AMD Engineer Comments On Radeon HD 5870 Performance

AMD has, unofficially, responded to the somewhat disappointing performance of the new Radeon HD 5870. Below is reproduced the post by user "CarrelK", which is apparently Carrel Kilbrew, the lead engineer of the excellent RV770 design. Carrel has a good relationship with AnandTech lately, as you can see from this story. With no more delays, the post(top of page 17):

If a game ran almost entirely on the GPU, the scaling would be more of what you expect. You can put in a new GPU, but the CPU is no faster, main memory is no faster of bigger, the hard disk is no faster, PCIE is no faster, etc.

The game code itself also limits scaling. For example the texture size can exceed the card's memory footprint, which results in performance sapping texture swaps. Each game introduces different bottlenecks (we can't solve them all).

We do our best to get linear scaling, but the fact is that we address less than a third of the game ecosystem. That we do better than 33% out of a possible 100% improvement is I think a testimony to our engineers.

The card isn't horrible but giving it 1600 shaders and not enough bandwidth isn't really something worth praise. You have a decent card, not one as balanced as the previous ones and people, me included, were expecting a similarly well engineered card as the RV770. But I diverge, let's have a look at something:


Does it look like the test platform needs a new CPU, faster memory, hard-disk of faster PCI-Express slots? No, it doesn't. It still can go up to almost 35 FPS with an additional card. Extra memory? No, CrossFire doesn't double the framebuffer, you have 1GiB per card but it still only counts as 1GiB for the game.
If the Radeon HD 5870 could've really deliver double the performance, it should have at least went to 34.8 FPS, which is still less than double of what the HD 4890 can deliver and with less CPU/driver overhead from the second card. That kind of slight difference wouldn't surprise me, 40-60% of missing performance does.
What I was expecting to read but didn't were the words: "the driver is still immature". Well, we probably won't be seeing big boosts until they actually strap 6 GHz GDDR5 to the thing.

The card would be excellent if it would've fit the same $199/$299 bracket as before but it's not so surprising at $259/$379. The price was well known to be higher than $299 before the launch and that brought some higher expectations about performance. The magic die size target was missed and that messed plans for AMD, which still has to charge more for the bigger die, regardless of the actual performance.