1 |
On Tue, Nov 15, 2011 at 2:33 AM, Raffaele BELARDI |
2 |
<raffaele.belardi@××.com> wrote: |
3 |
> I have two gentoo boxes, X has an ASUS M2NPV-VM with AMD64 3500+ CPU, Y |
4 |
> has a AMD64 X2 5600+ CPU. Since I need more juice on X I thought I could |
5 |
> swap CPUs. |
6 |
> |
7 |
> After updating X's BIOS the system with the 'new' CPU boots up to the |
8 |
> MythTv screen with no error but does not respond to the USB keyboard nor |
9 |
> to the remote control keypresses. More precisely: |
10 |
> |
11 |
> - keyboard is fine at grub boot, I can select up and down or edit the |
12 |
> entries |
13 |
> - keyboard is no longer responsive at the init scripts start (when you |
14 |
> can press 'I' to select services) |
15 |
> |
16 |
> My first thought was that X (and Y) WERE compiled with '-march=native' |
17 |
> GCC flag and maybe the 5600+ does not execute properly the 3500+ code. |
18 |
> But a quick search on wikipedia shows that 5600+ has a superset of the |
19 |
> 3500+ so I should have problems putting the 3500+ in the Y box, not |
20 |
> vice-versa. |
21 |
> |
22 |
> Any suggestions? |
23 |
|
24 |
Play with your BIOS settings. Look for things like legacy USB support. |
25 |
Also, double-check that all the relevant USB drivers (UHCI, EHCI, |
26 |
XHCI, HID, etc) are either built-into the kernel, or are loaded as |
27 |
modules. Consider rebuilding your kernel. |
28 |
|
29 |
Just because one processor has a superset of the instructions of the |
30 |
other doesn't mean there may not be other compatibilities. Some time |
31 |
back, a thread on here discussed how to find out what -march=native |
32 |
becomes, in terms of -march and a bunch of other parameters. I've |
33 |
noticed parameters like cache line sizes and cache sizes, among a |
34 |
couple others. I imagine a bungling of, e.g., cache line sizes could |
35 |
break code that has heavy dependency on data locality and/or memory |
36 |
models; it might have broken your USB drivers, for example. |
37 |
|
38 |
But, really, I think BIOS settings and driver presence are the more |
39 |
likely culprit. |
40 |
|
41 |
-- |
42 |
:wq |