Slashdot Mirror
Virginia Tech to Build Top 5 Supercomputer?
hype7 writes "ThinkSecret is running a story which might explain exactly why the Dual 2GHz G5 machines have been delayed to the customers that ordered them minutes after the keynote was delivered. Apparently, Virginia Tech has plans to build a G5 cluster of 1100 units. If it manages to complete the cluster before the cut-off date, it will score a Top 5 rank in the Linpack Top 500 Supercomputer List. Both Apple and the University are playing mum on the issue, but there's talk of it all over the campus."
I got the following email the other day:
Virginia Tech is in the process of building a Terascale Computing Cluster which will be housed in the Andrews Information Systems Building (AISB). For those who are interested in learning more about this project, we will host an information session on Thursday, September 4th from 11 a.m. to noon in the Donaldson Brown Hotel and Conference Center auditorium.
We look forward to seeing you there
Terry Herdman Director of Research Computing.
I'll try to remember to take notes on this and let you all know if there's anything interesting...
Friends help you move. Real friends help you move bodies.
Altivec. Certain types of vector code when compiled to only run on a G4 outperform a pentium even at 3+x the ghz range (i.e. a 800 mhz G4 beating a 3ghz PIV). Assuming similar numbers for the G5 and the increase across the board on all the non vector operations + the fact that the 970 work together so much better....
I can see it making a lot of sense. NASA and lots of bio companies use the G4s this way.
>> A box designed to be separate just will not have the latency advantage of a supercomputer designed from the ground up.
I suggest you look at the list of the top supercomputers in the world. Most are clusters, ie. separate, distinct machines (just a quick glance shows the top 25 all are). It's just too darn hard to make a shared memory computer with 1000's of processors. So the common architecture is to make a cluster of smaller shared memory machines.
Besides, most clusters built utilize special interconnects like Myrinet that offer low latency connections. They're more expensive than ethernet, but it's a supercomputer so you spend it.
>> All this "the internet is one giant distributed computer" doesn't acknowledge this.
On the contrary... people know this very well. That's why we see rendering and SETI processing as distributed. They don't really need to communicate with others often.
There are tradeoffs actually. This isn't like distributed.net or seti@home, this is a controlled network. They have complete control over the network switches, technology, and topology used and can design the network to accomodate tho problems the cluster will be designed to solve.
For example, you could use Myrinet to get 2 Gigabit, super low latency connectivity, or Quadrix, or Infiniband, or just a well laid out Gigabit Ethernet with high end switches.
In multiple processors in a box, the processors have to fight for the resources that box has to offer. NUMA alleviates demand on the memory, but IO operations (when writing to disk or to network) in a multiprocessor box block a good deal as the processor count in a node rises.
The idea with clusters is that inter-node communication in most cases can be kept low. Each system can work on a HUGE chunk of a problem on its own, with its own dedicated hard drive, memory subsystem, and without having too much competition for the network card. A lot of problems are really hard to solve computation wise, but are *very* well suited to distributed computing. A prime example of this is rendering 3D movies. Perhaps oversimplifying things, but for the most part, a central node divides up discrete parts (a segment of video), and each node works without talking to others until done, so the negative impact is minimal. Certain problems (i.e. nuclear explosion simulations where time and spacial chunks interact more with one another) are much more sensitive to latency/throughut. Seti@Home and distributed.net are *extremely* apathetic to throughput/latency issues (not much traffic and very infrequent communication).
XML is like violence. If it doesn't solve the problem, use more.
As Zack pointed out, iWalk was not a Think Secret report; in fact, we debunked it. For WWDC, we reported that Apple would announce 64-bit Power Macs as well as a videoconferencing camera that we said would be called "iSight," -- I think we're in the clear there. iWorks? I maintain that it is still a future Apple release. As for 12-inch and 17-inch PowerBooks, while we raised the possibility of a release that week, we specifically said we couldn't confirm the delivery date: "It's unclear when Apple plans to announce the upgrades..."
Bottom line? Like any other news organization, Think Secret has occasional misses. But those misses don't appear to include any of the items mentioned here. I think our record speaks for itself.
Nick dePlume
Publisher and Editor in Chief
Think Secret
1. The PPC970 draws from the Power4 lineage, which I have used for a long time. The PPC970 has 2 double precision FPUs, each capable of fused multiply add instructions leading to 4 flops/cycle/processor (2 units*2flops/cycle). This is identical to the Itanium2 FPU microarchitecture. The Opteron on the other hand can only do 2 double precision flops/cycle, which makes it only half as powerful on matrix heavy scientific computations, when compared to the PPC970 or the Itanium 2. The PPC970 should really be compared in FP terms to the Itanium2 at 1/10th of the cost, and at 2GHz it is clocked higher than the top-end 1.5GHz Itanium2 Madison. Moral of the story, read thy arstechnica. 2. The standard benchmarking process (LINPACK) only uses double precision FP. If this rumor is true, then this machine is capable of an Rpeak (LINPACK) of 17.6 Teraflop, which those of you who follow top500 will realize is quite substantial. 3. If they are really using Infiniband, this should be a nice machine. Infiniband provides 10 Gbps (20 Gbps full duplex) of bandwidth, which is much faster than either Myrinet or Quadrics. Also Infiniband latency is 10us and the benchmarking process is bandwidth not latency sensitive. On the other, this stuff is really expensive. If all of this is true, this would be a major engineering endeavor. Also, it is probably cheap. However, all in all, this could well just be a rumor (come on it is thinksecret - remember iWorks). If not, this should be a fairly substantial machine.