AMD teams with Oracle on Java acceleration, Qualcomm on HSA

AMD has teamed up with Oracle to work on Project Sumatra, a bid by the OpenJDK community to explore how Java can be accelerated by graphics processors. The project will investigate adapting not only the Java virtual machine, but also the APIs and the language itself. Here’s a snippet from the official press release:

“Affirming our plans to contribute to the OpenJDK Project represents the next step towards bringing heterogeneous computing to millions of Java developers and can potentially lead to future developments of new hardware models, as well as server and cloud programming paradigms,” said Manju Hegde, corporate vice president, Heterogeneous Applications and Developer Solutions at AMD. “AMD has an established track record of collaboration with open-software development communities from OpenCL™ to the Heterogeneous System Architecture (HSA) Foundation, and with this initiative we will help further the development of graphics acceleration within the Java community.”

The Heterogeneous System Architecture is AMD’s intiative to allow software to exploit the combined CPU and GPU resources offered by chips like its A-series APUs, and another big name signed on in support this week. Qualcomm has joined the HSA Foundation as a Founder member. At the end of August, the HSA group also welcomed Samsung into its Founder fold, which includes ARM, Imagination Technologies, MediaTek, and Texas Instruments. If the graphic on the HSA site is correct, only one Founder slot remains.

AMD has a distinct advantage over Intel in the graphics department. With GPU-accelerated applications few and far between, the payoff has been restricted largely to games. Through efforts like Project Sumatra, OpenCL, and HSA, AMD should be able to convince more developers to exploit the graphics horsepower of its APUs. Let’s hope it doesn’t take too long.

Comments closed
    • Scrotos
    • 7 years ago

    What with all the recent JVM exploits hitting the news, I have visions of JAVA now somehow mangling my video card.

    • ronch
    • 7 years ago

    I bet they wouldn’t be doing all these collaborations on HSA and OpenCL if they had a very strong CPU architecture. The good part is that somehow, because of Bulldozer AMD is doing work on these other technologies which, overall, advances computing technology. Who knows, Bulldozer’s misfortune could actually be a blessing in disguise for AMD.

    • Bensam123
    • 7 years ago

    Imagine the performance benefits in Minecraft!

    • provoko
    • 7 years ago

    Where’s my reverse hyperthreading!

      • ronch
      • 7 years ago

      If I recall correctly AMD talked about something like this back when Bulldozer was in its early stages of development. The way I understood it, a single thread can run across multiple cores, which did seem really cool (and really, really difficult to pull off). Nowadays though, it looks like AMD has pretty much kept silent about it and ironically, it’s Intel who’s trying to make it easier for programmers to work with multiple cores with (Acura) TSX. TSX is nowhere near as cool as reverse HT IMHO, but hey, at least Intel’s doing something, compared to AMD who just wants to throw more and more cores at you. On the good side at least AMD is working with HSA and OpenCL. That field is also quite exciting.

      In the end, we have two tech companies working on new ways to make computing faster. That’s a great story in my book.

      • Bensam123
      • 7 years ago

      Yup, I remember hearing about both Intel and AMD talk about this like 7-8 years ago when duel core chips were first coming out… They were designing something at the hardware level that would allow the processor to multithread a single thread. It wasn’t hyperthreading.

    • vaultboy101
    • 7 years ago

    Has anyone noticed that AMD has been cute and left a space for Nvidia to join as a founding member of the HSA Foundation?

    Truly Nvidia have to join as all their major ARM competitors are members.

    It would be neat if Apple joined HSA as well.

      • Alexko
      • 7 years ago

      I think neither Intel nor NVIDIA is likely to join this initiative from AMD, precisely because it’s from AMD, but yes, apart from those two, Apple is the only big name missing.

      Still, AMD has managed to get a lot of big players on board, which bodes very well for HSA.

      • ronch
      • 7 years ago

      The proliferation of OpenCL and HSA couldn’t come at a better time for AMD, and I suppose if you were AMD’s boss you would want to spearhead efforts in those fields because you have a comparatively weak CPU architecture and would like to leverage your strengths, trying to attract as many members as you can so you won’t be left playing alone in your sandbox. However, not allowing Nvidia to join and instead being “cute” for leaving Nvidia a seat gives others the impression that AMD is not spearheading all this for its own gain, but rather, for the entire industry as a whole. I don’t think they have a choice on whether to leave a seat for Huang and the Gang on this one or not.

      • Wirko
      • 7 years ago

      Either that, or they are waiting for Google to join. Even if that wouldn’t benefit AMD, it would benefit ARM, Qualcomm and others.

      Funny, though, that Founders can join long after the foundation has ben founded.

    • brucethemoose
    • 7 years ago

    For anyone interested in this subject.

    [url<]http://forum.beyond3d.com/showthread.php?t=58195[/url<] There are some hidden gems in that thread that really use GPUs well. I found Musemage there, for example, and I use it over Photoshop 90% of the time. These sort of programs are more common than you think. EDIT: Why don't AMD/Nvidia post a list like this? AMD's "Vision" site is useless (Musemage has "Vision" branding, but wasn't even listed on AMD's site), Nvidia's isn't targeted at consumers. EDIT2: Seems AMD finally improved their Vision page. By improved, I mean Musemage at least shows up. Links are still broken though.

      • Ringofett
      • 7 years ago

      Well, they did fire their entire marketing department this past year, didn’t they? If thats a symptom of that or the cause of that, I don’t know.

      Edit: Thanks for that link though, wasn’t aware of some of those, and for some others I wasn’t aware GPU acceleration had been added. Might change some of my go-to apps.

    • jdaven
    • 7 years ago

    One of these two companies should buy AMD and use their technology to do something awesome.

      • destroy.all.monsters
      • 7 years ago

      You’ve been around TR long enough to know that x86 won’t transfer and that without it their value is minimal. Oracle would just butcher it like it did Sun and leave whatever parts weren’t in their immediate plans to die and Qualcomm has no interest in the desktop/laptop space and again no x86. The last thing that Qualcomm cares about is the enthusiast market.

      What it sounds like you’re wishing for is an AMD that actually has money and a forward thinking CEO as well as marketing department. All of which is just wishful thinking.

        • NeelyCam
        • 7 years ago

        [quote<]You've been around TR long enough to know that x86 won't transfer [/quote<] Do we know that's still the case after the settlement..?

          • eofpi
          • 7 years ago

          Did the settlement modify the terms of the cross-licensing agreement?

      • ronch
      • 7 years ago

      Well, I think AMD is still pretty awesome.

    • MadManOriginal
    • 7 years ago

    [quote<]AMD should be able to convince more developers to exploit the graphics horsepower of its APUs[/quote<] Unfortunately, AMD has a horrible track record of their APIs becoming important in the mainstream. It's usually only when other companies pick them up that they begin to matter. At least their are some other important companies pushing this.

      • ptsant
      • 7 years ago

      Unfortunately, this is true. I do believe, however, that openCL in a few key applications (Adobe suite, Handbrake, some codecs, maybe Office) would be sufficient to change the experience for the end user and substantially change benchmark results. Already, the acceleration in ZIP is quite meaningful and it appears that x264 with OpenCL might be coming soon. I certainly hope it catches on, it would justify my next GPU purchase 😉

      • shank15217
      • 7 years ago

      quite untrue, most amd standards have either been extremely influential or game changing.. hypertransport, amd64 were wildly successful.. 3dnow pushed intel to release sse and brought about the demise of x87 instruction set finally.. just to name a few.. also SOI tech with IBM put AMD and IBM years ahead of Intel in the 90nm and 65nm node (Intel used more capital and won back in 45nm and beond). AMDs fab automation systems was widely regarded and is now industry standard. AMD/Ati also played as big part in DX9/10/11 standards because they released the first parts and worked with Microsoft on all three standards.

        • Meadows
        • 7 years ago

        Those are not APIs.

          • shank15217
          • 7 years ago

          how do you know HSA is an API?

          • MadManOriginal
          • 7 years ago

          Not the right term, I know, but whatever, you get the idea.

        • MadManOriginal
        • 7 years ago

        “when other companies pick them up that they begin to matter”

        Funny, you prove this in your post while trying to disagree.

    • NeelyCam
    • 7 years ago

    [quote<]AMD has a distinct advantage over Intel in the graphics department.[/quote<] Not in performance/power. First

      • sschaem
      • 7 years ago

      Benchmarks prove otherwise.

        • NeelyCam
        • 7 years ago

        Links?

          • Arclight
          • 7 years ago

          Yes…..he meant you got links to prove it?

      • BobbinThreadbare
      • 7 years ago

      [url<]https://techreport.com/review/22932/amd-a10-4600m-trinity-apu-reviewed/13[/url<] Faster GPU, lower power draw.

        • NeelyCam
        • 7 years ago

        That test doesn’t measure GPU power consumption; it measures the whole system. Also:

        [quote<]"The Intel systems had larger panels than the AMD ones, though, so that might have impacted power consumption."[/quote<] and [quote<]"It's no surprise to see the Trinity whitebook pulling off longer run times than the two Intel notebooks, since those have bigger displays and more power-hungry CPUs. "[/quote<]

          • Meadows
          • 7 years ago

          So you’re reinforcing the position that intel is less power-efficient.

            • NeelyCam
            • 7 years ago

            Intel is more power-efficient; that review doesn’t compare apples to apples.

            • Meadows
            • 7 years ago

            You quoted it:

            [quote<]"[...] longer run times than the two Intel notebooks, since those have [...] more power-hungry CPUs."[/quote<]

            • NeelyCam
            • 7 years ago

            Power efficiency has two components – power consumption [i<]and[/i<] performance. You forgot the second one.

        • bcronce
        • 7 years ago

        Haswell is coming out next year, 60%(under load) to 2000%(idle) less power draw then Ivy Bridge.

          • BobbinThreadbare
          • 7 years ago

          Ok, well then next year things might be different.

          • jensend
          • 7 years ago

          2000% less power draw, wow! In other words, when idle it will [b<]generate[/b<] ~190W of power which can be used by the rest of the system! That's a real accomplishment! (Homer: LIS-AAAA!!! IN THIS HOUSE, WE OBEY THE LAWS OF THERMODYNAMICS!!)

            • bcronce
            • 7 years ago

            20x = 2000%

            1/20th

            How I meant it anyway.. bit tired

            • willmore
            • 7 years ago

            No, that’s not how it works. 100% less would mean zero. Think about it for a bit.

            • bcronce
            • 7 years ago

            I think my problem is in both EvE and WoW, they refer to lowering cycle/cast times in percentages that can be greater than 100%.

            Your cast time is now 200% shorter!!
            You cycle time is reduced by 120%!!

            • Meadows
            • 7 years ago

            Those games say no such things. A cast time that’s “200% shorter” would [i<]finish twice before you even press the button[/i<]. 1/20 power draw means it's reduced by 95%.

            • willmore
            • 7 years ago

            Exactly, using percent confuses the issue. Use factors and it makes sense that you have to do the reciprical for decreases.

          • NeelyCam
          • 7 years ago

          [quote<]2000%(idle) less[/quote<] Love it! <3

      • cal_guy
      • 7 years ago

      [url<]http://www.anandtech.com/show/5771/the-intel-ivy-bridge-core-i7-3770k-review/20[/url<] In exchange for 17W the HD 5570 gets between 20-30% gains in Metro 2033 (other games show much greater gains) despite the HD 5570 having an extra GB of DDR3 memory, a small fan, and separate VRM and is also built on a far inferior 40nm process.

        • NeelyCam
        • 7 years ago

        Interesting. I guess I was wrong

          • Unknown-Error
          • 7 years ago

          HD 4000 of i7 3770K is trigate 22nm, now compare to HD 7660D of A10-5800K is 32nm,

          Metro 2033 : [url=http://www.anandtech.com/show/6332/amd-trinity-a10-5800k-a8-5600k-review-part-1/2<]Link-1[/url<], [url=http://www.anandtech.com/show/6332/amd-trinity-a10-5800k-a8-5600k-review-part-1/8<]Link-2[/url<] 1366 x 768 LQ DX9 - [b<]39.5[/b<] (HD 7660D) Vs. [b<]32.5[/b<] (HD 4000) 1366 x 768 LQ DX11 - [b<]41.3[/b<] (HD 7660D) Vs. [b<]29.5[/b<] (HD 4000) 1680 x 1050 LQ DX9 - [b<]28.3[/b<] (HD 7660D) Vs. [b<]20.5[/b<] (HD 4000) GPU Load Power Consumption - [b<]133.6[/b<] (HD 7660D) Vs. [b<]93.4[/b<] (HD 4000)

            • NeelyCam
            • 7 years ago

            And the plot thickens..

            • BobbinThreadbare
            • 7 years ago

            Kind of cherry picking there.

            the 7560D is withing 7% of the performance of the 7660, yet cuts 20% off power consumption (down to 107 watts).

            Also the AMD GPU creams the Intel one in the other graphics tests (getting for instance double the performance in Total War: Shogun 2).

            The AMD GPU seems to benefit from DX11 too, but they didn’t run 1680×1050 in DX11, which is kind of disappointing.

            • Unknown-Error
            • 7 years ago

            Cherry picking? I just used the best iGPUs from the respective companies. I am too lazy to type the whole bunch. But it does show how well the AMD iGPU does even on an inferior manufacturing process. I am looking forward to the 65W A10-5700 reviews.

            • BobbinThreadbare
            • 7 years ago

            Well yeah, AMD doesn’t need their best iGPU to beat Intel in performance, and we’re trying to find the best performance/power.

            Ideally, one would use an AMD iGPU that is as close to identical to Intel’s as possible and see how much power each one takes.

            The article itself had more problems than your post did, and I don’t think either one of you was biased, just not using the best information.

Pin It on Pinterest

Share This