Bookmark and Share


ScaleMP, a provider of virtualization solutions for high-end computing, has announced a strategic collaboration with AMD for vSMP Foundation to support AMD Opteron processor-based servers. The collaboration enables users to scale AMD-based servers beyond four processors and create AMD Opteron processor-based virtual systems with up to 512 sockets (up to 8192 cores) and 64TB of memory.

"The ability for our customers to aggregate individual AMD Opteron processor-based servers into a virtual high-end symmetric multiprocessor (vSMP) computer with vSMP Foundation is a game-changer. Our customers can now easily build a cost-effective large memory and CPU system out of up to 128 servers to help meet their high performance computing needs," said Margaret Lewis, director, product marketing, AMD.

Traditional multi-processor systems require the creation of a custom chipset and board to enable processor communication and shared memory. The larger the system is, the more complex and costly the required solution. vSMP Foundation creates a single virtual symmetric multiprocessing system (vSMP) by combining multiple commodity servers and enabling multiple physical systems to function as a single logical system. It is unknown how efficient such virtual SMP systems are compared to multi-socket custom machines.

vSMP Foundation will support both AMD Opteron 6100- and 6200-series chips with up to twelve and up to sixteen cores, respectively.

“Supporting AMD servers has been a goal of ours. AMD has a long history of offering powerful, yet cost-effective solutions and we have always seen a natural fit. Through this collaboration, ScaleMP enables AMD customers to create large virtual SMPs that will lower the TCO for organizations – enabling customers to focus their time and budgets on their applications and workloads," said Shai Fultheim, founder and chief executive officer of ScaleMP.

The upcoming release of vSMP Foundation will support aggregating up to 128 AMD processor-based servers, providing customers with up to 64TB RAM and 512 sockets, offering thousands of cores. A limited release of vSMP Foundation will be available to customers starting October 1, 2011, followed by a general availability November 21, 2011.

Tags: ScaleMP, AMD, Opteron, Interlagos, Bulldozer, Orochi, Valencia, 32nm


Comments currently: 15
Discussion started: 09/21/11 08:33:53 AM
Latest comment: 07/13/16 10:58:24 AM
Expand all threads | Collapse all threads


Nice to see AMD constantly getting more industry and enterprise support for their products. It's all good for consumers.
4 0 [Posted by: beenthere  | Date: 09/21/11 08:33:53 AM]

This is interesting. From a software development point of view, the code no longer needs to worry about using special libraries and compiler directives to do massively parallel workloads. I wonder how fast x264 could rip a DVD using 8192 cores...
0 1 [Posted by: user99  | Date: 09/21/11 08:35:37 AM]

Like that.
0 0 [Posted by: Pouria  | Date: 09/21/11 10:15:40 AM]

...or faster.
1 0 [Posted by: beenthere  | Date: 09/21/11 11:19:55 AM]

who cares abouth ripping a dvd the question is can in run crysis.
but yea it is a good idea to drive the prices down of a system.
0 0 [Posted by: massau  | Date: 09/21/11 01:09:35 PM]
- collapse thread

I don't think some people see the humour. Obviously you'd be ripping BluRay not DVD with that much compute power.
0 0 [Posted by: user99  | Date: 09/22/11 11:42:07 AM]

Probably the idea of running 1 application using a pool of 8000 threads to spread the job could be a small hell if not done in the right way: all this CPUs has to share data, exchange messages and the bottleneck will always be the same: the interface betweens servers.
Of corse would be easier to aministre such kind of system (specialy if it would be allowed to plug/unplug servers without halting the system). But the comunication between a thread/application running fisically in a server to and a thread/application running in another would never be as faster as in a single fisical server. Probably the bigger advantage would be to share easely and coerentely a common filesystem, and to distribute automatically the work between different servers (hopefully linux kernel scheduler would do that for us).
0 0 [Posted by: Serenico  | Date: 09/21/11 03:10:23 PM]

Looks like the re-birth of the age of the mainframe CPU. Who needs powerful CPUs in your workstations when your server has 8000 cores. I can imagine quite a rennaisance in AMD server CPUs with this technology. Hopefully, nothing equivalent will be easily implementable on the Intel side for a while.
2 0 [Posted by: anubis44  | Date: 09/21/11 03:28:23 PM]

This is just what SGI have been doing with Intel Xeons for years.
0 0 [Posted by: Andys  | Date: 09/21/11 09:32:41 PM]
- collapse thread

no, it's not. first of all, SGI was doing it (numalink, years ago) with ia64, not Xeons. second, numalink is higher bandwidth and lower latency, and has direct hardware assist. scalemp is basically middleware that handles page-based coherency, not cachelines.
0 0 [Posted by: markhahn  | Date: 09/22/11 10:57:50 AM]

wasn't it on news about another japanese SUPER computer archived over 8 pflop. if vsmp and AMD ARE so sure about their products than let it run that LINPACK and prove itself to the world
1 0 [Posted by: idonotknow  | Date: 09/21/11 11:11:02 PM]

unfortunately, scalemp doesn't public performance numbers (the ones that matter, latency, rate and bandwidth). without those numbers, it's impossible to judge whether their middleware is worth looking at for other projects. (also, nothing about this kind of above-hardware-based coherency depends on cluster size - doing it on 128 servers just means you can put together a working IB network of that size. which is nothing to write home about.)
0 0 [Posted by: markhahn  | Date: 09/22/11 11:01:08 AM]

yeah get that one up i will install the cheapest/free os and software pack and away the world goes without wasting millions. of course i mean linux/ubuntu + virtualbox just divide those 8192 cores into 128 seperate system with each having 64 cores. that should be plenty for any corporate size system all in one location/data center.

OMFG where was that system went all of a suden that host server got a BLUE screen. i think all 128 server are gone for good.

wasn't it that 8192 cores keep locking on. one core got the clear to go do it's task. it announced a lock on ram and fp pipe. every other cores just keep waiting. oops! that core that generayed th initial lock did not know it was the one who lock the whole thing. it keep waiting for that lock to be clear before it will execyte it's task/ now we have a whole line of cores waiting for that locked bit.

thanks for waiting.

--i do not know.
0 0 [Posted by: idonotknow  | Date: 09/22/11 02:55:40 PM]

Excellent article!

Many chip analysts are forgetting to mention that Socket G34 motherboards will support these new Bulldozer Opteron 6272 processors from AMD.

For example, the dual cpu, Asus KGPE-D16 Dual Socket G34 motherboard will benefit greatly for this new 32nm architecture!
0 0 [Posted by: rgathright  | Date: 10/05/11 06:55:11 AM]


Add your Comment

Related news

Latest News

Wednesday, November 5, 2014

10:48 pm | LG’s Unique Ultra-Wide Curved 34” Display Finally Hits the Market. LG 34UC97 Available in the U.S. and the U.K.

Wednesday, October 8, 2014

12:52 pm | Lisa Su Appointed as New CEO of Advanced Micro Devices. Rory Read Steps Down, Lisa Su Becomes New CEO of AMD

Thursday, August 28, 2014

4:22 am | AMD Has No Plans to Reconsider Recommended Prices of Radeon R9 Graphics Cards. AMD Will Not Lower Recommended Prices of Radeon R9 Graphics Solutions

Wednesday, August 27, 2014

1:09 pm | Samsung Begins to Produce 2.13GHz 64GB DDR4 Memory Modules. Samsung Uses TSV DRAMs for 64GB DDR4 RDIMMs

Tuesday, August 26, 2014

10:41 am | AMD Quietly Reveals Third Iteration of GCN Architecture with Tonga GPU. AMD Unleashes Radeon R9 285 Graphics Cards, Tonga GPU, GCN 1.2 Architecture