AMD Preps For Server Graphics Push 41
Nerval's Lobster writes "AMD named John Gustafson as senior fellow and chief product architect of AMD's Graphics Business Unit, the former ATI graphics business unit. Gustafson, known for developing a key axiom governing parallel processing, will apply that knowledge to AMD's more traditional graphics units and GPGPUs, co-processors that have begun appearing in high-performance computing (HPC) systems to add more computational oomph via parallel processing. At the Hot Chips conference, AMD's chief technical officer, Mark Papermaster, also provided a more comprehensive look at AMD's future in the data center, claiming that APUs were the keystone of the 'surround computing era,' where a wealth of data — through sensors, gestures, voice, augmented reality, metadata, and HD video and graphics — will need to be contextualized, analyzed, and either encrypted or assigned privacy policies. That, of course, means the cloud must shoulder the computational burden."
Good (Score:2)
Nvidia could really use some competition in the server space. The render farms and most GPGPU (or CUDA) are pretty much completely dependent on Nvidia.
A Wealth of Data, Huh? (Score:2, Informative)
So it has nothing to do with graphics then... (Score:2)
Re:So it has nothing to do with graphics then... (Score:5, Informative)
Except a modern GPU is basically a coprocessor that, 99% of the time, is used to run a library that primarily does graphics. Rendering, shading, transformation, those are now all done "in software". The only things still done "in hardware" are texture lookups and video output (turning an int[4][1080][1920] into a DVI or HDMI or VGA or whatever signal.
They're also a pretty high-volume market, so you get them much cheaper than you would a custom-built coprocessor or even FPGA, and they're *probably* better-designed than the one you would make, as they have entire teams of professionals working on them.
Also, both nVidia and AMD already make "compute-only" cards - nVidia under the brand "Tesla", AMD under the brand "FireStream".
Re: (Score:1)
Re: (Score:3)
You'd be surprised at how useful servers with GPUs are these days. When you're talking about clients like iPad and android devices, often the rendering is done server side and then sent to the client. A (CAD related) product I work on renders thumbnails using a server GPU. There is also a game service that does all rendering server side and sends it to a display (often a TV).
That oughta work (Score:4, Insightful)
Geez, didn't we have this stuff years ago, only it was called mainframes and minicomputers?
Someone refresh my memory as to why we fled those for PC's? Oh yeah, it cost too much to centralize, the 'one size fits all' solutions actually fit no one, and it took too long to wait for someone to fix things or come up with new tools.
Same problem with "the cloud". Good luck with it.
Re: (Score:2)
Agreed. Absolutely nothing has changed since the 70s. Nope. Nothing.
I wish it had anything at all to do with technology or things that have improved over time, but it doesn't. Centralizing resources and having those under one persons charge while many others want to use it or use it in a currently unsupported way costs more and gives less flexibility. That human issue has been the case for a lot longer than the 70's.
Re: (Score:1)
Centralizing resources and having those under one persons charge
Soooooo... you want to fire your CTO? Not sure how cloud computing is somehow run by one person. Maybe you just have no idea what you are talking about.
Re: (Score:3, Informative)
Having watched half an hour of the Republican Convention last night, I'll have to agree.
Re:That oughta work (Score:5, Insightful)
We have things we didn't have last time.
Massive central storage
Enormous bandwidth
Excellent frameworks for distributed processing (no RPC does not count)
Long ago.. your Cloud had to be custom built for the app. EC2 doesn't have that restriction. I know the people who developed S3. They had no idea they'd be hosting thier Killer App. (Netflix) at design-time. It's that flexible.
Plus, we now have PCs. No one is saying we have to go back to thin clients, you can keep your PC, and you the cloud where it excels. Gmail and Netflix Streaming are both things for which I've done the equivalents on home servers, and they don't hold a candle to the cloud versions.
Re: (Score:1)
Yeah, the virtualization instruction flagset in the Intel/AMD processors that were invented only a couple of years in the past have nothing to do with any of this.
(/sarcasm)
Can they just not work on their device drivers? (Score:1)
Re: (Score:2, Interesting)
Didn't you know? They're open source now. Fix the problem yourself!
Sarcasm aside, I feel AMD open sourcing the drivers was more because they're throwing up their hands in surrender; they can't manage it themselves, so they're asking for outside help.
AMD also provides a library that makes it easy to write a userspace program to disable all fans and thermal throttling on the GPU - melt the thing; maybe even start a fire... useful feature, that.
The beauty is that if a user can run a GL program (or even a GPU c
Not for the foreseeable future (Score:5, Interesting)
There's a big problem, however: http://developer.amd.com/sdks/AMDAPPSDK/assets/App_Note-Running_AMD_APP_Apps_Remotely.pdf [amd.com]
To run apps that use AMD's GPU's remotely (ie. not from a local X11 session - and I mean "Local X11 session"), you have to open a security hole so big you can fit Rush Limbaugh's ego through it.
* Log into the system as root. /dev/ati/card*
* Add "Xhost +" to your X11 startup config (so every X session allows anybody to access it... with root permissions)
* chmod ugo+rw
I asked a group of devs from X.org how stupid it was... the short answer is "how stupid is giving root access to everybody?"
So, I asked AMD when they were planning on fixing the problem.
Short answer: Not for the foreseeable future.
I seem to recall a similar issue where CERT told users not to use AMD drivers for Windows, because it forces Windows to disable many of its security features.
I'm sensing a trend...
Do you want this kind of irresponsibility in the datacenter? EVER?
Re: (Score:2)
The problem is that to use OpenCL (or ATI's Stream SDK) to offload work from the CPU, you have to do the "xhost +" breakage, which is a serious problem for anybody who actually cares about security.
Re:Not for the foreseeable future (Score:4, Interesting)
Which Windows' security features? I wasn't aware of this. :(
Re: (Score:2)
IIRC It's DEP. My wintel crashed for ages without it disabled. Had to be uber vigilant but I don't think it's an issue now.
Re: (Score:2)
DEP for all programs or just essential Windows programs and services?
Re: (Score:2)
You have to disable DEP entirely or the drivers will fail. At least at one point you couldn't even open CCC.
Re: (Score:2)
Wow. I only have essential programs and services in my old Windows XP Pro. SP3 machine. No problems.
Re: (Score:2)
http://tech.slashdot.org/story/12/06/07/1653206/amdati-video-drivers-unsafe-at-any-speed [slashdot.org]
ASLR appears to be one of them, as is DEP.
Re: (Score:2)
Thanks. :)
Re: (Score:2)
Re:Not for the foreseeable future (Score:4, Interesting)
I think AMD is jumping into the arena because they feel they have to:
- NVIDIA is already making quite a splash in big data processing with their many-core GPGPU offerings
- AMD already offers their FirePro line to compete with NVIDIA's Tesla and Quadro
- Intel is entering the arena with their MIC/Xeon Phi product line (http://en.wikipedia.org/wiki/Intel_MIC)
AMD apparently feels they have to go down a similar path. Hopefully they will do it in a way better than is possible with their competition's offerings; NVIDIA doesn't build a full CPU on-die with their GPU, and Intel appears to have chosen not to.
Additionally, NVIDIA's and presumably Intel's many-core offerings can easily swamp the latest PCIe Gen3 bus with the number of cores they have. The total memory per core on the GPU or Phi device isn't that high, so it's very easy to become bounded by PCIe's I/O bandwidth - they have to transfer boatloads of data over the PCIe bus.
For some workloads, you can get some great performance gains; it's also important to remember that while NVIDIA (for one) likes to trumpet their 20-30x performance increase, the fact is they're cherry-picking workloads that are well-suited to their product. In my experience, it's 3-5x in the general case, because of their fundamental limitation in memory bandwidth between the PCIe card and other sources of memory - be it the "main" system memory, RDMA via InfiniBand, etc.
I'm confident AMD will design decent hardware - they might even turn a corner and make great hardware again. Without the software to drive it, however, it's a lost cause - and I have zero confidence in AMD's ability to develop that software.
Re: (Score:1)
NOT your $100 graphics card. I'd imagine what they are going to produce will go into server farms with custom made software.
I couldn't care less about desktop graphics; it just isn't interesting to me.
My original post was about my experiences with their $1-2k FirePro boards that compete with nVIDIA's Tesla & Quadro, to be slotted into several hundred nodes of a supercomputing cluster. If that isn't a server environment, then what is?
I hate to break it to you, but AMD's attention to software, even for a
I have waited patiently (Score:1)
I have waited patiently for Intel's offering 'Knights Corner' now rebranded as "Xeon Phi". We have been tempted with 64+ cores and 4 threads per core. We have been tempted with 'run existing software'. But I don' see anything available in stores. That they aren't pushing product into stores means the thing will be gawd-awful expensive, production is limited, and they don't want people to spend a grand or two, and create amazing software around the hardware. Instead, the word 'Xeon' means in general 'no