Questions and Answers : Windows : What is the fastest CPU in climateprediction?
Message board moderation
Author | Message |
---|---|
Send message Joined: 8 Mar 05 Posts: 17 Credit: 114,685 RAC: 0 |
AMD 64/XP? Intel P4/Xeon HT Enable/Disable? I want to run \"hadsm\" in new PC. Thanks. |
Send message Joined: 5 Feb 05 Posts: 465 Credit: 1,914,189 RAC: 0 |
Right now mostly Sulfur units are being sent out. As for fastest, go to stat sites and examine what crunches the fastest. There are plenty of them out there, and they show a lot of good information. Sulfur units take a lot longer than HADSM units, because they are 5 sections, and a lot more data, but the credits refelct that, also. |
Send message Joined: 7 Aug 04 Posts: 2187 Credit: 64,822,615 RAC: 5,275 |
Tough question. Up until the last month or so, the answer for a single model would be a highly clocked Pentium 4 running Linux. For total throughput it would have been a highly clocked multiple Xeon workstation running Linux. For the Pentium 4, running hyperthreading helps throughput. For a multiproc Xeon, it does not appear to help much if at all. In Windows only, the above would also be true. Recently, with the latest compilation of sulphur, in WinXP this has changed some. A highly clocked Athlon64 would probably have the fastest single model runtime while a multi CPU highly clocked Opteron system would have the best throughput (most model years in a given period of time). Of the dual core CPUs, the Athlon64 X2 4800+ would probably beat the high end dual core Intel CPUs. In Linux, the new version of sulphur is problematic. It\'s a little faster than the old version on AMD, but much slower on Intel than it used to be. It\'s going to be recompiled so that the Intel performance will again be good. But since this is in the Windows forum, the 2nd paragraph applies to your question. |
Send message Joined: 27 Aug 05 Posts: 156 Credit: 112,423 RAC: 0 |
To get you started Running a Dell 840ee HT enabled 3.2 3 gig 667m ram win xp Time Sent (UTC) Host ID Result ID Result Name Phase Timestep CPU Time (sec) Average (sec/TS) 19 Dec 2005 17:59:52 257345 1361519 sulphur_eopw_000685220_0 1 172,832 728,287 4.2138 19 Dec 2005 15:01:47 268890 1435915 sulphur_fnqr_000730611_0 1 140,426 631,403 4.4963 19 Dec 2005 10:49:54 268890 1299058 sulphur_ddae_000623750_0 2 183,634 1,876,103 4.2361 19 Dec 2005 09:21:01 257345 1309209 sulphur_dl3m_000633874_0 2 75,614 1,426,239 4.2592 19 Dec 2005 06:20:12 268890 1298737 sulphur_dd1h_000623429_0 2 129,624 1,664,991 4.2816 19 Dec 2005 05:20:42 257345 1361519 sulphur_eopw_000685220_0 1 162,030 684,713 4.2258 19 Dec 2005 05:20:35 268890 1306022 sulphur_dinf_000630699_0 2 43,208 1,302,023 4.3048 19 Dec 2005 01:19:55 268890 1435915 sulphur_fnqr_000730611_0 1 129,624 582,381 4.4928 18 Dec 2005 21:20:55 268890 1299058 sulphur_ddae_000623750_0 2 172,832 1,828,771 4.2325 18 Dec 2005 20:27:39 257345 1309209 sulphur_dl3m_000633874_0 2 64,812 1,382,406 4.2659 18 Dec 2005 16:49:09 257345 1361519 sulphur_eopw_000685220_0 1 151,228 641,329 4.2408 18 Dec 2005 16:48:58 257345 1265032 1j56_100092200_1 3 259,248 1,744,707 2.2433 18 Dec 2005 16:20:31 268890 1298737 sulphur_dd1h_000623429_0 2 118,822 1,617,133 4.2773 18 Dec 2005 15:20:18 268890 1306022 sulphur_dinf_000630699_0 2 32,406 1,254,361 4.3009 18 Dec 2005 10:49:17 268890 1435915 sulphur_fnqr_000730611_0 1 118,822 533,862 4.4930 18 Dec 2005 09:48:54 257345 1265032 1j56_100092200_1 3 248,446 1,719,921 2.2426 18 Dec 2005 07:48:57 257345 1309209 sulphur_dl3m_000633874_0 2 54,010 1,337,101 4.2684 18 Dec 2005 07:20:04 268890 1299058 sulphur_ddae_000623750_0 2 162,030 1,780,982 4.2276 18 Dec 2005 02:57:32 268890 1298737 sulphur_dd1h_000623429_0 2 108,020 1,569,365 4.2731 18 Dec 2005 02:49:10 257345 1265032 1j56_100092200_1 3 237,644 1,694,989 2.2416 BOINC Wiki |
Send message Joined: 19 Apr 05 Posts: 53 Credit: 6,325,436 RAC: 0 |
other than crazy XEON/Opteron multi-processor systems...... a O/C\'d DC A64 or socket 939 opty would be the fastest my DC opty 175 is the fastest machine on the spin-up model currently but it is O/C to 2.85 gHz................... |
Send message Joined: 8 Mar 05 Posts: 17 Credit: 114,685 RAC: 0 |
http://climateapps2.oucs.ox.ac.uk/cpdnboinc/cpu.php I saw this page. AMD Athlon 64 3000+ is the fastest? (without MAC) (Macintosh\'s CPU is very fast in integer and float!!!) And, Vendor is AMD Athlon 64 Model is Pentium What does it mean? \"Pentium\" is Intel\'s Trademark. Thank you for the answer. |
Send message Joined: 7 Aug 04 Posts: 2187 Credit: 64,822,615 RAC: 5,275 |
The results you saw at that link were run under an old version of BOINC (thus the ridiculous labelling of Pentium for AMD processors), with an old version of climateprediction application (the relative speeds of one CPU vs. another has changed recently). Also, at the top of the list are a bunch of PCs that weren\'t really that fast. Some type of error. As for the BOINC benchmark numbers (MFlop, MIop), older versions of BOINC gave higher numbers than newer versions, and Windows gives higher numbers than Linux. Yet, those benchmark numbers have little to nothing to do with cpdn performance. |
Send message Joined: 8 Mar 05 Posts: 17 Credit: 114,685 RAC: 0 |
Thank you, geophi. AMD is faster than Pen4 in float. Which power does it need. float or integer? |
Send message Joined: 7 Aug 04 Posts: 2187 Credit: 64,822,615 RAC: 5,275 |
AMD is faster than Pen4 in float. Float, but memory bandwidth and latency are also very important. In the early days of BOINC CPDN, Intel was faster because the Intel compiler wouldn\'t allow SSE and SSE2 on AMD CPUs. The latest versions now do and AMD is faster. |
Send message Joined: 31 Oct 04 Posts: 336 Credit: 3,316,482 RAC: 0 |
For Pentium M (not Pentium 4M), CPDN is the ideal project, they perform much better than expected from the benchmark results. They are probably not the top performers but sure amongst the best - and they must have a very good Trickle/Watt ratio. My not overclocked Banias 1600 needs 2.38 sec/TS for a HadSM3 Slab model and that\'s one of the slow Pentium M. |
Send message Joined: 5 Sep 04 Posts: 7629 Credit: 24,240,330 RAC: 0 |
Lyrical-Nanoha No one has mentioned it yet, but there aren\'t any hadsm at the moment. It\'s all sulphur_um. Longer and slower. And soon to come: the coupled ocean model. |
Send message Joined: 5 Sep 04 Posts: 7629 Credit: 24,240,330 RAC: 0 |
What I should have added was, make sure that the computer is stable. A large heatsink/fan, lots of airflow through the case, and a large power supply so that the voltages don\'t sag under continuous use. All just as important as speed. |
Send message Joined: 8 Mar 05 Posts: 17 Credit: 114,685 RAC: 0 |
Thank you. I\'ll buy new PC and I\'m going to run CPDN. |
©2025 cpdn.org