Sat Nov 10, 2007 10:24 pm
danny, since you asked, I'll answer, but...
I dunno what's going on there, difference-wise, between those two reviews. That's ancient history, pretty much, so it's hard to say why Wolf ET w/8X aniso at 16x12 is so much slower on the 9800 XT in the second review. The biggest differences hardware-wise include varying:
-CPUs
-Motherboards
-Chipset drivers
-Catalyst drivers
In the test configs. Two strong possibilities, given the era, include AGP performance/compatibility issues between the ATI card and the VIA chipset causing a slowdown and the possibility that ATI is cheating less on texture filtering in the latter case.
Another possibility is that I changed the texture size or something when testing the game, although that's not documented in the article and I don't think I have my notes.
It's quite likely both results are valid for how they were tested, for what it's worth. Of course, that highlights an inherent weakness in what you're trying to do with your comparo chart, which is unfortunate but probably unavoidable.
Even if we pulled out all those old cards and tried to test again, getting a really valid comparison would be difficult since feature sets have changed so much. DX9 PS 2.0 cards are running into compatibility problems these days, even if you skip over DX10. And really, no GeForce FX was ever worth much in true DX9 apps. I'm not sure how you quantify "can't do it" vs. "does it X fast."
I would like to include a broader scope of older cards in our GPU reviews, but even that is problematic when today's $200 cards can play games at settings and resolutions that are pretty much unplayable on, say, a GeForce 7800 GTX. You can do it with timedemos and testing resolution scaling to some degree, but teasing out the differences at a single set of quality options at a display res that's playable on older cards and still shows a meaningful performance advantage (i.e., not 100 FPS vs. 250 FPS) is tough.
The comparison easier to do with basic math. Have a look at the texel fill rate and memory bandwidth on a 7800 GTX vs. an 8800 GT.
7800 GTX: 10.3 Gtex/s and 38.4 GB/s
8800 GT: 33.6 Gtex/s and 57.6 GB/s
Even that doesn't account for the advancements shader power and efficiency in the G92, let alone the very noticeable improvements in image quality, mathematical precision, anti-aliasing capabilities, HD video decode assist, or all the rest. Nor does it take into account the two-megapixel buffer-sizing limitations in the 7800 GTX. Newer chips handle higher resolutions much more gracefully than older ones.
So much comes down to what you want to do with the card. If you want to play DX9 games on a smallish monitor and don't care much about image quality, the 7800 GTX may be fine. But if you want to drive a bigger display, play HD movies, try 8X AA, future-proof for when DX10 actually matters, or are picky about image quality, the 8800 GT would be a huge upgrade.
Ok, I'm going to stop now. This comparison thing is hard!
Scott Wasson - "Damage"