iMac G5 1.6 ghz - 5 hours 41 minutes average
Pentium 4 3.4 ghz - 11 hrs 16 minutes per thread. (2 threads) 5hrs 38 min average
Pentium D840 dual core 0 7 hrs 22 min per core or 3 hrs 41 min average.
AMD 64 X2 4200 - 5 hrs 42 min per core or 2 hrs 51 min average.
Athlon 2400 - 6 hrs 25 min average
Athlon 2800 (previous because Motherboard just died) - 6 hrs 7 min average
(looking for a replacement mobo then will upgrade to Athlon 3200)
iBook 600, 600mhz G3 - about 27 hours average.
PowerMac G5 2.0 GHz ... looks like 4.32 hours for most of them. Used to be 7 to 10 hours. I was not paying attention and had not noticed the speed increase ...
PowerMac G5 2.0 GHz ... looks like 4.32 hours for most of them. Used to be 7 to 10 hours. I was not paying attention and had not noticed the speed increase ...
That was the benefit of the beta test. An altivec enabled app thanks to Bernd.
> And I should not have said that hyperthreading doubled the throughput. I have
> never run it with hyperthreading turned off & expect it would look quite
> bad.
It does not look that bad. HT does not double the performance. It does give a boost to throughput between 10% to as high as 70% more work done per unit time.
We did fairly extensive tests during the late stages of the Beta test program with HT on and off and it does surprisingly well at improving the total production. And I say surprising in that our usage of the processors is very atypical ... we heavily use the math portions and do not have a very "standard" mix of instruction streams.
I mean, my G5 does more work because of the two CPUs than does any of my HT systems, but it is a cheap way to "see" two CPUs.
What is your opinion: I have a choice in the next couple of days to get either a P4 3.4 Gh with HT, or an AMD 64 3800 computer. Which would you choose?
It may be that the P4 with HT does more work per unit time, because it works on 2 E@h data sets at once.
What is your opinion: I have a choice in the next couple of days to get either a P4 3.4 Gh with HT, or an AMD 64 3800 computer. Which would you choose?
Those are the same two options I was looking at... but then I happened to read about the newer "San Diego" core AMD 64 3700+ (socket 939). Chip cost was less than the 3800 or the P4, while cache is larger (more important to SETI than Einstein, but still nice to have) and every indication is that it will overclock to be faster than the 3800 while still running cooler. HT doesn't mean much to me for the other purposes this computer will have, so CPU choice was based on BOINC.
The P4 seems to take 20-21K seconds per WU (40-42K but doing 2 at a time). Without overclocking, this is exactly what the 3700+ is giving me, 20-21Ksec/single WU. The 3800 might be very slightly faster "stock" for Einstein, but I believe it's slightly slower for SETI. Overclocking on the P4 is limited by heat; the 3800 by chip design. The 3700 seems to have much looser limits. I don't know yet how far I'll be able to push the 3700, but others have gotten into the 4000 or even 4200 range, with fans, nothing fancy like liquid cooling. Not being a PC guru, I bought an ASUS motherboard that has "overclocking for dummies" built in, and a case with extra fans - "stock", my CPU is running at 40-41 degrees, so there's a lot of room left to play - I'll hopefully know more in a few days.
Hm... just realized I didn't answer your question, but instead added a third option to the mix!
What is your opinion: I have a choice in the next couple of days to get either a P4 3.4 Gh with HT, or an AMD 64 3800 computer. Which would you choose?
Those are the same two options I was looking at... but then I happened to read about the newer "San Diego" core AMD 64 3700+ (socket 939). Chip cost was less than the 3800 or the P4, while cache is larger (more important to SETI than Einstein, but still nice to have) and every indication is that it will overclock to be faster than the 3800 while still running cooler. HT doesn't mean much to me for the other purposes this computer will have, so CPU choice was based on BOINC.
The P4 seems to take 20-21K seconds per WU (40-42K but doing 2 at a time). Without overclocking, this is exactly what the 3700+ is giving me, 20-21Ksec/single WU. The 3800 might be very slightly faster "stock" for Einstein, but I believe it's slightly slower for SETI. Overclocking on the P4 is limited by heat; the 3800 by chip design. The 3700 seems to have much looser limits. I don't know yet how far I'll be able to push the 3700, but others have gotten into the 4000 or even 4200 range, with fans, nothing fancy like liquid cooling. Not being a PC guru, I bought an ASUS motherboard that has "overclocking for dummies" built in, and a case with extra fans - "stock", my CPU is running at 40-41 degrees, so there's a lot of room left to play - I'll hopefully know more in a few days.
Hm... just realized I didn't answer your question, but instead added a third option to the mix!
Thanks for your response, and it is nice to hear of all the options!
The only thing is, this computer is for me at work, so I may need to be a bit careful with how much I can play around with it in terms of overclocking. I overclock my AMD Thunderbird 1.2 Ghz to 1.368 GHz at home, and that has been great to speed up an old system that still has a lot of life left in it.
The only thing is, this computer is for me at work, so I may need to be a bit careful with how much I can play around with it in terms of overclocking.
No results or WU timings back yet, but since my last posting I hit the "105%" button on the ASUS overclocking utility - went from 2.200 GHz to 2.310 GHz. Reran benchmarks, they went from 2095/3930 to 2185/4090 (4%)... CPU temp went from 40 to 42.
I tried the "110%" button but it locked up after a minute or so. I suspect the "for dummies" part of the program doesn't adjust voltages and multipliers and whatnot just right for the increased clock speed, or the cheap RAM I have is a problem - will have to do some reading and try again manually. You (and everyone else here probably, I'm a Mac guy) I'm sure know more about overclocking than I do. From what I've read about 3700 "San Diego"s though, 10% (2.42 GHz) should barely be pushing it. The limit on air seems to be around 30% in fact... although I value accuracy and CPU lifetime too much to push it that far.
The problem with overclocking an "office" computer is the case and fans - I looked for a conservative-looking case that would have plenty of airflow and eventually gave up. All the ones that really move air are for gamers. The red and blue LEDs really light up my home office at night! Reasonably, though, as cool as the 3700 runs, and with only a 2 degree increase at 5%, stock "boring case" fans should work fine.
All this effort to shave a few seconds off our crunching times... is this a hobby or an obsession? :-) I'll have at least one or two result times at 105% to report back soon.
You (and everyone else here probably, I'm a Mac guy) I'm sure know more about overclocking than I do. From what I've read about 3700 "San Diego"s though, 10% (2.42 GHz) should barely be pushing it. The limit on air seems to be around 30% in fact... although I value accuracy and CPU lifetime too much to push it that far.
All this effort to shave a few seconds off our crunching times... is this a hobby or an obsession? :-) I'll have at least one or two result times at 105% to report back soon.
I am probably newer than you at overclocking! I just started the weekend before this after reading an article about it in PC World magazine. Then my power supply failed and I was sure it was my pedestrian efforts at overclocking, but the guy at CompUSA said this was probably due to dirty power from the utility company.
And yes, shaving off E@h result speeds does seem an obsession at times! It's one reason I am asking for an upgrade of my office computer (although it does have other issues, too, but I am asking for the fastest processor I can get for a price that I think management will find acceptable). ;)
iMac G5 1.6 ghz - 5 hours 41
)
iMac G5 1.6 ghz - 5 hours 41 minutes average
Pentium 4 3.4 ghz - 11 hrs 16 minutes per thread. (2 threads) 5hrs 38 min average
Pentium D840 dual core 0 7 hrs 22 min per core or 3 hrs 41 min average.
AMD 64 X2 4200 - 5 hrs 42 min per core or 2 hrs 51 min average.
Athlon 2400 - 6 hrs 25 min average
Athlon 2800 (previous because Motherboard just died) - 6 hrs 7 min average
(looking for a replacement mobo then will upgrade to Athlon 3200)
iBook 600, 600mhz G3 - about 27 hours average.
Team MacNN - The best Macintosh team ever.
iBook G4 1.33 GHz, 7.7 hours
)
iBook G4 1.33 GHz, 7.7 hours aaverage
PowerBook G4 1.67 GHZ, 6.3 hours average
Join Team MacNN
Mac Mini (G4 1.25GHz) - 8
)
Mac Mini (G4 1.25GHz) - 8 hours, 8 min avg
iBook (G3 900MHz) - all results are gone, haven't used in a while; guessing 20 hours?
AMD 3700+ "San Diego", WinXP - 5 hours, 37 min avg
This last one is my new "cruncher", not yet overclocked, only a couple of results to get a baseline.
PowerMac G5 2.0 GHz ... looks
)
PowerMac G5 2.0 GHz ... looks like 4.32 hours for most of them. Used to be 7 to 10 hours. I was not paying attention and had not noticed the speed increase ...
RE: PowerMac G5 2.0 GHz ...
)
That was the benefit of the beta test. An altivec enabled app thanks to Bernd.
Team MacNN - The best Macintosh team ever.
RE: > And I should not have
)
What is your opinion: I have a choice in the next couple of days to get either a P4 3.4 Gh with HT, or an AMD 64 3800 computer. Which would you choose?
It may be that the P4 with HT does more work per unit time, because it works on 2 E@h data sets at once.
Thanks!
RE: What is your opinion:
)
Those are the same two options I was looking at... but then I happened to read about the newer "San Diego" core AMD 64 3700+ (socket 939). Chip cost was less than the 3800 or the P4, while cache is larger (more important to SETI than Einstein, but still nice to have) and every indication is that it will overclock to be faster than the 3800 while still running cooler. HT doesn't mean much to me for the other purposes this computer will have, so CPU choice was based on BOINC.
The P4 seems to take 20-21K seconds per WU (40-42K but doing 2 at a time). Without overclocking, this is exactly what the 3700+ is giving me, 20-21Ksec/single WU. The 3800 might be very slightly faster "stock" for Einstein, but I believe it's slightly slower for SETI. Overclocking on the P4 is limited by heat; the 3800 by chip design. The 3700 seems to have much looser limits. I don't know yet how far I'll be able to push the 3700, but others have gotten into the 4000 or even 4200 range, with fans, nothing fancy like liquid cooling. Not being a PC guru, I bought an ASUS motherboard that has "overclocking for dummies" built in, and a case with extra fans - "stock", my CPU is running at 40-41 degrees, so there's a lot of room left to play - I'll hopefully know more in a few days.
Hm... just realized I didn't answer your question, but instead added a third option to the mix!
RE: RE: What is your
)
Thanks for your response, and it is nice to hear of all the options!
The only thing is, this computer is for me at work, so I may need to be a bit careful with how much I can play around with it in terms of overclocking. I overclock my AMD Thunderbird 1.2 Ghz to 1.368 GHz at home, and that has been great to speed up an old system that still has a lot of life left in it.
Thanks again!
RE: The only thing is, this
)
No results or WU timings back yet, but since my last posting I hit the "105%" button on the ASUS overclocking utility - went from 2.200 GHz to 2.310 GHz. Reran benchmarks, they went from 2095/3930 to 2185/4090 (4%)... CPU temp went from 40 to 42.
I tried the "110%" button but it locked up after a minute or so. I suspect the "for dummies" part of the program doesn't adjust voltages and multipliers and whatnot just right for the increased clock speed, or the cheap RAM I have is a problem - will have to do some reading and try again manually. You (and everyone else here probably, I'm a Mac guy) I'm sure know more about overclocking than I do. From what I've read about 3700 "San Diego"s though, 10% (2.42 GHz) should barely be pushing it. The limit on air seems to be around 30% in fact... although I value accuracy and CPU lifetime too much to push it that far.
The problem with overclocking an "office" computer is the case and fans - I looked for a conservative-looking case that would have plenty of airflow and eventually gave up. All the ones that really move air are for gamers. The red and blue LEDs really light up my home office at night! Reasonably, though, as cool as the 3700 runs, and with only a 2 degree increase at 5%, stock "boring case" fans should work fine.
All this effort to shave a few seconds off our crunching times... is this a hobby or an obsession? :-) I'll have at least one or two result times at 105% to report back soon.
RE: You (and everyone else
)
I am probably newer than you at overclocking! I just started the weekend before this after reading an article about it in PC World magazine. Then my power supply failed and I was sure it was my pedestrian efforts at overclocking, but the guy at CompUSA said this was probably due to dirty power from the utility company.
And yes, shaving off E@h result speeds does seem an obsession at times! It's one reason I am asking for an upgrade of my office computer (although it does have other issues, too, but I am asking for the fastest processor I can get for a price that I think management will find acceptable). ;)