AMD hits the peak of performance in gaming and productivity

Subject: General Tech | August 7, 2014 - 09:45 AM |
Tagged: HPC, amd, firepro, S9150, S9050, opencl

The new cooling on the 290X tends to have it at the top of the gaming charts and with the impending release of two new FirePro HPC cards AMD looks to take the productivity title away from the Tesla K40.  The higher end S9150 boasts 16GB GDDR5 memory with a 512-bit memory interface, 44 GCN compute units with 64 stream processors each there is a total of 2816 stream processors on board.  That equates to 5.07 TFLOPS peak single-precision  2.53 TFLOPS peak double-precision performance with theoretical memory bandwidth of 320GB per second.  AMD expects the S9150 to have support for OpenCL 2.0 drivers by the end of the year, which the lower priced and specced S9050 will not though both will support AMD Stream technology and OpenCL 1.2.  Check them out at The Register.

View Full Size

"The company's new big gun is the FirePro S9150 card, which maxes out at a blistering 5.07 TFLOPS peak single-precision floating-point performance and 2.53 TFLOPS peak double-precision performance."

Here is some more Tech News from around the web:

Tech Talk

Source: The Register
August 7, 2014 | 10:18 AM - Posted by Anonymous (not verified)

If AMD expects to compete with Nvidia, in the Future, for this HPC business, then AMD is going to have to get its GPUs integrated on a mezzanine module with the CPU! Currently AMD has its x86 products, and AMD's version of HSA, but for HPC AMD needs to get a HPC/server APU going, or it needs to License the Power8, or other high performance CPU designs, and connect their HPC GPUs with a coherent interconnect fabric to whatever high performance CPU IP that AMD has, and AMD really needs to get their new x86 microarchitecture rework done, or just license Power8, and do what Nvidia is doing, connecting HPC CPUs with its GPUs on a ultra wide/fast data bus, traced out on a mezzanine module, PCI is just not fast enough, and the Nvidia Pascal type systems will have the advantage. These Mezzanine modules systems will allow for 1024 Bit, or wider buses, and utilize CAPI(Coherent. Accelerator Processor. Interface) types of interconnects. AMD is doing good by releasing this HPC accelerator, but future HPC systems will require that the GPU accelerators be connected by faster than PCI, speeds, and have the ability to maintain coherency with massive amounts of data, that will be processed by the GPU/CPU HPC systems. Nvidia has the lead, as it is already developing such systems with IBM, for use with Power8 systems, and everything is going mezzanine module, coherent fabric, and stacked RAM, all on the same module for the exascale computing systems under development.

August 7, 2014 | 11:43 PM - Posted by Anonymous (not verified)

Man, just shut up with that Power8, mezzanine crap. You're really annoying with that broken record style. We all get it; the CPU and GPU need to reside on the same substrate and share high speed RAM, cache and other resources in order to really shine. But just think for a second:

YOU ARE PREACHING TO THE CHOIR.

We are not the unwashed masses that you seem to think we are. And re-stating it over and over wont change a thing. It will happen when AND ONLY WHEN they decide to make it happen. Furthermore, did it ever occur to you that if you know it, then AMD's engineers must also be aware of it? Or maybe you think you've suddenly stumbled onto something new?

Chill out and make the most of what you have ATM.

August 9, 2014 | 03:01 PM - Posted by Anonymous (not verified)

Hay Duck wanker man, Power8 is going to force Intel to lower its Xeon prices, Power8 is going to replace Xeon in a lot of high power servers, at Google, and other server installations! You do not speak for the WE, Intel damage control, but the Intel Technology milking is about to end, via some real Power8 competition, and Maybe Nvidia will tape out some Power8, with Mezzanine Nvidia GPUs, with its own Power8 license, Nvidia does not need to beg Intel for an x86 license anymore, Nvidia is already working with IBM for the server variants with its GPUs, so Nvidia get some Power8/Pascal based gaming rig monsters for the home gaming server market. For sure Apple, and their PA semiconductor Boffins, are doing the same, I need 12 cores, and 8 treads per core, and rock bottom prices, just like the mobile market gets its ARM SOCs, competition is good, Chipzilla is scared! AMD will have to get in the market for power8, just like it is in the market for ARM, its compete or die, once those third party Power8's begin to flood the market, starting in 2015, Bwa! Ha ha Ha! AMD better get some of that action, for its server business, instead of selling seamicro servers with Xeon, sell them with Power8, power8, and AMD GPUs, and some Mezzanine module goodness, Innovate or die! Power8 Power8, x86 has learned its fate!

Post new comment

The content of this field is kept private and will not be shown publicly.
  • Lines and paragraphs break automatically.
  • Allowed HTML tags: <a> <em> <strong> <cite> <code> <ul> <ol> <li> <dl> <dt> <dd> <blockquote><p><br>
  • Web page addresses and e-mail addresses turn into links automatically.

More information about formatting options

By submitting this form, you accept the Mollom privacy policy.