Compute & Tessellation Performance

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. Note that this is a DX11 benchmark.

While Turks and the 6670/6570 are quite a bit faster than the 6450, both are still so-so as compute cards. Although the unit of measurement (fps) is effectively arbitrary, what isn’t arbitrary is that the 5750 is nearly 33% ahead of the 6670. Juniper (5750/5770) seems to be the floor for good compute performance on AMD cards.

Our second compute benchmark of the day 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. AMD in turn recently hit a milestone by offering the OpenCL runtime with their drivers, so SmallLuxGPU will now run out of the box without the AMD Compute SDK.

At 3200K rays/sec, the 6670 actually beats the Radeon HD 4870 by about 10%, showcasing just how far AMD’s compute performance has come compared to the 4000 series. The 6670 also enjoys a similar lead over the 5670 thanks to its additional shaders, but the 720SP 5750 looms large above it.

At the other end of the spectrum from GPU computing performance is GPU tessellation performance, used exclusively for graphical purposes. Barts’ tessellation improvements are inherited by Turks, which should give it an edge over the 5670.

What ended up surprising us the most with our DX11 Detail Tessellation Sample test wasn’t so much the raw performance of the 6670, it was the performance relative to some other AMD cards. At a high tessellation factor the Radeon HD 5870 isn’t all that far ahead, and meanwhile the 5750 is edged out by the 6670 in all cases. Given that our most tessellation-intensive games struggled at 1680x1050 on the 6670 it may not be powerful enough to really use its tessellation abilities in full on top of a heavy graphical workload, but the basis for respectable tessellation performance is there.

Wolfenstein Power, Temperature, and Noise
Comments Locked

53 Comments

View All Comments

  • Ryan Smith - Tuesday, April 19, 2011 - link

    Yes, they do. So does the 6450, which is why I suspect 6450 is going to be the best HTPC card for most people.
  • SteelCity1981 - Tuesday, April 19, 2011 - link

    I got a 5670 and the 10% increase in performance isn't worth me spening an extra 100+ dollars on a 6670. See what AMD has in stores with the Radeon 7 series.
  • tomoyo - Tuesday, April 19, 2011 - link

    Yep, it's not for anyone with a current 5xxx. For me, I may go for this card because I want something low power for htpc + casual gaming, and it should be silence-able as well with the low idle power usage. Also hdmi will be nice with audio when connecting to tvs :)
  • G-Man - Tuesday, April 19, 2011 - link

    Ryan

    Are there any plans to do articles where you "revisit" old conclusions and see if they still hold true? Like for Radeon 5670, it wasn't priced well at launch and suffered for it in the review, but after price cuts it's pretty much recommended everywhere in the price segment. It's often hard to understand exactly what Anandtech recommends at the moment, as opposed to at launch.

    Thanks!
  • tomoyo - Tuesday, April 19, 2011 - link

    Sounds like one of the only useful things from tomshardware, the video card price guide. I'd say you could always look there if you want to know the best current price/performance cards.
  • 789427 - Wednesday, April 20, 2011 - link

    Ding, Ding...
    I enjoy both sites as both provide me with news and different perspectives but Mr Smith, kindly refrain from bad-mouthing your colleagues.

    Let me provide you some insight that seems to refute Anandtech and Tomshardware *shared* claims about the latest and best in graphics...

    Both sites focussed on noise as being the decision criteria for adopting one top end card over the other. Availability is more pertinent in the case of the high end cards.

    To focus on noise as the deal-breaker seemed quite moot when Nvidia doesn't intend shipping 590 cards in volume.

    Furthermore, a reference design at this level is simply a speed showcase - I don't for a second think that the majority of cards will be sold based on the reference design but at this level, manufacturers will listen to their client's desires (or to the noise the card actually makes and make some changes;))

    In this particular instance, I was amazed at how much your opinions overlapped.

    The reason that I've gone to such lengths with this particular example is to highlight what the most useful thing about Anandtech and Toms actually is - a difference of opinion that better highlights the truth often shaded by opinions and preconceptions.

    Your comment above should seriously be reconsidered as I believe that as a genre, both sites have actually contributed to each other's success far more than any individual, including yourself, has.

    My 2c.

    Finally, about the cards... I find both fill a market. new purchasers of 1080 screen non-gamers will be satisfied with the lower offering and people that just want good performance without changing PSUs, heating a room and generally want good bang for the buck. How many people pay MSRP online? So the launch price tends to be higher... Just state the price that you'd recommend it and save us from reading the same old comments over again...

    So, the 5750 is going to become an outmode as soon as the 7 series is out. and there are cards to replace it that are cheaper to make and run cooler. Not quite news but good to see in the flesh. Wasn't the real story of the 6 series that AMD finally got crossfire to scale right? ah well...
  • Soulkeeper - Tuesday, April 19, 2011 - link

    I'd like to see the stats tables include the memory bandwidth so I don't have to calculate it myself each time I read one of these.

    thanks
  • AstroGuardian - Thursday, April 21, 2011 - link

    +1
  • Arnulf - Tuesday, April 19, 2011 - link

    This not only goes to show how much hardware has improved over time but it also gives owners of older cards, perhaps with a rattling fan or some other issues, plenty of incentive to upgrade. My X1950Pro gave up on me recently so I went for Vapor-X version of HD5770. New card is virtually silent compared to that old beast.

    Keep up this practice and consider including comparable models from BOTH manufacturers ! :)
  • vavutsikarios - Tuesday, April 19, 2011 - link

    There are 2 things I want to say, both a bit sideways the topic.

    First, about the paperlaunched 6450, I dont get it. The card is for sale, retail, at various shops here 2 weeks now. Doesnt look like a paperlaunch from where I stand, and this is not the most central part of the world (i live in greece)

    Second, and more important, there is an upcoming game I d like to see in reviews. I am talking about might and magic heroes 6, scheduled to launch on June 23.
    There is an interesting twist about it, most probably. I mean, if the game is anything like its predecessors, its going to be much more useful as a CPU than as a GPU benchmark. I mean, you can easily play a turn based strategy while on 15-20 fps. They are enough. But when I run the 6 years old heroes 5, with all the graphics details on high, no problem there, I still have to wait for the AI to complete their turn. The game is CPU limited SIX YEARS after it was released!!

Log in

Don't have an account? Sign up now