Follow-up to NVIDIA’s shady trick on 9600GT
Page 2: No more linkBoost in nForce mobos
Page 4: 9600GT runs better on NVIDIA mobo?
Page 5: 9600GT’s GPU and PCIe clock relations
Page 6: Other cards also have this issue?
9600GT’s GPU and PCIe clock relations
Quoted from TPU’s post
TPU said, if PCIe clock raise 10% then 9600GT’s performance will also raise 10%. We are sure when PCIe clock increased, 96GT GPU will also increase, but is it really 1:1?
When we change PCIe clock, from Rivatuner/GPU-Z we can not read any GPU clock changes, they still keeps the same number. So we can only make some guesses from the test result.
MT score changes: 9600GT runs with different GPUclocks or PCIe Clocks.(Intel P35)
Comparison
We increased Intel P35′s PCIe bus clock via BIOS, each time steps up 1%, and run the Multi-Texture test. The truth is surprising: each time we increase 1%, MT score rise 1%.
According to TPU, if PCIe clock increase 1%, then GPU clock increase 1%. That means GPU increased 650*1%＝6.5MHz, to 656.5MHz.
But when we keep PCIe Clock stable @100MHz and increase GPU clock by 1%, the score differs. It is weird that GPU clock and MT score are not 1:1. So TPU’s conclusion is not 100% correct. PCIe clock affects GPU clock, but it is not 1:1.
When we run the test on 780i SLI the graphic is exactly the same.
MT score changes: 9600GT runs with different GPU clocks or PCIe Clocks.(780i)
Comparison (nForce 780i)
What about 3Dmark06 total score?
3Dmark06 score: 9600GT running on different PCIe clock and GPU clock(780i)
Comparison(780i)
When PCIe clock increases, the performance changes. But these kind of change not 100% equal to GPU clock change. Also the ratio isn’t fixed. So we guess when PCIe clock increased, it will affect some part inside GPU which still unknown.
We got anther proof: when we increase PCIe to 122MHz, we can pass 3Dmark06 test but when we increase GPU clock to 793MHz(650*122%), it is unstable, can not pass 3DM06 test.
March 11th, 2008 at 6:15 pm
thanks,Original linkboost has been cancelled, all speculations are unfounded.
March 11th, 2008 at 7:07 pm
[...] GPU clock rise when PCIe clock frequency rise? Other GeForce also have this issue? Source __________________ Please ignor my spelling and grama! FAQ:s I have a slow computer can i run [...]
March 11th, 2008 at 8:24 pm
nvidia is the best, but Ati is more better n yet cheaper than nvidia, just my opinion
March 11th, 2008 at 8:36 pm
linkboost is not with us anymore but if you manually increase pcie clock, then there is hidden 9600gt boost
so people should know about it. imagine oc version of 9600gt and you will increase pcie clock %10. What will be result? maybe GPU will start crashing. thanks for article.
March 11th, 2008 at 9:12 pm
We already know about Asus. What we want to know about is XFX and Evga. The EVGA and XFX boards are rebadged boards while Asus is not.
March 11th, 2008 at 11:14 pm
Another site talked about this before you guys posted it here. And he did a much more detailed test to see that it IS overclocking and it IS cheating on the tests. Because your comparing an overclocked card to non overclocked cards. It can’t get much more unfair than that. Why you decided to defend Nvidia on this is beyond me.
March 12th, 2008 at 1:16 am
TC: are you a dork? Did you even read the article? It clearly said that they wanted to get to the bottom of what TechPowerUp discovered so far. What is beyond me is your clear inability to read and understand the damn article before spewing crap out of your mouth.
March 12th, 2008 at 10:14 am
@ anonymous,
no need to flame
@TC, we are not defending… we just said what we know.
March 13th, 2008 at 12:42 am
You should do an actual benchmark test with the changes and see if changing the PCIe clock increases fps performance at all. Since the theory is that the PCIe bus is not bandwidth bottlenecked there should be very little performance gains, but if 3dmark scores are worth any value today then there should be a notable difference in some aspect of performance.
Your results make me nervious about Nvidia’s intentions with PCIe 2.0. Also has anyone tested this with ATI cards?
March 13th, 2008 at 12:49 am
Edit: I meant benchmarks* with the 9600GT since it has this issue (where 7000 and 8000 series don’t). If there is NOT a performance gain, then we can say that something else other than clock speed and bandwidth is changing with the frequency.
March 14th, 2008 at 5:35 pm
nvidia’s magic once more
March 15th, 2008 at 10:56 am
[...] performance: techPowerUp! Review :: NVIDIA’s shady trick to boost the GeForce 9600GT :: Page 1 / 4 Follow-up to NVIDIA
March 20th, 2008 at 8:12 pm
[...] Related:Follow-up to NVIDIA’s shady trick on 9600GT [...]