can one of you guys with a 8800GTX do me a favour?

W

wizzackr

[H]ard|Gawd
Joined
May 5, 2000
Messages
1,579
Last year Fallguy helped me get professional 3D-application benchmarks for the 7900GTX, based on the rather common SpecViewPerf benchmark. The original thread can be found here.

Now, with g80 released, it would be really cool to see how it fairs in professional applications. Therefore it would be really great if one of you could download the SpecViewPerf-benchmark here and run it on nVidia's latest and greatest offering.

As in the original thread, it'd be interesting to see if the last generation's fireGLs/Quadros still outperform the g80-based cards. Thanks already for your help, guys,
J
 
Highendtoys said:
I will do it on a GTS for you. Give me a sec.
Click to expand...

Thanks a bunch already, mate! I am really anxious to see how this one turns out! :)
 
Run All Summary

---------- SUM_RESULTS\3DSMAX\SUMMARY.TXT
3dsmax-04 Weighted Geometric Mean = 17.04

---------- SUM_RESULTS\CATIA\SUMMARY.TXT
catia-02 Weighted Geometric Mean = 12.33

---------- SUM_RESULTS\ENSIGHT\SUMMARY.TXT
ensight-03 Weighted Geometric Mean = 15.97

---------- SUM_RESULTS\LIGHT\SUMMARY.TXT
light-08 Weighted Geometric Mean = 14.51

---------- SUM_RESULTS\MAYA\SUMMARY.TXT
maya-02 Weighted Geometric Mean = 31.92

---------- SUM_RESULTS\PROE\SUMMARY.TXT
proe-04 Weighted Geometric Mean = 14.66

---------- SUM_RESULTS\SW\SUMMARY.TXT
sw-01 Weighted Geometric Mean = 15.61

---------- SUM_RESULTS\TCVIS\SUMMARY.TXT
tcvis-01 Weighted Geometric Mean = 3.815

---------- SUM_RESULTS\UGNX\SUMMARY.TXT
ugnx-01 Weighted Geometric Mean = 4.682

System Specs Here

Please note that I am back on the phase and the proc was at 3.6.
 
That is pretty sad, let me switch the drivers out for the public drivers, I am still using the early reviewer set,
 
Highendtoys said:
That is pretty sad, let me switch the drivers out for the public drivers, I am still using the early reviewer set,
Click to expand...

I hear you - still, thanks a bunch, highend, helped a lot. It's too sad how good nVidia is at crippling their power-houses when it comes to professional applications :D - luckily ATI is not as good at it, so I have hopes for r600 to deliver at least equally as much punch as my good'ol V5200...
 
wizzackr said:
I hear you - still, thanks a bunch, highend, helped a lot. It's too sad how good nVidia is at crippling their power-houses when it comes to professional applications :D - luckily ATI is not as good at it, so I have hopes for r600 to deliver at least equally as much punch as my good'ol V5200...
Click to expand...

That's why they have Quadro FX...
 
just wait, someone will probably come up with a rivatuner patch script to allow the Quadro drivers to be installed on the 8800 as soon as a G80-based Quadro is available......
 
Silus said:
That's why they have Quadro FX...
Click to expand...
Obviously - but that can easily mean you have to pay 4 grand for it. I was hoping that even though it is crippled in the drivers it wouldn't be as bad - with THAT much of a performance improvement over the 7900GTX. I just want a card that can run all the latest games AND still give me decent viewport performance for my computer at home ;)


coolie_d said:
just wait, someone will probably come up with a rivatuner patch script to allow the Quadro drivers to be installed on the 8800 as soon as a G80-based Quadro is available......
Click to expand...
Not going to happen IMHO - nVidia effectively disabled that workaround on the hardware level for their entire last generations of cards already IIRC. With ATI I wouldn't know for sure, though...
 
wizzackr said:
Obviously - but that can easily mean you have to pay 4 grand for it. I was hoping that even though it is crippled in the drivers it wouldn't be as bad - with THAT much of a performance improvement over the 7900GTX. I just want a card that can run all the latest games AND still give me decent viewport performance for my computer at home ;)



Not going to happen IMHO - nVidia effectively disabled that workaround on the hardware level for their entire last generations of cards already IIRC. With ATI I wouldn't know for sure, though...
Click to expand...
Wow, didn't know that..... as far as ATI, speaking for myself, I know the softfiregl mod works at least through the x1800xt level, as I am personally running one as a FireGL v7350 card..... scores:

Run All Summary

---------- SUM_RESULTS\3DSMAX\SUMMARY.TXT
3dsmax-04 Weighted Geometric Mean = 23.76

---------- SUM_RESULTS\CATIA\SUMMARY.TXT
catia-02 Weighted Geometric Mean = 28.32

---------- SUM_RESULTS\ENSIGHT\SUMMARY.TXT
ensight-03 Weighted Geometric Mean = 21.25

---------- SUM_RESULTS\LIGHT\SUMMARY.TXT
light-08 Weighted Geometric Mean = 25.65

---------- SUM_RESULTS\MAYA\SUMMARY.TXT
maya-02 Weighted Geometric Mean = 81.37

---------- SUM_RESULTS\PROE\SUMMARY.TXT
proe-04 Weighted Geometric Mean = 21.79

---------- SUM_RESULTS\SW\SUMMARY.TXT
sw-01 Weighted Geometric Mean = 35.81

---------- SUM_RESULTS\TCVIS\SUMMARY.TXT
tcvis-01 Weighted Geometric Mean = 10.78

---------- SUM_RESULTS\UGNX\SUMMARY.TXT
ugnx-01 Weighted Geometric Mean = 18.74
 
wizzackr said:
Obviously - but that can easily mean you have to pay 4 grand for it. I was hoping that even though it is crippled in the drivers it wouldn't be as bad - with THAT much of a performance improvement over the 7900GTX. I just want a card that can run all the latest games AND still give me decent viewport performance for my computer at home ;)
Click to expand...

As I said so many times, in this forum, these companies are in it for the money. Why would they provide you amazing gaming performance and amazing viewport performance in one card, when they can do it with two and get more money ?
If ATI's cards actually let you do this, then you should try one instead. But I'm not seeing ATI's advantage in it. So all I can say is (and It's my own guess), I don't think R600 will let you mod it, to a FireGL card...
 
coolie_d said:
just wait, someone will probably come up with a rivatuner patch script to allow the Quadro drivers to be installed on the 8800 as soon as a G80-based Quadro is available......
Click to expand...

I'm a video card nit-

So if (well, when) a G80 based Quadro Card comes out, will there be any instance where a comsumer level 8800 series (G80) card would have an advantage?

Do they put any limitations in the Quadros to give the consumer based cards an advantage, or is the Quadro the all-stops-pulled solution?'

I see a really awesome thing with that Quadroplex box they have out there- imagine sticking some G80's in that!.... (and imagine me getting away with robbing a bank or casino..allright- wakeup!!) :D

Holy $hit on a stick
 
if installing the softmod in an ATi card, will it lower gaming performance? or it just simply increases 3d app performance?

as i am trying to get into 3d work and got an x1900XT......
 
MrGuvernment said:
if installing the softmod in an ATi card, will it lower gaming performance? or it just simply increases 3d app performance?

as i am trying to get into 3d work and got an x1900XT......
Click to expand...

I can only comment on older ATI hardware, as I owned a 9700pro soft-modded to fireGL X1 a couple fo years ago. I did not affect the gaming-performance by much (2-5% max.), but the viewport performance in 3D-packages skyrocketed.

Definitely worth the try, if you ask me, as you can always revert back to using normal drivers after cleaning the old ones in case you don't like it.

The only drawback I found using a softmodded card was the fact that you cannot just update drivers that often - it becomes a royal pain in the buttocks to have to softmod new drivers whenever they come out, just to find out they changed the protection mechanism, go search for new soft-mod tools, re-install everything etc etc. - you get the point.
 
Highendtoys said:
Run All Summary

---------- SUM_RESULTS\3DSMAX\SUMMARY.TXT
3dsmax-04 Weighted Geometric Mean = 17.04

---------- SUM_RESULTS\CATIA\SUMMARY.TXT
catia-02 Weighted Geometric Mean = 12.33

---------- SUM_RESULTS\ENSIGHT\SUMMARY.TXT
ensight-03 Weighted Geometric Mean = 15.97

---------- SUM_RESULTS\LIGHT\SUMMARY.TXT
light-08 Weighted Geometric Mean = 14.51

---------- SUM_RESULTS\MAYA\SUMMARY.TXT
maya-02 Weighted Geometric Mean = 31.92

---------- SUM_RESULTS\PROE\SUMMARY.TXT
proe-04 Weighted Geometric Mean = 14.66

---------- SUM_RESULTS\SW\SUMMARY.TXT
sw-01 Weighted Geometric Mean = 15.61

---------- SUM_RESULTS\TCVIS\SUMMARY.TXT
tcvis-01 Weighted Geometric Mean = 3.815

---------- SUM_RESULTS\UGNX\SUMMARY.TXT
ugnx-01 Weighted Geometric Mean = 4.682

System Specs Here

Please note that I am back on the phase and the proc was at 3.6.
Click to expand...

Damn - the card is still heavily capped in pro apps. :( It's kinda sad that a softmodded 6800GT is almost three times as fast as the 8800GTX in some pro apps. I did a search through the RivaTuner site and found a guy running SpecViewperf 9 on his stock Core Duo E6300, 6800GT softmodded to Quadro 4000 and here are his scores. Who sets this apparent performance cap on consumer cards, and why can't they raise it between generations? Look at the SpecViewperf scores, it looks like little has changed performance wise between the consumer GF6, GF7, and GF8 series of cards when running pro level apps. I can understand having performance and application optimizations for those willing to pay for Quadro series of cards, but artificially crippling the consumer series to slow them down just doesn't make any sense to me. The 3d workstation market is changing quite a bit as of late, with softeware applications dropping in price so that the average joe can get into the market, and the new dual core processors and other high end consumer level components equaling and in some case out performing the high end workstations of just last year. Yet companies such as Nvidia and ATI still expect us to drop more for a non-crippled video card than it costs to put together the rest of the system componenets.

Core E6300 @ stock / 6800GT

POV 6800GT PCIe @ Quadro FX4000
Tested on C2D E6300 1Go PC5300 DDR2, drivers Quadro 91.36
3dsmax-04 = 22.58
catia-02 = 29.43
ensight-03 = 16.62
light-08 = 25.56
maya-02 = 72.13
proe-04 = 21.58
sw-01 = 34.11
ugnx-01 = 8.734
tcvis-01 = 6.944

Lots of good info in this thread.

http://forums.guru3d.com/showthread.php?t=172822&page=7&highlight=3dsmax-04
 
Here's my scores on a FX 4500 SLI system for comparision sake:

Run All Summary

---------- SUM_RESULTS\3DSMAX\SUMMARY.TXT
3dsmax-04 Weighted Geometric Mean = 30.20

---------- SUM_RESULTS\CATIA\SUMMARY.TXT
catia-02 Weighted Geometric Mean = 37.69

---------- SUM_RESULTS\ENSIGHT\SUMMARY.TXT
ensight-03 Weighted Geometric Mean = 23.50

---------- SUM_RESULTS\LIGHT\SUMMARY.TXT
light-08 Weighted Geometric Mean = 36.73

---------- SUM_RESULTS\MAYA\SUMMARY.TXT
maya-02 Weighted Geometric Mean = 59.93

---------- SUM_RESULTS\PROE\SUMMARY.TXT
proe-04 Weighted Geometric Mean = 33.11

---------- SUM_RESULTS\SW\SUMMARY.TXT
sw-01 Weighted Geometric Mean = 43.65

---------- SUM_RESULTS\TCVIS\SUMMARY.TXT
tcvis-01 Weighted Geometric Mean = 10.87

---------- SUM_RESULTS\UGNX\SUMMARY.TXT
ugnx-01 Weighted Geometric Mean = 16.49

I just have it set up for split frame rendering, what's the best SLI config for this test?
 
ND40oz said:
Here's my scores on a FX 4500 SLI system for comparision sake:

Run All Summary

---------- SUM_RESULTS\3DSMAX\SUMMARY.TXT
3dsmax-04 Weighted Geometric Mean = 30.20

---------- SUM_RESULTS\CATIA\SUMMARY.TXT
catia-02 Weighted Geometric Mean = 37.69

---------- SUM_RESULTS\ENSIGHT\SUMMARY.TXT
ensight-03 Weighted Geometric Mean = 23.50

---------- SUM_RESULTS\LIGHT\SUMMARY.TXT
light-08 Weighted Geometric Mean = 36.73

---------- SUM_RESULTS\MAYA\SUMMARY.TXT
maya-02 Weighted Geometric Mean = 59.93

---------- SUM_RESULTS\PROE\SUMMARY.TXT
proe-04 Weighted Geometric Mean = 33.11

---------- SUM_RESULTS\SW\SUMMARY.TXT
sw-01 Weighted Geometric Mean = 43.65

---------- SUM_RESULTS\TCVIS\SUMMARY.TXT
tcvis-01 Weighted Geometric Mean = 10.87

---------- SUM_RESULTS\UGNX\SUMMARY.TXT
ugnx-01 Weighted Geometric Mean = 16.49

I just have it set up for split frame rendering, what's the best SLI config for this test?
Click to expand...

These are pretty low scores if you look at what was posted before. The POV 6800GT PCIe @ Quadro FX4000 (softmodded) scores a lot higher in maya. somethings wrong? are these even the same revisions of specViewperf? Something seems way off, as wto FX4500s should wipe the floor with a single 4000.


loafer87gt said:
Yet companies such as Nvidia and ATI still expect us to drop more for a non-crippled video card than it costs to put together the rest of the system componenets.
Click to expand...

Amen to that, I just wanted to point out that fireGLs are a lot cheaper than qudros, though - granted, performance-wise there often is a pretty wide gap, but in rare situations that would justify such a difference in price. Also, I find radeons to be slightly less crippled that gFs - lets see if that still holds true for r600...
 
wizzackr said:
These are pretty low scores if you look at what was posted before. The POV 6800GT PCIe @ Quadro FX4000 (softmodded) scores a lot higher in maya. somethings wrong? are these even the same revisions of specViewperf? Something seems way off, as wto FX4500s should wipe the floor with a single 4000.
Click to expand...

Yeah, I wonder if it doesn't like running in XP x64 and it runs windowed, is it supposed to run windowed or full screen?

When I look at the viewper.txt file, it shows unknown for CPU count, primary cache and secondary cache size. It also lists my memory size as 0 which is strange as I have 8 gigs. It also appears to run single threaded, is that right?
 
EVGA 8800 GTX + E6600

Run All Summary

---------- SUM_RESULTS\3DSMAX\SUMMARY.TXT
3dsmax-04 Weighted Geometric Mean = 17.25

---------- SUM_RESULTS\CATIA\SUMMARY.TXT
catia-02 Weighted Geometric Mean = 13.00

---------- SUM_RESULTS\ENSIGHT\SUMMARY.TXT
ensight-03 Weighted Geometric Mean = 23.99

---------- SUM_RESULTS\LIGHT\SUMMARY.TXT
light-08 Weighted Geometric Mean = 14.40

---------- SUM_RESULTS\MAYA\SUMMARY.TXT
maya-02 Weighted Geometric Mean = 38.33

---------- SUM_RESULTS\PROE\SUMMARY.TXT
proe-04 Weighted Geometric Mean = 14.48

---------- SUM_RESULTS\SW\SUMMARY.TXT
sw-01 Weighted Geometric Mean = 12.52

---------- SUM_RESULTS\TCVIS\SUMMARY.TXT
tcvis-01 Weighted Geometric Mean = 4.803

---------- SUM_RESULTS\UGNX\SUMMARY.TXT
ugnx-01 Weighted Geometric Mean = 5.665
 
Just so some comparison for others

i had 2 IE windows open, not sure if that affects this or not?

Intel Pentium D 840 (dual core 3.2ghz pressy based)
2g (2 x 1g) DDR2 667
ATi X1900 XT @ stock
74g Raptor


Run All Summary

---------- SUM_RESULTS\3DSMAX\SUMMARY.TXT
3dsmax-04 Weighted Geometric Mean = 8.812

---------- SUM_RESULTS\CATIA\SUMMARY.TXT
catia-02 Weighted Geometric Mean = 9.729

---------- SUM_RESULTS\ENSIGHT\SUMMARY.TXT
ensight-03 Weighted Geometric Mean = 10.78

---------- SUM_RESULTS\LIGHT\SUMMARY.TXT
light-08 Weighted Geometric Mean = 8.896

---------- SUM_RESULTS\MAYA\SUMMARY.TXT
maya-02 Weighted Geometric Mean = 11.64

---------- SUM_RESULTS\PROE\SUMMARY.TXT
proe-04 Weighted Geometric Mean = 6.311

---------- SUM_RESULTS\SW\SUMMARY.TXT
sw-01 Weighted Geometric Mean = 13.21

---------- SUM_RESULTS\TCVIS\SUMMARY.TXT
tcvis-01 Weighted Geometric Mean = 3.342

---------- SUM_RESULTS\UGNX\SUMMARY.TXT
ugnx-01 Weighted Geometric Mean = 7.853



nice to see some of my scores are higher then others, on one or two :)

Shall have to try the softmod drivers fr the x1900 Xt if i can find any.
 
I'm in professional DCC, and need to point out that most DCC is done in software and depend more on your cpu and system memory than your gpu, exspeaclly if the gpu does not off load any work from cpu such as all consumer videocards! The professional cards are setup to offload certain tasks that help, but if maya, 3dmax, metal ray, renderman, etc do not know the hardware is they will ignore the option and use only software rendering. The g80 series has a new option if you are running cuda based dev drivers to start off loading c++ code, which should be interesting to see what happenes with stable driver as it is possible to run shader code on the cuda plaform but right now that is limited to devs with a relationship with nvidia. So all specview is doing is telling how well your system renders in software. The only thing the g80 do in maya and 3dmax is display your image and so opengl rendering. If you want to do DCC go spend the money on a quadcore and as much ram as possilbe, on the other hand if it is a gameing machine then enjoy your blazing fast card(s).
I'm not saying the g80 are bad cards my gaming rig loves my 8800GTX, it just has no place (currently) in my renderfarm, unless cuda become avalible to more of the smaller devs, ie make less than 6 figures a project...
 
hey drakken, 'cuda' sounds interesting and i must admit i never even heared of it. do you have any links or further information on it?

drakken said:
I'm in professional DCC, and need to point out that most DCC is done in software and depend more on your cpu and system memory than your gpu
Click to expand...

i'm with you, but that does not make specViewPerf less relevant IMHO. once scenes include really heavy geometry or large textures or lots of displacement maps etc. the viewport performance needs to be up to snuff in order to not hamper productivity vastly. when we are working on larger jobs and manupulating objects on-screen becomes more of a slide-show than a productive process the amount of time to get stuff done skyrockets - and that costs money. so if you want to get an idea about what vid-card can give you what kind of oGL/D3D performance within the viewports, specviewperf does a decent job.

it does, of course, NOT give you any clue about hardware render capabilities (a la gelato, maya's native hardware renderer etc.), software render power (MR, RMFM, PRMan or whatever floats your boat) or whatever else comes later down the pipe.
 
Think compiler for your videocard... this generation nvidia decided to make a cpu for all intents and purposes it has local cache, it has brach prediction, it has 128 registers you can access and all the gpu functions are software this time around. That and with special cuda drivers
http://developer.nvidia.com/object/cuda.html
which are not even avalible under dev section of nvidia site you have to sign an application and proablaly an nda, that will alllow you to compile c++ code on the gpu. How this effects rendering, well all shader code is based on c++, so things like renderman and mental ray which are more parallel than in-order due to the nature of the compares and the math, means we are looking at cutting render time, as 128 registers vs 40 something on a x64 cpu is very nice, exspecially if I can use the cpu and gpu in tandem.

One of my conserns about specview is that it is using older versions of the apps maya 6.5, max 7, etc... many code updates that deal with opengl were addressed in maya 7.01 and maxons new version just replaced the entire gui, which means spec-view 903 the current version do not reflect it all, and vue was not even in the list and that was used in pirates of the carribbean, which would suggest that it is not a toy.

Oh I have to ask what pkg has the displacements effect anything other than the render? and the g80 has issues loading the texures into memory but once there is very smooth. The odd thing is spec-view did not reflect actually performence in maya 7 unlim, as I can load up the same scene file on the workstation and on my gaming rig and both deal with geometry the same, and my gaming rig only has a 4400 X2 which specview kills, yet the 8way op blasts through specview but still takes a hit when large texture 2048 or larger are loading into memory at first, then again I find animations generally cause more slugging than geomtry bay itself, I get the feeling the pkg decides many of issues.
 
FYI, I had been looking for this info a week ago when I decided against purchasing an 8800GTX and instead ordered Quadro FX3500 cards for my workstations. I had really been hoping that the new card architecture of the 8800 would provide enough of a boost to justify the purchase. Here are my Quadro specviewperf 8.1 results:

CPU: E6600 2.4GHz 1066 MHz FSB
Memory: 4GB Corsair DDR2-6400 800 MHz
HDD: 150GB WD Raptor

Run All Summary

---------- SUM_RESULTS\3DSMAX\SUMMARY.TXT
3dsmax-03 Weighted Geometric Mean = 51.12

---------- SUM_RESULTS\CATIA\SUMMARY.TXT
catia-01 Weighted Geometric Mean = 47.28

---------- SUM_RESULTS\ENSIGHT\SUMMARY.TXT
ensight-01 Weighted Geometric Mean = 32.41

---------- SUM_RESULTS\LIGHT\SUMMARY.TXT
light-07 Weighted Geometric Mean = 39.42

---------- SUM_RESULTS\MAYA\SUMMARY.TXT
maya-01 Weighted Geometric Mean = 67.07

---------- SUM_RESULTS\PROE\SUMMARY.TXT
proe-03 Weighted Geometric Mean = 61.84

---------- SUM_RESULTS\SW\SUMMARY.TXT
sw-01 Weighted Geometric Mean = 40.88

---------- SUM_RESULTS\UGS\SUMMARY.TXT
ugs-04 Weighted Geometric Mean = 40.55

I also ran 3dMark06 and only scored 4500 marks with the Quadro. Can I use RivaTuner and get any extra gaming performance out of this card or is that a decent score for the G71 GPU?
 
You must log in or register to reply here.
Back
Top