Hi, I can report that this problem has now been resolved. While trying (and actually failing due to piss-poor SATA chipset driver-support!) to install Windows 7 on another partition earlier today, I was struck by similar slow-speeds as the 2.6.30-1-amd64 kernel. Googling revealed that my chipset for some reason has great trouble when all 4 memory banks are used, and L1/L2 cache speeds deteriorate. ( http://forums.anandtech.com/messageview.aspx?catid=83&threadid=2059915&enterthread=y&STARTPAGE=2 , post at "01/13/2008 09:26 PM" - I didn't confirm this) http://windows7forums.com/windows-7-installation-upgrade/1388-windows-7-slow-install-wont-boot.html http://communities.intel.com/thread/3197 ) This had been addressed by my mainboard manufacturer a while ago ( http://www.gigabyte.com.tw/Support/Motherboard/BIOS_Model.aspx?ProductID=2546 ), but since I'd never had problems before I had not updated my BIOS. An upgrade from F2 to F8 have resolved the issue. Not knowing for certain that L1/L2 speed degradation was the exact cause of the problem, it's interesting how it obviously did not occur when I ran the 2.6.26-2-amd64 kernel. I do notice another type of difference though. The 2.6.26-2-amd64 kernel gave me somewhere along 7600000 KB of available memory, according to top. 2.6.30-1-amd64 now report the full 8 GB (4x2GB) I have. I do not know if it's related. Perhaps there was some memory mapping error of some sort, when attempting to use all memory. Regards, -- Martin Millnert <martin@millnert.se>
Attachment:
signature.asc
Description: This is a digitally signed message part