Linux - HardwareThis forum is for Hardware issues.
Having trouble installing a piece of hardware? Want to know if that peripheral is compatible with Linux?
Notices
Welcome to LinuxQuestions.org, a friendly and active Linux Community.
You are currently viewing LQ as a guest. By joining our community you will have the ability to post topics, receive our newsletter, use the advanced search, subscribe to threads and access many other special features. Registration is quick, simple and absolutely free. Join our community today!
Note that registered members see fewer ads, and ContentLink is completely disabled once you log in.
If you have any problems with the registration process or your account login, please contact us. If you need to reset your password, click here.
Having a problem logging in? Please visit this page to clear all LQ-related cookies.
Get a virtual cloud desktop with the Linux distro that you want in less than five minutes with Shells! With over 10 pre-installed distros to choose from, the worry-free installation life is here! Whether you are a digital nomad or just looking for flexibility, Shells can put your Linux machine on the device that you want to use.
Exclusive for LQ members, get up to 45% off per month. Click here for more info.
After searching around many of the hardware sites and Linux sites out there, I wasn't able to find anyone talking about using AMD's new APUs under Linux, so I figured I'd finally make an account and start a conversation! (I'm a very long time reader lol...).
I am curious to know if anyone is playing around with OpenCL for general computing and how the graphics driver support is for AMD these days. Also, are there any optimizations that take into account the single shared floating point unit between two "cores"?
I'm personally still using a Core2 Q8200 and Nvidia GTS450 video card as my main computer, which I am happy with, so I can't personally make use of OpenCL. However, AMD's "heterogeneous computing" concept seems like a really smart way to bring excellent application performance on modest, well balanced hardware. From the benchmarks that I have seen on the web, Handbrake and Winzip OpenCL performance on the A10-5800K APU is very good, which makes me wonder what magic the minds in the Linux community can work with it.
...I can't personally make use of OpenCL. However, AMD's "heterogeneous computing" concept seems like a really smart way to bring excellent application performance on modest, well balanced hardware.
I think that is the easy part; it does sound like a smart idea, but, essentially, no one is using it for ordinary, general purpose, programs. You might find something obscure in the areas in which it has most potential (maybe CFD or weather forecasting or something), but for ordinary desktop activities, you'll find almost no ordinary desktop programs that are open source and run natively on Linux that do.
Quote:
Originally Posted by Tatwi
Also, are there any optimizations that take into account the single shared floating point unit between two "cores"?
I can only think of one general purpose optimisation that would make overwhelming sense and that would be adjust scheduling so that, as far as possible, fp-heavy apps would be scheduled onto different modules, and given that apps don't in any simple way declare how fp-heavy they are, or expect to be, I can't see how that happens in any easy way and I don't know of any efforts to make that happen (not that anyone would tell me, of course).
Outside of the 'obscure app division', there is one intra-app case that is interesting and that concerns games, and there I have to admit that I just don't know, but much magic happens there.
Outside of games and the 'obscure' stuff, fp-heavy isn't common anyway, so maybe it is being treated as a non problem. Except for compute servers?
Quote:
Originally Posted by Tatwi
After searching around many of the hardware sites and Linux sites out there, I wasn't able to find anyone talking about using AMD's new APUs under Linux, so I figured I'd finally make an account and start a conversation! (I'm a very long time reader lol...).
Phoronix is the only place I can think of that would be likely to.
I am curious to know if anyone is playing around with OpenCL for general computing and how the graphics driver support is for AMD these days. Also, are there any optimizations that take into account the single shared floating point unit between two "cores"?
As far as I know, no, there isnt anything that will use the GPU section of the APU that wont run the same or better with a standalone video card.
Quote:
Originally Posted by Tatwi
I'm personally still using a Core2 Q8200 and Nvidia GTS450 video card as my main computer, which I am happy with, so I can't personally make use of OpenCL.
OpenCL does run on GTS450 cards.
Quote:
Originally Posted by Tatwi
However, AMD's "heterogeneous computing" concept seems like a really smart way to bring excellent application performance on modest, well balanced hardware. From the benchmarks that I have seen on the web, Handbrake and Winzip OpenCL performance on the A10-5800K APU is very good, which makes me wonder what magic the minds in the Linux community can work with it.
A10-5800K is pointless IMO.
Get a FX-4XXX (or better) and a video card, or an intel. Its only a tiny amount more than buying a A10-5880K, is more upgradable, more flexable, and you arent buying a GPU on a CPU that probably wont have enough guts to play the games that you bougth the GTS450 to play (the HD 7660D in the A10 has less power than a GTS450, and even a $70 HD 6670 will easily outrun the HD 7660D)
There are a number of people using Llano and Trinity under Linux with no problems, though I agree not much onlien chatter about it. I have an A10-5800K siting on the desk here that I won that will likely find its way into a media machine or perhaps a small home server.
As the centre of a low power general use machine I think they'de do fine for most things. Intensive computer or graphics requirements would be better served with other processors and GPU's - but of course that's not the niche the AMD APU's are aimed at.
Thanks for pointing that out about OpenCL, Cascade9. I had not looked into it, assuming it special hardware level support, but it does work on Nvidia's CUDA GPUs with recent drivers. Nifty (and handy for me).
I feel kind of bad for AMD, in that they are willing to innovate with OpenCL, but Intel and Nvidia are right there doing it too. Apart from good value for the dollar on the +/- $100 cpu market, they just don't have much to offer anymore. I was hoping that OpenCL would give them a bit of a leg up if folks got on board to program for it, but I have the feeling the end result is that it will just end up working better on Nvidia video cards with Intel cpus. Poor AMD...
Still, it would be neat to see some useful apps, like GIMP, get some speed boosts when an AMD APU is available. May as well use those resources if they are there, right?
I feel kind of bad for AMD, in that they are willing to innovate with OpenCL, but Intel and Nvidia are right there doing it too. Apart from good value for the dollar on the +/- $100 cpu market, they just don't have much to offer anymore.
Intel isnt really doing much with OpenCL. IIRC MIC does support openCL, but who is going to buy one of them?
AMD still holds down Intel CPU prices, and even if thats all they had to offer its still A Good Thing.
Quote:
Originally Posted by Tatwi
I was hoping that OpenCL would give them a bit of a leg up if folks got on board to program for it, but I have the feeling the end result is that it will just end up working better on Nvidia video cards with Intel cpus. Poor AMD...
You might be suprised. Yes, I know, that is high end GPUs, AFAIK AMD has a similar advantage on lower end GPUs.
Still, it would be neat to see some useful apps, like GIMP, get some speed boosts when an AMD APU is available. May as well use those resources if they are there, right?
Yes, but a video card can do the same job as the onboard GPU.
LinuxQuestions.org is looking for people interested in writing
Editorials, Articles, Reviews, and more. If you'd like to contribute
content, let us know.