how do you OC these dualies (X5680) Jcool?
how do you OC these dualies (X5680) Jcool?
Not at all, if those idiots at German customs keep drooling over my evga SR-2
Now can we stay on topic, please? I know it's hard
I'd like to keep this thread easy-to-read for once, so people can gather all the relevant data quickly.
hhrm...12-Cores: Dual Westmere-EP X5680 (SR-2 missing...) @ 4.3
dual MB?
then we can go on topic.
Am I misunderstanding yor results jcool? I was wondering since you found that wondows was getting higher credit, but the linux system was finishing faster. So in the end do you thing the linux system finishing WU's faster would ever catch up, equal out, or pass the actual credit of the windows maching over time?
I am just not sure i'm quite understanding all the little nuances at work between the 2 systems maby.
Cheers, Mike.
Custom Avi By: Martijn, Custom Calvin V.S. EN by:RSmura
In short: It seems that under Linux, you can compute an identically structured Work Unit (=equally important piece of science) in about half the time it takes to compute an equally big WU under Windows.
In other words, Linux processes Help Conquer Cancer WUs twice as fast as Windows does.
The different credits awarded are due to Linux clients comparing only with other Linux clients, while Windows compares only with Windows. The vast majority of users are running Windows, hence the higher claimed/granted credit (at least that's what we assume). Even though technically Linux should score way better than Windows for granted credit, IF the credit system was fair and IF my findings are correct.
sometimes, it is so easy, just to answer a simple question and raise a smile on a person face and a gratitude in his heart,
what is it with people that they decide just to ignore?
is it nice?
does they like when people do it to them?
do they think of the person involved?
sometimes it is so easy, just to be humble and kind.
Sometimes, just sometimes people actually manage to FIND and USE the forum's very own search function.Originally Posted by onex
sometimes, it is so easy, just to answer a simple question and raise a smile on a person face and a gratitude in his heart,
what is it with people that they decide just to ignore?
is it nice?
does they like when people do it to them?
do they think of the person involved?
sometimes it is so easy, just to be humble and kind
http://www.xtremesystems.org/forums/...ight=evga+SR-2
jcool, jcool,
it's not about the SR board...
never mind though
Can the difference in points, perhaps also be explained by using different boinc versions?
(6.10.45 in Windows vs 6.10.17 in Ubuntu)
I did already noticed that in Windows boinc 6.10.43 was claiming (and granted) more points than 6.10.18.
I don't use linux (or ubuntu), so i don't know if it would be easy to upgrade to latest stable version of boinc 6.10.56 and then compare windows vs linux using the same boinc version.
Last edited by pfm3136; 05-29-2010 at 05:36 PM.
I can kind of follow that line of thought, but why should it make a difference for the claimed credit? Shouldnt that be based on the benchmark score and computing time alone? It used to be like that anyway. I remember some crazy benchmarks in Linux a couple of years ago, maybe thats why they are puishing Linux point wise now.
It could be because Linux machines tend to get the work done faster but on the whole benchmark lower because they're handling server duties etc as well. Less time * lower benchmark (in operation) = lower scores.
[SIGPIC][/SIGPIC]
If they're doing server tasks..
Jcool, did you grab benchmarks for the machine between OS's?
All along the watchtower the watchmen watch the eternal return.
Well yes, that's a point. Though I suspect that most Linux machines (over the whole grid) aren't dedicated crunchers.
[SIGPIC][/SIGPIC]
Did you guys see this post ?
Linux doesn't score lower overall @ BOINC benchmark, on the contrary.
@pfm3136: It's possible, but I don't think so. Plus I've heard people saying BOINC 6.10.17 on Linux gets rather good scores.
Here's another thought: Maybe I (we) get lower credit on Linux, because most Unix machines are non-overclocked, non-dedicated crunchers (ie servers and stuff). Image stock systems being loaded to 30-40% on average with stuff other than BOINC, compared to 100% dedicated and highly OCed machines. A stock i7 920 that has other stuff to handle as well, vs. a dedicated i7 at close to 4Ghz doing nothing but BOINC.
My guess is that there are more people running overclocked machines on Windows; maybe team XS members even get teamed up with other team XS members (= almost all overclocked machines), netting a better score
That's weird, I get 3220/9859 in win7 x64, and 3034/16981 in an Ubuntu 10.04 VM (because I can't be bothered to reboot). It scores higher again if I boot the OS properly. Ie, linux absolutely destroys windows at Integer performance on a Q6600. Perhaps this has something to do with the time taken to complete the work units?
I think its just poor scoring methods.
All along the watchtower the watchmen watch the eternal return.
well, they did it for security reasons. there are other ways to do secure wu's. f@h has their wu's encrypted.
Jcool's benchmark did show a difference in integer speed (a little less fpu, but more integer performance), but nothing like almost doubling integer performance.
Maybe Kentsfield cpu get more performance benefit from linux than westemere.
It would seem that HCC uses more the integer part of the cpu, and thats may lead to crunch the work unit in almost half the time.
So it seems that i'm going to have to learn how to install and use ubuntu in my Q6600, it's almost a dedicated cruncher, so it will be a install and forget.
Not really worried about the points, as i'm going to mount another pc very soon, just a few more parts needed.
Here's some more data for you guys to chew on.Here's another thought: Maybe I (we) get lower credit on Linux, because most Unix machines are non-overclocked, non-dedicated crunchers (ie servers and stuff). Image stock systems being loaded to 30-40% on average with stuff other than BOINC, compared to 100% dedicated and highly OCed machines. A stock i7 920 that has other stuff to handle as well, vs. a dedicated i7 at close to 4Ghz doing nothing but BOINC.
My guess is that there are more people running overclocked machines on Windows; maybe team XS members even get teamed up with other team XS members (= almost all overclocked machines), netting a better score
This is an unscientific comparison between two very similar machines using the first page of valid results for FAAH.
Credit was granted based on 1 machine (no wingman) except for the last one in each list.
Both run 64-bit versions.
Both are overclocked to 4.0GHz.
Memory speed is 1603MHz.
Win7 is P6X58D Premium.
Win7 benchmarks: 3631/11470
Win7 is my main PC used for browsing, email, etc. and nothing really heavy-duty.
Fedora is P6X58D-E.
Fedora benchmarks: 3386/12638
Fedora is dedicated cruncher.
Based on this, fedora did the tasks 6.3% faster, claimed 13.07% fewer points, received 17.03% fewer points, and ended up with 10.64% fewer points/hr.Code:Machine Time Claimed Granted %ofClmd Points/Hr Win7-980X 4.20 124.30 105.10 84.55% 25.02 Win7-980X 3.69 109.70 92.80 84.59% 25.15 Win7-980X 3.88 114.90 96.30 83.81% 24.82 Win7-980X 3.63 106.30 87.80 82.60% 24.19 Win7-980X 3.69 108.70 85.60 78.75% 23.20 Win7-980X 2.76 81.60 64.20 78.68% 23.26 Win7-980X 3.64 109.50 84.90 77.53% 23.32 Win7-980X 3.62 108.30 112.20 103.60% 30.99 Win7-980X 3.57 106.10 100.10 94.34% 28.04 Win7-980X 3.48 100.30 83.10 82.85% 23.88 Win7-980X 5.73 168.50 167.50 99.41% 29.23 Win7-980X 4.46 131.40 106.80 81.28% 23.95 Average 3.86 114.13 98.87 86.00% 25.42 Fedora-980X 3.89 109.80 86.60 78.87% 22.26 Fedora-980X 3.91 101.40 91.20 89.94% 23.32 Fedora-980X 3.41 91.30 73.70 80.72% 21.61 Fedora-980X 3.70 99.80 81.10 81.26% 21.92 Fedora-980X 3.38 90.00 77.50 86.11% 22.93 Fedora-980X 3.44 91.70 78.20 85.28% 22.73 Fedora-980X 3.41 92.50 77.00 83.24% 22.58 Fedora-980X 3.33 89.50 75.60 84.47% 22.70 Fedora-980X 3.38 90.80 85.80 94.49% 25.38 Fedora-980X 4.73 125.10 104.60 83.61% 22.11 Fedora-980X 3.52 97.80 78.30 80.06% 22.24 Fedora-980X 3.55 104.00 77.30 74.33% 21.77 Fedora-980X 3.25 91.70 77.70 84.73% 23.91 Fedora-980X 3.68 103.60 83.80 80.89% 22.77 Fedora-980X 4.06 114.10 87.40 76.60% 21.53 Average 3.62 99.22 82.03 82.92% 22.72 Difference 0.24 14.92 16.84 3.08% 2.71 Percentage 6.30% 13.07% 17.03% 3.58% 10.64% Difference
Who says the point system isn't perfect?
I'm not at home right now, but I did a 14 day test between 2 identical computers with the same exact BIOS settings, and using WIndows 7 x64 Ultimate and Ubuntu 10.4. I came up with numbers pretty close to other posters. That is, Linux does more work but benchmarks a lower value and therefore receives less points based on the way WCG calculates points. It's complete BS that LInux can do more work but get less PPD. Their system is broke.
Here's my devil's advocate comment: We're all spending money on computer parts and electricity to do this work. We have proven that their system for calculating points isn't entirely accurate, but we're supposed to trust that the results that we are all providing and paying for (via electricity, heat, and computer parts) actually provide accurate scientific data?
If I went into my previous posts I asked for any comments that they data WCG is using is actually scientifically valid and being used around the world for great studies. I can read lots of posts from people linking to WCG's pages and a study or two using WCG data, but I am beginning to have serious doubts as to how useful this project REALLY is.
I switched from folding at home to WCG about 2 months ago, and I'm starting to having second thoughts on my decision, which is quite disappointing. I'm starting to question the validity of this project and how useful it really is to mankind. Stay tuned because I've decided that I want to do some deep level research on the net and figure out WTH is really going on and if these work units really are as useful as we all have been told.
WCG is supposed to be looking at reworking the points system, but the science always takes priority. Which to me, is completely understandable. Even after reworking it, there is no guarantee that it will ever be perfect.
While the points are a fun way to keep track of the crunching, it is not why I do it.
<rant>In some respects the points are bogus anyway, especially for overall BOINC points. I mean what's the point of having 20 5890s blasting away at endless, useless calculations figuring out if Collatz was right or not? So someone can say they have the most BOINC points or highest BOINC PPD or used the most KWH last month?</rant>
To figure out if Collatz was right or not.
It scores lower at fpu, higher at integer.. Could it be that points are mostly based on the fpu part of the benchmark?
That would explain less points and more work done at the same time, as it seems that hcc can really use well the better integer performance of linux.
This could explain the granted credit, but not the claimed credit, i think...
so what you say actually, is that if one cpu does the same work faster then a second cpu, it would get less points,I'm not at home right now, but I did a 14 day test between 2 identical computers with the same exact BIOS settings, and using WIndows 7 x64 Ultimate and Ubuntu 10.4. I came up with numbers pretty close to other posters. That is, Linux does more work but benchmarks a lower value and therefore receives less points based on the way WCG calculates points. It's complete BS that LInux can do more work but get less PPD. Their system is broke.
they seem to grant more credit per time crunched and not per work done..
E: or either pfm is right.
that is confusing.
can't it be actually tested, i.e - a 980 going 4GHz should be granted less credit then a 980 going 3GHz?
Last edited by onex; 05-30-2010 at 08:18 AM.
I think the points system is actually a mix (or better, a mess), and it does take runtime into account.
The problem is when the computer is doing something else, it'll take more time to complete the work unit, and so will claim more credit, hence some of the measures they have taken towards the granted credit.
The problems with the points system, is that it isn't very fair, and the benchmark in boinc seems to change a little from version to version.
There isn't a way to test the exact same work unit on linux and windows, but (for example) when someone in linux does 2 wu in 2 hours and 1 wu in windows in the same 2 hours, it doesn't seem very fair that they give more credit for the windows system than the linux one... Unless they are giving credits because of windows price...
Last edited by pfm3136; 05-30-2010 at 09:10 AM.
Posting Permissions
Reply With Quote
Bookmarks