Printable View
errr ... dunno really :lol: I just keep 'em coz I find 'em interesting. I'm not the thread originator, though. I've just stayed interested. If you have a look you'll see that with the benchmark you can get an idea what a given machine will claim assuming 24/7 production. with some creative sorting you can get an idea which OS/client version/ 32-64bit etc SHOULD give you better output on given hardware.
my E8400 @ 4GHz in Vista Ultimate 64bit
19/10/2008 8:06:03 PM||Starting BOINC client version 6.2.19 for windows_x86_64
19/10/2008 8:06:04 PM||Processor: 2 GenuineIntel Intel(R) Core(TM)2 Duo CPU E8400 @ 3.00GHz [Intel64 Family 6 Model 23 Stepping 6]
19/10/2008 8:06:04 PM||Processor features: fpu tsc pae nx sse sse2 pni
19/10/2008 8:06:04 PM||OS: Microsoft Windows Vista: Ultimate x64 Editon, Service Pack 1, (06.00.6001.00)
19/10/2008 8:06:04 PM||Memory: 4.00 GB physical, 8.16 GB virtual
22/10/2008 6:53:40 PM|| Number of CPUs: 2
22/10/2008 6:53:40 PM|| 3944 floating point MIPS (Whetstone) per CPU
22/10/2008 6:53:40 PM|| 11293 integer MIPS (Dhrystone) per CPU
22/10/2008 6:53:41 PM||Resuming computation
same CPU in 6.3.12 64bit I get:
22/10/2008 7:28:43 PM|| 3938 floating point MIPS (Whetstone) per CPU
22/10/2008 7:28:43 PM|| 11682 integer MIPS (Dhrystone) per CPU
AMD 4400+X2 @ 2.2GHz WinXP Pro SP3 32bit
17/10/2008 3:09:24 AM||Starting BOINC client version 6.2.19 for windows_intelx86
17/10/2008 3:09:25 AM||Processor: 2 AuthenticAMD AMD Athlon(tm) 64 X2 Dual Core Processor 4200+ [x86 Family 15 Model 43 Stepping 1]
17/10/2008 3:09:25 AM||Processor features: fpu tsc pae nx sse sse2 3dnow mmx
17/10/2008 3:09:25 AM||OS: Microsoft Windows XP: Professional x86 Editon, Service Pack 3, (05.01.2600.00)
17/10/2008 3:09:25 AM||Memory: 2.00 GB physical, 3.85 GB virtual
22/10/2008 6:55:26 PM|| Number of CPUs: 2
22/10/2008 6:55:26 PM|| 2103 floating point MIPS (Whetstone) per CPU
22/10/2008 6:55:26 PM|| 3516 integer MIPS (Dhrystone) per CPU
22/10/2008 6:55:27 PM||Resuming computation
Q6600@3.2GHz WHS 32Bit running same version as all the others.
19/10/2008 4:40:39 PM|| Number of CPUs: 4
19/10/2008 4:40:39 PM|| 3108 floating point MIPS (Whetstone) per CPU
19/10/2008 4:40:39 PM|| 6699 integer MIPS (Dhrystone) per CPU
19/10/2008 4:40:40 PM||Resuming computation
Q6600@3.3GHz 64Bit Vista Ultimate BOINC client version 6.2.19 for windows_x86_64
22/10/2008 7:04:37 PM|| Number of CPUs: 4
22/10/2008 7:04:37 PM|| 3307 floating point MIPS (Whetstone) per CPU
22/10/2008 7:04:37 PM|| 7286 integer MIPS (Dhrystone) per CPU
22/10/2008 7:04:38 PM||Resuming computation
Oviously I need to clock these some more right? ;) So far I am getting almost 50k WCG points a day with these 5 comps according to your spreadsheet. If only I could afford another Q6600!!
p.s. Hopefully that wasnt too much info!
Nice, my Harper owns your list :D :wasntme: :up:Quote:
Originally Posted by D_A
Best single machine I've gotten hold of the details for. :up:
Say, who goes by "DCManiac" anyway? I know they're here somewhere but I'm not sure who it is.
That is SerlvQuote:
Originally Posted by D_A
I just tracked him down, but you beat me to it.
jcool, that machine of yours ranks 4th fastest overall at WCG, and so far THE fastest I can find that someone wants to admit to owning.
THE hottest system I can find is a 32 core Opteron based system.
Ya, I used to have 2nd place with my E5420's a while ago... guess it's only down from there :down:
I'd like to know more about that 32 core box, but there's not a lot of details about it aside for it's running Win Server 2008. No owner, no bench scores ... :(
Maybe some kind of cluster? You know, like that Helmer machine? Several rigs put together to appear as one?
Makes no sense to do that for WCG though, as you loose quite a bit of firepower linking them.
I've seen hardware before that would take 8 quad core chips ... might be one of those.
Thanks to Asus releasing a new Bios for the Z7S that now sports Vcore adjustment, the team gained an additional 1,5Ghz (8*190Mhz :D )
X5450's at 451*9=4,06Ghz, primed them a few hours at 1,22V Vcore and 1,07V VTT :shocked:
<cracks out the spreadsheet>
edit: updated and uploaded.
Them calculations can't be real tho.. so far they did an average of maybe 40k/d... your calc says 45k. Maybe now it's a true 44K.. I hope, but we'll see.
Those numbers indicate what the machine will CLAIM. What comes out the other end of the quorum, however ...
As a good example: I have an XP2400+ system here. (overclocked of course) that typically gets awarded HALF what it claims, but it claims about 1700ppd. It benches much better than the P4 2.4 sitting beside it but takes longer to process the same work, so it effectively over-claims. This makes no sense to me, but the numbers don't lie (this time). I suspect some of the science apps (HCC for one) are optimised for one processor type or the other.
Hm, well. It never threw any errors, and I only got 2 "Inconclusive" WUs once. :shrug:
Oh errors are separate to what I'm talking about. We all suffer from the tyranny of the quorum. If you and I both do the same work unit but I claim say 180 for it and you only claim 50, but both check as valid, I'm gonna get pwned on the returning numbers while you'll get AT LEAST what you've claimed.
Didn't know that. Blows. Anything you can do to ensure you get what you claim?
I bet they discard half of what my Harper is claiming cause they think it can't be that fast lol.
It's a factor of how much benchmark and how much time. My XP2400 takes too long to get through work based on it's benchmark (I agree with that, I wish I knew why it's slower than the P4 coz it shouldn't be) so it over claims.
How long does your 2400+ take?
I just recently set up a 1700+.. running 210x10,5 of course ;)
It takes about 8:45h for one hcc unit.. sooo slow :(
XP 32, 6.3.12
XP2400+ (@2160, unstable any higher), 1GB RAM, stripped down Ubuntu 8.04 32bit BOINC 6.2.15 takes maybe 13hrs. I've been suspicious there's something wrong here for a while now.
edit: I should mention that the board isn't supposed to run the 2400
Wth? 13h sounds seriously wrong. And I thought mine was slow :p:
Mine's only running 1 stick of mem, so not even DC.
It's an Abit NF7 mobo tho with CR1 modbios.. FSB 210 synced, but only 2,5-3-3-8 on the mem :/
CPU can't do more than FSB 210 either. Also, it hits a wall beyond 2,2Ghz. That's rockstable at 1,88V Vcore, but not even >2V would get me 2,3Ghz, oh well.
At least it's not that power-hungry at 146W boinc...
I have 2 sticks of RAM in there, but no dual channel (I don't think). It's a MSI MS-6390 mATx board which is only listed to take up to 2200+ chips. It's currently sitting at 42°C (slightly less than my Opty 185 @2.86). It's just too slow for it's benchmark, if that makes sense. For what it's claiming it should be working faster. I'm looking into compiling a custom kernel for it after this semester finishes and I have more time. There's a TON of stuff a headless machine just doesn't need in the kernel that could save me some overhead ... but there's a limit to what can be gained just the same. No matter what I strip out, it just ain't gonna be a X5450. If the science apps are compiled for Intel then I'm kinda wasting my time anyway.
Well, it should do sth between 9 and 10h anyway. If you ain't got time to compile, just throw Win XP on it like the rest of us, for heaven's sake :p:
Lessee ... I don't actually own a legal copy for one, I don't think that's really the problem, and I have the whole thing running on only a 6GB HDD, so after the install I have my doubts there'll be enough room for the OS to scratch itself let alone do much useful work. Don't get me wrong, I think XP is the best version of Windows out there ... but I walked away from it ages ago partly because of how it holds the hardware back. Even that old thing cold boots to a fully functional desktop in under a minute, with everything loaded, the hdd inactive and BOINC crunching away happily. XP on my daughter's XP3000+ ... try about 5 minutes.
Say ... what's the bench on that 1700+ anyway?