Kabini vs CT/ARM: GPU Performance

I pulled out 3DMark and GFXBenchmark (formerly GL/DXBenchmark) for some cross platform GPU comparisons. We'll start with 3DMark Ice Storm and its CPU bound multithreaded physics benchmark:

3DMark—Physics

The physics test is a bit unreasonably multithreaded, which is why we see a 75% uplift compared to AMD's E-350. For FP heavy game physics workloads however, Jaguar does quite well. While a big Ivy Bridge is still going to be quicker, AMD's A4-5000 gets surprisingly close given its much lower cost.

The 3DMark graphics test is more of what we're interested in seeing here. Two GCN compute units (128 SPs/cores) running at 500MHz will really put the old Radeon HD 6310 in Brazos to shame:

3DMark—Graphics

The results are quite good. Kabini manages a 61% performance advantage over AMD's old Brazos platform, and actually gets surprisingly close to Intel's HD 4000 in performance. As we discovered earlier, this isn't really enough performance to play modern PC games but casual (and especially tablet) gaming workloads should do wonderfully here.

3DMark—Ice Storm

The overall Ice Storm score just incorporates both physics and graphics test components. As expected, Kabini continues to lead over everything other than the i5-3317U.

Finally we have the GFXBenchmark T-Rex HD test. I threw in a handful of older PC GPUs, although keep in mind that T-Rex HD isn't very memory bandwidth intensive (penalizing some of the old big PC GPUs that had good amounts of memory bandwidth). The test is also better optimized for unified shader architectures, which helps explain the 8500 GT's excellent performance here.

GL/DXBenchmark 2.7—T-Rex HD (Offscreen)

Kabini does very well in this test as well. If we look at the tablet-oriented Temash part (A4-1200) we see that the number of GCN compute units remains unchanged, but max GPU frequency drops to 225MHz from 500MHz. If we assume perfect scaling with GPU clock speed, Temash could offer roughly the same graphics performance as the 4th generation iPad. AMD claims the A4-1200 Temash APU carries a TDP of only 3.9W, a potentially very interesting part from a GPU perspective if our napkin math holds true.

OpenCL Performance

For our last comparison we're looking at the OpenCL performance of these on-die GPUs. We're using a subset of Ryan's GPU Compute workload, partially because many of those tests don't work properly on Kabini yet and also because some of those tests are really built for much more powerful GPUs. We've got LuxMark 2.0 and two CLBenchmark 1.1.3 tests here. Their descriptions follow:

SmallLuxGPU is an OpenCL accelerated ray tracer that is part of the larger LuxRender suite. Ray tracing has become a stronghold for GPUs in recent years as ray tracing maps well to GPU pipelines, allowing artists to render scenes much more quickly than with CPUs alone.

CLBenchmark contains a number of subtests; we’re focusing on the most practical of them, the computer vision test and the fluid simulation test. The former being a useful proxy for computer imaging tasks where systems are required to parse images and identify features (e.g. humans), while fluid simulations are common in professional graphics work and games alike.

OpenCL GPU Performance
  LuxMark 2.0 CLBenchmark—Vision CLBenchmark—Fluid
AMD A5-5000 (Radeon HD 8330) 18K samples/s 1041 1496
AMD E-350 (Radeon HD 6310) 23K samples/s 292 505
Intel Core i5-3317U (HD 4000) 107K samples/s 819 1383

LuxMark is really a corner case here where Kabini shows a performance regression compared to Brazos. The explanation is simple: some workloads are better suited to AMD's older VLIW GPU architectures. GCN is a scalar architecture that is usually going to net more efficient usage, but every now and then you'll see a slight regression. Intel's HD 4000 actually does amazingly well in the LuxMark 2.0 benchmark.

Our two CLBenchmark tests however paint a very different picture. Here Kabini not only significantly outperforms its predecessor, but is faster than Ivy Bridge as well. 

Kabini vs. Clover Trail & ARM Kabini Windows 8 Laptop Performance
POST A COMMENT

130 Comments

View All Comments

  • takeship - Thursday, May 23, 2013 - link

    True, price will keep most budget buyers out of Haswell powered ultrabooks. Not so for the now-on-clearance-sale Ivy Bridge Dell & HPs though. And in that market Kabini loses most of it's price advantage, while still giving worse performance & marginal battery life improvements. There it's the new $500 plastic Kabini laptop, vs. the $600 aluminum IVB Lenovo. I just don't see that being a win. Reply
  • Gaugamela - Thursday, May 23, 2013 - link

    The HP Pavilion 11 Touchsmart costs 400$. It has a 10-point touchscreen. So, the 400$ touch enabled Kabini seems mighty atractive compared to the 600$ Lenovo now. Reply
  • axien86 - Thursday, May 23, 2013 - link


    Touche! Thanks to AMD for providing alternatives to Intel's Atoms and higher priced CPUs.
    Reply
  • kyuu - Thursday, May 23, 2013 - link

    Besides the price issue, you seem to be quite missing the point of a low-power architecture like Jaguar. Let me know when ULV Ivy Bridge can scale down to 3.9W.

    Also, a lot of people seem to have trouble comprehending the fact that TDP doesn't really have much in common with the actual power draw of the chip (or the heat output). We've already seen what happens when you try to cram even the lowest wattage Ivy Bridge into a modern tablet form factor: high temperatures w/ fans, low battery life, high price.
    Reply
  • Gaugamela - Thursday, May 23, 2013 - link

    It competes with an Sandy Bridge ULV i3 and it gets quite close to a i3 Ivy, while offering better battery life.
    Seems like a clear proposition to me: if Pentium like performance is what you need this offers you that and better GPU performance. Perfect for the low-end of the ultrathin/ultrabook market.
    Reply
  • andrewaggb - Thursday, May 23, 2013 - link

    yeah, pretty much how I read it as well. Assuming you can afford to pay an extra $200 dollars there's not much reason (other than possibly graphics drivers) to not get haswell instead. Reply
  • aryonoco - Thursday, May 23, 2013 - link

    Very interesting article. Just a few notes:

    I'm not sure that Exynos 5250 is the best representative of Cortex A15. For one, it's dual core without any HT so at a massive disadvantage in a test like Kraken. Secondly, it came out in actual device in the market about 7 months ago. By the time Kabini shows in products, it will probably be over 8 months later. We'll have other Cortex A15 parts by then, and if the Tegra 4 Kraken scores that I'm hearing (~6000) is right, and if something Tegra 4 is using about half as much power as this 15W Kabini, then Kabini suddenly doesn't look that competitive.

    Of course as you say, the success of Kabini will depend on what OEMs do with it, but traditionally these cheap AMD parts find themselves in devices that are compromised in everyway. Even if we get a 1080p screen this time around (which I'm hopeful), we'll still have to deal with sub-optimal keyboards, trackpads that work only half the time, and other cost-cutting measures we are familiar with. For this to succeed, someone needs to put it in a Ultrabook-style laptop, with good display, good keyboard and no software glitches, and price it under $500. That's a tall task.

    HP just announced a very interesting Slatebook X2 running Android (not skinned) on Tegra 4 with 1080p IPS screen and a Transformer-style detachable dock for $479, including the dock (sidenote, I was disappointed there was no coverage of this on AT). If the performance claims for Tegra 4 (and other future more optimized Cortex A15 parts) are accurate, with such devices on the horizon, will there even be a cheap Ultrabook-style market left for Kabini to compete in? Can Kabini find its way into similar designs by major OEMs? I hope, for everyone's sake that AMD can succeed, but I am doubtful.
    Reply
  • JarredWalton - Thursday, May 23, 2013 - link

    I certainly think Kabini can go into the same designs that we're seeing Tegra 4 target. Which will end up being faster? That's a different question, and I'm not sure we have enough information to come to a conclusion right now. If Kabini/Temash can get into tablets sooner rather than later, they've got a chance. If it takes 6-8 months, you're right: it may not matter at that point. Reply
  • Gaugamela - Thursday, May 23, 2013 - link

    HP just announced their back to school products and there is a successor to the HP dm1 that will cary A4's and A6's. Reply
  • Exophase - Friday, May 24, 2013 - link

    No Cortex-A15 has HT, and Kraken is single threaded anyway. I think Kraken just isn't a great benchmark. Although its makers say it's a lot better than Sunspider it shares a lot of the same sorts of problems:

    1) It's Javascript which has its own unique (pretty severe) overheads which can dominate run-time to the extent that it drowns out a lot of the variation from the type of JS code you're running. JS is used and it's important, but even today on good JITs most JS code is several times (think 5 times) slower than an equivalent written in something like C++ or even Java, making it a bad representation of performance of more efficient software.
    2) Much of what it does - path finding, signal processing, and crypto - is not the kind of stuff Javascript is usually used for.
    3) Its test durations are really short, which is bad for benchmarks in general but can be especially bad for JITs where it
    4) The variation in current browsers is extreme (http://arstechnica.com/civis/viewtopic.php?f=8&... where you can see some tests are substantially faster and others substantially slower. This again highlights the big overheads of Javascript and the impact of different JIT strategies, but also that the state of performance is still pretty volatile. A similar sort of variation could manifest between different backends (ie, x86 vs ARM) even for the same browser.

    Cross-architecture comparisons are hard and I don't blame people for using JS when there's not a lot else available (although at least some other inclusions would be nice).. but you shouldn't draw a very broad conclusion from a Kraken comparison alone. And if you did that it'd also make the Kabini vs i5 comparison look a lot worse than a lot of other tests show.
    Reply

Log in

Don't have an account? Sign up now