It is almost the norm to see overclocking results in your typical video card review. We have have been doing it since the 3DFX Voodoo, and usually you see the same thing. A card will be overclocked until it locks up or visual distortion begins, and a few benchmarks are run. You always have to think if the card tested was a golden sample.  Would this happen with a card you bought off the shelf? With that in mind we decided that a test of more than just one card was necessary to see how a model would perform in the same testing environment. Would each card overclock similarly? Would there be major differences in top speeds?

Do do this we turned to our recent upgrade at our gaming center, Tek Republik, from XFX GeForce 7900GTs to XFX GeForce 9600GTs. It was an excellent chance to get overclock numbers from not just one card, but from 24. The model chosen was the PVT96GYDF4 clocked at the standard 650MHz for the GPU, 900MHz for memory, and 1625MHz shader clock. Each card was tested on the same system with a Athlon 64 X2 4600+, 2GB of Corsair DDR2 800MHz on the Asus 590 SLI based CrossHairwith an 850 watt Antec Truepower Quattro. We pushed each card to its limits using Riva Tuner, and to test for stability Crysis was used. It made quick work of a clock speed it didn't like usually with a crash quickly into the built in GPU benchmark.

We topped out near 800/1100/2000 for many of these cards? Crysis made some pay that tried to?make it through the GPU benchmark? Here are our top clocks for each card to make it through stabley?

In each graph we show each overclocked card results, and the stock number as card zero for reference? As you can see results for almost all overclocks were very similar with 19 topping out at 790MHz? Two made it to 801MHz, and 3 managed just 708MHz? On average a 20% overclock was accomplished?

Moving on to memory overclocking we see much of the same as most cards, 18 in fact, reach 1066MHz.  Two hit 1086, and 4 manage just 1016MHz.   Stock is 900MHz so the most common overclock is just over 18% over stock. 

We wrap up with shader testing.  Here we see the same results, and it is clear the lower clocks are from the same cards, and the higher clocks are also from the same cards.  We get as high as 1984MHz, or 359MHz over stock, but most hit 1977MHz.  Four cards only made it to 1770MHz.  At 1977MHz you are looking at almost a 22% overclock.

Putting all the clocks for each card into a single graph we get an easier visual of which card did the best.   Cards 17 and 18 were the top overclockers, with 14, 15, and 5 being the worst. 

After getting these results we took a 9600GT at stock clocks, and then overclocked to 790/1066/1977 and tested with Crysis at 1680X1050 on high settings. 

At stock speeds we got 22.68 frames per second.  Overclocked 26.5 was the result as almost 4 frames per second of an increase.  That equates almost a 17% performance bump which is just a bit under the roughly 20% overall clock improvement.

So we conclude our overclocking test of the GeForce 9600GT.   As you can see the results for many cards were very similar, and almost all of them had close to 20% overclocks.  We hope you enjoyed these tests, and we expect we will do more large scale GPU overclocks in the future.  Please check out our most recent graphics card reviews, and check back for Battlefield Heroes benchmarks later today.

Foxconn GeForce 9500 GT
Asus GeForce GTX 260 TOP
ATI Radeon HD 4870X2
ATI Radeon HD 4870 Single and Crossfire
ATI Radeon HD 4850 Single and CrossFire Mode Review