Just volt modded 7900GT.. performance difference not much?!?!

[lektrix]

Just volt modded 7900GT.. performance difference not much?!?!

I'm currently benching my 7900GT and I just volt modded the card to 1.4V and shaded the resistor for the memory..

Did I do something wrong, or are these results really that miniscule?

3DMark03
7900GT 550/790 (Default) - 19828
7900GT 579/855 (Auto-detect) - 20791
7900GT 650/890 (1.4V mod) - 22276

Going from 550/790 (Default) to 650/890 is roughly a 18% increase in core clock and 13% increase in memory clock. In 3DMark03, I got a 12% increase in scores.

3DMark05
7900GT 550/790 (Default) - 9376
7900GT 579/855 (Auto-detect) - 9867
7900GT 650/890 (1.4V mod) - 10689

Going from 550/790 (Default) to 650/890 is roughly a 18% increase in core clock and 13% increase in memory clock. In 3DMark05, I got a 14% increase in scores.

3DMark06
7900GT 550/790 (Default) - 4674
7900GT 579/855 (Auto-detect) - 4870
7900GT 650/890 (1.4V mod) - 5213
7900GT 685/915 (1.4V mod) - 5257 12

Going from 550/790 (Default) to 685/915 is roughly a 25% increase in core clock and 16% increase in memory clock. In 3DMark06, I got a 12% increase in scores.

After looking at my scores, can you guys tell me if this is bad, normal, or excellent? Cause I sure as hell can't tell.

Here is my test bed:

DFI NF4 SLI-DR
Opteron 144 @ 2.7GHz
G.Skill 2GB ZX @ DDR500 3-3-2-5
WD Raptor 74GB x 2 RAID-0
eVGA 7900GT CO SC w/ VF700Cu
Audigy X-Fi
OCZ GameXStream 600W

Futuremark was installed on the RAID-0 raptors and they were defragged once last night before running the tests. All applications were shut down except MBM5 (for temperature monitoring) and anti-virus. Windows was heavily tweaked to run only on essential processes. All tests were run on default settings.

Now I have some questions and concerns about the 7900GT card. Initially after volt modding the card, I used nVidia's integrated Coolbits to "detect optimal frequency", but that immediately made my computer artifact and blink. I'm guessing it's the memory chips on the card overheating so I manually increase the clocks and used the test function in Coolbits to test. I got up to 685/915 but I think I can do a bit more. Haven't tried yet because I'm taking a break writing this post.

Anyways, is there any way to verify that the mod has been accomplished successfully WITHOUT taking the card and mobo out and using the multimeter to check for 1.4V on the GPU and 2.25V on the memory? And is there any way to run this thing cooler? I have a VF700Cu with 8 RAMsinks , an additional 2 RAMsinks for the memory regulators on the back, AND a 120mm side fan blowing on part of the card. Temperatures still at 52C idle and shot up to the 70s on full load! I'm really surprised because it was only 41C idle/55C load on my non-modded, non-OCed 7900GT w/ VF700Cu without the 10 RAMsinks.

 

[Nightmare225]

I think it was a good overclock. You should do some actual gaming benchmarks because synthetics aren't worth ****.

 

[Ike0069]

Yea, do some FPS checks at high resolutions with AA and check the difference. I'd be very interested in the results.

 

[Bobthelost]

I think the reason you didn't get a 1:1 performance increase is mostly because 3Dmark has a CPU component, you haven't increased that between voltmoding the GPU so it's not going to change. (ie CPU makes up 20% of the score while the GPU covers the other 80%)

 

[lektrix]

Are there any FPS recorders out there that calculate your minimum, maximum and average? It would be better than me staring at the FPS and recording it every few seconds.

I think my CPU hit a wall (it might be able to hit 2.8GHz) but I'm sure the memory could be tightened.

Another thing is, I just compared my benches to other Futuremark benches using their project comparison tool and it seems that some people are getting the same score/higher score with a similar overclock but using only 1GB system memory!!! wTF!

 

[JBT]

umm increasing clock speed is never going to be 1:1 with performance increases. I'm actually pretty impressed with the increases you got from that.

 

[Ichigo]

Benchmark a game. Honestly.

 

[lektrix]

Are there any FPS recorders out there that calculate your minimum, maximum and average? It would be better than me staring at the FPS and recording it every few seconds.

And is there any way to check the actual voltage of the GPU and memory without taking the motherboard + card out?

 

[darkswordsman17]

Originally posted by: lektrix
Are there any FPS recorders out there that calculate your minimum, maximum and average? It would be better than me staring at the FPS and recording it every few seconds.

And is there any way to check the actual voltage of the GPU and memory without taking the motherboard + card out?

FRAPS, and also some games have built in stress tests (Doom3 TimeDemo, HL2 has a stress test, etc.) that you could use.

I think that RivaTuner, and there's another program that I can't remember that should tell you voltages and anything else. This might not be totally accurate as it might just be getting its info from the card's BIOS, so the only way to really check would be with something like a multimeter.

 

[cpacini]

Originally posted by: lektrix
Are there any FPS recorders out there that calculate your minimum, maximum and average? It would be better than me staring at the FPS and recording it every few seconds.

And is there any way to check the actual voltage of the GPU and memory without taking the motherboard + card out?

Well that?s certainly not possible.

Seriously, go down to radio shack and buy a multi meter, or borrow one from a friend.

Physically modifying voltage levels on a $300 video card + No way to test changes = bad times.

 

[Howard]

Just not handy with my multimeter, eh?

It's not too great, though.

 

[Avalon]

FRAPS will let you record FPS, although it may be a little tricky to repeat your same exact test twice.

 

[Powermoloch]

Originally posted by: lektrix
Are there any FPS recorders out there that calculate your minimum, maximum and average? It would be better than me staring at the FPS and recording it every few seconds.

I think my CPU hit a wall (it might be able to hit 2.8GHz) but I'm sure the memory could be tightened.

Another thing is, I just compared my benches to other Futuremark benches using their project comparison tool and it seems that some people are getting the same score/higher score with a similar overclock but using only 1GB system memory!!! wTF!

FRAPS :thumbsup:

 

[buzzsaw13]

Try to voltmod it to 1.5V, your VF-700 should be able to handle it. I'm running mine off the stock cooler and it gets up to 691/901 3dmark stable

 

[JM Aggie08]

i don't trust anything but FRAPS.

 

[russki]

the performance gain is going to be greater going from stock to auto detect, than from auto detect to vmod, this is because the speed of the gpu is increasing, but the speed of the cpu is the same. its not really worth it to mod it any higher than 1.4volts and from what I can tell your results are consistent and expected.

 

[lektrix]

Is anyone else getting high temperatures after volt modding to 1.4V?

I'm using a VF700Cu, 8 RAMsinks and 2 additional RAMsinks for the memory chips on the back.

Here are my temps with the card on stock voltages with a minimal overclock (579/855): 42 idle / 60 load
Temps with 1.4V volt mod + mem mod on default clocks (550/790): 53 idle / 75 load
Temps with 1.4V volt mod + mem mod on highest clocks (693/930): 57 idle / 79 load

The drastic increase in temperatures is pretty bad for the card. There's no way I can run it on such high temps 24/7. How can I lower the temps to more reasonable numbers? I already reseated the heatsink w/ new AS5 paste twice in the past 60 hours.

 

[DidlySquat]

Your score of around 5200 in 3dmark06 is just about right for a 7900 GTX with FX-57, which is what your overclocked system is equivalent to. If you get a dual core CPU your score should increase to around 6000.

 

[ashishmishra]

If you have F.E.A.R. you can try it's built in benchmark. I usually use it to show the difference a Video/CPU/Memory overclock makes on my gaming performance. I like it because it a modern game in which overclocks might have actual gameplay related FPS improvements not just higher numbers from which you get no tangible gaming improvements.