AMD's Radeon HD 6990: The New Single Card King
by Ryan Smith on March 8, 2011 12:01 AM EST- Posted in
- AMD
- Radeon HD 6990
- GPUs
Compute Performance
Moving on from our look at gaming performance, we have our customary look at compute performance. With AMD’s architectural changes from the 5000 series to the 6000 series, focusing particularly on compute performance, this can help define the 6990 compared to the 5970. However at the same time, neither benchmark here benefits from the dual-GPU design of the 6990 very much.
Our first compute benchmark comes from Civilization V, which uses DirectCompute to decompress textures on the fly. Civ V includes a sub-benchmark that exclusively tests the speed of their texture decompression algorithm by repeatedly decompressing the textures required for one of the game’s leader scenes.
New as of Catalyst 11.4, AMD’s performance in our Civilization V DirectCompute benchmark now scales with CrossFire at least marginally. This leads to the 6990 leaping ahead of the 6970, however the Cayman architecture/compiler still looks to be sub-optimal for this test. The 5970 has a 10% lead even with its core clock disadvantage. This also lets NVIDIA and their Fermi architecture establish a solid lead over the 6990, even without the benefit of SLI scaling.
Our second GPU compute benchmark is SmallLuxGPU, the GPU ray tracing branch of the open source LuxRender renderer. While it’s still in beta, SmallLuxGPU recently hit a milestone by implementing a complete ray tracing engine in OpenCL, allowing them to fully offload the process to the GPU. It’s this ray tracing engine we’re testing.
There’s no CrossFire scaling to speak of in SmallLuxGPU, so this test is all about the performance of GPU1, and its shader/compute performance at that. At default clocks this leads to the 6990 slightly trailing the 6970, while overclocked this leads to perfect parity with it. Unfortunately for AMD this is a test where NVIDIA’s focus on compute performance has really paid off; coupled with the lack of CF scaling and even a $240 GTX 560 Ti can edge out the $700 6990.
Ultimately the take-away from this is that for most desktop GPU computing workloads, the benefit of multiple GPU cores is still unrealized. As a result the 6990 shines as a gaming card, but is out of its element as a GPU computing card unless you have an embarrassingly parallel task to feed it.
Facebook
Twitter
RSS
130 Comments
View All Comments
EmmetBrown - Tuesday, March 8, 2011 - link
Nice, but what about the Radeon HD 6450, 6570 and 6670?http://en.wikipedia.org/wiki/Comparison_of_ATI_Gra...
Why they are available for OEM only? They looks interesting, especially the 6670, which with its 480 SP should be faster than the 5670 which has 400 SP and lower frequency. Do you plan to review them?
Ryan Smith - Tuesday, March 8, 2011 - link
As you note, they're OEM only. AMD will release them to the retail market eventually, but clearly they're not in a hurry. It's unlikely we'll review them until then, as OEM cards are difficult to come by.misfit410 - Tuesday, March 8, 2011 - link
I have to ask, if you bring up the price and say that you might as well do two 6950's in SLI when this thing doubles the performance of the GTX580, I mean would it also not be the better solution than a GTX580 which is $500 while two 6950's can apparently double it for $550 being they can be found for $225 after rebates these days.Figaro56 - Tuesday, March 8, 2011 - link
You sound a little confused. You can't run ATI cards in SLI, they run in what is called crossfire (or crossfirex which is the same thing). Two 6950's don't equal GTX580 in SLI. You need two HD 6970 cards in crossfire to nearly equal two GTX580 in SLI.In my opinion, why limit your performance with 2 HD 6950 cards, why not just bye the 2 HD 6970 cards and never have to second guess if you should have or not? But... That's just me. I have a job.
silverblue - Tuesday, March 8, 2011 - link
Totally unnecessary closing comment there, considering most people here do actually have jobs. Not everyone who has a job can afford such gear as there's more important things to spend money on.Thanny - Tuesday, March 8, 2011 - link
You sound confused, too. He miswrote SLI, but you misunderstood his point entirely. He's saying that two 6950's are significantly faster than a single 580 for almost the same price.Loiosh - Tuesday, March 8, 2011 - link
Hey guys, you forgot one other usage case that would necessitate this card: ATI+physx setup: http://www.shackpics.com/viewer.x?file=DumbVideoca...I'm currently running one and it requires a dual-GPU card. :/
In my case I'm waiting for a watercooled version. BTW, you didn't say the release date for this?
nanajuuyon - Wednesday, March 9, 2011 - link
Funny after reading this review I went into town (Tokyo) to buy a new hard disk and saw this card for sale. So in Japan at least it is already on the market..... price was ridiculous though, 79,000YEN or $945 US..... I'm sure it will be available everywhere soon.Waterblocks on the other hand could be a couple month or so away I guess...
Vinas - Tuesday, March 8, 2011 - link
If you buy this you better have it on water. 'nuff said about all this tri slot cooler talk.JPForums - Tuesday, March 8, 2011 - link
First off, nice article Ryan.Good data, relevant commentaries on said data, and conclusions.
You mention in the article that you believe some of the shortcomings of the 6990 to be a lack of PCIe bandwidth. This got me thinking that perhaps it is a good time to revisit the effect of PCIe bandwidth on modern cards. Given the P67 only natively supports 16 lanes, I'm curious to see what effect it has on CF/SLI. It could make big difference in the recommended hardware for various levels of gaming systems.
Typically, someone looking for a CF/SLI setup will get a board that supports more lanes. However, I have seen a situation where a friend built a budget i5 system and about 4 months later was in a position to acquire an HD5970 on the cheap (relatively speaking). Clearly, two HD5850s/HD5870s would have been an option.
If newer cards are effectively PCIe bandwidth limited, then a 6990 may perform more closely to an HD6970 CF setup in such a system than it does in these graphs. This would be even more of a consideration at the high end if the rare boards with support for 4x8 lane (spaced) PCIe give you no real benefit over a more common 2x16 lane board (comparing 4 HD6970s to 2 HD6990s).