Yes, SETI issues WUs with a quite broad range of estimated work content and deadlines.
Just in the last couple of days my main host has been issued SETI results with deadlines less than ten days in the future, a little over five weeks in the future, about six weeks out, about seven weeks out, and over ten weeks out.
When they made the multi-beam transition a couple of months ago their deadlines pushed out quite a bit. At the short end this was a relief, as their extremely small units pushed people's machines into deadline thrashes, but at the long end it does seem quite a lot more generous than needed.
Somewhat OT: Does SETI still like its CPU cache so very much? I'd rejoin, but if it still needs CPUs with large cache, it would be a waste of crunching time on my boxes (an AMD and a T2060).
Somewhat OT: Does SETI still like its CPU cache so very much? I'd rejoin, but if it still needs CPUs with large cache, it would be a waste of crunching time on my boxes (an AMD and a T2060).
I am using an optimized application for MMX, with no graphics, and have no problem with my 512K L2 cache of the PII. Other CPUs use apps optimized for SSE, SSSE. SSE2 and so forth.
Tullio
Oh right... the optimized apps... never really familiarized myself with that. Is it really that much of a difference?
That varies with the course of time. Before multibeam started the best optimized aps had a really, really large advantage. Many of the coding advantages of that generation got included in the production ap for multi-beam. Meanwhile the optimizing folks have found some new advantages. The advantage at the moment for many mainstream machines is appreciable, but varies quite a bit with the computation content of a particular result, which varies quite a lot in character there compared to here at Einstein. As a consequence, the proportion of advantage is much more for some Angle Ranges than others.
I think the automatic choice package sometimes offered from Simon's site is not current, so you have to figure out which package fits each CPU you want to upgrade (mostly a matter of which level of instruction set extensions it supports).
The sticky thread on new optimized aps is a good place to start looking into this over there.
The x440 IBM was delivered yesterday morning. Has eight 1.9Ghz MP Xeons with HT. Short 2GB of memory for the second processor module so only have one running at present. Another 2GB of memory on the way for the second cpu module. Swapped the existing memory to the second module to insure it worked, and it does. When the memory arrives we'll see if Win2k Advanced Server can run 8 cpu's with HT turned on. That should increase the number of WU done per day by quite a bit. :)
It's really loud on start up for a few seconds, after that it quiets down. Under normal conditions my Z Pro is louder than the 440, the 360's, or the 255. The 440 uses 4 150mm fans, 2 are about 1.5 inches thick and the other 2 that cool the cpu's are 3 inches thick.
One thing I really like about these rack mounts is how easy they are to work on. Everything is modular in design and a lot of the stuff is hot swap, or at least you can shut that section down and replace whatever needs replacing without shutting down the complete system.
The x440 IBM was delivered yesterday morning. Has eight 1.9Ghz MP Xeons with HT. Short 2GB of memory for the second processor module so only have one running at present. Another 2GB of memory on the way for the second cpu module. Swapped the existing memory to the second module to insure it worked, and it does. When the memory arrives we'll see if Win2k Advanced Server can run 8 cpu's with HT turned on. That should increase the number of WU done per day by quite a bit. :)
Windows 2000 Advanced Server
* up to 8 CPUs (license-upgrade to 32 CPUs possible)
* Pentium-compatible CPU with 133 MHz or more
* min. 64 MB RAM (max. 8 GB)
* min. 2 GB Harddisk with at least 1 GB free
* CD-ROM drive
* VGA-card or better
but W2K counts HT CPUs, I think you will have to run your box HT disabled...
The installation instructions for W2K on a x440 seem to require you to disable Hyper-Threading first and install a special Hardware Abstraction Layer Driver anyway, after that you should be able to re-enable HT. Unfortunately the download link for the HAL driver was moved, I hope this one is included with the box or can be found somewhere else.
RE: Does SETI have shorter
)
Yes, SETI issues WUs with a quite broad range of estimated work content and deadlines.
Just in the last couple of days my main host has been issued SETI results with deadlines less than ten days in the future, a little over five weeks in the future, about six weeks out, about seven weeks out, and over ten weeks out.
When they made the multi-beam transition a couple of months ago their deadlines pushed out quite a bit. At the short end this was a relief, as their extremely small units pushed people's machines into deadline thrashes, but at the long end it does seem quite a lot more generous than needed.
Somewhat OT: Does SETI still
)
Somewhat OT: Does SETI still like its CPU cache so very much? I'd rejoin, but if it still needs CPUs with large cache, it would be a waste of crunching time on my boxes (an AMD and a T2060).
RE: Somewhat OT: Does SETI
)
I am using an optimized application for MMX, with no graphics, and have no problem with my 512K L2 cache of the PII. Other CPUs use apps optimized for SSE, SSSE. SSE2 and so forth.
Tullio
Oh right... the optimized
)
Oh right... the optimized apps... never really familiarized myself with that. Is it really that much of a difference?
RE: Oh right... the
)
That varies with the course of time. Before multibeam started the best optimized aps had a really, really large advantage. Many of the coding advantages of that generation got included in the production ap for multi-beam. Meanwhile the optimizing folks have found some new advantages. The advantage at the moment for many mainstream machines is appreciable, but varies quite a bit with the computation content of a particular result, which varies quite a lot in character there compared to here at Einstein. As a consequence, the proportion of advantage is much more for some Angle Ranges than others.
I think the automatic choice package sometimes offered from Simon's site is not current, so you have to figure out which package fits each CPU you want to upgrade (mostly a matter of which level of instruction set extensions it supports).
The sticky thread on new optimized aps is a good place to start looking into this over there.
The x440 IBM was delivered
)
The x440 IBM was delivered yesterday morning. Has eight 1.9Ghz MP Xeons with HT. Short 2GB of memory for the second processor module so only have one running at present. Another 2GB of memory on the way for the second cpu module. Swapped the existing memory to the second module to insure it worked, and it does. When the memory arrives we'll see if Win2k Advanced Server can run 8 cpu's with HT turned on. That should increase the number of WU done per day by quite a bit. :)
Impressive! That box must be
)
Impressive! That box must be quite noisy as well?
CU
Bikeman
It's really loud on start up
)
It's really loud on start up for a few seconds, after that it quiets down. Under normal conditions my Z Pro is louder than the 440, the 360's, or the 255. The 440 uses 4 150mm fans, 2 are about 1.5 inches thick and the other 2 that cool the cpu's are 3 inches thick.
One thing I really like about these rack mounts is how easy they are to work on. Everything is modular in design and a lot of the stuff is hot swap, or at least you can shut that section down and replace whatever needs replacing without shutting down the complete system.
RE: The x440 IBM was
)
Windows 2000 Advanced Server
* up to 8 CPUs (license-upgrade to 32 CPUs possible)
* Pentium-compatible CPU with 133 MHz or more
* min. 64 MB RAM (max. 8 GB)
* min. 2 GB Harddisk with at least 1 GB free
* CD-ROM drive
* VGA-card or better
but W2K counts HT CPUs, I think you will have to run your box HT disabled...
Udo
Hi! The installation
)
Hi!
The installation instructions for W2K on a x440 seem to require you to disable Hyper-Threading first and install a special Hardware Abstraction Layer Driver anyway, after that you should be able to re-enable HT. Unfortunately the download link for the HAL driver was moved, I hope this one is included with the box or can be found somewhere else.
CU
H-B