Dual-GPU Comparison - 3850x2 vs 3870x2 vs 9800GX2
Test Setup & Specifications
Published: 7th June 2008 | Source: ASUS | Price: N/A |
Test Setup
A common mistake made when benchmarking graphics cards is that the rest of the PC system isn't sufficient enough to test the GPU to its limits. This results in a bottleneck situation, where the system can only run at the speed of its slowest component. For this reason, the test configuration chosen below has been specially selected to give each of the graphics cards on test the headroom they require in order to produce the best results.

A selection of games and benchmark suites has also been chosen to test each of the cards with several game engines. Each of the cards will be run at both low and high resolutions with varying levels of texture filtering to represent the use of the card with both small and large screen sizes.
3DMark05
1024x768 / 0xAA / 0xAF (Default)
1920x1200 / 4xAA / 0xAF
1920x1200 / 4xAA / 0xAF
3DMark 06
1280x1024 / 0xAA / 0xAF
1920x1200 / 4xAA / 0xAF
1280x1024 / 0xAA / 0xAF
1920x1200 / 4xAA / 0xAF
Quake4
1280x1024 / Ultra / 4xAA / 4xAF
1920x1200 / Ultra / 4xAA / 4xAF
1920x1200 / Ultra / 4xAA / 4xAF
Unreal Tournament III
1280x1024 / DX10 / High / 0xAA / 0xAF
1920x1200 / DX10 / High / 0xAA / 0xAF
1920x1200 / DX10 / High / 0xAA / 0xAF
Call of Duty 4
1280x1024 / Max / 4xAA / 4xAF
1920x1200 / Max / 4xAA / 4xAF
1920x1200 / Max / 4xAA / 4xAF
F.E.A.R
1280x1024 / Max / 4xAA / 4xAF
1920x1200 / Max / 4xAA / 0xAF
1280x1024 / Max / 4xAA / 4xAF
1920x1200 / Max / 4xAA / 0xAF
BIOSHOCK
1280x1024 / DX10 / High / 0xAA / 0xAF
1920x1200 / DX10 / High / 0xAA / 0xAF
1280x1024 / DX10 / High / 0xAA / 0xAF
1920x1200 / DX10 / High / 0xAA / 0xAF
Crysis
1280x1024 / DX10 / High / 0xAA / 0xAF
1920x1200 / DX10 / High / 0xAA / 0xAF
1920x1200 / DX10 / High / 0xAA / 0xAF
To ensure that our testing was as fair as possible, the stock-clocked version of all three cards was chosen from a single manufacturer. Similarly the price of the cards listed above was also extracted from a single retailer - LambdaTek on the 29/05/08. This is extremely important as over the next few pages we will be using these figures to directly calculate how much of your hard earned cash goes on each and every one of those frame refreshes.

With almost £100 difference between the NVidia 9800GX2 and the ATI 3870X2, the GX2 is certainly going to have to pull some extremely good numbers in order to get the upper-hand on the more affordable ATI counterparts. Let's get down to business...
Most Recent Comments
kool - should be a good read as ive just bought a 9800GX2 and already own a 3870X2Quote
I`m always keen to see comparisons on a per-cost basis. I think it`s very important for people who intend to build the best pcs they can, with the budget they have available.
It`s not merely the case that with limited funds ur going to end up with a crap pc, but more that u have to be wizer about the parts u use for what. Value.
Same is to be said for cpus.
The only thing about these comparisons is that they can change so quickly. Next week or month, the prices change and the performance per cost changes - and yet u `can` be left with the memory of what the previous review said.
I`d like to see a push for a sale based on what grocery supermarkets have for products.
There are enough benches about, and as long as all the factors are taken into consideration - maybe at the start of each year, some1 like Futuremark can name a "performance" pc, afa they concerned. Thereafter in the year, the products for sale can have $4.50/frame08 ($3.75/frame08OC) - or something similar attached to them - based on their results within that PC setup.
Crazy idea, but it only takes some1 in someplace like Futuremark to mention it at a meeting and we`ll all be seeing it in the ads. Don`t think they necessarily need to take to AMD/nVidia about it, and have their legal people draw up a waterproof disclaimer.
Meh, I`d like it. Same for cpus.Quote
It`s not merely the case that with limited funds ur going to end up with a crap pc, but more that u have to be wizer about the parts u use for what. Value.
Same is to be said for cpus.
The only thing about these comparisons is that they can change so quickly. Next week or month, the prices change and the performance per cost changes - and yet u `can` be left with the memory of what the previous review said.
I`d like to see a push for a sale based on what grocery supermarkets have for products.
There are enough benches about, and as long as all the factors are taken into consideration - maybe at the start of each year, some1 like Futuremark can name a "performance" pc, afa they concerned. Thereafter in the year, the products for sale can have $4.50/frame08 ($3.75/frame08OC) - or something similar attached to them - based on their results within that PC setup.
Crazy idea, but it only takes some1 in someplace like Futuremark to mention it at a meeting and we`ll all be seeing it in the ads. Don`t think they necessarily need to take to AMD/nVidia about it, and have their legal people draw up a waterproof disclaimer.
Meh, I`d like it. Same for cpus.Quote
Except a 3850x2 is like impossible to find and 3870x2 are cheaper anyway 
Whatever happened to the GDDR4 models then :S
Nice review though, I think cards should be based on frames relative to money a lot more.Quote

Whatever happened to the GDDR4 models then :S
Nice review though, I think cards should be based on frames relative to money a lot more.Quote
good to see that ATi have actually worked their way into a comfortable spot in the market