A> A. P. Kennedy wrote: >> ****************** Turion below but with 32 bit kernel *********** >> # /usr/lib/atlas3-test/xzinvtst >> NREPS ORDER UPLO N LDA TIME MFLOP RESID >> ===== ===== ===== ===== ===== ======== ======== ============ >> 0 Col GE 100 100 0.008 996.37 6.066196e-03 >> 0 Col GE 200 200 0.045 1419.78 5.083446e-03 >> 0 Col GE 300 300 0.154 1401.06 4.577686e-03 >> 0 Col GE 400 400 0.353 1449.29 5.339619e-03 >> 0 Col GE 500 500 0.684 1461.11 4.390384e-03 >> 0 Col GE 600 600 1.172 1473.71 4.389517e-03 >> 0 Col GE 700 700 1.865 1470.75 4.753297e-03 >> 0 Col GE 800 800 2.793 1466.06 5.002779e-03 >> 0 Col GE 900 900 3.912 1490.02 5.070376e-03 >> 0 Col GE 1000 1000 5.278 1515.10 5.078879e-03 >> 10 cases: 10 passed, 0 skipped, 0 failed >> # cat /proc/cpuinfo processor : 0 >> vendor_id : AuthenticAMD >> cpu family : 15 >> model : 36 >> model name : AMD Turion(tm) 64 Mobile Technology ML-32 >> stepping : 2 >> cpu MHz : 1800.307 >> cache size : 512 KB >> fdiv_bug : no >> hlt_bug : no >> f00f_bug : no >> coma_bug : no >> fpu : yes >> fpu_exception : yes >> cpuid level : 1 >> wp : yes >> flags : fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca >> cmov pat pse36 clflush mmx fxsr sse sse2 syscall nx mmxext fxsr_opt lm >> 3dnowext 3dnow pni lahf_lm >> bogomips : 3601.38 >>
A> It would be great is you were able to run these tests using a 64 bit A> kernel and a 64 bit version of atlas3. It seems that there is a huge A> performance win for atlas under 64 bit, as I cannot imagine that the A> difference in 2. level cache (one half of that of a turion MT) can A> give this huge difference in MFlops. A> Gathering some of the data (Dual system excluded): A> MFLOPS MHz MFLOPS/MHz System A> 3113.9 2010 1.55 Opteron A> 2879.13 2009 1.43 Std amd64 A> 3367.67 2556 1.33 Newcastle A> 2148.33 1607 1.34 Turion MT A> 3317.38 2411 1.38 Std amd64 A> 1515.1 1800 0.84 Turion ML - 32 bit A> Reading this table (disregarding the 32 bit test) it seems that the A> Turion processors is just as fast as a standard opteron in terms of A> MFlops/MHz. The opterons beat every other AMD based CPU in this A> regard, which would be expected. A> One should note that the Turion MT-30 has a max power consumption of A> 25 Watt, where as the Opteron tops at 89 Watt IIRC - Huge gain in A> FLOPS / Watt when using a turion processor. A> Regards A> Anders Fugmann I saw that also. Even though this is my wife's computer, I will have to install 64bit also. The funny part was I paid 700 bucks US dollars for this acer 5000 series laptop, but the first one was broken which had an ML-30 @1024 Kb cache, 1.6 Ghz) processor. The store gave me the next model up which had the ML-32 (512 Kb cache, 1.8 Ghz) for the same price. Looked like a great deal. Now I wonder with those benchmarks. I will install 64 bit debian this weekend and get back to this list. I did turn off powernow before running the tests. For the record it is plenty fast for what my wife is doing (reading e-mail and the newspaper). Heck it never goes above the lowest cpu speed with powernow turned on. Now I wonder if a celery-M laptop would have been good enough ;-) Later, Alan -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]
