Performance Summary

On the whole, Nvidia’s GeForce 8800 GT is about 17 percent faster than both the GeForce 9600 GT and Radeon HD 3870 when they’re all running in their stock configurations. With many Radeon HD 3850 512MB cards coming clocked higher than the predetermined reference clock speeds, we decided that it was fair to include a higher-clocked Radeon HD 3850—in this case we’ve used PowerColor’s Xtreme PCS 512MB card—and that turned out to be around 6.5 percent slower than Nvidia’s GeForce 9600 GT.

When you factor in price, the GeForce 9600 GT is good value for money, but then so is the Radeon HD 3850 512MB – in fact, the 3850 (based on the numbers obtained from the PowerColor card at least) make it slightly better value for money. Obviously, performance is slightly lower, but it’s not that much slower if you’re willing to save around £10 (or eight percent).

The GeForce 8800 GT also represents great value for money and, with the price changes, the Radeon HD 3870 is left a little redundant if we’re honest. It makes sense to save the £20 that the 3870 costs over a Radeon HD 3850 512MB, and just overclock the 3850 if you’re feeling a little short on the performance front.




Where things get interesting is when you take anti-aliasing into account because, as I’m sure you know, the current Radeons don’t perform as well when anti-aliasing is enabled. When AA is disabled, the 3870 is actually faster than the 9600 GT, while the 3850 is almost in a dead heat with Nvidia’s newest mid-range card. When AA is enabled though, the 3870 and 3850 are around eight and 13 percent slower than the GeForce 9600 GT.

That’s not the only thing we can take from these results though, because the GeForce 8800 GT has its own anti-aliasing related story to tell. You see, in order to make the G92 graphics chip affordable, Nvidia chopped out G80’s 384-bit memory bus and left the new chip with a 256-bit wide bus.

From our performance data, the GeForce 8800 GT is, on average, around 20 percent (and sometimes up to nearly 30 percent) faster than the 9600 GT when AA is disabled. However, when AA is enabled, the performance difference drops down to just 13 percent – it’s clear, then, that the GeForce 8800 GT is being held back by some of the design decisions made to keep the cost of the chip down.

If anything, that makes the GeForce 9600 GT look better than it actually is in our opinion, because, as has always been the case in the world of 3D graphics, the reliance on more complex shaders increases over time. It’s quite feasible, then, that the GeForce 8800 GT will extend its performance lead over the GeForce 9600 GT over time. The only question is by how much – and that’s something that is pretty hard to quantify.

Final Thoughts...

There are a lot of great products in the performance mainstream sector that all appear to hit price points that are so finely balanced that it’s hard to quantify which is better value for money. Indeed, it was one thing that cropped up during my initial GeForce 9600 GT testing and it was what prompted me to write this article.

Could it be that there are maybe too many good products in this segment? On one front, this is obviously a good thing for the consumer, but has it got to the stage where the consumer is actually left in uncertain waters because there is no definitive answer to the ‘what graphics card should I buy?’ question.

We’re interested to hear your thoughts on this subject in the forums.