Ephud:
Testing a future, if not cancelled, CPU against current available CPUs on contrived tasks that don't measure all work done at current prices matching some future price, is not called being trounced, but sheer cheating. head to head, that future CPU in the common benchmarks loses most of the tests, on true multitasking as done by real people, loses more tests. And they didn't even test against dual Opteron, either two socket, which is available now, or dual core single socket which will be available in less than two weeks. And, oh look, given current pricing trends, it should have been tested against a A64 3800+ or 4100+ (2.6GHz 512KB L2) since 2.8 GHz, and even possibly, 3.0GHz gets released by then.
Oh, that future CPU would be beaten so soundly. Given that, the reviewers were kept from doing anything of that sort. Kind of makes those reviews quite valueless. As others have stated that. A rant by a well known person was posted earlier which, of course, you deliberately forgot 5 seconds after reading it. And it wouldn't do to repost it, because you will forget to look, as it trashes your views.
And I did notice that the standard "trounced" didn't include your standard SPECint_base or SPECfp_base, even though the peak scores really matter. What happened, did it get bad 2 core numbers? Perhaps all the numbers not included were ones where it was so bad, that Intel couldn't bear or was too embarrassed to have them see the light of day.
Pete |
