You are not logged in.
My motherboard is an Asus A8N-SLI with nforce4 chipset, and the problem is that everytime the HDD is working my CPU gets a very low performance because (it seems) DMA is not working.
I've seen this:
http://http.download.nvidia.com/XFree86 … blems.html
But it's strange, because it says this only happens in old distributions, so I'm not sure if this could be related.
hdparm -Tt /dev/sda
/dev/sda:
Timing cached reads: 1220 MB in 2.00 seconds = 609.67 MB/sec
Timing buffered disk reads: 152 MB in 3.01 seconds = 50.57 MB/sec
lspci | grep -i ide
00:06.0 IDE interface: nVidia Corporation CK804 IDE (rev f2)
00:07.0 IDE interface: nVidia Corporation CK804 Serial ATA Controller (rev f3)
00:08.0 IDE interface: nVidia Corporation CK804 Serial ATA Controller (rev f3)
"lspci | grep -i sata" outputs nothing.
hdparm /dev/sda
/dev/sda:
multcount = 1 (on)
IO_support = 0 (default)
readonly = 0 (off)
readahead = 256 (on)
geometry = 30401/255/63, sectors = 488397168, start = 0
lsmod|grep ata
pata_amd 15652 0
pata_acpi 5824 0
ata_generic 6244 0
sata_nv 29292 5
libata 193612 4 pata_amd,pata_acpi,ata_generic,sata_nv
scsi_mod 133592 5 usb_storage,sg,sr_mod,sd_mod,libata
Is anybody facing the same problem?
Last edited by lostevil (2009-11-16 18:32:38)
Offline
50MB/s seems like DMA is working... if it wasn't you'd probably be getting 5MB/s.
hdparm -i /dev/sda
dmesg|grep -i dma
You can try sudo hdparm -D1 -Xudma5 for example, but some drives can't be set to the highest mode they show with the -i option.
Offline
50MB/s seems like DMA is working... if it wasn't you'd probably be getting 5MB/s.
You are right about the hdparm -Tt, it seems to be a normal output, but every time the hard drive is accessed, my CPU usage peeks and stays at 100% until the HDD operation ends, and it takes too much to end.
hdparm -i /dev/sda
dmesg|grep -i dma
You can try sudo hdparm -D1 -Xudma5 for example, but some drives can't be set to the highest mode they show with the -i option.
I will be away from that computer for a few days, so I'll be more concrete and post the outputs when I return.
Offline
Earlier than I thought:
hdparm -i /dev/sda
/dev/sda:
Model=WDC WD2500JD-00HBB0, FwRev=08.02D08, SerialNo=WD-WCAL72647665
Config={ HardSect NotMFM HdSw>15uSec SpinMotCtl Fixed DTR>5Mbs FmtGapReq }
RawCHS=16383/16/63, TrkSize=57600, SectSize=600, ECCbytes=74
BuffType=DualPortCache, BuffSize=8192kB, MaxMultSect=16, MultSect=1
CurCHS=16383/16/63, CurSects=16514064, LBA=yes, LBAsects=488397168
IORDY=on/off, tPIO={min:120,w/IORDY:120}, tDMA={min:120,rec:120}
PIO modes: pio0 pio1 pio2 pio3 pio4
DMA modes: mdma0 mdma1 mdma2
UDMA modes: udma0 udma1 udma2 udma3 udma4 udma5 *udma6
AdvancedPM=no WriteCache=enabled
Drive conforms to: Unspecified: ATA/ATAPI-1,2,3,4,5,6
* signifies the current active mode
hdparm -D1 -Xudma5 /dev/sda
/dev/sda:
setting drive defect management to 1
HDIO_DRIVE_CMD(defectmgmt) failed: Input/output error
setting xfermode to 69 (UltraDMA mode5)
HDIO_DRIVE_CMD(setxfermode) failed: Invalid exchange
So dma is working... but I'm still experiencing the CPU problem.
Offline
Isn't DMA irrelevant on Sata drives?
Btw, i'm also having this problem. Have been for months. It seems to be a kernel issue though, from what i've been reading anyway. Many people have resolved it by running BFS/BFQ schedulers. Neither of which have worked for me.
Here's some discussion about it at Gentoo forums http://forums.gentoo.org/viewtopic-t-482731.html
Oh more here http://bbs.archlinux.org/viewtopic.php?id=70525 and here http://bbs.archlinux.org/viewtopic.php?id=79319
Last edited by Mountainjew (2009-11-18 21:47:52)
Offline
So if it really is an I/O scheduler problem, none of the existants solve the problem to me.
http://www.redhat.com/magazine/008jun05 … chedulers/
I tried all of them and had the same problem.
@Mountainjew: Is your chipset a nforce4 one?
Offline
As I recall, BFS is a cpu scheduler, which are not pluggable like the I/O schedulers, hence te needed kernel patch.
The kernel devs don't want to make the cpu scheduler modular, rather they want to keep a single solution for all machines. This is part of the reason Con left.
P.S. There has historically been trouble with the nforce drivers, maybe you have experienced some of these before.
Last edited by fphillips (2009-11-19 18:04:09)
Offline