If it does, how fast will the machine crunch workunits compared to an identical one using Windows XP?
I presently have WIN2k Pro running an Athlon-64 dual core, so that will run. In the distant past I have run a dual cpu Intel Pentium Pro MP on WIN2k Pro.
I have run a Core 2 quad on WIN-XP. I tried the quad with WIN2k but could not get it to run, but I cannot say it is impossible.
[FWIW, I have also run 1-core, 2-core & 4-core Intel cpus on Linux SuSE 10.1.]
As far as I can determine, the crunching speed is not discernably different on WIN-XP vs WIN2k. In fact, there is a thread here about Linux which says that there is no longer even any discernable difference in crunching speed between WIN vs Linux because all the important loops in E@H are now coded in assembly & every machine instruction in those loops is now the same on both platforms.
That speed claim is consistent with my own experiments.
It is still possible that other programs you run on your computer might be slower on one OS than another, & might then encroach more into time that would otherwise be spent on E@H, but I cannot say anything about that comparison.
In fact, there is a thread here about Linux which says that there is no longer even any discernable difference in crunching speed between WIN vs Linux because all the important loops in E@H are now coded in assembly & every machine instruction in those loops is now the same on both platforms.
Old thread. The S5r1 app was hand coded assembly in the hot loops. s5r2 is a new app and asm tweaking is being deferred until the last handful of rare bugs are found. While reasonably close realtive windows/*nix performanace has shifted by decent amounts between app versions.
While reasonably close realtive windows/*nix performanace has shifted by decent amounts between app versions.
On the other hand, as to the original poster's question, comparing XP and 2000, I don't think the application itself relies enough on system calls for the difference between those two OSs to matter that way. As to irreducible system overhead, probably there is a difference, but on my XP quad, non-BOINC stuff routinely is quite far under 1% when I'm not actually doing anything, and I'm a non-fanatic with a wireless network connection, a copy of BoincView, and other stuff allowed to run. My biggest loss is to XXcopy when my backup to a network drive on another system runs, and I don't regard that as optional, given the non-BOINC uses of this system.
I'd be pretty surprised if there would be an appreciable difference systematic to the OS choice compared to the impact of user setup choices altering overhead.
What is the best operating system (?way) to get the maximum BOINCing from a DL-580 4 X Piii 700MHz 2GB?
Robert
Hi Robert,
Please don't double-post, it will be confusing if people post answers to the same question in many different threads. I think the perfect place to discuss your question is this thread where an 8 (!!) Xeon PIII Proliant box under Linux was discussed. As you can see there, that PC performs rather impressively under Linux.
Does Windows 2000 support dual (or even quad) core CPUs?
)
AFAIK profesional supports 2 CPU's, Server 4 CPU's and Advanced Server 8 CPU's.
Advanced Server also provides the option to set up a cluster.
How much faster (or not) it will let you crunch??
I wouldn't know...
;-)
RE: See title. If it
)
I presently have WIN2k Pro running an Athlon-64 dual core, so that will run. In the distant past I have run a dual cpu Intel Pentium Pro MP on WIN2k Pro.
I have run a Core 2 quad on WIN-XP. I tried the quad with WIN2k but could not get it to run, but I cannot say it is impossible.
[FWIW, I have also run 1-core, 2-core & 4-core Intel cpus on Linux SuSE 10.1.]
As far as I can determine, the crunching speed is not discernably different on WIN-XP vs WIN2k. In fact, there is a thread here about Linux which says that there is no longer even any discernable difference in crunching speed between WIN vs Linux because all the important loops in E@H are now coded in assembly & every machine instruction in those loops is now the same on both platforms.
That speed claim is consistent with my own experiments.
It is still possible that other programs you run on your computer might be slower on one OS than another, & might then encroach more into time that would otherwise be spent on E@H, but I cannot say anything about that comparison.
ADDMP
RE: In fact, there is a
)
Old thread. The S5r1 app was hand coded assembly in the hot loops. s5r2 is a new app and asm tweaking is being deferred until the last handful of rare bugs are found. While reasonably close realtive windows/*nix performanace has shifted by decent amounts between app versions.
RE: While reasonably
)
On the other hand, as to the original poster's question, comparing XP and 2000, I don't think the application itself relies enough on system calls for the difference between those two OSs to matter that way. As to irreducible system overhead, probably there is a difference, but on my XP quad, non-BOINC stuff routinely is quite far under 1% when I'm not actually doing anything, and I'm a non-fanatic with a wireless network connection, a copy of BoincView, and other stuff allowed to run. My biggest loss is to XXcopy when my backup to a network drive on another system runs, and I don't regard that as optional, given the non-BOINC uses of this system.
I'd be pretty surprised if there would be an appreciable difference systematic to the OS choice compared to the impact of user setup choices altering overhead.
What is the best operating
)
What is the best operating system (?way) to get the maximum BOINCing from a DL-580 4 X Piii 700MHz 2GB?
Robert
RE: What is the best
)
Hi Robert,
Please don't double-post, it will be confusing if people post answers to the same question in many different threads. I think the perfect place to discuss your question is this thread where an 8 (!!) Xeon PIII Proliant box under Linux was discussed. As you can see there, that PC performs rather impressively under Linux.
CU
BRM