Message boards :
Number crunching :
Slowest Number Cruncher?
Message board moderation
Author | Message |
---|---|
jimmyhua Send message Joined: 16 Apr 05 Posts: 97 Credit: 369,588 RAC: 0 |
What is the slowest machine you've got running BOINC and seti@home? I have 1. Sempron XP2500 512MB RAM 2. Duron 800MHz 128 MB RAM 3. Sun Ultra 2 Enterprise Service 256MB RAM with Dual 200 Mhz Ultra sparc CPUs 4. Sun Sparc 5 128 MB RAM The benchmark says in MFLOPS we are looking at: 1. 1300 MFLOPS Sempron XP2500+ 2. 400 MFLOPS Duron 800 Mhz 3. 90 MFLOPS Single 200 Mhz UltraSparc2 4. 20 MFLOPS Single 60 Mhz Sparc processor in WU's average turn around is: 1. 4 hours 2. 8 hours 3. 32 hours (2 WU's per turn around though) 4. 144 Hours or about 5 days. I'm a bit surprised that the Duron runs a bit faster than it's MFLOP rating. not sure why, maybe cuz it's running Linux instead of XP. The Sun computers are running Solaris9. And back in the day, these were THE BEST computers out there, bar none! Now it seems like they are SLOW, I was expecting them to keep up with the Duron at least. However, 5+ days to finish a single WU, you might actually miss the duedate!!! After seeing how SLOW that thing is, I have 2 thoughts. Is it worth the electricity bill to keep that Sparc 5 going? I think I'm going to turn it off... Anyone running 386 or 486 crunchers? I imagine the turnaround for an average seti WU would be close to the time limit of 14 days!!! |
jimmyhua Send message Joined: 16 Apr 05 Posts: 97 Credit: 369,588 RAC: 0 |
> in WU's average turn around is: > > 1. 4 hours > 2. 8 hours > 3. 32 hours (2 WU's per turn around though) > 4. 144 Hours or about 5 days. Well it turns out my Sun Ultra 2 Enterprise Dual 200Mhz Ultra Sparc 2 processes with 256MB RAM may be slower than I thought it was! The average credits claimed by my fastest machine is 30-40 credits. 2 results were spit out by my Ultra 2 (FINALLY!), and it claimed 0.7 credits for one and 10 credits for the other!!!! This is after, 19 hours for one, and 33 hours for the other!!! So if my Ultra were to actually work the credits that the server is sending out to my fastest machine, I'd be sitting here for 90 hours average per WU. Am I correct to assume then, that the server is smart enough to send the WU's with little to do on to the slower machines, and the ones with alot to do on to the faster ones???? Jimmy |
N/A Send message Joined: 18 May 01 Posts: 3718 Credit: 93,649 RAC: 0 |
If I can find eight 16MB 30-pin 100ns SIMMs, I can out-slow you. |
lanky Send message Joined: 26 Sep 03 Posts: 5 Credit: 663,641 RAC: 1 |
|
jimmyhua Send message Joined: 16 Apr 05 Posts: 97 Credit: 369,588 RAC: 0 |
Well, I decided that my cut-off point will be 24hours. If the machine can't finish your typical WU in 24 hours. It ain't worth the electricity bill :-). Once that sparc5 finishes working that 1 WU it downloaded 2 days ago. I'll be taking it offline also. I've also have access to an IBM RISC6000 machine running AIX. This sounds fast yes? Unfortunately, it is running a single 604e 233Mhz Processor (IBM G3 Processor? It can take 6 of these). I am guessing it runs the same speed as the Ultra2. I'm too lazy to find out though, as it looks like another loser. It turns out that the Ultra2 was claiming WAY LESS CREDIT than any other computer. So probably the benchmarks should have been better for it. I've got access to a Dell laptop that runs off a Pentium M processor also. I'm curious how it will run too, now that I see some guys laptop runs faster than a server! Geez! You gotta love those Computer sellers though. Most servers need way less resources than your typical workstation. Yet, servers are usually WAY more expensive. If you're paying for a server, and it doesn't have any fault tolerance, you're paying too much for it, get a regular PC and put server software in it. Jimmy Huang A strange thing happened. My Ultra2 finished working all 3 WU's issued to it. I have taken it offline. However, website only shows that it got 2 of 3. I guess I'll just let that other one expire. Dunno what else to do. It's not on my Ultra anymore, and it isn't registered on the website as received either. |
N/A Send message Joined: 18 May 01 Posts: 3718 Credit: 93,649 RAC: 0 |
I've also have access to an IBM RISC6000 machine running AIX. This sounds fast yes? Unfortunately, it is running a single 604e 233Mhz Processor (IBM G3 Processor? It can take 6 of these). I am guessing it runs the same speed as the Ultra2. I'm too lazy to find out though, as it looks like another loser. It'll most likely be a ppc604e because I don't recall AIX being written for G3. Still, the MHz sounds G3-likely. |
Contact Send message Joined: 16 Jan 00 Posts: 197 Credit: 2,249,004 RAC: 0 |
|
Everette Dobbins Send message Joined: 13 Jan 00 Posts: 291 Credit: 22,594,655 RAC: 0 |
> What is the slowest machine you've got running BOINC and seti@home? > > I have > > 1. Sempron XP2500 512MB RAM > 2. Duron 800MHz 128 MB RAM > 3. Sun Ultra 2 Enterprise Service 256MB RAM with Dual 200 Mhz Ultra sparc > CPUs > 4. Sun Sparc 5 128 MB RAM > > The benchmark says in MFLOPS we are looking at: > > 1. 1300 MFLOPS Sempron XP2500+ > 2. 400 MFLOPS Duron 800 Mhz > 3. 90 MFLOPS Single 200 Mhz UltraSparc2 > 4. 20 MFLOPS Single 60 Mhz Sparc processor > > in WU's average turn around is: > > 1. 4 hours > 2. 8 hours > 3. 32 hours (2 WU's per turn around though) > 4. 144 Hours or about 5 days. > > I'm a bit surprised that the Duron runs a bit faster than it's MFLOP rating. > not sure why, maybe cuz it's running Linux instead of XP. > > The Sun computers are running Solaris9. And back in the day, these were THE > BEST computers out there, bar none! Now it seems like they are SLOW, I was > expecting them to keep up with the Duron at least. > > However, 5+ days to finish a single WU, you might actually miss the duedate!!! > > > After seeing how SLOW that thing is, I have 2 thoughts. Is it worth the > electricity bill to keep that Sparc 5 going? I think I'm going to turn it > off... > > Anyone running 386 or 486 crunchers? I imagine the turnaround for an average > seti WU would be close to the time limit of 14 days!!! > > > > I have a Pentium-III 700MHz/100MHz FSB/128KbL2 Average wu time is 11Hrs/52min its slow but has been running since it was new. I just cant seem to pull the plug on it. |
N/A Send message Joined: 18 May 01 Posts: 3718 Credit: 93,649 RAC: 0 |
I think I might have found the slowest cruncher on record: 639525 at 10/10. |
Ron Roe Send message Joined: 28 Feb 02 Posts: 156 Credit: 24,124 RAC: 0 |
> I think I might have found the slowest cruncher on record: <a> href="http://setiweb.ssl.berkeley.edu/show_host_detail.php?hostid=639525">639525[/url] > at 10/10. > It appears this user is using the -skip_cpu_benchmarks command line switch which gives the artificially low benchmarks of 10M flops/iops. Worse is the very low claimed credits that this causes - 55195562. |
N/A Send message Joined: 18 May 01 Posts: 3718 Credit: 93,649 RAC: 0 |
But if you go back some 700 WUs in its history, you start finding "normal" credit being claimed. Out of curiosity, can anyone explain why there'd be over 600 downloaded WUs for one box? |
Holmis Send message Joined: 1 Jun 99 Posts: 30 Credit: 951,184 RAC: 0 |
> But if you go back some 700 WUs in its history, you start finding "normal" > credit being claimed. > > Out of curiosity, can anyone explain why there'd be over 600 downloaded WUs > for one box? How about this for an answer: Number of CPUs 64 |
N/A Send message Joined: 18 May 01 Posts: 3718 Credit: 93,649 RAC: 0 |
But that's the number of CPUs in one host. I thought there was a limit on the number of WUs that can be downloaded based on the number of attached hosts, not CPUs. Am I wrong? Moreover, another IRIX box (same owner) has only about 30 WUs cached. Still, if you look at the crunch times for the host (starting at the 700th on back), it looks as if the average crunch time is around 18Ksec/WU to 22Ksec/WU. For the sake of argument, let's say that it's 18Ksec/WU. If there are 600WUs listed as "In Progress", and it's 18Ksec/WU, then it'll be 10.8Msec ([b]125 days</B>) before all of the WUs are completed! Maybe it's a storm in a teacup, but IMO this is one heckuva tea cup. Credit where it's due, though: This guy's a serious Crucnher! |
jimmyhua Send message Joined: 16 Apr 05 Posts: 97 Credit: 369,588 RAC: 0 |
I'm not familiar with that box at all. SGI IP35 with 64 CPUs at 32GB RAM. Running IRIX64 version 6.5 When I was in college, they used to have a few Sequent Machines in Berkeley. These ran off a multiprocessor kernel version of BSD. Usually had anywhere between 100 to 200 386 or 486 processors!!! That would make their turnaround time close to 10 days, however, they'd spit out 100 to 200 WUs in that one turn around!!! Seeing the recent average credits, Each processor is probably faster than a Pentium 3-800Mhz, with 512MB RAM for each processor. :) Pretty modern if you ask me. That box is WAY out of my league. Jimmy |
Kajunfisher Send message Joined: 29 Mar 05 Posts: 1407 Credit: 126,476 RAC: 0 |
I'm wondering if he is setting "connect to..." at a very large number, and crunching them on the other faster systems, then reporting the results back through the same system. Would that be possible? I've looked at the other systems he has also, wondering if there isn't a loophole of sorts to report results that were never sent to pc "a" but crunched on "b"... your thoughts? No matter where you go, there you are... |
jimmyhua Send message Joined: 16 Apr 05 Posts: 97 Credit: 369,588 RAC: 0 |
> I've also have access to an IBM RISC6000 machine running AIX. This sounds > fast yes? Unfortunately, it is running a single 604e 233Mhz Processor (IBM G3 > Processor? It can take 6 of these). I am guessing it runs the same speed as > the Ultra2. I'm too lazy to find out though, as it looks like another > loser. > It'll most likely be a ppc604e because I don't recall AIX being written for > G3. Still, the MHz sounds G3-likely. > I didn't know there was a difference between the ppc604e and the iMac G3. I thought they used the same exact processor? You are right, labeling on CPU says ppc604e 233Mhz. Jimmy |
jimmyhua Send message Joined: 16 Apr 05 Posts: 97 Credit: 369,588 RAC: 0 |
> >...can't finish your typical WU in 24 hours. It ain't worth the > electricity bill :-). > > > > > Still runnin' a few of these. > But could be worse. Could be powered by gas! > Hey, I don't know if you realize this, but if you live in the good 'ole USA. Or most other parts of the world except France or Washington State, the electricity coming into your house can be directly traced to some fossil fuel burning powerplant somewhere. Jimmy |
N/A Send message Joined: 18 May 01 Posts: 3718 Credit: 93,649 RAC: 0 |
I used to know the differences between the PowerPCs... now it's a bit of a blur. The G3 had a smaller trace, higher bus multipliers, a larger on-chip cache, and couldn't be used for multiprocessing (without intensive mobo design work, but it could be done). This is why the Power Mac 9600 was still king of the hill when the blue & white towers came out - it could still outperform the G3 tower. Again, IIRC, IBM produced the 601, 604, G3, and G5, but only used the 604 and G5 on their own boards and systems. |
Ingleside Send message Joined: 4 Feb 03 Posts: 1546 Credit: 15,832,022 RAC: 13 |
> But that's the number of CPUs in one host. I thought there was a limit on the > number of WUs that can be downloaded based on the number of attached hosts, > not CPUs. Am I wrong? Moreover, another IRIX box > (same owner) has only about 30 WUs cached. Daily quota is per computer, multiplied by how many cpu's in the computer but with upper cpu-limit 4. In seti, this means 100-400 wu/day per computer, depending on how many cpu's. > > Still, if you look at the crunch times for the host (starting at the 700th > on back), it looks as if the average crunch time is around 18Ksec/WU to > 22Ksec/WU. For the sake of argument, let's say that it's 18Ksec/WU. If there > are 600WUs listed as "In Progress", and it's 18Ksec/WU, then it'll be 10.8Msec > ([b]125 days</B>) before all of the WUs are completed! 18ks/wu = 5h/wu. This gives 4.8 wu/day per cpu, or 307.2 wu/day. So can look like he's using a cache-setting around 2 days. |
N/A Send message Joined: 18 May 01 Posts: 3718 Credit: 93,649 RAC: 0 |
Once again, Ingleside, you enlighten me. |
©2024 University of California
SETI@home and Astropulse are funded by grants from the National Science Foundation, NASA, and donations from SETI@home volunteers. AstroPulse is funded in part by the NSF through grant AST-0307956.