Re: Your post is complete FUD. The 2nd sentence is a complete lie: "there are not enough peak scores to do a thorough comparison." LOL, on the AMD side, DELL is the only vendor to not give peak scores. They are not the fastest AMD machines. On the Intel side, 143 of 164 scores include PEAK, including 100% of the 3 GHz results. The first missing PEAK score is ranked #33 out of the 2.66 GHz SPECint_rate scores! QED.
More FUD from you, Petz. It wasn't a lie. I started my post with the intention of keeping all the scores related to a single vendor, and for SPECfp_rate, many of the submissions were base only. But it turns out either way (base or peak), you end up with the same result! Check your math before shooting your mouth off, next time.
Re: You pick and choose the SPEC benchmarks you compare. In 4 of 8 comparisons, you even compare dissimilar systems. I picked the fastest scores. I picked identical systems, submitted on the same date.
No you didn't. You showed Clovertown scalability in isolation and with a single benchmark (SPEC_CPU2006) and proclaimed that it sucked. You didn't bother to compare it to Woodcrest or to dual core Opteron, which I bothered to do. And the results show that AMD doesn't have an advantage in SPECint_rate, and only a small advantage in SPECfp_rate. The only conclusion is that SPEC_CPU2006 doesn't scale well with frequency - ON ANY PLATFORM!
Re: I never claimed that AMD will scale perfectly, but scaling 50% in SPECint_rate and 36% in SPECfp_rate is quite poor.
Take your foot out of your mouth and see for yourself what AMD scores. |
