RAC dropping

Message boards : Number crunching : RAC dropping

To post messages, you must log in.

1 · 2 · 3 · 4 · Next

AuthorMessage
truckpuller

Send message
Joined: 5 Nov 05
Posts: 40
Credit: 229,134
RAC: 0
Message 25897 - Posted: 2 Sep 2006, 16:53:48 UTC

One of my computers has dropped from 245 Rac down to like 220 Rac. This is all this machine does is crunch Rosetta, no gaming or anything else it's a 1.6 AMD duron overclocked to 2.0 and 512MB ram was just wondering what would cause this to happen.

Thanks in advance
Visit us at Christianboards.org
ID: 25897 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
FluffyChicken
Avatar

Send message
Joined: 1 Nov 05
Posts: 1260
Credit: 369,635
RAC: 0
Message 25899 - Posted: 2 Sep 2006, 17:30:46 UTC - in response to Message 25897.  
Last modified: 2 Sep 2006, 17:34:52 UTC

One of my computers has dropped from 245 Rac down to like 220 Rac. This is all this machine does is crunch Rosetta, no gaming or anything else it's a 1.6 AMD duron overclocked to 2.0 and 512MB ram was just wondering what would cause this to happen.

Thanks in advance


Most probably the new credit system, I doubt RAC will be as steady as before, that and afaik we don't get credit for 'failed' tasks.
(Anyone know of the crediting stsatus of failed tasks ?)
P.S. It's not failed task in your case but the new credit system, looking at your systems the Sempron and AthlonXP will go up slightly looking at the result returned.
Team mauisun.org
ID: 25899 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
J D K
Avatar

Send message
Joined: 23 Sep 05
Posts: 168
Credit: 101,266
RAC: 0
Message 25902 - Posted: 2 Sep 2006, 17:40:43 UTC

ID: 25902 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
truckpuller

Send message
Joined: 5 Nov 05
Posts: 40
Credit: 229,134
RAC: 0
Message 25907 - Posted: 2 Sep 2006, 20:42:07 UTC - in response to Message 25899.  

One of my computers has dropped from 245 Rac down to like 220 Rac. This is all this machine does is crunch Rosetta, no gaming or anything else it's a 1.6 AMD duron overclocked to 2.0 and 512MB ram was just wondering what would cause this to happen.

Thanks in advance


Most probably the new credit system, I doubt RAC will be as steady as before, that and afaik we don't get credit for 'failed' tasks.
(Anyone know of the crediting stsatus of failed tasks ?)
P.S. It's not failed task in your case but the new credit system, looking at your systems the Sempron and AthlonXP will go up slightly looking at the result returned.


Well most all of my systems are running right around the same overclock which is around 2.oGhz give and take.Just seemed strange that they are all running around the same speed and to have that much differance in RAC's. So then apparently overclocking doesn't matter then?? which i know has been a heated debate.

Thanks again TP

Visit us at Christianboards.org
ID: 25907 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Mod.Tymbrimi
Volunteer moderator
Avatar

Send message
Joined: 22 Aug 06
Posts: 148
Credit: 153
RAC: 0
Message 25911 - Posted: 3 Sep 2006, 0:18:26 UTC

I read descriptions of the Duron calling it a crippled AthlonXP. If your AthlonXP at the same clock is out producing the Duron, they were telling the truth. Is there a source for AthlonXPs anymore? Other than eBay?


Rosetta Moderator: Mod.Tymbrimi
ROSETTA@home FAQ
Moderator Contact
ID: 25911 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
truckpuller

Send message
Joined: 5 Nov 05
Posts: 40
Credit: 229,134
RAC: 0
Message 25912 - Posted: 3 Sep 2006, 0:43:24 UTC - in response to Message 25911.  

I read descriptions of the Duron calling it a crippled AthlonXP. If your AthlonXP at the same clock is out producing the Duron, they were telling the truth. Is there a source for AthlonXPs anymore? Other than eBay?



According to Boinc Benchmarks

1.6 Duron
1787 double persision MIPS
3040 integer MIPS

Athlon xp 2500+

1715 Floating point MIPS
2882 integer MIPS

Maybe it is all due to the points system but just seemed odd that RAC dropped like it did.My sempron64 2800@2.0 ghz has a RAC of 281

I dont know of any other source for the Athlon xp's
Visit us at Christianboards.org
ID: 25912 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Ananas

Send message
Joined: 1 Jan 06
Posts: 232
Credit: 752,471
RAC: 0
Message 25914 - Posted: 3 Sep 2006, 1:00:47 UTC

Not many on stock anymore but some dealers still have a few : Athlon XP 2400+ Tray
ID: 25914 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
BennyRop

Send message
Joined: 17 Dec 05
Posts: 555
Credit: 140,800
RAC: 0
Message 25917 - Posted: 3 Sep 2006, 1:27:56 UTC
Last modified: 3 Sep 2006, 1:33:05 UTC

seconds old credit new credit new credit/hr
627,860.40 1,757.09 1,507.24 8.642150389 Duron
588,995.38 1,571.26 1,723.99 10.5372032 AthlonXP

200 credit difference a week between the Duron and AthlonXP. Looks like they've switched scores with the new system.

Do the 462 pin Semprons act more like the Duron or the AthlonXP?

Here's a list of companies selling their collections of cpus, including 462 pin chips. http://www.pricewatch.com/cpu/

ID: 25917 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
FluffyChicken
Avatar

Send message
Joined: 1 Nov 05
Posts: 1260
Credit: 369,635
RAC: 0
Message 25934 - Posted: 3 Sep 2006, 8:36:00 UTC - in response to Message 25917.  

seconds old credit new credit new credit/hr
627,860.40 1,757.09 1,507.24 8.642150389 Duron
588,995.38 1,571.26 1,723.99 10.5372032 AthlonXP

200 credit difference a week between the Duron and AthlonXP. Looks like they've switched scores with the new system.

Do the 462 pin Semprons act more like the Duron or the AthlonXP?

Here's a list of companies selling their collections of cpus, including 462 pin chips. http://www.pricewatch.com/cpu/


for each of the CPU's mentioned the only thing different is the cache size.
Duron = 128k
Sempron = 256k (not Sempron64 aka s754/939's of course)
AthlonXP = 256/512kb (the 2500+ 'barton' mentioned is 512k)

the Socket-A Semprons are just rebranded AthlonXP


It is often possible to modify the Duron and unlock to extra cache ;-)

If the speeds are both at 2.0GHz (i.e. both overclocked to it) then it's probably the Cache size of 128k of the Duron showing it's effect with Rosetta Work against the AthonXP's 512k, somthing you wouldn't see in the benchmarks.
Team mauisun.org
ID: 25934 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Profile dcdc

Send message
Joined: 3 Nov 05
Posts: 1829
Credit: 114,410,030
RAC: 54,612
Message 25964 - Posted: 3 Sep 2006, 21:14:46 UTC - in response to Message 25934.  

seconds old credit new credit new credit/hr
627,860.40 1,757.09 1,507.24 8.642150389 Duron
588,995.38 1,571.26 1,723.99 10.5372032 AthlonXP

200 credit difference a week between the Duron and AthlonXP. Looks like they've switched scores with the new system.

Do the 462 pin Semprons act more like the Duron or the AthlonXP?

Here's a list of companies selling their collections of cpus, including 462 pin chips. http://www.pricewatch.com/cpu/


for each of the CPU's mentioned the only thing different is the cache size.
Duron = 128k
Sempron = 256k (not Sempron64 aka s754/939's of course)
AthlonXP = 256/512kb (the 2500+ 'barton' mentioned is 512k)

the Socket-A Semprons are just rebranded AthlonXP


It is often possible to modify the Duron and unlock to extra cache ;-)

If the speeds are both at 2.0GHz (i.e. both overclocked to it) then it's probably the Cache size of 128k of the Duron showing it's effect with Rosetta Work against the AthonXP's 512k, somthing you wouldn't see in the benchmarks.


Yeah - i've noticed a similar drop on my Duron 1.6 (OCd) - the cache size doesn't affect the benchmark but does affect Rosetta throughput.
ID: 25964 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
_heinz

Send message
Joined: 30 Jun 06
Posts: 24
Credit: 38,697
RAC: 0
Message 26218 - Posted: 6 Sep 2006, 23:32:28 UTC

since 08/22 my RAC is falling like a rock ;-)
from 511 to 361 now
crunching 7/24

ID: 26218 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Mats Petersson

Send message
Joined: 29 Sep 05
Posts: 225
Credit: 951,788
RAC: 0
Message 26248 - Posted: 7 Sep 2006, 12:15:57 UTC

Reducing RAC is probably a combination of things:
1. If you have a processor with small cache (Duron/Sempron/Celeron), then cache-size will negatively affect the credit given by the new credit system (as it gives credit based on the actual number of models processed, which will be higher with a big-cache processor than a small-cache processor, at least to some extent).

2. Pentium 4 processors on Windows may actually get slightly less credit now than before, for the same reason of "actual work done vs. benchmark result". However, I looked at seti_britta's results, and on a quick scan, I didn't see any credit granted (i.e. new credit) that was lower than the claimed credit (credit in old crediting system). By the way, your OVERALL RAC is around 353 at the moment, with the Rosetta being the largest contributing factor on it, maybe it's some other project that has changed crediting or giving you less credit for the same amount of work?)

--
Mats
ID: 26248 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
BennyRop

Send message
Joined: 17 Dec 05
Posts: 555
Credit: 140,800
RAC: 0
Message 26285 - Posted: 7 Sep 2006, 17:31:58 UTC

https://boinc.bakerlab.org/rosetta/forum_thread.php?id=2212#25188

The P4s I compared were getting 1.6 to 1.86 times the standard client's benchmarks; while the one machine with HT on got 1.98 times the standard client's benchmarks.

Running on Linux or Windows didn't seem to matter. (Averaging all results turned in for a week.)
ID: 26285 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Profile dcdc

Send message
Joined: 3 Nov 05
Posts: 1829
Credit: 114,410,030
RAC: 54,612
Message 26287 - Posted: 7 Sep 2006, 18:13:54 UTC - in response to Message 26248.  
Last modified: 7 Sep 2006, 18:15:20 UTC

2. Pentium 4 processors on Windows may actually get slightly less credit now than before, for the same reason of "actual work done vs. benchmark result". However, I looked at seti_britta's results, and on a quick scan, I didn't see any credit granted (i.e. new credit) that was lower than the claimed credit (credit in old crediting system). By the way, your OVERALL RAC is around 353 at the moment, with the Rosetta being the largest contributing factor on it, maybe it's some other project that has changed crediting or giving you less credit for the same amount of work?)

yeah- it looks like the P4s have shown an increase in credit becasuse their benchmarks are too low.

We could do with a comparison of the throughput of CPUs and OSs. Devs/Mods: can we get an xml (or does one already exist?) showing:

Computer ID
WU Name
WU ID
No decoys produced
Result ID
Time taken

It would also be useful to have this info, but isn't vital:
CPU Type (I know the reported CPU isn't always right, particularly for Athlons)
OS
BOINC ver
Any other info?
ID: 26287 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Profile Buffalo Bill
Avatar

Send message
Joined: 25 Mar 06
Posts: 71
Credit: 1,630,458
RAC: 0
Message 26335 - Posted: 8 Sep 2006, 4:15:38 UTC

I've read that some small cache machines are getting lower granted credits but my Celeron D is getting about 30% more granted credits than claimed. I didn't expect that when the new credit system started. I was sure I'd get less.

Bill
ID: 26335 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Whl.

Send message
Joined: 29 Dec 05
Posts: 203
Credit: 275,802
RAC: 0
Message 26338 - Posted: 8 Sep 2006, 4:33:06 UTC

It was stated early on that CPU type, architecture and cache size etc would'nt matter with this new credit system. What happened to all of that then ?
ID: 26338 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
BennyRop

Send message
Joined: 17 Dec 05
Posts: 555
Credit: 140,800
RAC: 0
Message 26341 - Posted: 8 Sep 2006, 5:58:02 UTC - in response to Message 26335.  

I've read that some small cache machines are getting lower granted credits but my Celeron D is getting about 30% more granted credits than claimed. I didn't expect that when the new credit system started. I was sure I'd get less.

The P4s are getting 1.6-1.86 times the claimed credits; so if you're getting 1.3 times the claimed credits you're getting less than the P4s. The smaller cache Durons are getting less than the full cache Athlons.




ID: 26341 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
tralala

Send message
Joined: 8 Apr 06
Posts: 376
Credit: 581,806
RAC: 0
Message 26345 - Posted: 8 Sep 2006, 7:51:32 UTC - in response to Message 26338.  

It was stated early on that CPU type, architecture and cache size etc would'nt matter with this new credit system. What happened to all of that then ?


It was stated that the reported specs won't matter, since those data are easy to fake. Credit is granted according real work done and here CPU type, architecture, cache size and speed do matter. If that wouldn't matter a lowly P1@90Mhz would get the same credit as a super-fast Kentsfield, which would be certainly not fair.
ID: 26345 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Profile Saenger
Avatar

Send message
Joined: 19 Sep 05
Posts: 271
Credit: 824,883
RAC: 0
Message 26346 - Posted: 8 Sep 2006, 7:57:37 UTC - in response to Message 26338.  

It was stated early on that CPU type, architecture and cache size etc would'nt matter with this new credit system. What happened to all of that then ?

They are gone with the decision to base the credit system on the measure of real work done by the machines instead of their theoretical crunching capabilities. So the amount of credits depends on how good the machine suits the certain demands of the application and will be at least architecture sensitive, perhaps even OS-sensitive.

Both systems are fair in itself,
- counting real work like done now, and thus showing only what's good for the project regardless of the unused capabilities, or
- measuring the CPU-time x capabilities (benchmark) of a puter, thus measuring the ressources donated to the project, regardless of the use of this ressources.

In an ideal world, where it's easy to optimize the application to any architecture, both will be the same. In the real world, where the validity of the results is more important than crunching speed, they may and probably will differ.
ID: 26346 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Profile dcdc

Send message
Joined: 3 Nov 05
Posts: 1829
Credit: 114,410,030
RAC: 54,612
Message 26347 - Posted: 8 Sep 2006, 7:58:33 UTC - in response to Message 26338.  

It was stated early on that CPU type, architecture and cache size etc would'nt matter with this new credit system. What happened to all of that then ?

It was stated that all platforms would get the credit they deserve without favouritism for any particular setup, including CPU, OS, cache size etc. I think the point made about the old credit system was that the benchmark was very small and so fit inside any size cache, with the result being that cache size made no difference to the benchmark result. It was always expected that a larger cache would increase real-world throughput, and this is showing now.

Buffalo Bill's Celeron has shown a 30% increase in scores because the P4 architecture was hard-done-by under the old system. However, a P4 of the same speed would probably show more of an increase still as it has the advantage of extra cache and so can get more work done than the Celeron.

HTH
Danny
ID: 26347 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
1 · 2 · 3 · 4 · Next

Message boards : Number crunching : RAC dropping



©2024 University of Washington
https://www.bakerlab.org