Skip to contentSkip to conversation

severe performance degradation on serverworks with high mem

DISCUSSION Brian C. Thomas 3 participants 4 messages Oct 18, 2001 – Oct 19, 2001

lore
Share & tools
lore
#0 lore raw

Hi

I have a Supermicro S2QR6 board, which has the "Serverworks serverset III HE" chipset. (Also, i have 8GB RAM, 4 Xeon P3 700MHz processors, and an Adaptec Ultra160 SCSI system - aix7xxx, built onto the motherboard)

In all the 2.4.x series kernels (even up to 2.4.12-ac3 or 2.4.13-pre4), as soon as the high memory, 64GB option is enabled, the server slows down to a crawl, with no load showing. (A "crawl" example is a kernel recompile time going from ~5 min to >2 hr!)

I have tried numerous kernel config options, removing scsi support, networking, and have narrowed the problem down to the serverworks chipset- on a SC440NX mb, also with 8GB and everything else in the system, all the new kernels work with no slowdown.

I've tried...
kernels: 2.4.1, 2.4.3, 2.4.5, 2.4.7-ac9, 2.4.9, 2.4.12 (-pre4, -ac3) with and w/o mtrr support kernel config

Any ideas?

Thanks,

BCT

____________________________
Brian C. Thomas
[email protected]

#1 lore raw

Hi

I don't know if this helps, but I seem to have stumbled onto a possible "fix" for this...

With 64GB high memory enabled in kernel 2.4.12-ac3, and mtrr turned on, I was able to see all 8GB RAM on my machine by using the old 'append="mem=8000M"' command in my lilo.conf file... AND WITH NO LOSS OF PERFORMANCE!

Does that help anyone with defining where this problem is coming from?

BCT

____________________________
Brian C. Thomas
[email protected]

Quoted 35 lines · Brian C. Thomas

Hi

I have a Supermicro S2QR6 board, which has the "Serverworks serverset III HE" chipset. (Also, i have 8GB RAM, 4 Xeon P3 700MHz processors, and an Adaptec Ultra160 SCSI system - aix7xxx, built onto the motherboard)

In all the 2.4.x series kernels (even up to 2.4.12-ac3 or 2.4.13-pre4), as soon as the high memory, 64GB option is enabled, the server slows down to a crawl, with no load showing. (A "crawl" example is a kernel recompile time going from ~5 min to >2 hr!)

I have tried numerous kernel config options, removing scsi support, networking, and have narrowed the problem down to the serverworks chipset- on a SC440NX mb, also with 8GB and everything else in the system, all the new kernels work with no slowdown.

I've tried...
kernels: 2.4.1, 2.4.3, 2.4.5, 2.4.7-ac9, 2.4.9, 2.4.12 (-pre4, -ac3) with and w/o mtrr support kernel config

Any ideas?

Thanks,

BCT

____________________________
Brian C. Thomas
[email protected]
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to [email protected]
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/

#2 lore raw
In reply to Brian C. Thomas
Quoted 5 lines · Brian C. Thomas

Hi

I don't know if this helps, but I seem to have stumbled onto a possible "fix" for this...

Quoted 7 lines

With 64GB high memory enabled in kernel 2.4.12-ac3, and mtrr turned on, I was able to see all 8GB RAM on my machine by using the old 'append="mem=8000M"' command in my lilo.conf file... AND WITH NO LOSS OF PERFORMANCE!

Does that help anyone with defining where this problem is coming from?

The real question is how to reproduce the > 2hr kernel compiles.

Are there any certain steps that would reliably create the problem, and if so what are they?

#3 lore raw
In reply to Brian C. Thomas
Quoted 11 lines · Brian C. Thomas

Hi

I don't know if this helps, but I seem to have stumbled onto a possible "fix" for this...

With 64GB high memory enabled in kernel 2.4.12-ac3, and mtrr turned on, I was able to see all 8GB RAM on my machine by using the old 'append="mem=8000M"' command in my lilo.conf file... AND WITH NO LOSS OF PERFORMANCE!

Does that help anyone with defining where this problem is coming from?

cat /proc/mtrr and see what the mtrrs look like. It feels like there you have a spot of RAM the BIOS doesn't cover with mtrrs.

Eric