AMD Demos x86 APU Server Running Fedora Linux

[piesquared]

AMD is showing off the new server chip, code-named "Berlin," at Red Hat Summit 2014 in San Francisco April 16. The Fedora Project is a community-based Linux distribution sponsored by Red Hat.

Having both Linux and Java running on the AMD APU platform will help showcase how HSA-based Opteron APUs can lead to server environments that efficiently process workloads in both general-purpose and multimedia environments, according to AMD officials. The chip vendor also will demonstrate software based on OpenCL and OpenGL running on Berlin.

Here it comes! I lol at people dismissing HSA.


http://www.eweek.com/servers/amd-demos-x86-apu-server-running-fedora-linux.html

 

[CHADBOGA]

Having both Linux and Java running on the AMD APU platform will help showcase how HSA-based Opteron APUs can lead to server environments that efficiently process workloads in both general-purpose and multimedia environments, according to AMD officials. The chip vendor also will demonstrate software based on OpenCL and OpenGL running on Berlin.

Company PR is supposed to impress anyone?

This post is like a rice cake. It looks impressive, but is in fact devoid of all content. In the future please write more productive posts.
-ViRGE

 

[Phynaz]

OMG! An x86 CPU that runs both Linux and Java!! What will AMD come up with next?!!!

 

[mrmt]

Those anemic APUs are being touted as flagship in the server market? AMD has reached new lows here, I really pity the marketing guys that have to spin this in the press.

 

[Sweepr]

Yawn. Wake me up when they come up with something that actually rivals a 14C/28T Haswell-EP (coming later this year) in most tasks.

 

[MisterLilBig]

This would be the first time HSA will be shown, no? Well, atleast something close to what HSA can do.

Not sure why the negatives, first ever HSA hardware and probably showing of some OpenCL 2.0 functionality in an app of sorts.

The gains, if any, should be a sign of things to come, and not just from AMD.

 

[mrmt]

This would be the first time HSA will be shown, no? Well, atleast something close to what HSA can do.

Not sure why the negatives, first ever HSA hardware and probably showing of some OpenCL 2.0 functionality in an app of sorts.

OpenCL 2.0 can be much better served with dGPU cards or Phi cards, and the rest of workloads will be much better with 8-12 cores Intel CPUs. These "server" chips are only enough for dime a dozen web hosting boxes, not even for light duties server or even workstations worthy of the name.

There may be a very narrow spectrum of workloads that will benefit from having the GPU and the CPU on the same die, but these results would be offset due to sheer brute force from other solutions (heavier dGPU and/or heavier CPU).

As I said, it's shameful that the company that created the 64 bit market for x86 servers is reduced to this.

 

[MisterLilBig]

I'm optimistic. An HSA Java showing, hopefully.

 

[mrmt]

I'm optimistic. An HSA Java showing, hopefully.

Optimistic based on what? On the cherry picked benchmarks AMD has been showing? Or in a real business case for that product? I really can't think of a business case for these things in a workstation environment, let alone in a server environment. It might be a nice poor student's toy, but once we start talking about workstations or servers, things does change. Workstations requires brute force, servers require brute force and scalability, AMD APU has neither.

 

[piesquared]

This would be the first time HSA will be shown, no? Well, atleast something close to what HSA can do.

Not sure why the negatives, first ever HSA hardware and probably showing of some OpenCL 2.0 functionality in an app of sorts.

The gains, if any, should be a sign of things to come, and not just from AMD.

[redacted] No matter though, HSA is coming regardless and has a very large part of the industry backing it. Good times ahead, the industry can finally crawl out from under intel's monopolistic foot.

Another article here from Ars

http://arstechnica.com/information-...mos-hsa-for-the-server-room-with-java-on-top/

Seattle is next up on the list of game changers.

You of all people should know our policy on meta commentary
-ViRGE

 

[Abwx]

Company PR is supposed to impress anyone?

You think that technical information can be released while being mum.??.

Better to look if there s not something that is of relevance and indeed you quoted an interesting information it s just that you missed the good line.

Berlin and other AMD chips—such as "Kaveri," for PCs—will be built on HSA Foundation specifications, which include the ability to view the GPU and the CPU as a single processor and to move the workloads to whichever one is needed. In addition, programmers will no longer need to be concerned with determining whether the software will run on the CPU or GPU; the system will make the decision. GPUs also will get more access to memory.

Having both Linux and Java running on the AMD APU platform will help showcase how HSA-based Opteron APUs can lead to server environments that efficiently process workloads in both general-purpose and multimedia environments, according to AMD officials. The chip vendor also will demonstrate software based on OpenCL and OpenGL running on Berlin.

 

[MisterLilBig]

Optimistic based on what? On the cherry picked benchmarks AMD has been showing? Or in a real business case for that product? I really can't think of a business case for these things in a workstation environment, let alone in a server environment. It might be a nice poor student's toy, but once we start talking about workstations or servers, things does change. Workstations requires brute force, servers require brute force and scalability, AMD APU has neither.

Which benchmarks??

Edit: Optimistic on performance improvements on Java and OpenCL applications using HSA hardware.

 

[Cogman]

HSA will only work with a good programming interface. Right now, our programming interfaces for HSA like work are, frankly, terrible.

The first big issue is that the model for writing programs looks something like this:

Write in C, load giant string make api call to compiler for giant string, pray that nothing bad happens, pass compiled code into magic API, query magic api for results from program". The giant strings are either CUDA, OpenCL, or some other monstrosity.

The next problem is that the languages are terribly restrictive and low level. Do you like playing with random memory addresses and pointers? because that is what you get, loads of memory addresses and pointer work (especially if you want to be fast).

And finally, the type of data which this is good for is pretty limited. It works great if you want to take every value in a big blob of data and multiply it by 8. It isn't so great if you need to perform a bunch of financial logic that has to access seemingly random portions of the data. In other words, it makes good short work of things like cryptography, compression, and maybe a couple of other things. It fails at handling a request for a web server, talking with a database, or several other mundane but all to common problems.

Even things that benefit from moer cores won't necessarily benefit from 1000 extra GPGPU cores. (A web server being a great example of this).

 

[CHADBOGA]

You think that technical information can be released while being mum.??.

You don't appear to be writing about anything I wrote.

Where did I say that technical information couldn't or shouldn't be released?

Better to look if there s not something that is of relevance and indeed you quoted an interesting information it s just that you missed the good line.

All it says is that software will be shown running.

They don't say they will be able to show any increased benefits to software running on other platforms.

 

[Ajay]

Does Berlin offer quad channel memory? If so, then there might be some good uses for this, especially if Berlin features a larger or faster iGPU. Otherwise, a desktop APU which is hamstrung by limited bandwidth will just become a server APU hamstrung by limited bandwidth. HSA isn't a cure for bandwidth problems.

 

[piesquared]

Another good article here. HSA sure seems to be gaining a lot of traction!

The compute and graphics cores in the HSA-enabled X2100 work together by means of heterogeneous unified memory architecture (hUMA) and heterogeneous queueing (hQ), which enable threads to run simultaneously and independently on all of the CPU and GPU cores, and sharing the same memory space so that the CPU doesn't have to get involved with feeding the GPU.

http://www.theregister.co.uk/2014/0...neous_system_architecture_opteron_at_red_hat/

 

[mrmt]

Another good article here. HSA sure seems to be gaining a lot of traction!

Yes, just like Bulldozer, Barcelona.... you know the drill.

 

[CHADBOGA]

Another good article here. HSA sure seems to be gaining a lot of traction!

That article refers to the same event as your opening post does.

How is that gaining any further traction?

 

[Vesku]

The next problem is that the languages are terribly restrictive and low level. Do you like playing with random memory addresses and pointers? because that is what you get, loads of memory addresses and pointer work (especially if you want to be fast).

Goal of HSA is exactly to minimize the amount of low level code that needs to be written. Shared memory space and such. Can't see any downsides to pursuing HSA yet somehow there are always posters raging about how silly it is.

Yet somehow the forums remain free of "omg why does Intel's Phi try to make coding for it closer to coding for CPUs".

 

[sefsefsefsef]

A few months ago I attended a keynote by the guy behind HSA who's now at nVidia (Norm Rubin), and while he couldn't talk specifically about how great HSA is because of his current affiliation, he did an excellent job of selling the notion that the future lies in ISA-independent languages, like Java and OpenCL. I'm personally aboard this particular hype train because of my own research, and how universal languages fit into it.

If you hate AMD, that's fine. I personally think it's silly to hate a company, but I won't stop you. AMD might even totally fail in this particular implementation and attempt at HSA, but I think the history books will show that something like HSA + high level languages are the way to go. The cool thing isn't how you can run code on a CPU or GPU (IMO GPUs are almost as lame as CPUs), but how this will maybe make it easy and practical to run our general purpose code on something way more powerful and efficient than either.

Sorry if there are mistakes, I wrote this out on my phone in a busy airport.

 

[frozentundra123456]

Just because someone doesnt accept unproven claims of a company spokesman about how widely accepted and beneficial their special software is going to be is a far cry from hating AMD. There might also have been less backlash if the OP had been made in a more analytical and less confrontational manner.

 

[mrmt]

If you hate AMD, that's fine. I personally think it's silly to hate a company, but I won't stop you. AMD might even totally fail in this particular implementation and attempt at HSA, but I think the history books will show that something like HSA + high level languages are the way to go.

Not believing AMD marketing material isn't the same as hating AMD, especially because that company has a poor track record of upholding their claims.

And as I said in other threads, execution is everything. AMD doesn't have what it takes to make something as big as HSA take off.

 

[piesquared]

Another article here on the subject.

“As servers adapt to new and evolving workloads, it’s critical that the software ecosystem support the requirements of these new workloads,” Suresh Gopalakrishnan, corporate vice president and general manager of the Server Business Unit, AMD said yesterday at the demonstration. “We are actively engaged with a broad set of partners in the data center software community who are bringing to market the software infrastructure to seamlessly enable x86 APU based servers.” In addition to Fedora and HSA, softwares based on Open-CL parallel programming open standard and OpenGL cross-language, multi-platform application programming interface (API) for rendering 2D and 3D vector graphics on the “Berlin” APUs are also expected to be demonstrated at the summit.

http://wccftech.com/amd-demo-fedora/#ixzz2z6YCRQLI

Looks like there is a broad based ecosystem in place, nice. Hmm and Red Hat is a member of the HSA foundation I believe, interesting. Well it looks like AMD won't be 'going it alone' after all, nice! HSA is almost certainly going to have a big impact.

 

[sandorski]

It will be interesting to see the results.

 

[Ajay]

A few months ago I attended a keynote by the guy behind HSA who's now at nVidia (Norm Rubin), and while he couldn't talk specifically about how great HSA is because of his current affiliation, he did an excellent job of selling the notion that the future lies in ISA-independent languages, like Java and OpenCL. I'm personally aboard this particular hype train because of my own research, and how universal languages fit into it.

Wow, NV pushing OpenCL - now that's a shocker! The specifically didn't implement OpenCL 1.2 on Kepler (which was annoying). I wonder what they'll do with Maxwell? Would love to see them implement v2.0. Off topic I know...