// DBOINCP-300: added node comment count condition in order to get Preview working ?>
pschoefer
Joined: 11 Feb 05
Posts: 12
Credit: 1266595385
RAC: 220
20 Jan 2007 15:45:44 UTC
Topic 192340
(moderation:
)
See this workunit.
My notebook with it's 1.5Ghz Intel Celeron processor finished it in about 7 hours, the other host with 3.0Ghz Intel Pentium IV needed 7.8 hours! How can that be?
Better architecture perhaps? I think Intel mobiles are more suited to this app than other Intels, at least that is the impression I have running my laptop, which actually does way better here than over at SETI (SETI being an Intel project whereas it is usually better to run Einstein on AMDs).
Plus, your CPU (presumably a Banias core like mine) is more efficient per clock anyway. Not all CPUs which run at the same clock speed are really the same in what they get done. The Netburst architecture (P4) runs at very high clock speeds but is no faster than other modern CPUs. My Athlon is called a 3500+ because it is about as fast as a 3.5 Ghz P4 would be (roughly) but the clock speed is really 2.2 Ghz (I didn't overclock).
But looking at that certain host, it really has low benchmarks for the CPU. My 1.3 Ghz Celeron M could do way better if it wasn't running Linux. Maybe sth is wrong with that host, like the CPU getting to hot and thermal throttling kicking in...
See this workunit.
My notebook with it's 1.5Ghz Intel Celeron processor finished it in about 7 hours, the other host with 3.0Ghz Intel Pentium IV needed 7.8 hours! How can that be?
3.0 GH Pentium IV CPUs are Hyperthreaded!
The P4 will finish 2 WUs in its 7.8h if the CPU is running hyperthreading enabled.
3.0 GH Pentium IV CPUs are Hyperthreaded!
The P4 will finish 2 WUs in its 7.8h if the CPU is running hyperthreading enabled.
In which case it should show up 2 CPUs, but this one says it has only one CPU.
Guess why Intel trashed the P4 arcitecture.
No, if you request only to use 1 processor, it shows as 1 processor. The person may want to only use 1 thread, and have the other thread open for other processing they do.
I think this example is exhibiting something special about the host in question.
My Gallatin 3.2 GHz P4 running a pair of the new Einstein WU's in hyperthreading is taking about 18,000 reported CPU seconds each.
The P4 paired with the original poster's unit reported times ranging from 3250 to 7256 CPU seconds for small units from the previous run processed the same day. That is a far wider range than I see for mine--even though my times get moved around when they get paired with an application that pushes their reported time up or down. Possibly that P4 spends some of its time running applications with a really large reporting skew impact.
By the way "Celeron" is just a brand name, not a model name, and over time that brand name conveys next to nothing about the CPUs sold under it save that they were meant to sell at a lower price than contemporaneous top-line ones. Some Celerons have been real dogs (the first one, for example), while others have been downright splendid (most of the P55's sold uner the Celeron brand were wonderful--not just a good buy, but in many months the most sensible desktop Intel CPU to buy for most purposes).
By the way, I'm not a fan of the Willamette or any of its descendents, most particularly not Prescott. Just trying to get the interpretation here to fit the data better.
Just for the record, I used core names here ;-) Well, back on topic, that really is a huge variance in finishing times. Most my WUs ever reach is a few hundred seconds, with my laptop taking between 3900 and 4400 secs for a short WU. And we're really talking a solid two-year-old workhorse here, nothing especially powerful or flashy. By comparison this really gives the impression that sth is wrong with the host in question. I stick to my earlier guess and say it's occasionally getting too hot ;-)
I have a pentium D 805, which is basically 2 prescotts slapped together and i finish up a 55 credit WU in 13 000 seconds for each core.
Its rated at 2.66Ghz and i OC'ed it at 3.0Ghz
Maybe the PIV that is really slow has a thermal throttling like my Pentium D has. I know i have a really hard time cooling down my Pentium D.
My guess is, that the user is working on the computer, while E@H runs in the background (according to those other results). I have experienced the same change of computing speed, when I play games or watch movies, while E@H is running (with the lowest priority).
- you'll only get hyperthreading if your sure it's enabled it on the motherboard ( via the BIOS menu during startup ), and I'm pretty sure it's not a default.
Cheers, Mike.
I have made this letter longer than usual because I lack the time to make it shorter ...
... and my other CPU is a Ryzen 5950X :-) Blaise Pascal
funny: Celeron 1.5GHz beats PentiumIV 3GHz!
)
Better architecture perhaps? I think Intel mobiles are more suited to this app than other Intels, at least that is the impression I have running my laptop, which actually does way better here than over at SETI (SETI being an Intel project whereas it is usually better to run Einstein on AMDs).
Plus, your CPU (presumably a Banias core like mine) is more efficient per clock anyway. Not all CPUs which run at the same clock speed are really the same in what they get done. The Netburst architecture (P4) runs at very high clock speeds but is no faster than other modern CPUs. My Athlon is called a 3500+ because it is about as fast as a 3.5 Ghz P4 would be (roughly) but the clock speed is really 2.2 Ghz (I didn't overclock).
But looking at that certain host, it really has low benchmarks for the CPU. My 1.3 Ghz Celeron M could do way better if it wasn't running Linux. Maybe sth is wrong with that host, like the CPU getting to hot and thermal throttling kicking in...
RE: See this workunit. My
)
3.0 GH Pentium IV CPUs are Hyperthreaded!
The P4 will finish 2 WUs in its 7.8h if the CPU is running hyperthreading enabled.
Udo
RE: 3.0 GH Pentium IV CPUs
)
In which case it should show up 2 CPUs, but this one says it has only one CPU.
Guess why Intel trashed the P4 arcitecture.
RE: RE: 3.0 GH Pentium
)
No, if you request only to use 1 processor, it shows as 1 processor. The person may want to only use 1 thread, and have the other thread open for other processing they do.
RE: RE: 3.0 GH Pentium
)
...you are right... I missed that..
but my 2.0 GHz Pentium IV only needs 21200 sec. for a WU with 53.8 credits...
Udo
RE: Guess why Intel
)
I think this example is exhibiting something special about the host in question.
My Gallatin 3.2 GHz P4 running a pair of the new Einstein WU's in hyperthreading is taking about 18,000 reported CPU seconds each.
The P4 paired with the original poster's unit reported times ranging from 3250 to 7256 CPU seconds for small units from the previous run processed the same day. That is a far wider range than I see for mine--even though my times get moved around when they get paired with an application that pushes their reported time up or down. Possibly that P4 spends some of its time running applications with a really large reporting skew impact.
By the way "Celeron" is just a brand name, not a model name, and over time that brand name conveys next to nothing about the CPUs sold under it save that they were meant to sell at a lower price than contemporaneous top-line ones. Some Celerons have been real dogs (the first one, for example), while others have been downright splendid (most of the P55's sold uner the Celeron brand were wonderful--not just a good buy, but in many months the most sensible desktop Intel CPU to buy for most purposes).
By the way, I'm not a fan of the Willamette or any of its descendents, most particularly not Prescott. Just trying to get the interpretation here to fit the data better.
Just for the record, I used
)
Just for the record, I used core names here ;-) Well, back on topic, that really is a huge variance in finishing times. Most my WUs ever reach is a few hundred seconds, with my laptop taking between 3900 and 4400 secs for a short WU. And we're really talking a solid two-year-old workhorse here, nothing especially powerful or flashy. By comparison this really gives the impression that sth is wrong with the host in question. I stick to my earlier guess and say it's occasionally getting too hot ;-)
I have a pentium D 805, which
)
I have a pentium D 805, which is basically 2 prescotts slapped together and i finish up a 55 credit WU in 13 000 seconds for each core.
Its rated at 2.66Ghz and i OC'ed it at 3.0Ghz
Maybe the PIV that is really slow has a thermal throttling like my Pentium D has. I know i have a really hard time cooling down my Pentium D.
My guess is, that the user is
)
My guess is, that the user is working on the computer, while E@H runs in the background (according to those other results). I have experienced the same change of computing speed, when I play games or watch movies, while E@H is running (with the lowest priority).
I'll throw in a really dumb
)
I'll throw in a really dumb gotcha:
- you'll only get hyperthreading if your sure it's enabled it on the motherboard ( via the BIOS menu during startup ), and I'm pretty sure it's not a default.
Cheers, Mike.
I have made this letter longer than usual because I lack the time to make it shorter ...
... and my other CPU is a Ryzen 5950X :-) Blaise Pascal