Page 1:Are Sixteen Lanes Enough?
Page 2:Test Hardware And Benchmarks
Page 3:Benchmark Results: 3DMark 11
Page 4:Benchmark Results: Aliens Vs. Predator
Page 5:Benchmark Results: Crysis
Page 6:Benchmark Results: F1 2010
Page 7:Benchmark Results: Just Cause 2
Page 8:Benchmark Results: S.T.A.L.K.E.R.: Call Of Pripyat
Page 9:PCIe Scaling Summary
Page 10:Bonus Page: Does CrossFire Work At x16/x4?
We found that Nvidia’s high-flying GeForce GTX 570 suffers around 2% performance loss when using an x8 slot, while the Radeon HD 6950 gives up around 4%. While noteworthy in respect to the negative impact on system value, those losses probably wouldn’t be noticeable in actual game play.
More significant are the losses when using a x4 slot, as expected. The GeForce GTX 570’s relatively significant 9% performance loss is far exceeded by the Radeon HD 6970’s 18% loss. And this is where some builders will hit an advice snag, otherwise known as self-proclaimed experts analyzing an effect and incorrectly deducing the cause.
Tom’s Hardware readers often rely on each other to provide the best technical advice, and the advisor offering the most technically-detailed description often gets the nod (hey, if it sounds fancy, it's more likely to be right, right?). Unfortunately, we often see well-meaning enthusiasts leading less-experienced readers astray. Case in point: we have one pervasive poster who offers exceptionally-detailed explanations of why Nvidia graphics cards need no more than four lanes of PCIe 2.0 bandwidth, citing something about internal PCIe to PCI-X conversion. We've also seen presumed experts tell curious gamers that AMD relies so heavily on its CrossFire bridge that a four-lane slot is ideal for hosting a second card. Today’s test results prove both suggestions completely wrong. We’re hoping that today’s article will be the impetus other advice-givers need to chime-in when the community calls for suggestions on how to best configure multi-card configurations.
Putting aside our concern for the honest knowledge-seekers, we can conclude that minimal performance losses in x8 mode set us up for the perfect logical next-step in our three-part series, where we compare X58 and NF200 solutions to today’s board. Will that 2% to 4% difference remain in SLI? Does the NF200 really hurt graphics performance by adding latency? Will the smaller performance gains of three-way CrossFireX mitigate the performance deficits of its x4 slot? Stay tuned to find out!
- Are Sixteen Lanes Enough?
- Test Hardware And Benchmarks
- Benchmark Results: 3DMark 11
- Benchmark Results: Aliens Vs. Predator
- Benchmark Results: Crysis
- Benchmark Results: F1 2010
- Benchmark Results: Just Cause 2
- Benchmark Results: S.T.A.L.K.E.R.: Call Of Pripyat
- PCIe Scaling Summary
- Bonus Page: Does CrossFire Work At x16/x4?