Great bunch of questions ! I think the following references will answer all your questions in their entirety - of course, each one with its own POV. They, unfortunately, don't get deeply technical.
1. The first is one from AnandTech on the performance of Intel and AMD CPUs for Half Life 2. The link was originally posted by Aces here (Thanks Mate). But that thread went down the obvious route of - well, you guessed it - Intel vs AMD party time . A section in that review which caught my attention was on the AMD 64 memory subsystem, particularly as it relates to the a) impact of bandwidth on Socket 754 vs Socket 939; b) limpact of loss of bandwidth when using a divider on a Winchester c) memory timings. Specifically look at the impact on the test at_c17_12 which is the most CPU dependent.
2. The second one is a long running thread in AnandTech forums by a person named Zebo. The evidence he presents and his conclusions are pretty damning. It's an interesting read for sure and it's here:
Also, evaluate (2) above in light of AnandTech's conclusions at the very end.
Hope this helps to some extent !
PS: A Big Hi to you too Crazy_Eddy
First post here. My 2.4C reaches 3 Ghz pretty easily and even with this overclock and while running super Pi, the temp does not cross 30C. See the first screenshot. Northwoods run cool and the prolonged winter of the northern part of the country helps
But the surprising part was that even with a clockspeed of 3 Ghz, the super pi 2 million takes about 1 minute 50 seconds. The AMDs are scoring much better here. It looked a bit odd and so I ran the windows task manager to monitor CPU performance while running the super PI. See the second screenshot. The screenie shows that the super-Pi is stressing only half the CPU. It appears to me that unless one disables Hyperthreading of HT enabled P4s, Super Pi will always give underbiased results for these processors.
Sorry about the confusion. Some differences there. In the second screenie, it is 1 million super Pi count and it was done without OC. The first one shows results for 2 million and with OC. I posted the 2 million count because earlier posts in this thread have used that benchmark. The results between the two screenshots are not comparable. HT was on in both cases.
The second screenie was more to show that only half the processor is stressed while running super Pi.
Bookmarks