Hard disks faster in suse than gentoo?

Message

murphydog · Post by **murphydog** » Mon May 03, 2004 11:47 am

I have two hardrives, hdb and hdd (They are on cable select and thats the only way the cable fitted so they are now slaves

)
Hdb has Suse 8.2 pro on it and is a Maxator 80GB 7200rpm.
hdd has gentoo (gentoo-dev-sources 2.6.3) and is (i think) a seagate 60GB 7200rpm.

These are the hdparm results from gentoo

Code: Select all

bash-2.05b# hdparm -tT /dev/hdb

/dev/hdb:
 Timing buffer-cache reads:   496 MB in  2.00 seconds = 247.54 MB/sec
 Timing buffered disk reads:   94 MB in  3.06 seconds =  30.71 MB/sec
bash-2.05b# hdparm -i /dev/hdb

/dev/hdb:

 Model=Maxtor 6Y060L0, FwRev=YAR41VW0, SerialNo=Y2SN9XHE
 Config={ Fixed }
 RawCHS=16383/16/63, TrkSize=0, SectSize=0, ECCbytes=57
 BuffType=DualPortCache, BuffSize=2048kB, MaxMultSect=16, MultSect=16
 CurCHS=16383/16/63, CurSects=16514064, LBA=yes, LBAsects=120103200
 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=yes: disabled (255) WriteCache=enabled
 Drive conforms to: (null):

 * signifies the current active mode

bash-2.05b# hdparm /dev/hdb

/dev/hdb:
 multcount    = 16 (on)
 IO_support   =  1 (32-bit)
 unmaskirq    =  1 (on)
 using_dma    =  1 (on)
 keepsettings =  0 (off)
 readonly     =  0 (off)
 readahead    = 256 (on)
 geometry     = 65535/16/63, sectors = 120103200, start = 0
bash-2.05b# hdparm -tT /dev/hdd

/dev/hdd:
 Timing buffer-cache reads:   476 MB in  2.01 seconds = 236.50 MB/sec
 Timing buffered disk reads:   44 MB in  3.02 seconds =  14.56 MB/sec
bash-2.05b# hdparm -i /dev/hdd

/dev/hdd:

 Model=ST340015A, FwRev=3.01, SerialNo=5LA06QSV
 Config={ HardSect NotMFM HdSw>15uSec Fixed DTR>10Mbs RotSpdTol>.5% }
 RawCHS=16383/16/63, TrkSize=0, SectSize=0, ECCbytes=4
 BuffType=unknown, BuffSize=2048kB, MaxMultSect=16, MultSect=16
 CurCHS=16383/16/63, CurSects=16514064, LBA=yes, LBAsects=78165360
 IORDY=on/off, tPIO={min:240,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
 AdvancedPM=no WriteCache=enabled
 Drive conforms to: device does not report version:

 * signifies the current active mode

bash-2.05b# hdparm /dev/hdd

/dev/hdd:
 multcount    = 16 (on)
 IO_support   =  1 (32-bit)
 unmaskirq    =  1 (on)
 using_dma    =  1 (on)
 keepsettings =  0 (off)
 readonly     =  0 (off)
 readahead    = 256 (on)
 geometry     = 65535/16/63, sectors = 78165360, start = 0

And these from suse

Code: Select all

linux:/home/homepc # hdparm -tT /dev/hdb

/dev/hdb:
 Timing buffer-cache reads:   128 MB in  0.96 seconds =133.33 MB/sec
 Timing buffered disk reads:  64 MB in  1.47 seconds = 43.54 MB/sec
linux:/home/homepc # hdparm -i /dev/hdb

/dev/hdb:

 Model=Maxtor 6Y060L0, FwRev=YAR41VW0, SerialNo=Y2SN9XHE
 Config={ Fixed }
 RawCHS=16383/16/63, TrkSize=0, SectSize=0, ECCbytes=57
 BuffType=DualPortCache, BuffSize=2048kB, MaxMultSect=16, MultSect=16
 CurCHS=16383/16/63, CurSects=16514064, LBA=yes, LBAsects=120103200
 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=yes: disabled (255) WriteCache=enabled
 Drive conforms to: (null):  1 2 3 4 5 6 7

linux:/home/homepc # hdparm /dev/hdb

/dev/hdb:
 multcount    = 16 (on)
 IO_support   =  1 (32-bit)
 unmaskirq    =  1 (on)
 using_dma    =  1 (on)
 keepsettings =  0 (off)
 readonly     =  0 (off)
 readahead    =  8 (on)
 geometry     = 7476/255/63, sectors = 120103200, start = 0

 linux:/home/homepc # hdparm -tT /dev/hdd
/dev/hdd: Input/output error

(For some reason suse keeps saying that hdd has a bad superblock when i try and mount it, but it works fine with gentoo.)

As you can see, under suse hdb is about 50% faster than under gentoo. Hdd is pitifully slow under gentoo, and doesn't work with suse.
I assume this is because i haven't configured gentoo properly.

Can anyone tell me where i've gone wrong?

Thanks in advance
Richard.

moocha · Post by **moocha** » Mon May 03, 2004 11:57 am

Wrong IDE chipset driver compiled into the Gentoo kernel? Or maybe correct driver compiled in, but you forgot to remove Generic IDE chipset support and that one's getting used?

murphydog · Post by **murphydog** » Mon May 03, 2004 12:32 pm

Thanks for the reply. I did have the generic IDE chipset compiled in, so i removed it, but it seemed to make little difference.
I have a Jetway 663AS Pro mother board, so now in my kernel config i have

Code: Select all

< >     generic/default IDE chipset support
[ ]       Generic PCI IDE Chipset Support
<*>         VIA82CXXX chipset support

The mother board manual says:
VIA VT8363/VT 82C686B KT-133 Chipset
Does that sound right?

Any other ideas?

Thanks
Richard

revertex · Post by **revertex** » Mon May 03, 2004 2:00 pm

why it does't work with cable select?

turn off your pc
disconect both disks
change the jumper to cs
conect only one hd at the end of the cable
power on(do not connect the second disk)
look at your bios if everithing is ok(udma,etc...)
run your benchmarks
repeat these steps with the other disk.

kinda weird, both disks has the same cache(2mb).
is your cable a 80 pin?

WaVeX · Post by **WaVeX** » Mon May 03, 2004 2:35 pm

Sorry to go a little off topic here. But what does cable select do anyways. I've wondered this for years. Never used it before.

moocha · Post by **moocha** » Mon May 03, 2004 3:12 pm

murphydog wrote:I have a Jetway 663AS Pro mother board, so now in my kernel config i have
Code: Select all
< >     generic/default IDE chipset support
[ ]       Generic PCI IDE Chipset Support
<*>         VIA82CXXX chipset support
The mother board manual says:
VIA VT8363/VT 82C686B KT-133 Chipset
Does that sound right?

Yup, that's exactly right...

murphydog wrote:Any other ideas?

Not at the moment, but I'll keep on it - it's made me curious. And it has to be software-related since the hardware configuration didn't change between Gentoo and SuSE. Very strange.

WaVeX wrote:Sorry to go a little off topic here. But what does cable select do anyways. I've wondered this for years. Never used it before.

http://www.pcguide.com/ref/hdd/if/ide/confCS-c.html

revertex · Post by **revertex** » Mon May 03, 2004 3:34 pm

hummm, try play with different hdparm configs, just search in these foruns for "hdparm" and you will find a lot of usefull settings.

Gentree · Post by **Gentree** » Mon May 03, 2004 4:17 pm

WaVeX wrote:Sorry to go a little off topic here. But what does cable select do anyways. I've wondered this for years. Never used it before.

not your main point but CS is for a cable with a crossover pair in the cable that sets the master or slave status by its position on the cable. Thus you put the two dirves on CS and let the hardware connection determine who is master.

More usually you chose which is Master and slave with jumpers and use a unbiased cable that connects both drives identically from an electrical point of view.

Be sure to select one MA and one SL in the latter case and verify that you cable does not have on pair or wires cross-over inbetween the two connecters to the disks.

I have had some disks which dont work with the jumper in certain positions . ie they are faulty but do work reliably if you let them have the position they want.

Each time you change something obviously go through the BIOS on reboot (this may explain hdd u/s on SUSE)

Getting back to murphydog's question:

In fact some of the performance figures were better under Gentoo for hdb: 247 Mb/s againt 133 !

Are these results repeatable or do you get largely varying results if you rerun hdparm?

It seems a bit odd that there is such a major diff but obvioulsy you dont have the same detailed control over your kernel options on SUSE so its hard to do an accurate comparison.

dmesg | grep DMA

will show you if DMA and UDMA are setting up correctly on boot. This may shed some light.

HTH

Nate_S · Post by **Nate_S** » Mon May 03, 2004 6:31 pm

I noticed that the drives are using different udma modes

hdb:
UDMA modes: udma0 udma1 udma2 udma3 udma4 *udma5 udma6

hdd:
UDMA modes: udma0 udma1 *udma2 udma3 udma4 udma5

* signifies the current active mode

the udma 5 versus the udma2 probly causes the speed difference you're seeing.

I think udma5 is ATA133 and udma2 is ATA66. If the software settings are all right, check your ide cables, be sure that they are both 80wire and also remember that they can get damaged pretty easily if they are very bent or twisted (i've had that happen before)

-Nate

moocha · Post by **moocha** » Mon May 03, 2004 6:40 pm

UDMA2 is ATA33 (the highest mode you can get without an 80 wire cable), and UDMA5 is ATA100.

MadEgg · Post by **MadEgg** » Mon May 03, 2004 7:44 pm

moocha wrote:UDMA2 is ATA33 (the highest mode you can get without an 80 wire cable), and UDMA5 is ATA100.

Would UDMA6 be ATA133 then?

moocha · Post by **moocha** » Mon May 03, 2004 8:24 pm

MadEgg wrote:Would UDMA6 be ATA133 then?

Yup.

Gentree · Post by **Gentree** » Mon May 03, 2004 8:36 pm

the udma 5 versus the udma2 probly causes the speed difference you're seeing.

I think the speed diff he was refering to was the diff between Gentoo and SUSE on hdb , not the diff between the drives which are physically very different.

murphydog · Post by **murphydog** » Tue May 04, 2004 11:54 am

First off thanks for all the replys.

Are these results repeatable or do you get largely varying results if you rerun hdparm?

They are all repeateable, but now and again (usually the first one or two times hdparm is run) they are lower. I've always assumed this to be because the disk is being used in the background, but i may be wrong

I tried dmesg | grep DMA in gentoo, but it didn't return anything, so i tried straight dmesg and it is full of crap like

Code: Select all

IN=eth0 OUT= MAC=00:50:ba:c6:99:1c:00:09:f3:04:18:c0:08:00 SRC=68.38.124.101 DST=10.0.0.10 LEN=52 TOS=0x00 PREC=0x00 TTL=113 ID=4550 DF PROTO=TCP SPT=4670 DPT=32805 WINDOW=17184 RES=0x00 ACK FIN URGP=0
IN=eth0 OUT= MAC=00:50:ba:c6:99:1c:00:09:f3:04:18:c0:08:00 SRC=83.31.158.91 DST=10.0.0.10 LEN=44 TOS=0x00 PREC=0x00 TTL=48 ID=0 DF PROTO=TCP SPT=411 DPT=33235 WINDOW=5808 RES=0x00 ACK SYN URGP=0
IN=eth0 OUT= MAC=00:50:ba:c6:99:1c:00:09:f3:04:18:c0:08:00 SRC=83.31.158.91 DST=10.0.0.10 LEN=44 TOS=0x00 PREC=0x00 TTL=48 ID=0 DF PROTO=TCP SPT=411 DPT=33235 WINDOW=5808 RES=0x00 ACK SYN URGP=0

I take it this is a seperate problem with eth0

(eth0 is connected to an adsl router, if that makes any difference)

I'm not sure why hdd is only at udma2. It has a new 80 wire cable, but it seems to be stuck at udma2.

If i put the jumpers to master and slave instead of cable select, will they still work at the same speed, or will they both revert to udma2?

I think the speed diff he was refering to was the diff between Gentoo and SUSE on hdb , not the diff between the drives which are physically very different.

Yes thats right. I was about to move gentoo to hdb to get the extra speed, when i noticed that the drives were faster in suse. Although if anyone can get hdd running faster please feel free, as theoretically it should be about the same speed as hdb

Richard

huw · Post by **huw** » Tue May 04, 2004 12:16 pm

the IP stuff in dmesg is logging from iptables, it's not a problem with your ethernet card.

Wabkebab · Post by **Wabkebab** » Tue May 04, 2004 12:54 pm

I have the same problem with the CDROMS, between gentoo and mandrake 9.2. On gentoo the performance is quite poor, and in the other hand, on mandrake it's excelent!

Initially when I did hdparm /dev/hda (My 16x DVD / 48x CDROM) I get the astoning 615 kB/s transfer rate, and playing with hdparm(seting up udma2 and PI0 mode 4, the settings supported by the motherboard and the drive) I reached to 1.30 Mb/s.

Note: My HD has a 38 Mb/s transfer rate.

Any ideas?

murphydog · Post by **murphydog** » Thu May 06, 2004 11:13 pm

Does anyone have any more ideas?
Are there any configeration files I can copy over from suse to gentoo, or is there any way of telling if its a kernel thing or some other problem. Does suse have some magic kernel patch that makes hard disks faster

?

Regards,
Richard

Rainmaker · Post by **Rainmaker** » Fri May 07, 2004 1:04 am

try hdparm -d 1 -A 1 -m 16 -u 1 -a 64 /dev/hdb

these are "Safe performance-enhancing options"

Try looking at man hdparm for tweaking the settings a bit for your drive

molander · Post by **molander** » Fri May 07, 2004 3:48 am

I may be missing something here.
Gentoo:

Code: Select all

bash-2.05b# hdparm -tT /dev/hdb 
/dev/hdb: 
 Timing buffer-cache reads:   496 MB in  2.00 seconds = 247.54 MB/sec 
 Timing buffered disk reads:   94 MB in  3.06 seconds =  30.71 MB/sec

Suse:

Code: Select all

linux:/home/homepc # hdparm -tT /dev/hdb 
/dev/hdb: 
 Timing buffer-cache reads:   128 MB in  0.96 seconds =133.33 MB/sec 
 Timing buffered disk reads:  64 MB in  1.47 seconds = 43.54 MB/sec

Looks to me like your gentoo system is faster than Suse for buffer-cache reads. The disk reads are a touch slower in gentoo but that could be due to the smaller file size. My suggestion is to copy over hdparm from one to the other and see if it will use the same test file sizes. Also curious is the read ahead is 256 in gentoo but only 8 in suse. You can specify that by using a "-a". That may also account for part of the difference.

moocha · Post by **moocha** » Fri May 07, 2004 4:45 am

molander wrote:My suggestion is to copy over hdparm from one to the other and see if it will use the same test file sizes.

That will not change anything since hdparm doesn't test by reading files.

murphydog · Post by **murphydog** » Fri May 07, 2004 10:48 am

I've tried messing about with hdparm settings from the man page and also other threads on this forum, and the drive seems to be fastest with those original settings from the first post.
I tried lowering the read ahead to 64 from 256 in gentoo on hdb and the disk reads stayed the same at approx 30mb/s. I then lowered it to 8 (like suse's) and the disk reads dropped to 19mb/s. I then booted into suse and raised the reaad ahead to 64 , expecting a performance increase like in gentoo, but the disk reads stayed at about 44. When i tried 256 in suse it said it was an invalid option.
Does this make any sense? Or is my computer haunted by gremlins....
Thanks for all your help,
Richard

jokernel · Post by **jokernel** » Fri May 07, 2004 12:45 pm

Are you using different kernels?
Are you using ACPI?

murphydog · Post by **murphydog** » Fri May 07, 2004 1:20 pm

Are you using different kernels?

Yes. Suse has the stock kernel it installed itself (8.2 pro) and gentoo has gentoo-dev-sources 2.6.3

Are you using ACPI?

No i don't think i am in gentoo. The only bits selected in the whole of the power management section are

Code: Select all

[*] Power Management support
[*]   Software Suspend (EXPERIMENTAL)

I think suse might use acpi because it turns itself off, but gentoo doesn't (

)
Would it make a difference?
Thanks
Richard

molander · Post by **molander** » Fri May 07, 2004 2:42 pm

moocha wrote:
molander wrote:My suggestion is to copy over hdparm from one to the other and see if it will use the same test file sizes.
That will not change anything since hdparm doesn't test by reading files.

Didnt say that it did. I am wondering if the suse version of hdparm prefers different test properties than the gentoo version.

jokernel · Post by **jokernel** » Fri May 07, 2004 3:42 pm

I think suse might use acpi because it turns itself off, but gentoo doesn't ( )
Would it make a difference?

On some boards it does.

On 2.6 kernel try
hdparm -a 4096 ...
or bigger
and
check the performance with
hdparm -t /dev/hdaX
hdaX your root partition