They chose the G5 processors because it was the only affordable 64 bit processor they could buy before their money vanished. This month, there's an equally good chance it would be an Opteron system.
Note: I am completely processor/OS agnostic. If I could get the best performance for my cluster by by running Windows 3.1 on a Z80, that suits me. "Religion" doesn't sway me, only real-world application results.
I have inside contacts on the VT cluster. The money for it came from state funds, and was going to vanish within a few weeks. When the deal with Dell for Itaniums fell through, they didn't have time to wait for Opteron systems to come out because the money would be gone before they came out, so VT went with G5's because they were the only other 64 bit system available at that time. If they had another month, there's just as good a chance that they would have gone for Opteron.
According to Pricewatch: 3 GHz Xeon $473 Opteron 242 $297
So your Xeon price wasn't quite accurate. Admittedly, you lowballed the Itanium. I couldn't find it anywhere for that low.:) But Intel is coming out with a "low-price" Itanium next year.
The G5 is a cool processor, but it isn't the reason the VT cluster is so fast, the Infiniband interconnect is. The LINPACK benchmark that is used to determine position on the Top 500 list depends very strongly on the latency of the network connection.
Infiniband has ~ 8-12 us latency (probably even less by now), while ethernet is an order of magnitude slower. In real-life applications it's actually worse than this suggests.
We have tested a real-life application (socorro) using both gigabit ethernet and Myrinet (slightly slower than Infiniband), and gigE took 600 seconds to finish a run, while Myrinet took 4.
VT's cluster is using the largest Infiniband network yet built (or at least announced). The previous largest Infiniband network was O(100) machines. VT could have built the cluster using Xeons, Itaniums, or Opterons and arrived at roughly the same level of performance.
I am the senior sysadmin for a large (> 200 nodes today, ~1200 in 3 years) academic compute cluster. We are currently evaluating new operating systems for the cluster, and we've looked at RedHat Advanced Server 3. I am curious about your plans for this market. It seems to be the best choice at the moment, but the current pricing structure makes it ludicrously expensive.
What I would like to see is a low-cost cluster version, that basically comes with a CD and access to the RPM updates folder, for $20/node. That would be a major sell around here.
Slashdot Mirror
Um, no...
They chose the G5 processors because it was the only affordable 64 bit processor they could buy before their money vanished. This month, there's an equally good chance it would be an Opteron system.
Ask them.
Note: I am completely processor/OS agnostic. If I could get the best performance for my cluster by by running Windows 3.1 on a Z80, that suits me. "Religion" doesn't sway me, only real-world application results.
:) But Intel is coming out with a "low-price" Itanium next year.
I have inside contacts on the VT cluster. The money for it came from state funds, and was going to vanish within a few weeks. When the deal with Dell for Itaniums fell through, they didn't have time to wait for Opteron systems to come out because the money would be gone before they came out, so VT went with G5's because they were the only other 64 bit system available at that time. If they had another month, there's just as good a chance that they would have gone for Opteron.
According to Pricewatch:
3 GHz Xeon $473
Opteron 242 $297
So your Xeon price wasn't quite accurate. Admittedly, you lowballed the Itanium. I couldn't find it anywhere for that low.
The G5 is a cool processor, but it isn't the reason the VT cluster is so fast, the Infiniband interconnect is. The LINPACK benchmark that is used to determine position on the Top 500 list depends very strongly on the latency of the network connection.
Infiniband has ~ 8-12 us latency (probably even less by now), while ethernet is an order of magnitude slower. In real-life applications it's actually worse than this suggests.
We have tested a real-life application (socorro) using both gigabit ethernet and Myrinet (slightly slower than Infiniband), and gigE took 600 seconds to finish a run, while Myrinet took 4.
VT's cluster is using the largest Infiniband network yet built (or at least announced). The previous largest Infiniband network was O(100) machines. VT could have built the cluster using Xeons, Itaniums, or Opterons and arrived at roughly the same level of performance.
I am the senior sysadmin for a large (> 200 nodes today, ~1200 in 3 years) academic compute cluster. We are currently evaluating new operating systems for the cluster, and we've looked at RedHat Advanced Server 3. I am curious about your plans for this market. It seems to be the best choice at the moment, but the current pricing structure makes it ludicrously expensive.
What I would like to see is a low-cost cluster version, that basically comes with a CD and access to the RPM updates folder, for $20/node. That would be a major sell around here.
See you at Supercomputing 2003...