I copied the result the CPU benchmark of Boinc reported me...
Yes, those are my results. Are they really bad?
By the way, I'm sorry about my English, I just know it a little!
If you copied that text from the message log in boincmgr on your host, it is curious that it differs so drastically from what is currently shown on the web page. If that means you recently ran a benchmark, but have not done an update, then the web page presumably has the result of a previous benchmark run, which for some reason gave a drastically worse result. I just checked for forcing both a benchmark and an update for one of my hosts, and aside from the descriptive text the results in the message log and on the web page match.
For Einstein the floating point benchmark is the one of primary interest.
Relying one the number automatically posted to the web site, here are some comparisons, old and new:
your host 289.16 (as posted on web page)
your host 761 (as you report in this thread)
my retiring Coppermine 864.14
my five year old Banias laptop 1292.38
a brand-new Q9550 2658 per core, so 10,632 total
The 289.16 previously posted is truly dreadful, while the 761 is not especially surprising for a machine of that vintage. I think the BOINC people have acknowledged that the benchmark results are particularly unreliable on Windows 98 machines, as both of Coppermines are.
The modern machines are actually ahead of the Coppermines for Einstein (and SETI) by more than the ratio of floating point benchmarks. Fast Coppermines are pretty memory-starved, which the benchmark pretty much does not notice, there are other architectural issues, and the current SETI and Einstein aps use some specialized parallel arithmetic hardware which is probably ignored by the benchmark.
Still, if 761 is the proper floating point benchmark for your machine, then Einstein results should take you closer to 50 than to 100 hours of CPU time to complete. Possibly something in its configuration (more likely software than hardware) is especially hurting it.
Regarding the English--no apology needed, just forgive me if I ask an extra clarifying question or two. Mistakes in my former profession were staggeringly expensive, so I got in the habit of trying pretty hard to be sure I knew what people meant.
Regarding the English--no apology needed, just forgive me if I ask an extra clarifying question or two.
Thank you!!!
Well, it takes about 100 hours to complete a task.
I've just runned again the CPU Benchmark and tis is the result:
29/10/2008 20.39.56||[error] FP benchmark ran only -231.120001 sec; ignoring
29/10/2008 20.39.56||[error] CPU benchmarks error
I'd rather restart the computer...
In the section of this thread discussing Coppermine P3s crunching Einstein. Am I wrong in reading that the new client einstein_S5R4_605_etc can run at a faster rate than the 604 client on such a Coppermine?
I have 1 dual P3 @933, under Win2K Pro, on a Super-Micro server board.
Shih-Tzu are clever, cuddly, playful and rule!! Jack Russell are feisty!
In the section of this thread discussing Coppermine P3s crunching Einstein. Am I wrong in reading that the new client einstein_S5R4_605_etc can run at a faster rate than the 604 client on such a Coppermine?
I have 1 dual P3 @933, under Win2K Pro, on a Super-Micro server board.
I would expect that 6.05 is indeed significantly faster than 6.04 on Coppermines (in fact on any SSE capable CPU). Certainly worth a try.
Oh...yes....indeed...sorry!!! You are absolutely right. It may not work on SSE-but-non-SSE2 hosts, too bad. Once it passes the beta test, I'm sure Bernd will generate a "switcher" version with "plain FPU", SSE and SSE2 variants just like the Linux version.
Sorry for the wrong info.
However, it should work fine on your Banias, (Pentium-M, which is SSE2 capable).
Oh...yes....indeed...sorry!!! You are absolutely right. It may not work on SSE-but-non-SSE2 hosts, too bad. Once it passes the beta test, I'm sure Bernd will generate a "switcher" version with "plain FPU", SSE and SSE2 variants just like the Linux version.
Sorry for the wrong info.
CU
Bikeman
That is what I thought, and why I asked.
I installed the 605 client on my Core 2 rigs and the dual Prestonia Xeon (with HT). These are all SSE2 capable or better. The Coppermine is SSE only.
So, I can only use the 605 on 3 of my rigs when I have reached my target on Milkyway and then swap back to Einstein.
Shih-Tzu are clever, cuddly, playful and rule!! Jack Russell are feisty!
Oh...yes....indeed...sorry!!! You are absolutely right. It may not work on SSE-but-non-SSE2 hosts, too bad. Once it passes the beta test, I'm sure Bernd will generate a "switcher" version with "plain FPU", SSE and SSE2 variants just like the Linux version.
It might be a while before that happens. I asked the question about a new "switcher app" for Windows a few weeks ago over in the 6.05 sticky thread. Bernd's response was quite interesting, particularly the bit about the possible early termination of S5R4 and replacing it with S5R5 with new (and incompatible) apps. We are now in the middle of the decision making time frame that he mentioned so I hope to see further news shortly.
I copied the result the CPU
)
I copied the result the CPU benchmark of Boinc reported me...
Yes, those are my results. Are they really bad?
By the way, I'm sorry about my English, I just know it a little!
RE: I copied the result the
)
If you copied that text from the message log in boincmgr on your host, it is curious that it differs so drastically from what is currently shown on the web page. If that means you recently ran a benchmark, but have not done an update, then the web page presumably has the result of a previous benchmark run, which for some reason gave a drastically worse result. I just checked for forcing both a benchmark and an update for one of my hosts, and aside from the descriptive text the results in the message log and on the web page match.
For Einstein the floating point benchmark is the one of primary interest.
Relying one the number automatically posted to the web site, here are some comparisons, old and new:
your host 289.16 (as posted on web page)
your host 761 (as you report in this thread)
my retiring Coppermine 864.14
my five year old Banias laptop 1292.38
a brand-new Q9550 2658 per core, so 10,632 total
The 289.16 previously posted is truly dreadful, while the 761 is not especially surprising for a machine of that vintage. I think the BOINC people have acknowledged that the benchmark results are particularly unreliable on Windows 98 machines, as both of Coppermines are.
The modern machines are actually ahead of the Coppermines for Einstein (and SETI) by more than the ratio of floating point benchmarks. Fast Coppermines are pretty memory-starved, which the benchmark pretty much does not notice, there are other architectural issues, and the current SETI and Einstein aps use some specialized parallel arithmetic hardware which is probably ignored by the benchmark.
Still, if 761 is the proper floating point benchmark for your machine, then Einstein results should take you closer to 50 than to 100 hours of CPU time to complete. Possibly something in its configuration (more likely software than hardware) is especially hurting it.
Regarding the English--no apology needed, just forgive me if I ask an extra clarifying question or two. Mistakes in my former profession were staggeringly expensive, so I got in the habit of trying pretty hard to be sure I knew what people meant.
RE: Regarding the
)
Thank you!!!
Well, it takes about 100 hours to complete a task.
I've just runned again the CPU Benchmark and tis is the result:
29/10/2008 20.39.56||[error] FP benchmark ran only -231.120001 sec; ignoring
29/10/2008 20.39.56||[error] CPU benchmarks error
I'd rather restart the computer...
In the section of this thread
)
In the section of this thread discussing Coppermine P3s crunching Einstein. Am I wrong in reading that the new client einstein_S5R4_605_etc can run at a faster rate than the 604 client on such a Coppermine?
I have 1 dual P3 @933, under Win2K Pro, on a Super-Micro server board.
Shih-Tzu are clever, cuddly, playful and rule!! Jack Russell are feisty!
RE: In the section of this
)
I would expect that 6.05 is indeed significantly faster than 6.04 on Coppermines (in fact on any SSE capable CPU). Certainly worth a try.
CU
Bikeman
RE: I would expect that
)
Are you sure it's safe to use on SSE cpus rather than SSE2? Bernd particularly described it as an SSE2 power app??
Cheers,
Gary.
RE: RE: I would expect
)
My concern echoes yours. I did not install it on either my Coppermine or my (considerably more modern) Banias.
Oh...yes....indeed...sorry!!!
)
Oh...yes....indeed...sorry!!! You are absolutely right. It may not work on SSE-but-non-SSE2 hosts, too bad. Once it passes the beta test, I'm sure Bernd will generate a "switcher" version with "plain FPU", SSE and SSE2 variants just like the Linux version.
Sorry for the wrong info.
However, it should work fine on your Banias, (Pentium-M, which is SSE2 capable).
CU
Bikeman
RE: Oh...yes....indeed...so
)
That is what I thought, and why I asked.
I installed the 605 client on my Core 2 rigs and the dual Prestonia Xeon (with HT). These are all SSE2 capable or better. The Coppermine is SSE only.
So, I can only use the 605 on 3 of my rigs when I have reached my target on Milkyway and then swap back to Einstein.
Shih-Tzu are clever, cuddly, playful and rule!! Jack Russell are feisty!
RE: Oh...yes....indeed...so
)
It might be a while before that happens. I asked the question about a new "switcher app" for Windows a few weeks ago over in the 6.05 sticky thread. Bernd's response was quite interesting, particularly the bit about the possible early termination of S5R4 and replacing it with S5R5 with new (and incompatible) apps. We are now in the middle of the decision making time frame that he mentioned so I hope to see further news shortly.
Cheers,
Gary.