Replies to post #42056 on Advanced Micro Devices Inc (AMD)

[pgerassi]

Dear Upc:

Emperical results trump on paper analysis. Scientists use experiments to prove and disprove theories. Michaelson and Morley set out to find ether drift by their inferometer experiment and ended up proving that ether doesn't exist. Intel's Prescott documentation states that Prescott doesn't use all that much power than AMD K8's stated power usage. But when looked at by experiment, the difference is significant and large. Whether one looks at idle or under load, Prescott based systems use more power.

How much is due to the CPU? Most of it. Power supply and NB do not take more than 40% or so worst case. CPU fan doesn't use more than 4W, so 68W*60%-4W = 30W minimum. More likely 68W*75%-4W = 47W typical. That is significant when compared to the 90-115W TDPtyp of Prescott. The actual TDPtyp of Athlon 64 would be in the range of 43-68W, well below the family power envelope of 89W.

I take emperical (experiment) based results over paper. Due diligence is needed for both anyway. Look at what happened at Anandtech for good reasons why.

Pete

[wbmw]

UPC, thanks for proving my point with regards to the AMD systems using a DBS systems. It seems to save the Athlon 64 35-45W. I bet it would save the Prescott system even more.

However, I still don't trust the results. The numbers claim a ~10W difference between Athlon 64 idle and "max load", but that doesn't sound right. Still no description of the tools used or the methodology. Just numbers in tables, but I'm sure there are morons out there that will take that as "empirical evidence".