Questions and Answers :
Unix/Linux :
CPU Benhmarks for RH Linux vs XP on Identical Machine Dual Boot
Message board moderation
Author | Message |
---|---|
Send message Joined: 13 Nov 05 Posts: 4 Credit: 87,385 RAC: 0 |
Running the Climate Prediction Models on Dell OPtiplex 733 mHz cpu, Windows XP reports a much higher bechmark than RH Linux. However, the work accomplished on Linux seems to be much more. The important difference between the machines is the disk drives - Windowns has a newer drive, Linux is using the older drive. Otherwise, the machine reesources are identical. Do you have data on the releative performance on the XP vs Linux platforms? Bottom Line - elecctricity costs and creates CO2 when running on different speed platforms. I would like to finish as expeditiously as possible. I gather you would also like to finish. |
Send message Joined: 7 Aug 04 Posts: 2167 Credit: 64,487,091 RAC: 4,506 |
On Intel platforms, Linux has always been faster for the same PC, at least until the sulphur 4.22 release a few days ago. The benchmarks have always been faster in Windows. BOINC benchmarks do not equate in any way to CPDN performance. |
Send message Joined: 18 Dec 05 Posts: 10 Credit: 5,184,083 RAC: 0 |
On Intel platforms, Linux has always been faster for the same PC, at least until the sulphur 4.22 release a few days ago. The benchmarks have always been faster in Windows. BOINC benchmarks do not equate in any way to CPDN performance. I have just benchmarked a dual-boot Athlon 2000+ with 512MB RAM, both are installed to different partitions on the same HDD. Win2k benchmarks are 1396 Whetstones & 2347 Dhrystones, but Kubuntu 5.10 gives me 767 Whetstones & 1330 Dhrystones. The predicted completion times are about 1492 hours for Win2k & about 2700 hours for Kubuntu. So it would seem that the benchmarks in my case DO seem to equate to predicted performance, or is there some basis for saying that benchmarks don\'t equate to CPDN performance ? I guess I\'ll get through more WUs & get more credits if I go back to Win2k. Shame really. |
Send message Joined: 7 Aug 04 Posts: 2167 Credit: 64,487,091 RAC: 4,506 |
On Intel platforms, Linux has always been faster for the same PC, at least until the sulphur 4.22 release a few days ago. The benchmarks have always been faster in Windows. BOINC benchmarks do not equate in any way to CPDN performance. The predicted time to completion from the BOINC clients since 4.7x has been terrible for climateprediction. There is no way that it would take twice as long in Linux as in Windows. The important measure is sec/TS on the built in cpdn graphics (or on the trickle listings). Comparing those will give you a relative measure of performance in Linux vs. Windows. But the latest executable (sulphur 4.22) does run faster on Windows than Linux. |
Send message Joined: 5 Aug 04 Posts: 1496 Credit: 95,522,203 RAC: 0 |
GreenBandit, For more than two years I\'ve run my dual-boot boxes in Linux (save one, for personal reasons)* despite the horrid numbers in Linux Benchmarks as compared to WinXP on a similar box. All that the Linux boxes did was consistently out-perform the similar M$ box with its higher Benchmarks. (In short, my worst-performing box was the one running XP, better Benchmarks notwithstanding.) Geophi is correct. Benchmarks are meaningless for CPDN. (Note that you are awarded credit for work done in CPDN, not for how your machine\'s work compares to some other other machine running the same Work Unit.) CPDN client 4.22 sets all that on its ear (Edit: for AMD vs. Intel, M$ vs. Linux) -- at least in the short term. Stay tuned. Hope your motor bikes are 4-stroke, not 2-stroke, pollution-wise... Seasons Greetings. * Also running Spinup in XP, but that\'s a different story. "We have met the enemy and he is us." -- Pogo Greetings from coastal Washington state, the scenic US Pacific Northwest. |
©2024 climateprediction.net