| View previous topic :: View next topic |
| Author |
Message |
beijingjj
Apprentice
Joined: 01 Jul 2006
Posts: 212
|
 Posted: Mon Sep 01, 2008 9:37 am Post subject: segfault when compiling with new motherboard |
 |
|
| I recently bought a new Asus P5E motherboard as my P5N bit the dust. All other hardware remains the same. My gentoo booted up fine, I had to use a new ethernet driver, but that was about it. However, I can't compile anything, gcc keeps giving me a segmentation fault. As I stated, all hardware: CPU, video, etc. is the same, only the motherboard has changed. Any suggestions? |
|
| Back to top |
|
 |
NeddySeagoon
Administrator
Joined: 05 Jul 2003
Posts: 27780
Location: 56N 3W
|
| Posted: Mon Sep 01, 2008 10:21 am Post subject: |
|
|
beijingjj,
Probably overheating from none/too much thermal compound between the CPU and heatsink.
If you make repeated attempts to compile the same package do you get identical errors in identical places or are the errors in random locations?
The former indicates a systematic issue, (not thermal paste) the latter points to a hardware problem (possibly thermal paste).
Why do you believe your RAM, CPU, etc, survived a motherboard failure?
_________________
Regards,
NeddySeagoon
Computer users fall into two groups:-
those that do backups
those that have never had a hard drive fail. |
|
| Back to top |
|
|
beijingjj
Apprentice
Joined: 01 Jul 2006
Posts: 212
|
| Posted: Mon Sep 01, 2008 2:20 pm Post subject: |
|
|
I was updating the BIOS and the update failed, so I don't believe anything else was affected.
I'm running memtest and getting some errors. I've got 2 sticks of 500MB and 2 sticks of 1GB DDR2 667. If I put it all in, in any order (I've tried 3 different configurations so far) I get errors. The 2 500MB sticks by themselves ran for a while without problem. Now I'm running the 2 500s with one of the 1GB sticks and we'll see. It could be a bad stick but I'm also wondering if it is a motherboard problem. |
|
| Back to top |
|
|
monsm
Guru
Joined: 26 Sep 2007
Posts: 449
Location: London, UK
|
| Posted: Mon Sep 01, 2008 3:34 pm Post subject: |
|
|
I had a similar problem. It turned out to be timing problems between CPU and memory. Memtest showed it. Search online for the recommended settings for your modules. One of the, I think is called the CT setting. It is typically set to 2.5, but my errors went away when I set it to 3. Turned out I had 2 different memory modules, the slowest was rated to 3...
Mons |
|
| Back to top |
|
|
NeddySeagoon
Administrator
Joined: 05 Jul 2003
Posts: 27780
Location: 56N 3W
|
| Posted: Mon Sep 01, 2008 3:54 pm Post subject: |
|
|
monsm,
That will be the CAS Cycle Time.
You must also have had a brain dead BIOS that read the timing info from the first memory stick, then used it for them all.
_________________
Regards,
NeddySeagoon
Computer users fall into two groups:-
those that do backups
those that have never had a hard drive fail. |
|
| Back to top |
|
|
Akkara
Administrator
Joined: 28 Mar 2006
Posts: 3715
Location: &akkara
|
| Posted: Mon Sep 01, 2008 11:12 pm Post subject: |
|
|
Are you perchance overclocking anything? Some motherboards have touchier timing than others so make sure everything's at normal speed first.
I had heard some Asus P5's have a problem with the ram voltage circuit - it runs the ram at a lower voltage than ideal. Perhaps try bumping it up 1 or 2 tenths and see if that helps. (Ideally check with a voltmeter if you know how and have a steady hand. Be *very* careful not to slip and short traces if you decide to try this.)
If after all that it still doesn't work, you can try running the system 10% slower and see if that solves the problem. If the problem goes away when running slower than rated you may have a iffy motherboard and might wish to exchange it. |
|
| Back to top |
|
|
platojones
Veteran
Joined: 23 Oct 2002
Posts: 1430
Location: Just over the horizon
|
| Posted: Mon Sep 01, 2008 11:22 pm Post subject: |
|
|
| I just went through a similar experience when I upgraded my CPU. Instead of segfaults, I was getting hung up during boot. In my case, my MB no longer just 'did the right thing' with everything turned on auto with the new CPU. I had to manually tweak the FSB/DRAM frequencies and voltages to get things stable again. I would suggest getting a copy of memtest86+ and running it to check your general setup stability. Also, mprime or Prime95 (Windows Version) are really good at finding tuning/hardware issues (good for testing overclock). Any failures at all by either of those, means you have more investigating or tuning to do. I would definitly run an all-night session of memtest86+ to test the memory stability before booting up under an OS, since memory errors can really screw up your system (corrupted HDs, etc). My gut feeling tells me things are pretty close to being right tuning-wise on your box, or you wouldn't be getting as far as you are. However, if it's some bad hardware, then better to find out with memtest that screw up your box. |
|
| Back to top |
|
|
beijingjj
Apprentice
Joined: 01 Jul 2006
Posts: 212
|
| Posted: Tue Sep 02, 2008 12:08 pm Post subject: |
|
|
Thanks to everyone: I was in fact running memtest86+ and it was giving me errors. I manually set the memory type to DDR2 533 (I thought it was 667 but it is in fact 533) and memtest ran without errors for an hour. Thanks for pointing me in the right direction.
Now, earlier I thought my memory was DDR2 667 and so manually set the motherboard to this. For the latency, etc. I went to manual mode which by default is the slowest, most conservative settings. Memtest seemed to run OK until I decreased some latency settings too much. So my question is am I better off with my current settings, which are memory type DDR2 533 (which is what my memory is) and AUTO for the speed settings, -=OR=- setting the memory type up to a faster type like 667 and then manually playing with the latency settings? Might I get better performance with the latter settings?
To answer one other posters question: no, I am not doing any overclocking intentionally, all the motherboard settings are default. |
|
| Back to top |
|
|
Akkara
Administrator
Joined: 28 Mar 2006
Posts: 3715
Location: &akkara
|
| Posted: Tue Sep 02, 2008 12:59 pm Post subject: |
|
|
| beijingjj wrote: | | So my question is am I better off with my current settings, which are memory type DDR2 533 (which is what my memory is) and AUTO for the speed settings, -=OR=- setting the memory type up to a faster type like 667 and then manually playing with the latency settings? Might I get better performance with the latter settings? |
That question has a complicated answer that I don't know how to express in a short way.
Basically: a memory transaction consists of several steps: precharge, present row address (if needed), present column address, read data. There needs to be a certain delay between each step, and between the last step and another first step.
In general, these delays are real numbers. But the memory timing is quantized to integer (or sometimes half-integer) clock cycles.
So to answer your question, the best performance is obtained when the combination of clock cycle time, along with the chosen integer delays, multiply out to as close to the minimums supported by the ram without going under the minimums.
I don't know a quick way of finding these. I had done it a few times, the procedure basically amounts to setting all parameters to their maximum values and then decreasing one by one to find the smallest value for each parameter that the machine still boots at (while keeping the others at max), and repeating this test for the various bus rates. With all that information you can then find the global minimum and get the best bus rate and timing numbers for your memory. (If you do try this, shoot for 5% or so above the minimum timings, since you're using "does it boot" as a measure of success instead of a full memtest which would be ideal test to use but then it'd take weeks to test all the combinations.)
Don't forget that reading the data itself takes a certain (fixed) number of clock cycles independent of the other delays (usually a number of consecutive locations are burst-read into cache). So slowing down the bus rate too much slows this data transfer even though the rest of the timing would otherwise be good.
However, there's one other complication, and that is synchronizer delay between the memory bus and the internal CPU logic. The data arriving from the memory needs to be synchronized with the internal CPU logic and depending on how the clock edges fall a delay might be introduced. Which means that - sometimes - you can get slightly better speed by running the memory bus at a integer multiple of the frontside bus clock, even if this rate is slower than the memory would otherwise be able to operate at. |
|
| Back to top |
|
|
NeddySeagoon
Administrator
Joined: 05 Jul 2003
Posts: 27780
Location: 56N 3W
|
| Posted: Tue Sep 02, 2008 2:20 pm Post subject: |
|
|
beijingjj,
The short answer is don't overclock.
Auto should be ok as that reads the timing information from the SPD read only memory on the RAM. Most but not all BIOSes set the timing per RAM stick, so if you have a mix, it still works properly. A few poor BIOSes only read the SPD in the first detected RAM stick and assume the rest are the same. Thats often a problem when they are not. e.g applying fast timings to slower RAM.
_________________
Regards,
NeddySeagoon
Computer users fall into two groups:-
those that do backups
those that have never had a hard drive fail. |
|
| Back to top |
|
|
|
|
You cannot post new topics in this forum
You cannot reply to topics in this forum
You cannot edit your posts in this forum
You cannot delete your posts in this forum
You cannot vote in polls in this forum
|
|
Kernel & Hardware
