- Oct 10, 2004
- 733
- 0
- 0
I was about to purchase an additional 6800 based video card for my gf's new computer (a 3.2E Prescott) when i noticed that 6800 GT derivatives with 128MB GDDR1 mem and 350/700MHz core respectively proliferating around. Compared to my 6800 Ultra with 256MB GDDR3 running at stock 425/1.1GHz, you;d think it would be immensely slower. Well think again! I did some benchmarks on my board, underclocking it per se, to see the difference (although I can't account for the 128MB difficiency), and here are my results
Baseline: 6800 Ultra + (425/1.1GHz)
Aquamark3 - 65496
Doom3 (1280x1024 default) - 85.8fps
3DM05 - 5179
GT128: 6800 GT (350/700MHz)
Aquamark3 - 59244
Doom3 (1280x1024 default) - 73.3fps
3DM05 - 3866
GT128+: 6800 GT (425/700MHz)
Aquamark3 - 63019
Doom3 (1280x1024 default) - 79.2fps
3DM05 - 4355
What we see is that, in the most synthetic of benchmarks (3DMark05), the performance delta between the Ultras and the stock "defunct" 6800 GT is huge (up to 25%), but is more miniscule as you go to more "real world" testing scenarios (AM3 - 9.5% / Doom3 14.5%). SHould you effectively raise the core to be comparable to the 6800 Ultra, the GDDR1 memory doesn't hinder you THAT MUCH: 16% in 3DMark05 / 7% in DOOM3 / 3.1% in AM3. So in essence, although theoretically an increase in GDDR1 to GDDR3 should net high performance, as evident in more synthetic situations, more real world tests don't show it as meaningful yet IMHO. I believe these "detuned" 6800 GT's could be a steal, provided the price is right of course and don't give much way to a "real" stock GT , with core clock being comparable.....
Note: I realize this is a small sampling of tests, but I believe DOOM3 to be indicative of the "next" generation in game engines and AQ3 as somewhat of an old defacto benchmark.....as always YMMV...
Baseline: 6800 Ultra + (425/1.1GHz)
Aquamark3 - 65496
Doom3 (1280x1024 default) - 85.8fps
3DM05 - 5179
GT128: 6800 GT (350/700MHz)
Aquamark3 - 59244
Doom3 (1280x1024 default) - 73.3fps
3DM05 - 3866
GT128+: 6800 GT (425/700MHz)
Aquamark3 - 63019
Doom3 (1280x1024 default) - 79.2fps
3DM05 - 4355
What we see is that, in the most synthetic of benchmarks (3DMark05), the performance delta between the Ultras and the stock "defunct" 6800 GT is huge (up to 25%), but is more miniscule as you go to more "real world" testing scenarios (AM3 - 9.5% / Doom3 14.5%). SHould you effectively raise the core to be comparable to the 6800 Ultra, the GDDR1 memory doesn't hinder you THAT MUCH: 16% in 3DMark05 / 7% in DOOM3 / 3.1% in AM3. So in essence, although theoretically an increase in GDDR1 to GDDR3 should net high performance, as evident in more synthetic situations, more real world tests don't show it as meaningful yet IMHO. I believe these "detuned" 6800 GT's could be a steal, provided the price is right of course and don't give much way to a "real" stock GT , with core clock being comparable.....
Note: I realize this is a small sampling of tests, but I believe DOOM3 to be indicative of the "next" generation in game engines and AQ3 as somewhat of an old defacto benchmark.....as always YMMV...
