PDA

View Full Version : Shader Clock's Impact on Performance


Pages : [1] 2

jAkUp
11-27-06, 02:35 PM
Here is a quick test to see how much the Shader clock impacts performance. Check out the HDR test, it increased by almost 100 points! My guess is that the results would be more dramatic in a more GPU limited environment. This was done by reflashing the BIOS.

Specs:
E6700 @ 2.66GHz
EVGA 680i Motherboard
8800GTX @ 1.35GHz shader clock and 1.45GHz shader clock

Here are the results:

8800GTX @ 1.35GHz Shader Clock:
http://www.nvnews.net/vbulletin/attachment.php?attachmentid=22398&stc=1&d=1164656003

8800GTX @ 1.45GHz Shader Clock:
http://www.nvnews.net/vbulletin/attachment.php?attachmentid=22399&stc=1&d=1164656003

Monolyth
11-27-06, 02:56 PM
Here is a quick test to see how much the Shader clock impacts performance. Check out the HDR test, it increased by almost 100 points! My guess is that the results would be more dramatic in a more GPU limited environment.

Sorry that's 50 points. :(
Hopefully they can start ramping that clock up some more though, given enough juice those numbers could really start coming out to something. Not to mention the immaturity of the drivers.

CaptNKILL
11-27-06, 03:00 PM
That doesn't even look worth it to me.

100Mhz overclock for 50 points on a shader intensive test? No thanks...

jAkUp
11-27-06, 03:02 PM
That doesn't even look worth it to me.

100Mhz overclock for 50 points on a shader intensive test? No thanks...

Well it probably equates to a few FPS in extremely GPU limited scenarios, hey, its free performance (nana2)

hirantha
11-27-06, 03:10 PM
wow just one GTX gives you that much in 3Dmark 06.. cant wait to get mine in!!

Mathas
11-27-06, 03:10 PM
How did you separately OC shader clock? With the latest RivaTuner?

jAkUp
11-27-06, 03:14 PM
How did you separately OC shader clock? With the latest RivaTuner?

It was done with a BIOS flash. NiBiTor allows you to adjust the Shader Clocks.

tacos4me
11-27-06, 03:19 PM
It was done with a BIOS flash. NiBiTor allows you to adjust the Shader Clocks.

Can we look forward to software doing it sometime soon? :)

Mathas
11-27-06, 03:19 PM
Clear:)

Monolyth
11-27-06, 03:30 PM
Well it probably equates to a few FPS in extremely GPU limited scenarios, hey, its free performance (nana2)

Nothing is ever 'free' ;) As with most over-clocking this has probably depreciated the life of the card somewhat, as well as potentially voiding the warranty.

CaptNKILL
11-27-06, 03:36 PM
I'd be more worried about heat and stability myself. My card runs rather hot and doesn't handle overclocks too well, so I'm back at stock for now.

Have you tried pushing it any further jAkUp? Maybe it scales better or maybe it overclocks like crazy.

I just don't think the added stress of a 100Mhz overclock on any component is worth a less than 1% increase in a benchmark score. :o

MaXThReAT
11-27-06, 03:44 PM
Thank you for the comparison. I've been wondering how much the shader clocks would effect performance since I got the card. What version of Nbitor did you use, V3.1? How did you go about clocking shaders independently. All my clocks show up blank, I only see the 3D clock and it's 576, Odd. Can you upload the bios you used? Have you kept it at this clock or was it just for testing?
Was there a change in temp?

jAkUp
11-27-06, 03:57 PM
Nothing is ever 'free' ;) As with most over-clocking this has probably depreciated the life of the card somewhat, as well as potentially voiding the warranty.

Of course, but who is going to be using this card in 3 years? :) Certainly not me.

I am not really sure about temps... nor about pushing the clock any further. I have not tried pushing it any further, 1.5GHz might be fine though. Also, I am using NiBiTor v3.1.

I think the just released version of RivaTuner allows you to increase it via software, but I am not sure yet.

Monolyth
11-27-06, 04:05 PM
Cards can die for all sorts of reasons and well within a 3 year period we will see a fair share of 8800GTX's die. :)

All preaching aside, increase seems minimal, would need a larger over-clock to see any large benefits (can we say nVidia's refresh? 8900GTX w/1.75-2Ghz shader clock?).

J-Mag
11-27-06, 04:13 PM
Did you try running any other tests? It would be interesting to see the results for FEAR and CoH, since they both have integrated performance tests.

SH0DAN
11-27-06, 05:36 PM
It does not seem worth the 100Mhz increase,and my card also runs hot.Mine idles around 62c,and under load (4 solid hours of Oblivion maxed @ 1680x1050 4xAA+16xAF) gets me a toasty 85c.I have two 120mm fans going full tilt most of the time.

jolle
11-27-06, 05:44 PM
Wonder how it would scale on resolution..
I assume you can change resolution in 3dmark06 if you´re registered, so perhaps you could try a run at, say 640x480, and one at 1600x1200 or higher.
I mean you got all the other components like ROPs and stuff running on the regular core clock that might be bottlenecking, or not..
I spose the shader load would scale aswell tho, but perhaps not to the same extent as the load on other parts? dunno really, but I assume that every shader isnt running on every pixel on the screen so..

Might show different levels of increase on the clock, I mean 3dmark06 prolly wasnt written for this sort of shader capacity...

DarkJedi664
11-27-06, 05:49 PM
50 points in 3dMark? Am I the only one in the world who thinks 3DMark is pathetic? All it is, is to show other people how big your e-penis is; it's good to see the first time, but who really cares what score you get? Synthetic benchmarks are worthless. It also makes me laugh when people on this forum go "OMG, this new driver made me lose 300 point in 3DMark, back to the old driver!". How sad is that?

jAkUp
11-27-06, 05:52 PM
50 points in 3dMark? Am I the only one in the world who thinks 3DMark is pathetic? All it is, is to show other people how big your e-penis is; it's good to see the first time, but who really cares what score you get? Synthetic benchmarks are worthless. It also makes me laugh when people on this forum go "OMG, this new driver made me lose 300 point in 3DMark, back to the old driver!". How sad is that?

Are lap times when racing worthless? Not for the racer. Same thing.

Everyone knows that 3DMark06 is not an actual game, just like a lap at a racetrack is not an actual drive on the freeway. If using a particular oil decreases your lap time, is that good? Or is it worthless?

It is just a quick way to get some performance numbers.

vmsein
11-27-06, 05:58 PM
It's good information to have just for comparison. NVIDIA will be releasing a newer version of nTune that will allow increasing the shader clock independantly of core and memory clocks, so this can serve as a preview of how performance will scale. So far though, it's not too impressive.

jAkUp
11-27-06, 05:59 PM
This new version of Rivatuner I think lets you increase in Windows:

http://www.nvnews.net/vbulletin/showthread.php?t=81262

DarkJedi664
11-27-06, 06:01 PM
It's not the same thing, because the racer can actually use those lap times; you can't use your 3DMark score at all, besides to brag. I also just cannot understand why people go ape-****-bannanas when they lose a few frames here and there, when they're already over 60FPS. I'm not trying to be negative, but I think people get way too overzealous about the lamest things. If you're over 60, why should you matter if you get 120+FPS besides to brag? Good for you, you get a million frames per second in Quake3; does that make you any more special? Gaming is optimal at 60FPS, anything else is just wasted. If you can turn on everything at high res, AND AA and still have 60FPS, that's great; but you shouldn't honestly care about getting anything over 60, not to mention that anything over 60 on an LCD causes tearing. Again, if you want to brag about getting a million frames in Quake3, fine, I just find it a bit pathetic.

CaptNKILL
11-27-06, 06:06 PM
Again, if you want to brag about getting a million frames in Quake3, fine, I just find it a bit pathetic.
No offense, but you may be in the wrong place if you think bragging about high framerates is pathetic.

Seriously, its been like this for a decade or more.

It may be pointless, but pathetic? Why the hell is it pathetic?

jolle
11-27-06, 06:07 PM
It's not the same thing, because the racer can actually use those lap times; you can't use your 3DMark score at all, besides to brag.
a 3dmark score on its own is perhaps not much, but in cases like these to try out a OC or something like that, it can be handy.
Its prolly more shader intesive than any game out there right now, and it was even more so when it was released and the time inbetween then and now.
But for real benchmarking, you need real game benches aswell, cause a 3dmark score doesnt always relate well to a game, and different games perform differently on different hardware.

Noones saying that 3dmark is the end all benchmark, but for comparision and stuff its handy.

tacos4me
11-27-06, 06:21 PM
This new version of Rivatuner I think lets you increase in Windows:

http://www.nvnews.net/vbulletin/showthread.php?t=81262

There's no slider for the shader clock. But I've noticed with the monitoring tool that the shader clock changes with the core clock. At 650 MHz core, the shader is @ 1.5GHz.