13s

hm, strange, why is mine faster?... http://erikalstad.com/backup/anims.php_files/aiwebs_029.gif
actually, correction, mine is a Q9550 2.83GHz

erikals
01-25-2010, 05:56 PM
btw, if anyone wonder what they are using, run this program (no install required)
http://www.cpuid.com/cpuz.php

OnlineRender
01-25-2010, 06:50 PM
this is strange,...

yours - Quad Core 2 Extreme 2.66Ghz - Win7, 64Bit, LW96-? > 6m 51s
mine - Quad Core 2 -------- 2.83Ghz - Vista, 64Bit, LW96-32 > 4m 13s

hm, strange, why is mine faster?... http://erikalstad.com/backup/anims.php_files/aiwebs_029.gif
actually, correction, mine is a Q9550 2.83GHz

vista is probably the key

and stop writting in white :P

erikals
01-26-2010, 03:38 AM
ok... ; )

i'm sort of shy to ask this, but could someone test this scene as well?
(on a i7 or Xeon or such)

rendered on QuadCore 2 Q9550 2.83GHz - 55sec

http://www.newtek.com/forums/attachment.php?attachmentid=81436&stc=1&d=1264502182

Twisted_Pixel
01-26-2010, 06:10 AM
Will do. when I get home.
It's interesting to see how the system compares.

probiner
01-26-2010, 06:36 AM
Intel Q9450 2.66Ghz
4Gb Ram (WinXP, so 3,2Gb)

Smoke
5m22s

Reflections
asks for a lot of files.

Cheers

COBRASoft
01-26-2010, 06:57 AM
My I7 -> 3m1s. This is with a lot of apps open (VisStudio, SLQ Manager, ...).

erikals
01-26-2010, 07:28 AM
Reflections
asks for a lot of files.

yeah,... it should have been cleaned up. sorry, missed that. zip file on previous page now replaced.
(link)
http://www.newtek.com/forums/attachment.php?attachmentid=81443&stc=1&d=1264515790
this one is easier.

shrox
01-26-2010, 09:06 AM
I did, once it passed 5 minutes I aborted...

Lightwolf
01-26-2010, 09:59 AM
Just for laughs: 14m49s on an Athlon X2, 2.4 GHz (and the odd task hogging some cycles in the background) ;)
LW 96 - 32bit, XP64.

Cheers,
Mike

BigHache
01-26-2010, 10:24 AM
I realized I didn't post some of my specs yesterday, sorry bout that. Yesterdays' render was in 9.6, OS 10.5.8.

Today's test with 9.6.1 -- 32.5s

Twisted_Pixel
01-26-2010, 11:21 AM
As promised earlier :)
Just ran the render, got a time of 1min12sec.

Traveler
01-26-2010, 01:15 PM
Rendertime rob-b2 scene : 33.1 secs
rendertime smoke scene : 3m6s (186.9s)
specs: Windows 7, i7 870 @ 2.93ghz, 4gb ram, Lightwave v9.6 32bit

kosmodave
01-26-2010, 03:04 PM
Thought I'd try different Lightwave versions with the smoke and got the following;

LW 9.6 64bit on Win 7 64 5mins 45 secs
LW HC2007 64bit on Win 7 64 5mins 43 secs

LW 9.6 Win 32 4mins 53 secs
LW HC2007 wIN 32 4mins 53 secs

Yikes! Think I will stick to Win XP for rendering anims :confused:

Dave

Specs;

Intel Q6600 at 3.068Ghz
4Gb ram
Win XP Pro 32 SP3.
or Win 7 Pro 64
System drives on plugin caddy so can choose which one to use.

hunter
01-26-2010, 05:24 PM
2m 28s - LW 9.6.1 - Sig has specs. (Hp Z800) @ work, not mine. but the i7, hmm... Now I know what to try to get at home. :hey:

btomer
01-26-2010, 06:34 PM
Rendered rob-b2 scene=: 27.5 secs

specs:
Windows 7
i7-920 @ 3.9ghz
6gb ram
Lightwave v9.6 32bit

Soth
01-26-2010, 06:47 PM
2m 28s - LW 9.6.1 - Sig has specs. (Hp Z800) @ work, not mine. but the i7, hmm... Now I know what to try to get at home. :hey:

Turn on HT in BIOS and you will have 16 threads. :)

zapper1998
01-26-2010, 07:59 PM
My machine did it in ........ 9 min 32 sec - 572.9 secs

81474

I AM BUYING A NEW MACHINE......


81475


...... I am at new egg ordering, NOW....



Michael

Mr Rid
01-26-2010, 09:10 PM
Velocity Micro ProMagix W160
Win7 Pro
i7 965 XE, quad 3.2GHz

Profetto Smoke:
x32 = 3m 2s
x64 = 3m 15s

Robot Reflection:
x32 = 33.7s
x64 = 35s

...what's the point of x64 again?

zapper1998
01-26-2010, 09:12 PM
My I7 -> 3m1s. This is with a lot of apps open (VisStudio, SLQ Manager, ...).


Antec 900-II
Win7 x64,
I7 920 @3.675GHz,
GB EX58-Extreme,
2x24" SyncMaster T240, NVidia 260OC,
12GB RAM,
300GB Raptor,
1TB WDBlack

Which power supply are you using, ??, with your i7 machine..??

Looking at a new system and wondering which power supply I should get..

Michael

splitpoint
01-26-2010, 10:50 PM
Rendertime rob-b2 scene : 38 secs
rendertime smoke scene : 3m54s
specs: Windows 7, dual opteron [email protected], 8gb ram, Lightwave v9.6 64bit

ivanze
01-26-2010, 11:22 PM
rob-b2 LW 9.6 x32 Win 7 i7 860 2.8Ghz = 36.3s
LW 9.6 x64 = 38.1

erikals
01-26-2010, 11:39 PM
Rendered on a i7 920 oc to 3.9Ghz = 2m 28s

this one sure looks to rock,
based on other tests this should render the RobotReflection scene in 22 seconds...

...darn good.

ok folks, you know what to get http://erikalstad.com/backup/anims.php_files/smile.gif

kosmodave
01-27-2010, 12:27 AM
Velocity Micro ProMagix W160
Win7 Pro
i7 965 XE, quad 3.2GHz

Profetto Smoke:
x32 = 3m 2s
x64 = 3m 15s

Robot Reflection:
x32 = 33.7s
x64 = 35s

...what's the point of x64 again?

Yes, I was wondering that as well. Apart from rendering larger scenes there's certainly no speed advantage as was shown in my results as well. Plus what's the point in Win 7 64 as that is almost 15% slower on my machine........ Is this what they call progress? :cursin:

Is anyone able to test this 32 v's 64 or Win 7 v's XP with other 3D programs to see if it's just a Lightwave thing?

Dave

Traveler
01-27-2010, 02:24 AM
Which power supply are you using, ??, with your i7 machine..??

Looking at a new system and wondering which power supply I should get..

Michael

That highly depends on the gfx card. Are you planning on using a multi-GPU solution (sli) for example. I bought this system a few months ago: i7, 4gb, ssd, radeon 4870, CoolerMaster Silent Pro M600 Modular 600Watt. I'm quite happy with it. Not only is it a lot faster than my older pc it's a lot more quiet too.

Anyway, there are a few sites that allow you calculate the number of watts required, like http://www.antec.outervision.com/ It should give you an idea of what is required.

(Sry for qoing slightly ot with this post... )

erikals
01-27-2010, 02:32 AM
---

HyperVoxels scene >

QuadCore 2.8GHz Vista32 > 5 minutes 5 seconds
QuadCore 2.8GHz Vista64 > 6 minutes 5 seconds


RobotReflection scene >

QuadCore 2.8GHz Vista32 > 55 seconds
QuadCore 2.8GHz Vista64 > 57 seconds

erikals
01-27-2010, 02:41 AM
Great link Traveler http://erikalstad.com/backup/anims.php_files/smile.gif

Soth
01-27-2010, 03:25 AM
HP Z600, 2x Intel Xeon E5520 2.27Ghz, 16 threads, LightWave HC (2007), looks like 9.6 is a bit faster, I will post screens soon.

http://www.newtek.com/forums/attachment.php?attachmentid=81484&stc=1&d=1264587883

http://www.newtek.com/forums/attachment.php?attachmentid=81485&stc=1&d=1264587895

Soth
01-27-2010, 03:47 AM
Here you are, look like HC is a bit slower.

http://www.newtek.com/forums/attachment.php?attachmentid=81487&stc=1&d=1264589188

http://www.newtek.com/forums/attachment.php?attachmentid=81486&stc=1&d=1264589178

...and HC again, with no render preview:

http://www.newtek.com/forums/attachment.php?attachmentid=81488&stc=1&d=1264589197

...a bit random, is it?

All on Windows 7 x64

erikals
01-27-2010, 08:24 PM
i was really impressed by the i7 920 with overclock. (4GHz)
here's a good page on how to do it if you have or want an i7...
http://www.clunk.org.uk/forums/overclocking/22106-core-i7-overclocking-guide-beginners.html

...now, how would the 6 Cored AMD istanbul be doing in this test... http://erikalstad.com/backup/anims.php_files/question.gif

Hopper
01-27-2010, 09:01 PM
Smoke: 3m:1s Just couldn't squeeze into the 2 minute zone.
------------------------
LW 9.6.1
Intel Q6600 OC'd to 4.1GHz
4GB Ballistix DDR3 1600
WinXP/32

Same hardware, but WinXP/64: 3m:12s What's up with that... :(

erikals
01-27-2010, 09:32 PM
i don't render in 64bit, it's not optimized,

use 64bit if you have memory hungry scenes, that's when it kicks.
(according to a studio, it kicks pretty good too)

erikals
01-27-2010, 09:37 PM
Intel Q6600 OC'd to 4.1GHz

you overclocked an Intel Q6600 to 4.1GHz ??

damn!... what's the original GHz speed?

interesting, http://hwbot.org/hardware/processor/core_2_q9550_2.83ghz

JonW
01-27-2010, 11:57 PM
XP64 LW64

Ava_Test-3
Mac Mini 2.0Ghz 16:25
940 @3.588 2:57
E5450 3:04
W5580 1:45

rob-b2
Mac Mini 2.0 Ghz 2:58
940 @ 3.588 31.6
E5450 28.9
W5580 17.9

erikals
01-28-2010, 12:14 AM
W5580 1:45

Holy Macaroni ! http://erikalstad.com/backup/anims.php_files/eek.gif

hunter
01-28-2010, 09:30 AM
Turn on HT in BIOS and you will have 16 threads. :)

I can't believe it wasn't enabled by our IT dept. Good catch.
Smoke
64 bit 9.6.1 - 1:50

32 bit 9.6 - 1:44

:thumbsup:

JonW
01-28-2010, 03:30 PM
Soth you have a keen eye to spot that one.

When I got my W5580 I gave it a test with HT turned off and It was about 30% slower.


W5580

Ave_Test-3
LW32 1:41
LW64 1:45

rob-b2
LW32 18.0
LW64 17.9

JonW
01-28-2010, 03:50 PM
With http://3dspeedmachine.com/?page=3&scene=39 LW 32 took longer.

W5580
LW 32 1:05
LW 64 1:03

ToMar
02-06-2010, 12:12 AM
Bookmarked this thread

Intel Celeron 2x 2,00 Ghz

Ava_Test-3
Run1: 16:43
Run2: 16:45
Run2: 16:34

Rob-b2
Run1: 3:24
Run2: 3:28
Run3: 3:32

AMD Opteron Series x2xx 8x 1.7 Ghz (cpus for 60 Bucks to start with something) will fallow asp I got money left over for Board and Memory (430€). My guess is performance will be between Intel Q6700 and Q9450.

Matt
02-06-2010, 03:55 AM
i don't render in 64bit, it's not optimized

Daaaaaaayyyyyyaaaammmmmm! I never knew that. That needs fixing ASAP.

Anyone have any experience overclocking Asus Striker Extreme NF680i / Intel Core2 Extreme Quad-Core QX6700 combo? Would you even try on air cooling?

Also, if you were to, do it in the BIOS or use some overclocking software? (If software, which one, there seems to be a few!)

Lots of questions, sorry!

erikals
02-06-2010, 04:31 AM
well, i assume, based on tests.
however i don't think it's worth the hazzle.
Core focus is more important imo. (or a FiberFX fix, needed badly)

overclocking,
i'm planning on overclocking my QuadCore 2 Q9550 2.83GHz
not just yet though.
the pc will be placed outside in a box to avoid noise.

air cooling,
air cooling can be good, though i have seen some cheap water cooling options.
but actually, quite a lot of people use air cooling
try to find some online English tests, kinda like this one (sorry, Norwegian text)
http://www.hardware.no/tester/?catId=100

not sure how to actually do it though... http://erikalstad.com/backup/anims.php_files/aiwebs_029.gif

linksys
02-06-2010, 04:50 AM
my test on win7x64 Core i7 920 @3.80ghz 3gb Ram EVGA GTX280

Layout 32bit:
rob - 29.0s
smoke - 2m36s

Layout 64bit:
rob -29.9s
smoke - 2m47s

Cageman
02-06-2010, 04:53 AM
on my QuadCore 2.6GHz it rendered in 4minutes 13seconds.
would you happen to have an i7 or Xeon? could you test render it for me?

No i7 or Xeon here, so this is just for fun. :)

LW9.6 x32, Vista x64

Workstation, [email protected], 8GB ram: 6m 52s
Mediastation, [email protected], 4GB ram: 6m 19s

Cageman
02-06-2010, 05:02 AM
...what's the point of x64 again?

64-bit is not about speed increase, but rather the ability to have almost unlimited RAM capabilities.

When talks about 64-bit computing started (way before there were such things avaliable for the consumer market), I talked to some computer-wizards about it and they put me on the right track when I told I hoped for way faster rendering capabilities, and they said it will not have a much impact on that, but rather the amount of data that can be processed.

So, I'm not at all surprised with the similar renderspeeds between LW x32 and x64.

linksys
02-06-2010, 05:10 AM
I think the upcoming gulftown 6 Core CPU´s will be a good investment, i wonder how they will perform. 12 Threads for rendering on one homebuild PC sounds good:thumbsup: get 3 of this and you will have a 36Threads Renderfarm at home:D
Arrgh i need more money;D

erikals
02-06-2010, 08:22 AM
what "worries" me is future needs to pay big buck for both CPU and GPU...

Mr Rid
02-06-2010, 04:32 PM
64-bit is not about speed increase, but rather the ability to have almost unlimited RAM capabilities....

Right, but what I meant is, what is the advantage of LWx64, if you can run LWx32 a tad faster on an x64 system with tons of RAM? Where does the LWx32 limitation come in?

Matt
02-06-2010, 05:11 PM
Right, but what I meant is, what is the advantage of LWx64, if you can run LWx32 a tad faster on an x64 system with tons of RAM? Where does the LWx32 limitation come in?

Running LW32bit on a 64bit system won't access the extra RAM, same limitation as if running on 32bit system.

Matt
02-06-2010, 05:13 PM
Just tried this on my home machine.

Custom build, Intel Quad Core 2 Extreme (QX6700) 2.66Ghz Windows 7 Pro, 64Bit, LightWave 9.6

Rendered frame 485 in 6m 51s

Right, time to upgrade I think, not happy with that speed at all.

Just tried this scene on the same box as above test, but using 32bit SSE version of LW.

Custom build, Intel Quad Core 2 Extreme (QX6700) 2.66Ghz Windows 7 Pro, 64Bit, LightWave 9.6 32Bit

Rendered frame 485 in 6m 03s

That's a reasonable amount faster.

64bit LW should NOT be any slower than 32bit on the same box, this is silly, needs fixing.

Mr Rid
02-06-2010, 05:21 PM
Running LW32bit on a 64bit system won't access the extra RAM, same limitation as if running on 32bit system.

I assumed that, but am wondering why I am able to load and render a x32 scene taking up 4.5gb ram. I thought the x32 limit was 2gb.

Lightwolf
02-06-2010, 06:44 PM
64bit LW should NOT be any slower than 32bit on the same box, this is silly, needs fixing.
Well, there's two reasons:
1) x64 bit apps use and need more memory than 32bit apps do by default - even if they both a use the exact same set of data (LWO/LWS), the internal storage requirements for the data structure as well as the code will bloat.
That in turn means more of the CPU cache is used, more RAM reads and writes are needed, etc...
2) Some CPUs (like all pre i3/i5/i7 Intel CPUs) don't have all CPU core optimizations available when operating on 64-bit code.

Cheers,
Mike

geo_n
02-07-2010, 07:51 AM
remember its not just your hardware , its your whole setup .

i would suggest amd over intel everytime

After athlonxp, intel has taken the lead price/perf wise.
Lightwave in particular performs very well on intel machines.
3dmax excels in amd machines but that was a long time ago.

geo_n
02-07-2010, 07:54 AM
I assumed that, but am wondering why I am able to load and render a x32 scene taking up 4.5gb ram. I thought the x32 limit was 2gb.

I'm able to load heavier scenes in 32bit lw in vista 64, than if I use 32bit lw in winxp 32.
Must be that 64bit handles memory management better than 32bit.
Some people say winxp 64 is better than vista 64. Best of vista and winxp is in winxp 64.

Soth
02-07-2010, 09:54 AM
my test on win7x64 Core i7 920 @3.80ghz 3gb Ram EVGA GTX280

Layout 32bit:
rob - 29.0s
smoke - 2m36s

Layout 64bit:
rob -29.9s
smoke - 2m47s

WOW... same speeds as my dual Xeon 2.27, how you overclocked i7 to 3.8?

erikals
02-07-2010, 10:22 AM
soth, check this guide...
http://www.newtek.com/forums/showpost.php?p=978858&postcount=41

Hopper
02-07-2010, 01:18 PM
you overclocked an Intel Q6600 to 4.1GHz ??

damn!... what's the original GHz speed?
The original speed is 2.4GHz. I had been stable at 3.8GHz for some time and every time I tried to push it further, everything worked great except rendering. It would get about 1/2 way through and lock up the system or blue screen. Then I lucked out and found a forum post by some guy that had the exact same setup - mobo, ram, etc... I used his settings and it worked like a charm.

There are records for this processor well into the 5GHz range. I've tried a couple of times and even got it to completely boot once or twice, but even with liquid cooling, the temps immediately went to the 70c range and locked up after a few minutes. As it stands, my temps are usually around 38-42c at idle. A cranked up render will get it to around 60 or so if it's a long one.

erikals
02-07-2010, 01:41 PM
it's strange that these can compete, as the Q6600 is a bit older...

Core 2 Quad Q6600
launch > nov 2006

Core 2 Quad Q9550
launch > feb 2008

i wonder though which one is the best to overclock...
you wouldn't happen to know, would you?...

3dWannabe
02-07-2010, 09:15 PM
Win7-64 (2) Intel XEON quad E5520 GTX 285

9.6.1 32 bit
Smoke: 2:14
robot: 25.7

9.6.1 64 bit
robot: 26.7

arrow1234
02-07-2010, 10:17 PM
if anyone has both windows xp and the recent windows(vista or 7). I really want to see the difference between the 2 windows. It bothered me so much about its performace that i was struggling if i should use win7 or winxp for days...
I am finally on windows 7 and if i know that winxp can make renders much faster, I will downgrade back to winxp in no time :((

anyone think different windows does a big deal on the time length of rendering ?

Hopper
02-07-2010, 11:59 PM
i wonder though which one is the best to overclock...
you wouldn't happen to know, would you?...
Nope. Not a clue with the 9550. I've never owned one or had one installed in one of the office systems to play with. The 6600's are known for being quite stable at 1.5x their normal speeds. They were also popular because of their price point vs. some of the newer chips (well.. "new" about 2 years ago anyway).

But I wouldn't discount the 9550. It might be a great CPU to OC. I'd check around on some of the OC forums (nVidia / ASUS geek boards). I've gotten a lot of rock solid information from some of the regulars there.

Soth
03-01-2010, 10:14 AM
I just got new processors to my new machine, have put one and here are test renders... will post with two tomorrow.


http://www.newtek.com/forums/attachment.php?attachmentid=82722&stc=1&d=1267463649
http://www.newtek.com/forums/attachment.php?attachmentid=82723&stc=1&d=1267463660

focus3d
03-02-2010, 02:16 PM
Interesting results on 27" iMac with i7 @ 2.8 and 8gb of ram. 32 bit LW 9.6

Smoke - 3m 35secs
Robo - 43.6

I must admit I was expecting better considering the other results.

jrandom
03-02-2010, 11:50 PM
Quad Core Intel i5 iMac 2.66 GHz

Smoke: 5 min, 10 sec

Urgh. Gah. Blech. I may just have to get that 12-core i7 Mac Pro when they come out...

Soth
03-10-2010, 11:37 AM
I got liquid cooling delivered today and I put my 2nd processor inside. :)

http://www.newtek.com/forums/attachment.php?attachmentid=82996&stc=1&d=1268246184

http://www.newtek.com/forums/attachment.php?attachmentid=82997&stc=1&d=1268246190

erikals
03-10-2010, 08:17 PM
:)

btw, 12 Core Magny-Cours Opteron looks to be interesting, but rather expensive.
$7000 for 4, so $1750 for 1.

http://www.theregister.co.uk/2010/02/22/opteron_6100s_on_ebay/
http://cgi.ebay.co.uk/MAGNY-COURS-set-of-four-G34-socket-AMD-12core-OPTERON_W0QQitemZ280464735612QQcmdZViewItemQQptZCO MP_EN_Networking_Components?hash=item414d003d7c

avkills
03-11-2010, 02:51 AM
I'll test these tomorrow on the Mac Pros at work; I don't want to know what my g5 does. :p

-mark

avkills
03-11-2010, 08:58 AM
Man those i7s smoke it. I have the 2008 Mac Pro models, pre Nehalem.

Here are my times.

Pepsmoke 2m50s

RobotReflections 32.3 Seconds

Still respectable though IMO.

-mark

Hieron
03-14-2010, 04:34 PM
remember its not just your hardware , its your whole setup .

i would suggest amd over intel everytime

You have got to be joking....


Great times with the I7's and the newer Xeons ofcourse.. A dual socket Xeon is tempting, but I'm waiting for the 6 core extreme edition coming up and just OC the crap out of it.... Soth, can you overclock that dual socket monster?

ToMar
08-14-2010, 11:14 PM
I think we had none of those on the list

Windows 7 Pro 64bit

Dual AMD Opteron 2344 HE (8x1.7 Ghz, 65nm from 2006) 8 GB Ram

Robo

LW9.6 64bit: 55s
LW9.6 32bit: 52,9s
LW9.6 32bit in one Segment (600 MB Segment memory): 50,3s

Ava_test-3 (Pepsmoke)

LW9.6 64bit: 5m18s
LW9.6 32bit: 4m35s

Red Teapots

LW9.6 64bit: 3m18s
LW9.6 32bit: 3m15s

VirtualFM
08-16-2010, 10:10 AM
I'm glad this thread reappeared, as I have missed it and I love benchmarks and comparing systems... it's a lost of time, I know, but what can I do?! :help:

Well, i was pretty surprised with mny results. I mean, I have this machine for almost 3.5 years and I was thinking I was in need to replace it.. even if I don't really feel it's slow... it's just a matter of "need to reinstall the system and need more disk space". Anyway...

I got 4'27" for the first scene (even with lots of programs opened: Maya, Photoshop, tons of utils like Mozy Backup and time trackers and DropBox and all that tiny little utils that do nothing most of the time but sit there spending cycles, Opera with about 100 tabs opened, etc!)

Then I closed most of what I could and got an amazing 4'01" (well, amazing for what I expected!)

I tried the same in LW64 and got 5'38", which sucks bigtime... until I saw the results in this thread I was erroneously convinced that LW64 would render faster than LW32... I've been loosing rendering time for a longtime!

As for the big picture robot scene I got 55.5 seconds.

Well, I still think I should upgrade the computer, but looks it's not old at all! Just feels like it :-)

To the user that said "i would suggest amd over intel everytime", well, I was of that opinion as well... but that was eons ago, when an AMD X2 was screaming, around 2003/2004, I guess. It's a longtime since they became near the power of an Intel!

erikals
08-16-2010, 02:41 PM
an i7, or better, an overclocked i7 still is the best thing, rather amazing really compared to the price :]
http://www.newtek.com/forums/showpost.php?p=978422&postcount=34

jrandom
08-16-2010, 02:57 PM
I can't wait to get my shiny new 12-core i7 Mac so I can throw these scenes at it.

erikals
08-16-2010, 03:19 PM
now That will be an interesting test :]

ToMar
08-17-2010, 07:51 PM
I'd rather like to see one of these: http://www.tyan.com/product_SKU_spec.aspx?ProductType=MB&pid=670&SKU=600000180

especially after reading the German "c't" No# 9/2010 "AMD 12 core versus Intel 8 core" (and my last power bill)


@VirtualFM. May we know what your system is?

jrandom
08-17-2010, 08:35 PM
I'd rather like to see one of these: http://www.tyan.com/product_SKU_spec.aspx?ProductType=MB&pid=670&SKU=600000180

Ooo! Maybe when I go to build my render cluster next year I'll get some of these. (I use a Mac as my primary system for many different reasons so I'll gladly pay the Apple tax for it. On a render cluster... not so much. :) )

ToMar
08-17-2010, 08:46 PM
Ooo! Maybe when I go to build my render cluster next year I'll get some of these. (I use a Mac as my primary system for many different reasons so I'll gladly pay the Apple tax for it. On a render cluster... not so much. :) )

"Next year" is a good call. I would not (2P yes, 4P no) buy one until bulldozer support is 100% confirmed. And that will not be until next year.

Soth
08-18-2010, 02:23 AM
I'd rather like to see one of these: http://www.tyan.com/product_SKU_spec.aspx?ProductType=MB&pid=670&SKU=600000180
I am wondering if the price will be similar to this system:
http://www1.euro.dell.com/uk/en/business/Servers/pedge_r905/pd.aspx?refid=pedge_r905&s=bsd&cs=ukbsdt1 :O

3dworks
08-18-2010, 02:42 AM
here the results with apple macpro (2009 model) 2 x xeon x5550 'gainestown' 4 [email protected] ghz, 12 gbyte ram:

OSX 10.6.3 with layout 9.6.0 build 1539 32 bit
ava_test-3 (smoke): 1 min 58 sec
rob-b2 (robohead): 24.1

OSX 10.6.3 with layout 9.6.1 open beta build 1553 64 bit
ava_test-3 (smoke): 1 min 40 sec
rob-b2 (robohead): 21.6 sec

windows 7 pro 64 (booted) with layout 9.6.1 open beta build 1553 64 bit
ava_test-3 (smoke): 2 min 00 sec
rob-b2 (robohead): 21.4 sec
sec

i took the fastest of 3 results per test.

cheers

markus

Ryan Roye
08-19-2010, 01:44 PM
Heh, if anyone's interested on how sluggish the first posted render is on yesterday's technology:

Athlon 64 +6000 (dual core)
2 gig ram

16 minutes to render.

ToMar
08-19-2010, 09:13 PM
I am wondering if the price will be similar to this system:
http://www1.euro.dell.com/uk/en/business/Servers/pedge_r905/pd.aspx?refid=pedge_r905&s=bsd&cs=ukbsdt1 :O

That s the base Price for an "old" 6 x 4 core system £6,739 = $10'516

Hmm... Top of the Line Twelve-Core AMD Opteron™ 6176 SE = $1,386 MSRP x4 = $5,544
Motherboard will be around = $900
Case (chieftec) will go for about $100
PSU with 2x EPS12V 8 Pin will be $ 250
___________________________________

That leaves you with $3722 for DDR III and GPU of your Choice.

Paddie
08-20-2010, 12:30 AM
Hi,

I´ve tried it just for fun on my Dual Xeon (Prestonia) 2.8 Ghz. (Dell Precission 650 must be about 9 Years old ;-)).

Pepsmoke: 8m 22s
Robot: 2m 58s

Greets

Paddie

VirtualFM
08-20-2010, 03:08 PM
@VirtualFM. May we know what your system is?

Sorry! Forgot about that!

It's a Quad Core Q9550 originally running at 2.8 Ghz but overclocked to 3.13Ghz (I don't like to overdo it, specially when running 24/24). Ram is 8Gb, not that it matters much.

VirtualFM
08-20-2010, 03:30 PM
Heh, if anyone's interested on how sluggish the first posted render is on yesterday's technology:

Athlon 64 +6000 (dual core)
2 gig ram

16 minutes to render.

Cool! :-)

Anyone has an old Pentium Pro 200Mhz at hand?! ;-)

I feel mesmerized when comparing our current technology with what we had when we first started! For example, an old Amiga 4000 rendered the "Textures" benchmark scene in about 29 minutes (1749 seconds). I bought a "Warp Engine" acceleration card (about $1500 at the time) and it was down to 10 minutes or so (628 seconds). Then I got a Pentium Pro at 200Mhz and it was down to 42 seconds. Now it takes 0.8 seconds and is no good for comparison anymore, as is is so fast! But in short this would mean my current machine is 2186 times faster than my first one (for that scene anyway!)! And costs less!

jrandom
08-20-2010, 03:36 PM
If the prices on the 12-core AMD chips come down in price enough next year:

12-core processor * 4 processors per motherboard * 4 motherboards put into a simple plexiglass case == 192-core render cluster in just over 1 square foot. With 2.22GHz processors that's nearly half a terahertz of rendering power.

zapper1998
08-21-2010, 04:21 PM
If the prices on the 12-core AMD chips come down in price enough next year:

12-core processor * 4 processors per motherboard * 4 motherboards put into a simple plexiglass case == 192-core render cluster in just over 1 square foot. With 2.22GHz processors that's nearly half a terahertz of rendering power.

wow thats some horse power wahoooo

ToMar
09-03-2010, 11:40 AM
last update on this machine:

Windows 7 Pro 64bit

Dual AMD Opteron 2425 HE (12x2.1 Ghz, 45nm from 2009) 8 GB Ram

Robo

LW9.6 32bit in one Segment (600 MB Segment memory): 29s

Ava_test-3 (Pepsmoke)

LW9.6 32bit: 2m33s

Red Teapots

LW9.6 32bit: 1m45s

all include frame display wich can take up to 5s.

K-Dawg
09-03-2010, 03:32 PM
How do you connect 4 Motherboards together so it runs as 1 System?

I'd like to know that. That would be one heck of a render machine.

ToMar
09-04-2010, 12:06 AM
How do you connect 4 Motherboards together so it runs as 1 System?

I'd like to know that. That would be one heck of a render machine.

Unless you are talking about jrandom's Plan:

Motherboard with HTX-Pro connector, the Propper Adapter (e.g. Pathscale InfiniPath), Linux. And there ends my klowlege allready.

Elmar Moelzer
09-04-2010, 07:14 AM
Just for fun, I tried rendering Pepsmoke on my laptop. It only has a Core2Duo with 2.1Ghz and 3GB Ram, running Vista32bit.
I got 13 mins 23sec, which is not all that bad considering how old the CPU is and even then it always was on the low end of the lineup.
It is quite amazing how much faster modern CPUs got though.

ToMar
09-04-2010, 10:47 AM
BTW, for those who did not read the fine print: Windows 7 may support 256 cores. BUT ONLY 2 (TWO) CPU sockets.

jrandom
09-04-2010, 02:07 PM
BTW, for those who did not read the fine print: Windows 7 may support 256 cores. BUT ONLY 2 (TWO) CPU sockets.

Well, how hard could putting 128 cores on a single die be? :P

I think I got 1min 17sec for pepsmoke, but I was compiling boost at the time as well so I'm not sure how accurate the time is. I'm guessing it's pretty close since Lightwave only uses 16 of the 24 available hardware threads.

This brings up another thing: I have 24 hardware threads because of hyperthreading, but how well does that actually work for floating-point intensive operations? Last I heard, hyperthreading was good for memory waits and the mixing of integer and floating point instructions but I may be a little out of date on my knowledge of today's processors.

ThriJ
09-08-2010, 06:44 PM
Haha, oh all you and your fancy machines, I rendered pepsmoke in only 44m an 36s on my Pentium 4!


Oh yeah still got it! :boogiedow

jrandom
09-08-2010, 07:31 PM
If only there was a patch for 9.6 that got rid of the 16-thread limitation. Why is that even there, anyways?

BTW, I re-rendered and found I got my numbers wrong. 16 threads: 1min 40sec.

ThriJ
09-15-2010, 06:22 PM
Ok, Got my new rig setup, so I did a serious test this time.:D

I7 975x processor 3.33ghz
12gb RAM
Rendered with 8 threads

LightWave 32Bit. pepsmoke: 2min 48sec

LightWave 64Bit. pepsmoke: 3min 1sec


This is with no Overclocking to the 975. At full load I am not coming even close to TjMax, so I may try a little overclocking a bit later on for fun.

Man, going from around 45 min to around 3 min is so incredible!:eek::thumbsup: There is so many artistic possibilities opening up, this is so exciting!!!:dance:



Man jrandom 1m40s is impressive! I bet the thread limit will be gone way before the LW10 cycle is over.

jrandom
09-15-2010, 09:29 PM
Man, going from around 45 min to around 3 min is so incredible!:eek::thumbsup: There is so many artistic possibilities opening up, this is so exciting!!!:dance:

You'll be back up to 45 minutes in no time, I guarantee it. Scenes grow in complexity to fill all available processing power. :)

lardbros
09-16-2010, 04:10 AM
BTW, for those who did not read the fine print: Windows 7 may support 256 cores. BUT ONLY 2 (TWO) CPU sockets.

I've looked for this everywhere, but cannot find this info... could you point me towards somewhere where it shows this limit?

You could always install Windows Server which supports whatever you want... don't know the limit. Although Lightwave isn't officially on the supported platforms, i'm sure it'd work, we don't have any problems on our 4 cpu machine and 3ds max

erikals
09-16-2010, 06:25 AM
strange, it seems to be right,
http://community.spiceworks.com/topic/109200

4dartist
09-16-2010, 08:25 AM
3:05 for smoke on 1st gen mac 8core 3.0ghz 16gb ram LW9.6 32bit.

OFF
09-16-2010, 08:56 AM
PhenomIIX6 1055

JonW
09-17-2010, 01:04 AM
Mac Dual 533Mhz

Rob-b2 - 16:02

erikals
09-17-2010, 08:51 AM
the rendertimes seems to vary for the Ava scene (http://www.newtek.com/forums/showpost.php?p=977395&postcount=1) for each render,
but according to my test Win7 seems to be twice as slow as Vista (i can't believe it...)


Quad Core 2 - 2.83Ghz - Win7Ult, 64Bit, LW96-32
7m 25s
9m 34s
9m 18s

Quad Core 2 - 2.83Ghz - VistaUlt, 64Bit, LW96-32
4m 13s

...the PC is the same... http://erikalstad.com/backup/anims.php_files/aiwebs_029.gif

erikals
09-17-2010, 09:01 AM
now this is strange, the Rob scene is 100% the same rendertime on Win7 and Vista...


Quad Core 2 - 2.83Ghz - Win7Ult, 64Bit, LW96-32
55.2s

Quad Core 2 - 2.83Ghz - VistaUlt, 64Bit, LW96-32
55.2s

erikals
07-31-2011, 06:48 AM
win7x64, i7 2600K 3.40GHz, Corsair H80 watercooler, 8GB ram

overclocked to 4.5GHz... so easy i could cry... just a press of a button in the Bios,
motherboard, Gigabyte GA-P67A-UD4-B3

Layout 32bit:
smoke - 2m 4s
roboto - 21s

just do it!
 

Soth
07-31-2011, 07:50 AM
4.5Ghz, wow :)

erikals
07-31-2011, 07:55 AM
with a bit better cooling (hot today) i can push it up to 5GHz :]

zapper1998
08-01-2011, 01:08 PM
will try with new cpu and water cooling

amath
08-02-2011, 07:56 PM
2m 19s
Smoke

2-6core opteron 4180 2.6
win-7 64

AmigaNewTek
08-03-2011, 03:14 AM
Maybe that CPU could help: http://www.youtube.com/watch?v=G0i3DB21rcg&feature=related

Soth
11-15-2011, 02:34 AM
[I am rendering big images now, 40-60 minutes each.]

Greetings from the world of rendering, where I render, render, render some more and sometimes - when there is nothing to render - I do some render tests... as I did today for example:

I am testing VirtualBox (http://en.wikipedia.org/wiki/VirtualBox) ATM, tried Parallels Workstation (http://en.wikipedia.org/wiki/Parallels_Workstation) but there is a limitation to 12 threads per VM so it is no use here.

dual Xeon X5650 2.66, Lightwave 9.6.1 x64

Ava test 3 / smoke



VirtualBox 32 virtual threads* 2008R2 SP1 1m 06.4s 31% faster than not virtualised**
as second scene loaded 1m 10.5s
after restarting lw 1m 1.4s
native, windows Server 32threadds in LW 1m 37.5s
native, windows Server... just 2nd go 1m 36.3s
native, windows Server 24threads in LW 1m 40.4s 45% slower than not virtualised**
VirtualBox 24 virtual threads* 2008R2 SP1 2m 20.9s
32threads in Lightwave 2m 03.5s

same scene, 10x bigger resolution
I am doing it now..

Rob-b2



VirtualBox 32 virtual threads 0m 13.2s
native, windows Server 32threadds in LW 0m 17.9s
native, windows Server 24threads in LW 0m 18.1s
VirtualBox 24 virtual threads* 2008R2 SP1 0m 24.3s
32threads in Lightwave 0m 23.5s


* - VirtualBox announces in RED that this is not optimal situation.
** - it might be just odd behaviour for something that renders so fast, I will try to render same scene 10x bigger to see if the (massive) difference will persist.

OK that is it for now, I am rendering same scenes 64 times bigger, because I do not think that the anomality will occur in that case.

JonW
11-15-2011, 03:58 AM
Short frame rendering times, using a SSD was quicker. Maybe RAID some SSDs is quicker still, & 10 gigabit network is worth looking into.

http://forums.newtek.com/showthread.php?t=123208

Soth
11-15-2011, 04:20 AM
ok

Markus Gröteke suggested using watch measuring the speed and apparently it does not work very well in virtual machine indeed :P

and virualising is just slower, much slower, cinebench drops from 14.5 to 8.9 if u render in virtualbox

but if u want nice cinebench score you know what to do (yes it works my cinebench on virtual computer is 20% higher)

madno
11-18-2011, 11:29 PM
Dual Xeon E5520 2.27 GHz / 12 GB Ram / Win7 Ult. 64 Bit / LW 10.1 64

_______ HT off (8 core) | HT on (16 core)
--------------------------------------------------
Smoke: ________3m 21.0s | 2m 34.0s
--------------------------------------------------
Robot: ___________32.9s |____26.4s

Danner
11-19-2011, 07:00 AM
I7 980 stock with no OC

Smoke 2m 23s

robot 25. 2s


(vista 64)

Danner
11-19-2011, 07:25 AM
I had the suspicion this benchmark was not really meaninful unless we all were using the same version of the sofware so I tested with other versions, I'm sure the OS has an influence on this too.

Versions tested:

Lw 10 - 64bits
-------Smoke__2m 23s
-------Robot______25.2s

lw 9.6 -64bits
-------Smoke__2m 12 sec
-------Robot _____23.1s

lw 9.6 -32bits
-------Smoke__2m 11 sec
-------Robot______23.2


The conclusion is that there is not that much difference. In this scenes at least. I should have minimized my background tasks to rule out variables.. but who can wait for a 2 minute render without surfing a bit? ;o)

Olof
12-01-2011, 02:52 AM
i7 2600k 4.8 mhz
smoke 2 min 18 sec
robot 21.2 sec

JonW
12-01-2011, 03:43 AM
That's a fast render for MHz!

geo_n
12-01-2011, 07:57 AM
love testing benchmarks.

lw 9.6 32bit

laptop win7 64 - spec on my sig
rob = 69 sec
smoke = 6min 46 sec

q6600 vista 64 - stock
rob = 59.6 sec
smoke = 5min 36 sec

corei3 540 win 7 64 - not bad for stock dualcore
rob = 68 sec
smoke = 6min 22 sec

JonW
12-01-2011, 02:08 PM
We really need something like Cinebench. A Lightwave version can't be that difficult to put together by someone who knows what they are doing, & it should be a sticky thread.

Lightwolf
12-01-2011, 02:12 PM
We really need something like Cinebench.
http://www.spec.org/gwpg/apc.static/lightwave_9.6_info.html

However, also more elaborate than Cinebench and 9.6 only.

Cheers,
Mike

JonW
12-01-2011, 03:07 PM
I have seen this before & couldn't be bothered. Although Cinebench Benchmark is simple, that is the point of it & it becomes a clever marketing tool by default. It's simplicity is probably some of the reason Techreport, Anandtech & others use it.

I think most of us want something that is easy gives a basic quick answer. It won't cover everything & won't be accurate as a system overall. But it will be adequate for the average punter!

geo_n
12-20-2011, 10:18 PM
win7x64, i7 2600K 3.40GHz, Corsair H80 watercooler, 8GB ram

overclocked to 4.5GHz... so easy i could cry... just a press of a button in the Bios,
motherboard, Gigabyte GA-P67A-UD4-B3

Layout 32bit:
smoke - 2m 4s
roboto - 21s

just do it!
 


i7 2600k 4.8 mhz
smoke 2 min 18 sec
robot 21.2 sec

The 2600k seems slow considering the high oc. The phenom 1055T which is cheaper rendered in 30 sec on stock
http://forums.newtek.com/attachment.php?attachmentid=88153&d=1284648949
Wonder if OFF can do the smoke scene.

Anyone have a corei5 2500,AMD FX6100 to test? They're in my budget for another rendernode.

Sensei
12-20-2011, 10:43 PM
The 2600k seems slow considering the high oc. The phenom 1055T which is cheaper rendered in 30 sec on stock

2600k 4.8 GHz * 4 = 19.2 * 1.5 (HT) = 28.8 GHz equivalent
Phenom 3.85 GHz * 6 = 23.1 GHz

Then 28.8 / 23.1 = 1.25 (25% faster)

So.. 30.1 sec render on Phenom 3.85 GHz/6 C should be 24 sec on 2600k 4.8 GHz/4 C/8 HT

21.2 sec is better than above estimation.. ;)
Probably better handling some float maths.

geo_n
12-20-2011, 11:00 PM
2600k 4.8 GHz * 4 = 19.2 * 1.5 (HT) = 28.8 GHz equivalent
Phenom 3.85 GHz * 6 = 23.1 GHz

Then 28.8 / 23.1 = 1.25 (25% faster)

So.. 30.1 sec render on Phenom 3.85 GHz/6 C should be 24 sec on 2600k 4.8 GHz/4 C/8 HT

21.2 sec is better than above estimation.. ;)
Probably better handling some float maths.

I like the ;) part.
Just saying 2600k with mobo is a lot more expensive than the 1100t setup, maybe. Depends on components.
But the rendertime for both an ocd 2600k is only +-30% faster. Hmmm..wonder what's the render time for stock 2600k

Sensei
12-20-2011, 11:14 PM
21.2 * 4.8 / 3.4 = 30 sec ;)

MarcusM
12-21-2011, 01:28 AM
You can set layout task priority on high in my case it give 1-2 sec better result. 100408

Intel Core i7 950 3.07GHz (Blomfield) OC - 3.3GHz
Windows 7 Pro
LightWave 10.1 64bit

rob - 35.6 sec
smoke - 3 min 29 sec

:beerchug:

geo_n
12-21-2011, 01:46 AM
21.2 * 4.8 / 3.4 = 30 sec ;)

How about pentium 3.0ghz with 2600k?;)

OFF
12-21-2011, 06:10 AM
LW 10.1 x64/AMD phenom II 1055T (4.1ghz OC)/smoke scene
2min 51sec

Sensei
12-21-2011, 07:09 AM
You can set layout task priority on high in my case it give 1-2 sec better result

I wrote plugin which is doing it automatically during rendering in Layout..
TrueArt's Render Optimizer
http://www.trueart.pl
http://www.trueart.pl/Products/Plug-Ins/RenderOptimizer/Graphics/RenderOptimizer_Panel_1.png

MarcusM
12-22-2011, 12:34 AM
Intel Core i7 950 3.07GHz (Blomfield, 8 threads) OC - 3.7GHz
Mainboard ASUS Sabertooth X58
Windows 7 Pro
LightWave 10.1 64bit

rob - 32.1 sec
smoke - 3 min 13 sec

mis
12-22-2011, 12:58 AM
Sensai

do you mind explaining this formular

2600k 4.8 GHz * 4 = 19.2 * 1.5 (HT) = 28.8 GHz equivalent

how do you get 1,5 x more cpu power out of HT

what i know of HT only split the core in half wich
as i see it dont give you more cpu power but a more fluid system
shouldnt the ghz be the same ith or without hyperthreading

care to explain

geo_n
12-22-2011, 04:56 AM
Formula is a joke. Sensei is just joking. You don't compare ghz from totally different platform directly like that.

Lightwolf
12-22-2011, 05:25 AM
how do you get 1,5 x more cpu power out of HT
It's more like an extra 15%-25% for the current generation by Intel.


what i know of HT only split the core in half wich
as i see it dont give you more cpu power but a more fluid system
shouldnt the ghz be the same ith or without hyperthreading

Not quite. What it does do is make more efficient use of the computing pipeline of a single core.
Basically, most of the time when running code, not all parts of a core will be utilised. They sit idle because they're either not used at all or need to wait for other parts of the core to finish.
This is where HT jumps in. By running a second thread at the same time, those idle parts of the core can be put into use.

A good (but simplistic) analogy is a factory with machinery. To produce goods, some of the machines are used more or less... or need to wait for other machines to finish their work first.
If you add a second assembly line that uses the same machines, the result will not be a production that runs twice as fast. But since the machines can be used more efficiently you're likely to still gain some productivity.

Cheers,
Mike

Sensei
12-22-2011, 07:48 AM
1.5 multiplier came from mine own experience. I rendered some scenes with and without using HyperThreading in mine Core i7 930 2.8 GHz and received such difference in render times. f.e. what rendered 100 sec with 4 threads, rendered 66 sec with 8 threads HT. Core i7 930 and 2600K are not so far different architectures after all (which means probably equal/close speed of executions of expensive math operations such as sqrt() or similar).

It's just estimation (not quite without sense- what I estimated as 24 sec, true machine counted in 21 sec).
It would be nice if somebody without overclocked 2600K came here and tell what is real render time of robot scene, at original 3.4 GHz.

Sensei
12-22-2011, 08:14 AM
It's more like an extra 15%-25% for the current generation by Intel.


I didn't know you have i7..

15-25% is way too small, IMHO

f.e. I just rendered scene with 8 TH 27 sec, with 4 threads 40 sec, so it's 48% (nearly mine multiplier 1.5)

ch.schrul
12-22-2011, 08:38 AM
Just got my new hardware together... i7 3960K (currently at stock speed 3.3GHz / 3.9GHz), Asus R.O.G. Rampage Extreme iV, 32GB Ram, Win7 64bit Ultimate, Lightwave v11 beta.

roboscene: 20.7 seconds
smokescene: 2m 8 seconds

Chris

Edit: I've tried @ 4.25GHz (on air)
roboscene: 17.8 seconds (16.3% faster)
smokescene: 1m 47 seconds (19.6% faster)

geo_n
12-22-2011, 08:51 AM
if somebody without overclocked 2600K came here and tell what is real render time of robot scene, at original 3.4 GHz.

+1
I'd like to know if the 100US difference with 2500 is worth it. I never oc. I think its just better to buy speed than oc because electricity isn't cheap and components wear down faster when oc.

JonW
12-22-2011, 02:21 PM
Add up all the components & OS. Divide the price of the whole box by the number of GHz to get a price of $ per GHz.

More expensive CPUs can work out cheaper. You also need to factor in, how long you want to keep the computer before you do the whole exercise again, the days spent getting the computer up & running maybe worth a lot to you. It maybe worth spending a bit more to get 3 years out of the computer before sending it to the farm.

This example is a few months old:

2 x X5690 $3650
Supermicro X8DAL-i $486
Supermicro 865w power supply $373
Chenbro SR11269 $218
24GB Ram Corsair1333 ECC REG 4GB x 6 $775
2 x Noctua heatsinks $160
Intel X25-M SSD 120Gb $211
2 x 1TB HD $150
Gigabyte GTX560 $269
CD/DVD $38
Win7 professional $162

Total price au$6492.00 (41.52GHz) $156.35 per GHz


We are going to have some massive price rises with electricity in the next few years so it's worth looking at the components you put in a new box. One may not need that GTX580!

If you calibrate your monitor & have a good low level light environment, it's brightness should be somewhere around 90 to 100. As a result you will save more electricity. One of my Dell monitors setting is on minimum brightness to be as close to proper calibration as possible & the other 2 are very low. As a result there is a lot less heat coming off these things.

mis
12-22-2011, 03:04 PM
well i seee

well i run a Q6600 2.4ghz wich is currently clocked to 2.91 ghz

and whati can se from intel is that it is a cpu supporting hyperthreading
and what i can see about my motherboard
is that "p5b plus vista edition" that it is intel hyperthreaded ready

but i dont see hyperthreading listet anyware in the bios

how am i shure that its used then

i saw on his video that his Task manager changed from 4 to 8 cpu windows
mine display 4 windows as it is now
does that meen that ime not running hyperthreadded
or does that just meen that windows Xp doesnt display the threads in the task manager ?


any clue on that

JonW
12-22-2011, 06:42 PM
I think the Q6600 & x53xx CPUs are from the same family. My 53 CPUs don't have HT.

A current model 6 core CPU will eat these old CPUs for breakfast, even more so with radiosity.

mis
12-23-2011, 12:52 AM
so you say perhaps we gotta give that up for a test :)
i got q6600 at 2,91 to butt in with
:)

though ofcause theres a few christmas days here where i dont have time do to family
selebrations :)

but what cpu should i compare with then you think or is there a meshure
of it here already

mis
12-29-2011, 01:10 PM
just did a test on the first smoke scene

rendering on win xp on a Q6600 quadcore 2.4 clocked to 2.9

in 9,6 it took 5 minutes 11 seconds

in 10,1 it took 6 minutes 49 seconds
so over a minute more for 10.1 sadly

mis
12-29-2011, 01:16 PM
the robotreflection3 tests on my Q6600 clocked to 2.9 running xp

Says

Lightwave 9.6 55,7 seconds
Lightwave 10.1 58 seconds

mis
12-29-2011, 01:30 PM
Hopper 4.1 ghz for a Q6600 sounds like alot
mine is currently running 2.91

i take it your not running with standart cooling on it am i right ?

what did you run when you did 3.8 ? was that standart cooling ?

JonW
12-29-2011, 08:07 PM
Smoke LW 64 bit W5580 (getting a bit long in the tooth now! A pair of X5690s would be nice!)

9.6 1:45
10.1 1:54
11.0 1:54

A bit slower, although unfortunate it's gone backwards, I would still rather have the new features.

Eroneouse
02-20-2012, 09:31 AM
got to page 5 and decided without reading the rest I win for slowest render time of 19 mins 50 secs for avalanche scene, forum browesing and DL going at same time, gonna turn DL and browse off to see if it speeds up by 10 seconds or so. Machine specs as in sig.

Edit - My dual channel may now be single channel as they werent match pair of ram sticks to start with and I changed the slots they were in some time back, didnt check to see if DC was on or off after last change will do now though.

Edit 2 - Even though they werent matched pair of ram sticks DC still worked and registered at 667 mhz instead of 333 mhz once ram sticks were put in the correct slots for DC useage. Will be a while before i switch off DL though as its been creeping along at 50kps for last few days but today its at 300+kps, typical isn't it when I want to stop it it speeds up. DL is water margin old show from 70's :D

Edit 3 - First render on LW11 btw rerendering now on LW10.1 seems faster already :/ will add to this edit speed on 10.1 when finished... 17 mins 49 secs in 10.1 so a little faster, 10.1 in discovery mode as dongle out to use LW11 in trial mode dunno if it makes a difference to render speed but will put dongle back in and rerun tests when DL finnished and post results to this post later tonight.

Soth
03-13-2012, 04:46 AM
9.6.1 on 4x Opteron 6272 2.1

http://forums.newtek.com/attachment.php?attachmentid=102576&stc=1&d=1331635375

http://forums.newtek.com/attachment.php?attachmentid=102577&stc=1&d=1331635363

JonW
03-13-2012, 06:31 AM
I didn't think 9.6 did 64 threads?

LW11 would be interesting?

Soth
03-15-2012, 10:17 PM
I didn't think 9.6 did 64 threads?
it does, i mean 9.6.1 does


LW11 would be interesting?
I will, just need some spare time.

lw 10.1 smoke on 3930k overclocked to 4000Mhz 1:52.. will overclock futher :)

JonW
05-13-2012, 07:38 PM
I just tried this scene with HT off:

Here are my benchmarks for LW11.0 1:54 (114) HT on, & 2:25 (145) HT off. (I haven't got around to updating LW11)

So if you have a 60 minute render with HT off. It's only going to take 47 minutes with HT on. Obviously it's will vary a bit from render to render. Having HT off, is not the end of the world, & will help in other ways!

Soth
05-14-2012, 01:00 AM
Tests from our new render node, I am still making some tests but looks like we will move to this new machines.

Ava_Test - 2:04 / 124.5
rob-b2 - / 24.5

If rest of the tests will go OK we will have 32 of them next month. It will be 10 cores per node only and super massive processor cache so most of your scenes should cost bit less per Ghz than on 12. :)