Pages: [1]
Pedro Mariano Yunes Garcia
BAM!ID: 7913
Joined: 2006-10-06
Posts: 20
Credits: 793,966
World-rank: 218,324

2007-01-30 14:16:48

I would like to see a list or rank of users or hosts with same processors. I think it would be usefull to determine the best processors for each project. The host CPU breakdown gives a idea, but the diferent time loads of each user can make things a bit blurry... It would also make possible to spot the stats of the best hosts and try to understand the reason for there diferent performance, if there is such a diference...
Honza
BAM!ID: 109
Joined: 2006-05-10
Posts: 154
Credits: 3,489,438,031
World-rank: 555

2007-01-31 07:59:13

I'm afraid this would be far from accurate.
First, some people do overclocked and frequency hence performance may vary abot 30 or even 50%
Second, credit is not 100% measurement of performance - nor claimed (see diference among Win, Linux and Mac version of BOINC over the time), nor granted (see different quora -> granted credit computation: average, lower etc.).

There comes more machine specific parts like memory configuration, which is not visible from BOINC. Lately, a user on CPDN gave an example that putting 4x1 memory instead of 2x1 memory in dual Clovertown gave the host +25% performance boots.

Having sayd that, you may look for typical procesor, says Intel P4 with HT, D series and Core 2 Duo and AMD 64 and X2 (perhaps mobile CPU). Some are covered in my table (see signature) to give some idea on how is each CPU doing.
It would be far from accurate saying that Core 2 Duo is close to best choice in overall effort.
Pedro Mariano Yunes Garcia
BAM!ID: 7913
Joined: 2006-10-06
Posts: 20
Credits: 793,966
World-rank: 218,324

2007-01-31 12:02:00
last modified: 2007-01-31 12:04:29

I'm afraid this would be far from accurate.


And I'm afraid I missed your point...

First, some people do overclocked and frequency hence performance may vary abot 30 or even 50%


So what? Then we could see how much the overclock can boost the crunching by watching the variaton within hosts with same processors...

Second, credit is not 100% measurement of performance - nor claimed (see diference among Win, Linux and Mac version of BOINC over the time),...


Ok, so it would be nice to be able to sort or display in that list/rank users OS...

... nor granted (see different quora -> granted credit computation: average, lower etc.).


Ok, here I failed to understand what you're saying...

There comes more machine specific parts like memory configuration, which is not visible from BOINC. Lately, a user on CPDN gave an example that putting 4x1 memory instead of 2x1 memory in dual Clovertown gave the host +25% performance boots.


Again, we could then, by identifying diferent performances in hosts with aparently same specs, start wondering and talking with each other (here in the forums) in order to uncover the real responsables for the possible performance spectrum within this hosts...

Having sayd that, you may look for typical procesor, says Intel P4 with HT, D series and Core 2 Duo and AMD 64 and X2 (perhaps mobile CPU). Some are covered in my table (see signature) to give some idea on how is each CPU doing.
It would be far from accurate saying that Core 2 Duo is close to best choice in overall effort.


By that last sentence you meant "It would NOT be far from accurate...", right?
PovAddict
BAM!ID: 115
Joined: 2006-05-10
Posts: 1013
Credits: 1,824,431
World-rank: 134,291

2007-01-31 15:45:44

First, some people do overclocked and frequency hence performance may vary abot 30 or even 50%


So what? Then we could see how much the overclock can boost the crunching by watching the variaton within hosts with same processors...

No you couldn't. The client isn't reporting the frequency at which the CPU is running, so the projects don't know that, and so there is absolutely no way for the stats sites to know it.
Pedro Mariano Yunes Garcia
BAM!ID: 7913
Joined: 2006-10-06
Posts: 20
Credits: 793,966
World-rank: 218,324

2007-01-31 16:00:47
last modified: 2007-01-31 16:02:18

First, some people do overclocked and frequency hence performance may vary abot 30 or even 50%


So what? Then we could see how much the overclock can boost the crunching by watching the variaton within hosts with same processors...

No you couldn't. The client isn't reporting the frequency at which the CPU is running, so the projects don't know that, and so there is absolutely no way for the stats sites to know it.


Ok, but then, the list will show two hosts (example) with same processors and all specs available, but with diferent outcomes, diferent performances (since one is O.C.ed and the other one is not). We can now get in touch with the best perf. host/user and ask him what he is doing that makes his numbers higher... With some luck, we can find out it is an O.C.ed sistem, and then, maybe, get ours O.C.ed too.

I think you're missing the point here. The idea isn't to prove which sistem/spec is the faster, or anything like that. The goal is to make it posible to people to compare performances between hosts with same parts...
Pedro Mariano Yunes Garcia
BAM!ID: 7913
Joined: 2006-10-06
Posts: 20
Credits: 793,966
World-rank: 218,324

2007-02-26 17:59:57

Willy, did you read this? What's your opinion on this suggestion?
Pages: [1]

Index :: Comments and suggestions :: List/rank of users/hosts with same processors
Reason: