NVIDIA GeForce GTX275 896MB vs. 1792MB Review - Conclusions

Submitted by skipclarke on 9 July, 2009 - 01:56

Article Type: 
Review

Other Benchmarks

I had hoped to use the benchmark tool in HAWX, as the game has excellent support for Widescreen and Wide Surround. But the game produced hugely variable numbers between runs. Where Trackmania and World in Conflict may produce variances of 3-4 fps against an average of 30-40 fps (roughly 10%), HAWX was producing variances of 20-30 fps against an average of 100fps (20%-30%). I simply didn't have the time to work it to consistent readings.

Also, I could not get Crysis Warhead to bench under Vista. Initially the game would run and bench, but would dump back to the desktop without producing or saving any results. I could watch the average fps count as the game finished each pass, but this would be a pain and defeat the purpose of a self-running demo. I attempted to fix the problem by deleting the game and the benchmark tool, and reinstalling both. The game wouldn't run at all after this.

Hopefully I can get these issues worked out for future iterations.


Conclusions

The GTX 275 performs extremely well, and is a hell of a value. However, there are some instances where the GTX 295 is a clear winner. For older titles based on the Source Engine, the GTX 275 is certainly in the sweet spot - and you avoid the issue of the GTX 295 whine.

The GTX 275 carries enough power to run a 30" panel, or a TH2Go rig in either Surround or Wide Surround. The larger frame buffer does provide a benefit, particularly at 5040x1050. At mid-range resolutions the benefit may simply be a smoother experience, with less hiccups for texture swapping.

I already own a GTX 295 (bought the day of release, well before the GTX 275 was available). However, if I were looking to purchase now, I would choose the GTX 275 1792MB over the GTX 295. Tweaking a few settings off of "max" should give you solid frame rates with the GTX 275 in most any situation, and you avoid the GTX 295 whine.


What's Next?

I've sent back the GTX 275 896MB card, and EVGA has agreed to send me a 2nd GTX 275 1792MB card. I will run through all of the benchmarks with the GTX 1792MB cards in SLI. After that I look forward to tackling the GTX 285 series, of which EVGA offers both 1GB and 2GB models. At the end, we should be able to provide a nice matrix of price and performance across Widescreen, Surround and Wide Surround.

Also, I will be dropping tests for 1680x1050, 1600x1200, 3072x768 and 4320x900.

  • We see consistent performance across 1680x1050, 1600x1200 and 1920x1200. There is no longer any need to run these separately, and we have already proven the impact of aspect vs. pixels in the original Widescreen and Surround Benchmarking Suite.
  • Additionally, we see comparable performance for the "Normal" Surround modes of 3072x768 and 3840x1024, so there is no added value in testing both.
  • Finally, while we do see some performance difference in the Wide Surround modes of 4320x900 and 5040x1050; most users are only interested in the seeing benchmarks for their native resolution of 5040x1050.

This will leave us with the native resolutions for the most popular panels 1920x1200 (24"+), 2560x1600 (30"), 3840x1024 (Surround) and 5040x1050 (Wide Surround). These changes will make the future charts easier to read (as we add more cards), and reduce the time it takes for each round of benchmarking.