| From: Mikkel L. Ellertson <mikkel@xxxxxxxxxxxxxxxx> | That sounds like a version of this problem: (from memory.txt in the | kernel docs...) | | 3) There are some motherboards that will not cache above | a certain quantity of memory. If you have one of these | motherboards, your system will be SLOWER, not faster | as you add more memory. Consider exchanging your | motherboard. As far as I know, that problem dates back to the Pentium 1 days. It probably isn't true of any modern systems. It might be the case for some CPUs designed for embedded applications and the like. I imagine that it only applies to systems with external (to the CPU module) cache. So it should not apply to Pentium II or later Intel chips and K7 or later AMD chips. (I'm typing this in on an AMD K6-200 system (roughly a decade old). Its motherboard uses an Intel TX chipset which has exactly this limitation: nothing over 64M would be cached. Luckily, I only have 64M. Motherboards of that era based on the Intel HX chipset sometimes had slots for specialized memory modules to extend the cache tags.) http://www.access-one.com/rjn/computer/cache64m.txt There are other odd memory limitations. For example, my notebook has a Broadcom 802.11g chipset that cannot DMA into memory above 1G. Undocumented (but what isn't undocumented about that chipset?). http://lists.pcxperience.com/pipermail/linuxr3000/2005-March/005085.html ================ Sometimes machines run slowly due to being swamped by unhandled interrupts. It would be interesting to monitor /proc/interrupts to see if this is happening. ================ My HP Pavilion a1250n desktop (Athlon 64 x2 3800+, ATI Radeon XPress 200 chipset) had some problems with Linux in the early days. The clock ran twice as fast due to some miswiring of the APIC (I think). Recent kernels have a workaround. As a purist, I really dislike the fact that Linux has to work around bugs rather than expect manufacturers to fix them. http://bugzilla.kernel.org/show_bug.cgi?id=3927
