Higher benchmark while underclocked?

[Guest]

I overclocked my GTS250 (foolishly) without benchmarking before hand because a new game I was playing was lagging a little. I noticed no improvement from a quite significant clock, so I went to download 3dmark06 and see if it was actually performing any better.

Score - 10.4k.

I then reset my gfx settings back to stock

Score - 11k

I then underclocked the core clock slightly

Score - 11.2k

What is this telling me about my graphics card/ set up? Too hot? Not enough voltage? I still feel this is a pretty poor score for my system, but I want to solve this problem first and then look at how to improve the rest.

Cheers!

 

[Nils]

Myself I find 3dmark not to be very consistent. I prefer game benchmarks above synthetic benchmarks. Do you have a game with a build in benchmark like Far Cry 2, Crysis, Doom3 or GTA4?

 

[Mousemonkey]

I overclocked my GTS250 (foolishly) without benchmarking before hand because a new game I was playing was lagging a little. I noticed no improvement from a quite significant clock, so I went to download 3dmark06 and see if it was actually performing any better.

Score - 10.4k.

I then reset my gfx settings back to stock

Score - 11k

I then underclocked the core clock slightly

Score - 11.2k

What is this telling me about my graphics card/ set up? Too hot? Not enough voltage? I still feel this is a pretty poor score for my system, but I want to solve this problem first and then look at how to improve the rest.

Cheers!

What's your CPU?

 

[Guest]

I have Crysis, though I can't run it at the moment.. how would you go about benchmarking using it?

And my CPU is a stock 2.4GHz Q6600. Will overclock that when I get round to it, trying to squeeze some performance out of this before i look to upgrade next year.

 

[Nils]

I have Crysis, though I can't run it at the moment.. how would you go about benchmarking using it?

And my CPU is a stock 2.4GHz Q6600. Will overclock that when I get round to it, trying to squeeze some performance out of this before i look to upgrade next year.

Go to Crysis installation directory. Click the bin32 or bin64 directory according to whether you have 32 of 64 bit os. There you will find a Benchmark_Gpu.bat file. Click it. There's also a Benchmark_Cpu.bat file. This test your cpu performance with physics, however nowadays, most of that work is done by the gpu so you can use it as a gpu test too.

 

[Mousemonkey]

Put the graphic card back to stock clocks and then OC the CPU and watch your 3D mark rise, when you can no longer get a stable run, back off the OC a little and then play with the graphic cards clocks and see if the results match those of when the CPU was at default clocks.

 

[Guest]

Alrighty. How does the CPU affect the speed of the GPU test, out of interest? Is the GPU rendering not independent of the CPU? Not particularly knowledgable here, I must admit

Is 87C under load hot for a graphics card as well - i.e. is an aftermarket cooler likely to improve performance?

 

[Mousemonkey]

Alrighty. How does the CPU affect the speed of the GPU test, out of interest? Is the GPU rendering not independent of the CPU? Not particularly knowledgable here, I must admit

In the case of 3D '06 it has everything to do with the scores as that bit software is so CPU dependent it's laughable at best as a GPU benchmark, still useful for testing OC's though.

Is 87C under load hot for a graphics card as well - i.e. is an aftermarket cooler likely to improve performance?

A decent aftermarket cooler can take 25 - 30c off the load temps, my 8800GT's rarely get above 60c under load whereas with the standard single slot HSF's on they could and did hit 90c once.

 

[Nils]

Alrighty. How does the CPU affect the speed of the GPU test, out of interest? Is the GPU rendering not independent of the CPU? Not particularly knowledgable here, I must admit

Is 87C under load hot for a graphics card as well - i.e. is an aftermarket cooler likely to improve performance?

Lower temps are more likely to improve stability than performance. if you plan on oc'ing the card, get a third party cooler unless you don't mind cranking up the gpu fan speed and thus also noise levels.

 

[jedimasterben]

In the case of 3D '06 it has everything to do with the scores as that bit software is so CPU dependent it's laughable at best as a GPU benchmark, still useful for testing OC's though.

A decent aftermarket cooler can take 25 - 30c off the load temps, my 8800GT's rarely get above 60c under load whereas with the standard single slot HSF's on they could and did hit 90c once.

My little demon of an 8800GT would sit at 85C idle ALL THE TIME, until I changed the fan speed from 30% to 60%, there was an almost inaudible change in noise (of course over my water cooling pump [outdoor fountain pump], pretty much anything else seems silent!), but the idle temp dipped down into the mid 50's. I was pretty much shocked at the difference. I still wish it had a dual slot heatsink, though.

 

[wolfram23]

Fairly peculiar.

What they said about 3DMark and the GPU scores is correct tho as far as it being CPU dependant. My Vantage GPU score dropped by 3k when my CPU score increased by 2k because I went from a 3.68ghz Turbo Boost OC (4.2ghz single core) to a straight 4ghz no Turbo. This tells me that for whatever reason Vantage relies on a single core (although in Windows performance it shows 25% usage across all 4 cores...).

I agree with trying to OC your CPU a bit. As for GPU OC, see what happens if you raise/lower your memory clock independantly of the core clock. It might be some odd issue with the core overruning the memory or something.

 

[Nils]

Fairly peculiar.

What they said about 3DMark and the GPU scores is correct tho as far as it being CPU dependant. My Vantage GPU score dropped by 3k when my CPU score increased by 2k because I went from a 3.68ghz Turbo Boost OC (4.2ghz single core) to a straight 4ghz no Turbo. This tells me that for whatever reason Vantage relies on a single core (although in Windows performance it shows 25% usage across all 4 cores...).

I agree with trying to OC your CPU a bit. As for GPU OC, see what happens if you raise/lower your memory clock independantly of the core clock. It might be some odd issue with the core overruning the memory or something.

Strange to see your GPU score drop that much.

When I OC'ed my i7 920 from 2.8G TM to 3.8G no TM, my CPU score increased by 10k, while the GPU score increased a little bit.

 

[wolfram23]

Strange to see your GPU score drop that much.

When I OC'ed my i7 920 from 2.8G TM to 3.8G no TM, my CPU score increased by 10k, while the GPU score increased a little bit.

I'm actually really scratching my head now, I got 30k GPU score initially and now I can only get 27k, even when I put my CPU back to the other OC setting I had. So yeah not sure what happened but apparently I was "lucky" and got 30k GPU score or something.
I'm not sure what my CPU score is at stock, I've only ran it with the 2 different OCs.

 

[Nils]

I'm actually really scratching my head now, I got 30k GPU score initially and now I can only get 27k, even when I put my CPU back to the other OC setting I had. So yeah not sure what happened but apparently I was "lucky" and got 30k GPU score or something.
I'm not sure what my CPU score is at stock, I've only ran it with the 2 different OCs.

Maybe it's not that important anyway. I don't find Vantage to be very consistent in CPU's scores when you have Physics on the GPU enabled. I have variations of 1-1.5k myself on cpu scores sometimes. Gpu score variates less but I don't have a score as high as yours.