It's time to take the card out for a test run. I will be benchmarking using Crysis Warhead, and FEAR 2: Project Origin. Crysis will show off how the card performs in the most demanding PC game on the market right now. FEAR 2 will showcase the sheer power of the card, and later give a good indicator how effective overclocks can be in game.

Test system specs used:

CPU - Intel Q6600 @ 3.2GHz

Motherboard - Gigabyte EP35 DS3L

Memory - 4GB DDR2 800

OS - Windows 7 Ultimate 64bit

Alright, lets start off by graphing the results at stock settings in FEAR 2. The test was a generic run incorporating the array of firefights and effect sequences commonly seen in game. The run was repeated in FRAPS for 3 loops. The three results were averaged to come up with the numbers shown in the graph.

What is there to note about these numbers? Well, at max settings and running at 1920x1200 FEAR 2 hardly challenges the 4890. During a high stress area the frames dipped to 50fps, but for the most part things run smooth as butter at an average of 83frames per second. There is no visual difference in performance at 1680x1050, so the huge numbers are nothing but overkill.

Crysis Warhead is a far more graphic intensive game. Testing involved using the Crysis Warhead benchmark utility on the Ambush test level. All settings were set at 'Gamer', and DX10 mode was enabled. Below are the graphed test results averaged from 3 looped runs.

The min FPS is a bit of an anomaly, but with the values so close they can be ignored. The maximum and overall average frames per second are pretty much what one would expect in such a demanding game.

Crysis is very playable at max settings, tweaking a couple settings, or maybe lowering to 2xAA would give enough of an fps bump to make it smoothly playable. Alternatively, an overclock could be tried. I tried bumping clocks for a bit better performance.

I am one who demands stability from my overclocks. I will accept a little bit less performance, so that I am not bothered by instability. My testing involves a loop of RTHDRIBL after setting clock speeds. If this proves stable after a 6 hour run, I then set up a loop of Crysis Warhead and see if any crashes occur.

The maximum stable clock I was able to reach was 945 on the core, and 4.1GHz ( 1024 x 4 ) on the memory. This passed my stability testing through both rounds. I was able to get 960 on the core, and 4.3GHz ( 1075 x 4 ) on the memory stable through RTHDRIBL, but in Warhead artifacts were a problem, that grew worse until a crash. I stepped down the frequencies until I was able to get Crysis entirely solid.

Once a stable overclock was reached I re-ran the tests with the higher clocks. Starting with Fear 2.

The minimum frames per second is unchanged. The average and maximum numbers show noted increase. The only real quip is that the stock numbers were already well within the acceptable and perfectly playable range. The increase is a bit useless for anything but bragging rights.

The numbers for Crysis may be a bit more interesting. An increase of a few fps can bring the numbers a little bit closer to perfectly playable at max settings.

Again minimum numbers are all but the same. Unfortunately, the results for the max and average don't increase enough to really give too much positive result. This is somewhat typical of small overclocks. With Crysis and it's odd ability to make more power amount to nothing, these numbers aren't surprising. All and all I was rather satisfied with the results. Looking around at others, the clocks I have reached are within range of common results.

Before this review turns into a novel, I think it's time to close up.

<-Back | | Next->