AMD two months ago released its Radeon HD 6800 graphics cards, and in doing so gave gamers some interesting choices. Preempting that launch, NVIDIA dropped the prices on a couple of its own models, so all in all, it was a great time to purchase a card. Last month, NVIDIA followed-up with its GTX 580, which succeeded the GTX 480 as being the fastest single-GPU model on the planet. And of course, we saw the launch of the GTX 570 last week.
Not wanting NVIDIA to have the last word for 2010, AMD has just released a couple of new parts that cater more towards the higher-end enthusiast, but at the same time, neither the HD 6950 or HD 6970 will break the bank. AMD is pricing these at $299 and $369, respectively, making them a competitor more towards NVIDIA’s GeForce GTX 570, not the GTX 580. Does that mean we won’t see some stellar performance from these cards? It depends on the viewpoint.
As is becoming an unfortunate theme as of late, we’re going to be a bit late on our launch article. In the meantime, we can talk a little bit about the latest offerings here. First, a table:
|
|
|
|
|
|
Radeon HD 6970 |
880
|
1350
|
2048MB
|
256-bit
|
1536
|
Radeon HD 6850 |
800
|
1250
|
2048MB
|
256-bit
|
1408
|
Radeon HD 6870 |
900
|
1050
|
1024MB
|
256-bit
|
1120
|
Radeon HD 6850 |
775
|
1000
|
1024MB
|
256-bit
|
960
|
Radeon HD 5870 |
850
|
1200
|
1024MB
|
256-bit
|
1600
|
Radeon HD 5850 |
725
|
1000
|
1024MB
|
256-bit
|
1440
|
Although AMD had hoped to base this launch on a 32nm process, it didn’t quite go according to plan, so for now, all of the HD 6000 series GPUs are built at 40nm. That doesn’t mean there aren’t new features, though, because the opposite is true. Continuing its goal for delivering the best power efficiency possible, an exclusive feature to the HD 6900 series is “PowerTune”, a tech that works to make sure the GPU never exceeds its max TDP. It does this by scaling the clocks during more strenuous use, and due to clever algorithms, overall performance shouldn’t be harmed.
Sound familiar? It should, because NVIDIA introduced similar tech with its GTX 500 series, as we talked a bit about in our launch article. In theory, the methods to keep the GPUs within their TDP limits are good, but it complicates things for us since it’s now much harder to get an accurate wattage reading when we test for it. We’re still in the process of figuring out the ideal test for that, since synthetic benchmarks is what this kind of scaling works against.
In addition to those improvements, the HD 6900 series also see a boost in tessellation performance, a refined architecture, improved anti-aliasing performance and a new anti-aliasing mode (EQAA (MSAA)). On the performance front, AMD touts the HD 6970 as being about 16~25% faster than the GTX 480, and in our tests so far, that seems reasonably accurate. According to 3DMark 11, the GTX 580 is about 8% faster, and at $500, it’s also about 35% more expensive. It’ll be interesting to see if NVIDIA plans to make another pricing move here, or if it is content where it is.
We’ll post our full look at AMD’s latest cards as soon as we’re able.