Best set up
-
- Posts: 1434
- Joined: Tue Jan 09, 2007 12:00 am
Best set up
Just a thought, and I'm not sure if it's been done before, but would it be worth starting a thread to establish the best crunching set up.
It could include a pole as to preference i.e.
Intel + Windows
Intel + Linux
AMD + Windows
AMD + Linux
And posts stating what set up produces how many credits on each project.
After a while the results could be collated and may be of use to anybody considering an upgrade or new system and in choosing which projects offer them the best returns.
Whilst I know this will become redundant as new hardware comes on the market, few of us can afford the latest greatest CPU's and opt for the best we can afford which are usually the superseded models.
Any thoughts anyone.?
Anyone wishing me to stop thinking please post here :oops:
Anonymous
It could include a pole as to preference i.e.
Intel + Windows
Intel + Linux
AMD + Windows
AMD + Linux
And posts stating what set up produces how many credits on each project.
After a while the results could be collated and may be of use to anybody considering an upgrade or new system and in choosing which projects offer them the best returns.
Whilst I know this will become redundant as new hardware comes on the market, few of us can afford the latest greatest CPU's and opt for the best we can afford which are usually the superseded models.
Any thoughts anyone.?
Anyone wishing me to stop thinking please post here :oops:
Anonymous
-
- UBT Forum Admin
- Posts: 9689
- Joined: Mon Mar 13, 2006 12:00 am
- Location: NW Midlands
- Contact:
Re: Best set up
Hi 'Mous,Rockinfroggi wrote:Just a thought, and I'm not sure if it's been done before, but would it be worth starting a thread to establish the best crunching set up.
It's a good idea, although I'm thinking that only few projects might give conclusive results, as most WUs are different to each other and hence we'd be talking about a range of credits gained on a specific project, and by a specific CPU/memory.
Of course, one easy way to do this is for each user to "grab" their stats from each project (via their "Your Account > View Computers" page) for each "host" that they have, such as this from my account on XtremLab:
GenuineIntel x86 Family 6 Model 8 Stepping 3 801MHz
Microsoft Windows NT Workstation Edition, Service Pack 1, (04.00.1381.00)
Code: Select all
11784480 11020904 22 Mar 2007 16:53:17 UTC 22 Mar 2007 17:13:36 UTC Over Success Done 596.20 0.68 0.68
11784479 11020903 22 Mar 2007 16:53:17 UTC 22 Mar 2007 17:13:36 UTC Over Success Done 600.00 0.69 0.69
11783704 11211098 22 Mar 2007 16:32:59 UTC 22 Mar 2007 17:13:36 UTC Over Success Done 596.20 0.68 0.68
11783703 11211097 22 Mar 2007 16:32:59 UTC 22 Mar 2007 17:13:36 UTC Over Success Done 599.00 0.69 0.69
11783702 11211096 22 Mar 2007 16:32:59 UTC 22 Mar 2007 16:53:17 UTC Over Success Done 598.91 0.69 0.69
11783460 11210880 22 Mar 2007 16:12:35 UTC 22 Mar 2007 16:53:17 UTC Over Success Done 597.23 0.69 0.69
11783459 11210879 22 Mar 2007 16:12:35 UTC 22 Mar 2007 16:53:17 UTC Over Success Done 597.84 0.69 0.69
11783458 11210878 22 Mar 2007 16:12:35 UTC 22 Mar 2007 16:53:17 UTC Over Success Done 597.89 0.69 0.69
11783040 11210476 22 Mar 2007 15:52:15 UTC 22 Mar 2007 16:32:59 UTC Over Success Done 597.58 0.69 0.69
11783039 11210475 22 Mar 2007 15:52:15 UTC 22 Mar 2007 16:32:59 UTC Over Success Done 598.94 0.69 0.69
11783025 11210461 22 Mar 2007 15:52:15 UTC 22 Mar 2007 16:32:59 UTC Over Success Done 598.17 0.69 0.69
11782439 11209938 22 Mar 2007 15:31:55 UTC 22 Mar 2007 16:32:59 UTC Over Success Done 598.98 0.69 0.69
11782438 11209937 22 Mar 2007 15:31:55 UTC 22 Mar 2007 16:12:35 UTC Over Success Done 596.91 0.69 0.69
11782437 11209936 22 Mar 2007 15:31:55 UTC 22 Mar 2007 16:12:35 UTC Over Success Done 597.98 0.69 0.69
11781995 11209522 22 Mar 2007 15:11:35 UTC 22 Mar 2007 15:52:15 UTC Over Success Done 599.11 0.69 0.69
11781969 11209496 22 Mar 2007 15:11:35 UTC 22 Mar 2007 16:12:35 UTC Over Success Done 599.13 0.69 0.69
I would suggest, initially, that maybe you call for stats from specific types of "hosts" such as Win XP with 512Mb running on say a Pentium 4 or Core class CPU.
This way, you can cut down on the initial variables, because I am pretty sure there are a lot of them....!
Some things will be obvious:
Slower CPUs generate higher credits (per WU, due to increase in time taken to finish WU)
Lower memory generates higher credits (per WU, due to more disk "paging", which slows down the CPU a bit, so it takes more time)
And of course, a faster CPU will perhaps generate more credits per period, as it will crunch more WUs in the time, but the credit it receives for EACH WU will be less than with a slower CPU.
Also: Please review this link to Honza's stats here: http://www.boincsynergy.com/forums/viewtopic.php?t=5231
regards
Tim
Re: Best set up
for seti@home you can do this http://setiathome.berkeley.edu/show_use ... format=xmlUBT - Timbo wrote:Of course, one easy way to do this is for each user to "grab" their stats from each project (via their "Your Account > View Computers" page) for each "host" that they have, such as this from my account on XtremLab:
which will give you an xml with all your hosts details in it.
I already do this for my hosts on seti and have a page very similar to the single project stats page here.
Not sure if it works for other projects but it would be a start.
Now if it didn't invlove KEYs, I could do it for anyone that wants it but it does, so I can't
-
- Posts: 1434
- Joined: Tue Jan 09, 2007 12:00 am
Thank you Tim,
You've raised some good points there and I think with a little bit of input from other more knowledgeable (than me that is) members the creases could be ironed out and a reasonably accurate list could be produced.
I'm happy to do any donkey work with a little bit of guidance but obviously it would require a good few active members with hidden hosts being willing to divulge their stats.
I'll see what response this thread gets but may make a start on collating those stats available to me at the moment as this is not going to be a 5 minute job.
Interesting link BTW, but I found it a little confusing and messy and would hope to produce something a little clearer.
Any other Ideas from anyone would be most welcome.
Gary.
You've raised some good points there and I think with a little bit of input from other more knowledgeable (than me that is) members the creases could be ironed out and a reasonably accurate list could be produced.
I'm happy to do any donkey work with a little bit of guidance but obviously it would require a good few active members with hidden hosts being willing to divulge their stats.
I'll see what response this thread gets but may make a start on collating those stats available to me at the moment as this is not going to be a 5 minute job.
Interesting link BTW, but I found it a little confusing and messy and would hope to produce something a little clearer.
Any other Ideas from anyone would be most welcome.
Gary.
-
- Posts: 1434
- Joined: Tue Jan 09, 2007 12:00 am
Re: Best set up
Thanks for that John, something along those line would certainly save a lot of work but not having a clue how you produce those pages I think I will have to do it the hard way.Temujin wrote:for seti@home you can do this http://setiathome.berkeley.edu/show_use ... format=xmlUBT - Timbo wrote:Of course, one easy way to do this is for each user to "grab" their stats from each project (via their "Your Account > View Computers" page) for each "host" that they have, such as this from my account on XtremLab:
which will give you an xml with all your hosts details in it.
I already do this for my hosts on seti and have a page very similar to the single project stats page here.
Not sure if it works for other projects but it would be a start.
Now if it didn't invlove KEYs, I could do it for anyone that wants it but it does, so I can't
I'll have a play around with things for now and adapt things as and if anyone has anything to add.
Gary.
-
- UBT Forum Admin
- Posts: 9689
- Joined: Mon Mar 13, 2006 12:00 am
- Location: NW Midlands
- Contact:
Feel free to take a look at my hosts - I've got an account on almost all the projects and I think all my hosts are NOT hidden....Rockinfroggi wrote:...I'm happy to do any donkey work with a little bit of guidance but obviously it would require a good few active members with hidden hosts being willing to divulge their stats. .....
And my "farm" is fairly constant, as currently, I can't afford to change much, (although I might buy a mobo or two soon, but am waiting for Intel to drop their CPU prices - see the other thread from today!).
So, I'll be carrying on with these four machines for a little while yet....they certainly need to be updated....but they can't really cope with "todays" BOINC WUs.
regards
Tim
-
- Posts: 1434
- Joined: Tue Jan 09, 2007 12:00 am
Been busy collating data and working out a draft page layout for my above project.
I would be grateful for any input on one area and that is, should the page show similar spec AMD next to Similar spec Intel i.e. Sempron next to Celeron which would make for rather a large page, or should I make 2 pages, one for AMD CPUs and one for Intel?
Gary.
I would be grateful for any input on one area and that is, should the page show similar spec AMD next to Similar spec Intel i.e. Sempron next to Celeron which would make for rather a large page, or should I make 2 pages, one for AMD CPUs and one for Intel?
Gary.
-
- Active UBT Contributor 15+ yrs
- Posts: 7515
- Joined: Wed May 03, 2006 1:00 am
- Contact:
I'd say two pages, more so if you intend to add further data.
I have started looking at credits per hour (per core where appropriate) on my pc's (all windows a mix of 3 Intel and 3 AMD) and so far from 8 projects AMD gives an average 68% more credit on 7 of them and Intel only wins on 1 (SAP by 20%). Be interesting to see how they measure up overall.
I have started looking at credits per hour (per core where appropriate) on my pc's (all windows a mix of 3 Intel and 3 AMD) and so far from 8 projects AMD gives an average 68% more credit on 7 of them and Intel only wins on 1 (SAP by 20%). Be interesting to see how they measure up overall.
-
- UBT Forum Admin
- Posts: 9689
- Joined: Mon Mar 13, 2006 12:00 am
- Location: NW Midlands
- Contact:
-
- Posts: 1434
- Joined: Tue Jan 09, 2007 12:00 am
Thank you both, that was the route I favoured. I will keep to the most popular CPUs currently available to start with and as Mark points out, this can be updated and added to over time as other projects and CPUs become available.
Once these pages are complete I may add a page for the benefit of those who may be stuck with older systems but would still like to see at a glance which projects would give them the best returns.
Back to work then
Gary.
Once these pages are complete I may add a page for the benefit of those who may be stuck with older systems but would still like to see at a glance which projects would give them the best returns.
Back to work then
Gary.
-
- Active UBT Contributor 15+ yrs
- Posts: 7515
- Joined: Wed May 03, 2006 1:00 am
- Contact:
-
- Posts: 1434
- Joined: Tue Jan 09, 2007 12:00 am
I hope mine will be much easier on the eye and really only focusing on CPU, platform and project combinations with credits claimed and credits granted.
All projects will be linked to their home page so I see little point in crowding the page with info that can be found there.
This is only meant to be an at a glance comparison of the above mentioned criteria but can be adapted and added to as and if others put forward ideas.
Gary.
All projects will be linked to their home page so I see little point in crowding the page with info that can be found there.
This is only meant to be an at a glance comparison of the above mentioned criteria but can be adapted and added to as and if others put forward ideas.
Gary.
-
- Active UBT Contributor 15+ yrs
- Posts: 7515
- Joined: Wed May 03, 2006 1:00 am
- Contact:
presumably not, I don't really follow that table. is average credit per cpu an hourly figure and on what project?
The figures I have for a few projects are below, PC's (all windows) are numbered 1 to 6 and output is per core per hour...while not a slave to credit clearly ABC is much better than Setibeta on the smallest machine but there is little difference for machines 2 & 3 and so on. I can match the machine that gives the best credit to the projects I want to run.
1 2 3 4 5 6
DOCKING 10.0 11.6 9.7 9.1 21.7 25.5
SETIBETA 6.4 15.7 13.2 8.8 17.4 19.0
XTREMLAB 9.3 9.0 5.9 7.5 13.3 15.7
ABC 15.5 16.3 12.8 13.7 22.2 n/a
1 AMD 2200+ 256MB RAM 40 GB HARD DRIVE
2 PENTIUM D 3.4 GHz 2 GB RAM 280 GB HARD DRIVE
3 INTEL T2050 1.6 GHz 1 GB RAM 60 GB HARD DRIVE
4 PENTIUM D 2.8 GHz 2 GB RAM 180 GB HARD DRIVE
5 AMD 64X2 4600+ 2 GB RAM 280 GB HARD DRIVE
6 AMD 64X2 5600+ 2 GB RAM 280 GB HARD DRIVE
-
- Posts: 1434
- Joined: Tue Jan 09, 2007 12:00 am
You've pretty much got where I'm going with this Mark and it will also show how the same system may vary between Windows and Linux.
Though I'm sure it will not be of any interest to some, as I stated in my first post, it may help some who are thinking of upgrading to improve their crunching output and those looking for projects that are going to at least make the effort worthwhile.
If at the end of the day if I'm the only one that looks at it then that's fine, the idea started as a project for my own benefit but thought I would share it once complete if there was any interest from others.
Gary.
Though I'm sure it will not be of any interest to some, as I stated in my first post, it may help some who are thinking of upgrading to improve their crunching output and those looking for projects that are going to at least make the effort worthwhile.
If at the end of the day if I'm the only one that looks at it then that's fine, the idea started as a project for my own benefit but thought I would share it once complete if there was any interest from others.
Gary.
Last edited by Rockinfroggi on Mon Mar 26, 2007 10:54 pm, edited 1 time in total.
Well said that man :thumbup:Rockinfroggi wrote:If at the end of the day I'm the only one that looks at it then that's fine, the idea started as a project for my own benefit but thought I would share it once complete if there was any interest from others.
It doesn't matter if others use it or what they think of it (but nice if they do use it)
its your project, go for it froggi
-
- Active UBT Contributor 15+ yrs
- Posts: 7515
- Joined: Wed May 03, 2006 1:00 am
- Contact:
-
- Active UBT Contributor 15+ yrs
- Posts: 7515
- Joined: Wed May 03, 2006 1:00 am
- Contact:
-
- Posts: 1434
- Joined: Tue Jan 09, 2007 12:00 am
I hope nobody is in a hurry for me to finish my stats project.
while I was just taking a break from collecting data I did a little maths.
So far I have only listed 35 CPUs over 40 projects for 2 OS platforms with 4 pages of results for each averaged out individually per CPU per project then totalled and averaged out again to try to allow for variations in WUs means I need to perform 50,400 calculations.
And there are many more CPUs to add over time
Cross eyed Frog.
while I was just taking a break from collecting data I did a little maths.
So far I have only listed 35 CPUs over 40 projects for 2 OS platforms with 4 pages of results for each averaged out individually per CPU per project then totalled and averaged out again to try to allow for variations in WUs means I need to perform 50,400 calculations.
And there are many more CPUs to add over time
Cross eyed Frog.