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
Figaro56 - Tuesday, March 8, 2011 - link
2 HD 6970 Cards for $640? I don't think so! These cards are over $300 everywhere. I purchased 2 for $710 shipped and I thought that was a deal. Maybe reviews like yours here inflated the price and I purchased after the price adjustment. I have the same luck with gasoline on days I fill my tank.ViRGE - Tuesday, March 8, 2011 - link
Looking at the Egg, there's 2 different 6970s at $320, which is probably where AT got $640 from.http://www.newegg.com/Product/Product.aspx?Item=N8...
Figaro56 - Tuesday, March 8, 2011 - link
All right, you got me there. I only buy XFX double lifetime warranty cards when I start spending this much on replacing my dual GPU solution.I seem to manage to actually re-sell my used video cards when I can offer then to a buyer with a lifetime warranty. XFX double lifetime warranty is not a sales gimic, it works. Heck, I would buy a used card if it had a lifetime warranty, it's kind of a no brainer given you actually want to buy that card int he first place.
Arbie - Tuesday, March 8, 2011 - link
Thanks for keeping the Crysis Warhead minimum FPS charts!! To me, Crysis/Warhead remains the defining game (and not only technically). I don't even look at the numbers on the other titles.Also of prime importance to me are the idle power and, to a slightly lesser extent, idle noise.
Of course, like most people reading your review, I wouldn't be buying a 6990 even if it were silent. In fact, given that PC graphics requirements are apparently ramping down to console levels, I wonder how AMD/Nvidia are going to sell any significant number of cards above midrange. My HD 5770 will run everything at 1920x1200, though not always with all sliders maxed. However, I don't see much if any difference (in DX9) when I do enable 4xAA vs 2xAA etc. Certainly not enough to double the price of this $140 card.
A nit on the Crysis Warhead minimum fps chart for 1920x1200 Frost Bench - Gamer Quality - Enthusiast Shaders + 4xAA: Your Dec 10 chart shows 6970CF at 66.2 fps but this Mar 11 chart shows 66.6. Can you believe anyone would actually notice this, much less comment on it? We are too absorbed in this tech stuff (ain't it grand...).
strikeback03 - Tuesday, March 8, 2011 - link
They did say the new drivers made a slight difference, that seems likely to be one of the configurations they retestedmorphologia - Tuesday, March 8, 2011 - link
That isn't portrait orientation in the picture...it's landscape.taltamir - Tuesday, March 8, 2011 - link
The card was measured at 77.3db in the article.1. At what distance was it measured?
2. What is its db measurement 1 meter away?
taltamir - Tuesday, March 8, 2011 - link
I just looked it up, gold is worth 1430$/ounce right now.I highly doubt a watercooled 6990 will weigh half an ounce.
ekrash - Tuesday, March 8, 2011 - link
The performance bottleneck is also seen in nvidia's dual gpu offerings. Dual GPU cards operating in X16 PCIe slots must have their data lanes divided between the gpu's, so they are effectively operating at X8 data rates, not at X16 data rates. Whereas single gpu cards will utilize all X16 express lanes, and even then the PCIexpress standard may soon be obsoleted. I hope we can look forward to Intel's fiber optic technology effectively replacing all data bus signalling with 10GB fiber optic bus and peripheral device signalling which can simultaneously and independently utilize all of the different data protocols used for inter-device and system bus communications. Imagine soon AMD and Nvidia will be producing video cards with fiber-optic data buses which may change requirements for power supplied to present day PCI express slots and may change the standards in power supply manufacturing to require that additional power connector to a video card since the 75 watt PCIe slot will be obsolete.But ATI and Nvidia may also have to work with motherboard manufacturers to see if Intel's "Thunderbolt" fiber optic data buses can increase or freely throttle the video data bandwidth through its 10GB interface and would be tantamount to increasing data lanes from X16 to X32. It would be almost unlimited video bandwidth which far exceeds any bandwidth limitations vs availability that is needed today. Dual GPU's cannot promise the performance with the limitation of the PCIe X16 slot being divided to dual X8 channels, but it would be nice to see how they perform with unlimited bandwidth potential over a single 10GB fiber-optic. And that would change the battlefield between ATI-AMD and Nvidia.
My 4870 X2's (Can run Quadfire) still rocks on enthusiast settings in Crysis and Warhead without any hiccups and I've not seen a slowdown of any sort on any level in Crysis.
The price to performance ratio is declining and may affect my decision to purchase another dual GPU card, opting instead for single GPU card CF solutions that can utilize all X16 lanes by the GPU.
BTW I did notice the lack of DATA on Crysis @1920x1200 with full enthusiast settings, so that data is missing from this review. Its Gamer plus enthusiast shaders.....not full enthusiast. As above the 4870 X2 runs full enthusiast settings, not one setting is scaled back, and not one hiccup....just smooth play throughout on a single 28" display.
cmdrdredd - Tuesday, March 8, 2011 - link
Why are we still using Crysis Warhead at "Gamer Quality"????? With cards like these why not turn everything maxed in game and then fidget with AA and the like? I don't get it.