Frank Cox wrote: > On Sat, 16 Feb 2008 16:09:28 -0500 > Bill Davidsen <davidsen@xxxxxxx> wrote: > >> It's a known limitation of some BIOS versions. I suspect it would happen >> with that other O/S as well, actually. > > After doing some Google-research, I think I have a bit of an understanding of > this issue. Someone who knows more about this, please correct me where my > understanding is wrong. > > Here is the contents of /proc/mtrr on the computer in question: > > reg00: base=0x00000000 ( 0MB), size=2048MB: write-back, count=1 > reg01: base=0x80000000 (2048MB), size=1024MB: write-back, count=1 > reg02: base=0xc0000000 (3072MB), size= 256MB: write-back, count=1 > reg03: base=0xcf800000 (3320MB), size= 8MB: uncachable, count=1 > reg04: base=0xcf400000 (3316MB), size= 4MB: uncachable, count=1 > reg05: base=0x100000000 (4096MB), size= 512MB: write-back, count=1 > reg06: base=0x120000000 (4608MB), size= 128MB: write-back, count=1 > reg07: base=0xd0000000 (3328MB), size= 256MB: write-combining, count=1 > > Now, it's my understanding that somehow, there is 64mb of ram that is not > properly accounted for in the above report, and that is what causes the problem > because this 64mb of ram is right at the top of the memory and the kernel > attempts to install itself at the top of the memory. This has the effect of > dragging everything that interacts with the kernel (meaning everything, period) > down to a snail's pace. > > (If someone could explain the meaning of the /proc/mtrr report, and how that > 64mb problem is derived from it I would appreciate it -- I've been looking it > over and don't understand this part of it at all.) > > The reason for this problem is that the bios is not correctly reporting or > accounting for the state of the last 64mb of ram in the computer. So what happens if you tell the kernel there is slightly less than 4GB of ram? A simple test would be to tell the kernel you have 3GB of RAM. If it runs fine, increase that number until it no longer runs fine. It its only 64MB of RAM at the top of the memory map, can you tell the system you only have 4095MB of RAM? > Obviously, the best fix for this would be to get a bios that works properly in > this regard. However, would there be a second-best fix where one could > possibly tell the kernel to ignore that last 64mb of ram and use everything > else instead? As this seems to be a problem in the very latest crop of Intel > motherboards, it will likely come up quite a bit over the next while as these > boards become more common in the market. > > Is there a way, perhaps, to draw this issue to Intel's attention and get it > fixed properly in the bios? I'm thinking I will bring it up up to the dealer > that I purchased this computer from -- he is an authorized Intel dealer and > perhaps will have some sort of a way to get some input back to Intel. Before I > do this, is my understanding of the problem correct and is there any other > information that should also be included here? This is an X86_64 capable CPU running in 64 bit mode, right? So you don't need a PAE kernel? -- Kevin J. Cummings kjchome@xxxxxxx cummings@xxxxxxxxxxxxxxxxxx cummings@xxxxxxxxxxxxxxxxxxxxxxx Registered Linux User #1232 (http://counter.li.org)
