What benchmarking program is most similar to Rosetta?

Message boards : Number crunching : What benchmarking program is most similar to Rosetta?

To post messages, you must log in.

AuthorMessage
far

Send message
Joined: 13 Jul 06
Posts: 6
Credit: 18,056,198
RAC: 0
Message 35367 - Posted: 23 Jan 2007, 10:21:07 UTC

Hi,

I have an overclocked E6600 running at 3.3GHz and very stable (well tested on Prime95). The memory was running at 1:1 = 366MHz. It was doing about 1100 units a day.

I uppped the ram voltage and changed the ratio to 2:3 which means the CPU is still at 3.3GHz but the memory is now running at 550MHz. It is stable as verifed by extended Prime95 tests. However it is now doing only about 900 units a day.

Does anyone have any ideas? Or maybe be able to suggest which benchmark program is closest to what Rosetta is doing so that I can do some testing myself and work out how this can be?

Thanks,
Herb
ID: 35367 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Astro
Avatar

Send message
Joined: 2 Oct 05
Posts: 987
Credit: 500,253
RAC: 0
Message 35379 - Posted: 23 Jan 2007, 12:15:32 UTC
Last modified: 23 Jan 2007, 12:16:56 UTC

UM, errrrrrr, The benchmark is done by Boinc, Not Rosetta. See Boinc website and search for benchmark.

More importantly, Rosetta is the only project where the user/participant PICKS the run time for each result (wu). If you click on "participants", then click on "rosetta prefs", you can pick your run times. I.E "not Selected" = 3 hours/ result, or 8 per day (16 if dual core).

Hmmm, by "Units", I assumed you meant "work units", did you mean credit?
ID: 35379 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Profile dcdc

Send message
Joined: 3 Nov 05
Posts: 1828
Credit: 107,774,731
RAC: 13,631
Message 35380 - Posted: 23 Jan 2007, 12:16:34 UTC

running memory out of sync often has a negative impact. It could also be due to the specific tasks you ran though.

At least, that certainly used to be the case, and I believe it still to be true, although modern prefetchers may have negated this?

As for a benchmark that is similar to Rosetta, it'd have to be something that is FPU intensive while using around 1MB cache. Don't have any idea what would be suitable, but standard whetstone/dhrystone tests are too small.
ID: 35380 · 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 35404 - Posted: 23 Jan 2007, 17:24:18 UTC

I agree with DCDC, that the most likely scenario is just that your machine is running various work-units, some of which are "easier" and others that are "harder" for your particular machine. My "system average credit" looks a bit like a cross-section of a mountain range, and I think that's perfectly natural.

Of course, running memory out of sync can add delays to the memory reads, so it may not give you any improvement.

--
Mats
ID: 35404 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
far

Send message
Joined: 13 Jul 06
Posts: 6
Credit: 18,056,198
RAC: 0
Message 35434 - Posted: 24 Jan 2007, 7:13:12 UTC

Its not a short term curiousity - its a consistent trend observed over the last few weeks compared with how it was running for the previous half year. I would expect the short term average credit to fluctuate a lot but not suddenly drop for several weeks. Even my monthly average has dropped 10%. The machine has 2G of ram of which it rarely uses more than 700Mb.

I have a feeling that Mats is right and its something to do with the memory being out of sync adding delays, tho I can't find any clear answer from searching.

Strange stuff.
ID: 35434 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Profile Who?

Send message
Joined: 2 Apr 06
Posts: 213
Credit: 1,366,981
RAC: 0
Message 35452 - Posted: 24 Jan 2007, 16:53:08 UTC - in response to Message 35434.  
Last modified: 24 Jan 2007, 16:53:29 UTC

When i look at the ASM code of BOINC BENCH, i see a very old code, nothing to take advantage of your Core 2 (even if it beat all the other)

I hope to see BOINC recompiled with modern compiler flags soon.

who?
ID: 35452 · 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 35460 - Posted: 24 Jan 2007, 19:21:10 UTC - in response to Message 35452.  

When i look at the ASM code of BOINC BENCH, i see a very old code, nothing to take advantage of your Core 2 (even if it beat all the other)

I hope to see BOINC recompiled with modern compiler flags soon.

who?


BOINC is just compiled with pretty normal Visual Studio.NET 2005 options. This is for compatability with all system, though when I compiled it with SSE or SSE2 and crosslinking, fpu fast (inlining I cannot remember) it all still works fine on my computers the benchmark jumps up and down but that is irrelevant to me. 'Boinc' doesn't do a lot to be honest it just keeps everything in check, boincmanager on the other hand is snappier with these compile options but then most people wouldn't notice or care about that since all it does is display a bit of info and give you some control options.

If you want to know which, just download the boinc source code http://boinc.berkeley.edu/source_code.php


Team mauisun.org
ID: 35460 · 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 35475 - Posted: 25 Jan 2007, 1:19:02 UTC

Back in the 2002 range, when we had 200/266Mhz buses (100/133 DDR) - there was a lot more talk about sync/async issues - along with discussions over which slots on a nVidia chipset motherboard offered the best performance with various memory options. There were noticeable changes on various memory benchmarks back then - between sync and async. So you might want to look at older reviews.

Perhaps taking a look at the DDR ram articles on the tech site (drawing a blank at the moment) that has wonderful articles on CPU makeup. i.e. showing the difference between the PIII, the P4, the Athlon XP, and the Athlon 64 cpus as each appeared on the market.


ID: 35475 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote

Message boards : Number crunching : What benchmarking program is most similar to Rosetta?



©2022 University of Washington
https://www.bakerlab.org