As a follow-up - it's one DDR3 channel - maybe 2. That puts it at about 1/30th of a K20.
People have tried to creep into Scientific Computing with processors like this (tile-based perf-per-watt SoCs). They haven't succeeded (see: Adapteva, Tilera, etc.). And they have much bigger budgets.:)
For the record, the Tesla K20x TDP numbers include the memory (it's for the entire card).
A comment below says that it uses DDR3 1333. Total bandwidth of that, being extremely generous and giving them 6 memory channels (unlikely) puts you in the neighborhood of about 1/10th the memory bandwidth of the K20.
Combine that with the "how do you connect this to other things" problem, and this chip has no chance in scientific computing.
In this case, they're operating the transistors in a sub-threshold voltage environment. A full channel never opens for the transistor, but energy will trickle through at different rates.
Instead of the typical "open/closed water pipe valve" model of the transistor, imagine having a leaky bucket, and then determining 1 vs 0 on how many drops get through.
It's a tough area to design circuits in because of the very delicate balance. It doesn't take many electrons (or much process variation) to bust up your circuit.
Burton was the co-founder of "Tera", the supercomputer company that purchased the old Cray division away from SGI in their 1999 restructuring.
Tera was founded to develop massively multithreaded machines. After their big purchase, they took the Cray name for continuity with Cray's old customers and products, along with the fact that it's a much more viable "commercial" supercomputing name.
Re:are you sure you remember seeing the Cray 3 ?
on
Cray XT-3 Ships
·
· Score: 1
wrong company - Cray 3 was being built by "Cray Computer", a different company that Seymour started after he left the original Cray Research.
He's most likely talking about the Cray 2, which even came with its own Waterfall
If I remember correctly, the Intel 300mm fabs are just beginning to produce commercial chips (I don't remember which process they run). So, I don't think we've seen the effect of their 300mm fabs yet.
From the same document the Mac proponents have been quoting from: Dondarra Doc
Table 3 - page 53:
Big Mac -> Rmax: 8164 Processors: 1936
Cray X1 -> Rmax: 2932.9 Processors: 252
Please be careful when making general statements. Thank you.
That said, yes, it has the highest per CPU performance of the machines with commodity processors. (that are listed, at least - including the year-old Xeons)
On the other side of the issue is that it places 4th in the current Top 500 list, which was released in June. We won't really know where it places on this "moving target" until the next list is released in November.
I was going to tack on something like "...which it won't." to the end of my post, but I thought it would get interpreted as an opinion as opposed to pointed out that the submitter was on crack.;)
Title is wrong - they get 80% efficiency on 128 nodes. The 14 TFlops number is if that efficiency is held through the full size of the machine (2000+ processors).
Re:Just out of curiosity....
on
Grid Processing
·
· Score: 1
The short answer is yes. The longer answer is that a large portion of the TRIPS project is devoted to compiler infrastructure, to make it manageable. From what I've seen at some of their presentations, they've made a decent amount of headway.
ok, NOW we've got some details.:) As a PhD student in hardware (microarchitecture), a PhD DOES open a lot of doors. In the hardware industry, the people doing the lower level implementation (i.e. HDL coding, verification, etc.) are BS and MS people. Most of the actual high-level designers have PhDs. They also do a lot of forward-looking studies on which direction their architectures should steer towards.
And you're right about being a professor - you won't have that option at all (especially tenure-track) if you don't have a PhD.
As far as I know, Cray's "Red Storm" system is the only one where they will be using the Opterons. (I don't think it will become one of their standard products). They have, however, signed a deal with Dell to collaborate on some cluster systems.
I think you've hit it on the head here. Google still wants Moore's law to continue. The plus side to it would be that they can get the same amount of performance per processor they have now (which is sufficient for them) for *much* less money.
Think about the "price shadow" of products - when a new product comes out, the older/slower/less sophisticated product becomes cheaper. If this happens *really quickly*, then the prices are likely to go down a lot, and very soon. If you've already got what you want, it's a great place to be in.
This doesn't happen much with industries where there aren't many advances (think electric range). A two year old stove is pretty close in price to a brand new one. Whereas, a two year old processor (and 50 cents) will get you a cup of coffee.
Intel actually purchased the Shrewsbury, MA Alpha EV8 design team (and facility). How many people left, I'm not sure, but I know several of the lead designers are still there (like Joel Emer, who's worked for 4 companies without ever changing his desk...) and that they're working on Itanium now.
Slashdot Mirror
IB connects through PCIe.
which isn't on this.
As a follow-up - it's one DDR3 channel - maybe 2. That puts it at about 1/30th of a K20.
People have tried to creep into Scientific Computing with processors like this (tile-based perf-per-watt SoCs). They haven't succeeded (see: Adapteva, Tilera, etc.). And they have much bigger budgets. :)
Thanks for saving me a lot of typing. :-)
For the record, the Tesla K20x TDP numbers include the memory (it's for the entire card).
A comment below says that it uses DDR3 1333. Total bandwidth of that, being extremely generous and giving them 6 memory channels (unlikely) puts you in the neighborhood of about 1/10th the memory bandwidth of the K20.
Combine that with the "how do you connect this to other things" problem, and this chip has no chance in scientific computing.
And interconnect them with what?
HDMI?
Man, your records don't back far enough for me, it seems. ;-)
5) ...and Cray is already installing a 20-ish PF machine at ORNL in the next year named "Titan".
"A New Concept in PC Case Mods"
In this case, they're operating the transistors in a sub-threshold voltage environment. A full channel never opens for the transistor, but energy will trickle through at different rates.
Instead of the typical "open/closed water pipe valve" model of the transistor, imagine having a leaky bucket, and then determining 1 vs 0 on how many drops get through.
It's a tough area to design circuits in because of the very delicate balance. It doesn't take many electrons (or much process variation) to bust up your circuit.
He didn't have much to work with - the press release (err..."article") was information free, too!
Burton was the co-founder of "Tera", the supercomputer company that purchased the old Cray division away from SGI in their 1999 restructuring.
Tera was founded to develop massively multithreaded machines. After their big purchase, they took the Cray name for continuity with Cray's old customers and products, along with the fact that it's a much more viable "commercial" supercomputing name.
wrong company - Cray 3 was being built by "Cray Computer", a different company that Seymour started after he left the original Cray Research. He's most likely talking about the Cray 2, which even came with its own Waterfall
Cray does have its own fully custom system still, its called the X1 - a highly scalable Vector machine (thats far from a NEC clone...)
I'd link it, but the site is down...
If I remember correctly, the Intel 300mm fabs are just beginning to produce commercial chips (I don't remember which process they run). So, I don't think we've seen the effect of their 300mm fabs yet.
From the same document the Mac proponents have been quoting from: Dondarra Doc
Table 3 - page 53:
Big Mac -> Rmax: 8164 Processors: 1936
Cray X1 -> Rmax: 2932.9 Processors: 252
Please be careful when making general statements. Thank you.
That said, yes, it has the highest per CPU performance of the machines with commodity processors. (that are listed, at least - including the year-old Xeons)
On the other side of the issue is that it places 4th in the current Top 500 list, which was released in June. We won't really know where it places on this "moving target" until the next list is released in November.
I thought the PDA was dead?!?
Jeez, make up your mind!
I was going to tack on something like "...which it won't." to the end of my post, but I thought it would get interpreted as an opinion as opposed to pointed out that the submitter was on crack.
Title is wrong - they get 80% efficiency on 128 nodes. The 14 TFlops number is if that efficiency is held through the full size of the machine (2000+ processors).
The short answer is yes. The longer answer is that a large portion of the TRIPS project is devoted to compiler infrastructure, to make it manageable. From what I've seen at some of their presentations, they've made a decent amount of headway.
ok, NOW we've got some details.
And you're right about being a professor - you won't have that option at all (especially tenure-track) if you don't have a PhD.
As far as I know, Cray's "Red Storm" system is the only one where they will be using the Opterons. (I don't think it will become one of their standard products). They have, however, signed a deal with Dell to collaborate on some cluster systems.
Only since January. This law was amended during 2002 (shown at the bottom of the page), so Engler signed it.
Think about the "price shadow" of products - when a new product comes out, the older/slower/less sophisticated product becomes cheaper. If this happens *really quickly*, then the prices are likely to go down a lot, and very soon. If you've already got what you want, it's a great place to be in.
This doesn't happen much with industries where there aren't many advances (think electric range). A two year old stove is pretty close in price to a brand new one. Whereas, a two year old processor (and 50 cents) will get you a cup of coffee.
Intel actually purchased the Shrewsbury, MA Alpha EV8 design team (and facility). How many people left, I'm not sure, but I know several of the lead designers are still there (like Joel Emer, who's worked for 4 companies without ever changing his desk...) and that they're working on Itanium now.