Advanced Micro Devices Inc (AMD)

[wbmw]

Neye, Re: As for IPF, I am still surprised that last year's chip continues to outperform AMD's yet to be released Opteron 250 on several benchmarks, still.

>> I surprised that a puny little Opteron continues to outperform the big bad 6MB cache 1.5Ghz I2 on the same software platform. what a disgrace.

I understand the sarcasm here, but you are missing the time element. When Madison launched, it creamed the Opteron 244. Later, it made the brand new Opteron 246 look pretty bad. Even later, the now current Opteron 248 got closer, but still loses. It actually took AMD a year and 600MHz worth of speed grades to beat Madison. And all it will take from Intel is one more speed grade, 1.7GHz with 9M of cache, to outperform anything that AMD has this year. That is not disgraceful at all.

Re: True, and aHT isn't supposed to increase in speed (oh, wait, it is), never use higher clocked memory (oh, wait, it is), and no apps will ever benefit from being compiled to 64bit native for x86-64 (oh wait, they will). sounds pretty ambiguous to me, not a clear cut win for IPF.

AMD isn't standing still, but however far they go with aHT and memory speeds, Intel will be able to match it eventually, with a chance of exceeding it. Right now, with a 400MT/s FSB and DDR266 memory, IPF is not just one generation behind, but more like 2 or 3. Yet in spite of that, the performance is quite good. I imagine a scenario where a future IPF CPU has interprocessor bandwidths at least as large as AMD with aHT 2.0, if not larger. I imagine it might make quite a bit of difference to go from 2+ generations behind to 1 generation ahead.