backtop


Print 79 comment(s) - last by PrezWeezy.. on Apr 3 at 1:57 PM


"Torrenza" systems will accept both multi-core, accelerators, or "Fusion" processors
AMD sheds more light on accelerated processing projects

This week at CeBIT 2007, AMD revealed more details about its "accelerated computing" platform, codenamed Torrenza. AMD's goal behind Torrenza is to create a platform where application-specific processors can interact cost effectively and offer better performance than a general purpose CPU, while remaining compatible with off the shelf platforms.

AMD guidance revealed this week that future processors will also have integrated "accelerators" embedded into them. A Torrenza system will have at least two sockets, and both will accept accelerators and accelerated CPUs.

One accelerated-processor project on AMD plate, slated for 2008 under the codename Fusion, and combines a dedicated GPU or GPU accelerator onto the same package or even the same silicon die as the main CPU. AMD has already set the ground-work for Fusion processing with its Stream Computing initiative -- utilizing ATI-based graphics adaptors for heavy number crunching. 

Other Torrenza ready projects are also coming to light.  Clearspeed announced its CSX600 math-coprocessor plug-in last year, with the stated intention of creating a socket plugin version for Torrenza.  Los Alamos National Labs is currently building the world's fastest supercomputer, Roadrunner, with Opteron and Cell processors on the Torrenza platform.

Torrenza is not just locked within the compounds of the CPU sockets. According to AMD, Torrenza systems will accept accelerators in a PCI-Express interface too, allow for multiple application specific accelerators to access system memory and processor functions directly.  Mercury systems announced a PCIe plug-in accelerator late last year.

While Torrenza is well on its way to seeing daylight, Intel is also working on its own open architecture platform. Notorious for keeping its CPU platform a closely guarded technology, Intel indicated that it was working on a competitive technology to AMD's HyperTransport, dubbed CSI, allowing direct CPU and memory access.

Intel guidance suggests the company will announce its Torrenza competitor sometime in mid-2008.


Comments     Threshold


This article is over a month old, voting and posting comments is disabled

By SexyK on 3/18/2007 8:38:31 AM , Rating: 3
lol, way to list off all the "old" technology Intel is "still" using. Too bad Intel's old, outdated platform is the highest performing available and in many cases leaves AMD's "new" and "superior" platform in the dust.


By zsdersw on 3/18/2007 8:45:01 AM , Rating: 2
Targon is also the same guy/girl who poo-poo'd Conroe when it first came out because it is basically a derivative of the P6.

He/she never mentions, though, that AMD's new chips are basically derivatives of older chips as well.


By Darkskypoet on 3/18/2007 4:05:00 PM , Rating: 2
Actually... Where platform matters, in higher socket count servers, AMD does still hold an advantage. The C2D superiority makes this difference completely unfelt in Single and dual socket servers. However, the higher end, 4, 8 socket, etc. is still dominated by AMD for performance. As well, it opens up for more possibilities for the near term future. The ability to simply drop semi autonomous processing units of varying design and manufacture compliant with an HT iteration makes the superiority of the platform for the future, while presently, it allows a much more linear progression of performance for higher socket / core count server implementations.

Its very easy to point to a situation where platform matters little (in the areas where AMD is superior) and say platform doesn't matter. It is much more useful to look at where the superiority is important, and compare implementations.

Even back with the Athlon MP, in many cases, it was better then the Intel conterpart because of the point to point nature of its connectivity. This, when leveraged properly, could make the inferior Athlon MP seem far better then its faster Intel competition on a tradition style FSB.

No doubt C2D > K8. However, this means that to continue the efficient feeding of such powerful CPUs, one must ensure higher bandwidth per socket in an Intel solution vs the AMD solution. This is not the case in Higher socket count implementations. As such, we castrate the higher IPC C2D by feeding it thru a smaller, higher latency tube. Whereas NUMA and HT continue to feed K8 at a higher rate as Socket count increases.

Thus, as I stated earlier, a much more linear performance curve is attained by AMD when adding sockets, then what Intel can manage with a similar socket count. This will only become more acute as each Intel proc's performance increases.

You are however correct when comparing a low number of sockets, as the limitations of Intel FSB tech is able to support full feeding in Single and some Dual Socket configs.


By zsdersw on 3/18/2007 4:32:34 PM , Rating: 2
4S+ servers are a relatively small market in terms of numbers. Dollar-wise it's big.. but as a share of the entire server market, it's pretty small.


By Darkskypoet on 3/18/2007 7:35:45 PM , Rating: 2
True... Numbers are small. But as you pointed out, profits are high. I feel more importantly for AMD is the legitimacy factor of leading a high end server market. This was unthinkable not too long ago.

Acceptance in Corporate / Scientific world is very important. Leading the 4S, 4S+ market gives AMD that very important legitimacy with the right people / firms. As did leading the workstation, desktop, and pretty much all aspects of the server market; during the Netburst era, got them where they are today.

As remarked upon further down, this increase by AMD has come at the expense of Intel. Intel, as of yet, has not responded to this 'platform' supremacy, and as the numbers show are continuing to lose ground in this Important mind share / profit sector.


By zsdersw on 3/18/2007 7:49:12 PM , Rating: 2
If Intel's solution (CSI) proves to be better or at least equal to that of AMD, there's no reason to expect any more losing ground to take place. At that point, the economies of scale that Intel's manufacturing advantage can provide will tip the scale in their favor.


By Darkskypoet on 3/19/2007 4:23:46 AM , Rating: 1
Economies of scale advantage Intel, only exists if Intel (itself plus partners) can best AMD and all consortium partners in the mfg capability challenge. As Intel is larger then AMD, it requires less partners then AMD to perhaps have some sort of mfg advantage. However, as AMD has many others behind them, Intel does not automatically get any "relative" economies of scale benefits, If everyone but Intel looks to be using, or licensing HT, advantage AMD.

This is why being first to the party with a new technology is best, you can choose to share it with many, and cement it as a standard .

But finally you are getting to the heart of the issue. Since AMD has the likes of many of the other semiconductor players behind them as part of the HT Consortium. Its not just Intel vs. AMD anymore. AMD has many friends, and they are all developing for HT. Not CSI.

Perhaps finally you'll get that Intel can have CSI all it wants, whenever it finally comes out, and it still doesn't instantly deprive AMD and partners of the years they have ahead of Intel in this field.

Again, to clarify for you; Not never catch up , Not no hope for Intel . Even when CSI debuts; Advantage AMD for having had a spec and mature platform to develop for, and many firms with shipping, installable products for, and implementations of, Hyper Transport .


You know zsdersw, you are starting to seem like an Intel Fanatic. Sigh.


By zsdersw on 3/19/2007 5:52:44 AM , Rating: 2
I'd definitely like to know how you know that Intel has no one developing for CSI who's also developing for HT.

Intel has partners developing for CSI. We haven't been told who, yet, because Intel has not talked at length about CSI. Do you honestly think Intel doesn't have its act together on this; have all the pieces in place to make CSI a success?


By Viditor on 3/19/2007 9:40:08 AM , Rating: 2
quote:
Intel has partners developing for CSI. We haven't been told who, yet, because Intel has not talked at length about CSI

The problem is that they need EVERYONE developing for CSI, not just a few partners...they should have learned this from the Itanium fiasco.


By zsdersw on 3/19/2007 10:15:00 AM , Rating: 2
How do you know that "everyone" isn't?


By zsdersw on 3/19/2007 5:59:43 AM , Rating: 3
If responding to notoriously one-sided pro-AMD posts is Intel fanaticism to you, then by all means think whatever you want. It doesn't matter to me.


By zsdersw on 3/19/2007 2:42:58 PM , Rating: 2
quote:
if you really think he is doing biased comments, plese explain and write something more


I have.. perhaps *you* should read the entire thread.


By PrezWeezy on 4/3/2007 1:57:08 PM , Rating: 2
You do realize that the BlueGrape3 I believe it is, which was reported on here if I'm not mistaken, is supposed to be 3 times faster than the RoadRunner...and it's built on Intel 7000 and 5000 sequence processors. Hundreds of them. FSB has advantages, and I realize you didn't post it, but the IMC is simply AMD's solution. Intel will be comming out with Nahlem soon, one version will have an IMC, the other wont. At that point we can get some actual benchmarks on if it truely makes a difference or if the platform is the difference.


"So, I think the same thing of the music industry. They can't say that they're losing money, you know what I'm saying. They just probably don't have the same surplus that they had." -- Wu-Tang Clan founder RZA

Related Articles
IBM's Next-Generation Supercomputer
February 28, 2007, 1:52 AM
AMD Reveals More "Fusion" Details
November 17, 2006, 10:01 AM
Intel Responds to AMD's "Torrenza"
September 27, 2006, 9:30 PM













botimage
Copyright 2014 DailyTech LLC. - RSS Feed | Advertise | About Us | Ethics | FAQ | Terms, Conditions & Privacy Information | Kristopher Kubicki