my harddrive performance is sucking

volospin · Posted: Tue Apr 29, 2003 2:51 am Post subject:

ehh... Bloody Bastard...
I said and mean the timing buffer-cache reads...
not the timing buffered disk reads...
does it mean my cache is slow or something?

Since the following appears, or how do I increase above my 116MB/sec

bzugda · n00b Joined: 19 Oct 2002 Posts: 39 Location: Barnsley

Sylgeist · n00b Joined: 04 Dec 2002 Posts: 36 Location: USA

I am getting 116.xx for the buffer and 38.xx for the disk amounts. I know that is not bad but I have two drives in my machine - both Western Digital, one is a special edition with 8MB cache and the other is a 2MB version. I get exactly the same numbers every time. On my other machine with the exact same drives (but different chipset) I get 900MB on buffer score and 45 on disk. Am I maxed out on my VIA chipset for speed? Why is the 8MB cache not giving better results?

AnvilDemon · n00b Joined: 12 Jul 2003 Posts: 21

Sylgeist · n00b Joined: 04 Dec 2002 Posts: 36 Location: USA

Thanks for the info - that makes sense, but why the big difference between my two machines?

AnvilDemon · n00b Joined: 12 Jul 2003 Posts: 21

ang · n00b Joined: 30 Jan 2003 Posts: 29

drekka · Tux's lil' helper Joined: 06 Jul 2003 Posts: 111

Hello everyone, read this thread with interest and thought I might ask your oinions on why I can't run higher than 25M/sec stable. I can get 41M/sec but the system crashses within a minute or two with lots of IO errors.

Asus A7N8X Deluxe with Sil3112 SATA raid controller.
2 x Seagate 120G SATA drives in raid 0.

ac-sources kernel.

hdparm gives:

Timing buffer-cache reads: 1612 MB in 2.00 seconds = 806.00 MB/sec
Timing buffered disk reads: 76 MB in 3.00 seconds = 25.33 MB/sec

Boot up hdparm settings are :

disc0_args="-c1 -A1 -d1 -X70 -m16 -u1"
disc1_args="-c1 -A1 -d1 -X70 -m16 -u1"

I've used the max_kb_per_request trick to get to 41M/sec, but as I said it was not stable. The default setting for these drives is 15, which is the setting everyone recommends. I had to go higher to 24 or greater to get faster speeds.

Any thoughts,
Derek.
_________________
Regards,
Derek

<nb>Reality is not what you think</nb>

AnvilDemon · n00b Joined: 12 Jul 2003 Posts: 21

drekka, I have two raptors. They used to be in a raid 0 but not anymore. I used the -d1 and -X70 command only. I have the Abit NF7-S with the silraid controller, same as you. My system seemed stable for about two weeks then one day it just locked up and I lost everything. So now I use the drives independent on the controller with the same settings and I have not had a problem yet.

I know this does not help much but atleast it is some information.

vbenares · Apprentice Joined: 13 Aug 2003 Posts: 205

Okay. Try and explain these numbers:

Timing buffer-cache reads: 1668 MB/sec
Timing buffered disk reads: 3.1 MB/sec

What the heck is going on?! And how do I fix it?

vbenares · Apprentice Joined: 13 Aug 2003 Posts: 205

I should have mentioned, it's a Seagate 80 GB, 7200 RPM, Serial ATA, w/ 8 MB Cache. Model number ST380023AS.

lamaistres · Posted: Wed Aug 20, 2003 8:42 pm Post subject:

Check this out:

# hdparm -T /dev/hda

/dev/hda:
Timing buffer-cache reads: 688 MB in 2.01 seconds = 342.29 MB/sec

hdparm -t /dev/hda

/dev/hda:
Timing buffered disk reads: 96 MB in 3.04 seconds = 31.63 MB/sec

using "hdparm -X68 -m16 -u1 -c1 -a64 /dev/hda"

688 MB in 2.01 seconds?
_________________
"Always do what you want, and say what you feel,
because those who mind don't matter, and those
who matter don't mind." -- Dr. Suess

TobiWan · Posted: Wed Aug 20, 2003 9:06 pm Post subject: hdparm guide

This thread really is interesting. That whole hdparm issue should be included in the installation manual at least for setting the dma and keep that when rebooting.

Maybe we can collect a couble of good hdparm settings for different purposes like:

I guess you get my point here. Make some more propositions and we can start collecting settings. I am even willing to write a hdparm summary afterwards and translate it into German and Dutch. Maybe French too if there's enough people interested in that language. :-)

Tobias
_________________
Killing for peace is like fucking for virginity.

viperlin · Veteran Joined: 15 Apr 2003 Posts: 1319 Location: UK

hey guys (and the few girls, there nees to be more girls here) , sorry but people in this thread should be able to help me with my problem it's similar ,but no it's not an unsearched simple thing (i hope to jebus it was though)

anyway details here

hopefully you can help me out, sorry for posing this type of thing but my comps goin a little slow because of it.

TobiWan · Posted: Wed Aug 20, 2003 10:16 pm Post subject: no offense meant

viperlin · Veteran Joined: 15 Apr 2003 Posts: 1319 Location: UK

erm u missunderstood it, i'm just trying to attract attention to my thread for help, since it's a chipset hdparm problem, darn chipsets, why cant it be hamster powered

fdavid · Posted: Sat Aug 30, 2003 12:44 pm Post subject:

I have a Quantum lct20 20G (ATA100, 5400rpm), and a Samsung SP P80 80G (ATA133, 7200rpm). My mainboard supports ATA100. On the ide channel of th Quantum there is also a CD-writer configured as slave.

hdparm recognizes the Samsung running in udma5, the Quantum running in udma2. When i disconnect the cd-writer the Quantum runs in udma5, as well.

So, here are my hdparm results:
Samsung (udma5):
-t: 53.78; 53.33; 54.24
-T: 124.27; 123.08; 125.49
Quantum (udma2 /cd-writer connected/):
-t: 18.99; 20.25; 19.88
-T: 120.75; 129.29; 129.29
Quantum (udma5 /cd-writer disconnected/):
-t: 17.07; 20.45; 20.38
-T: 110.24; 113.27; 115.32

Questions:
1. Why don't i measure significant difference between udma2 and udma5 in case of Quantum? Is it just because the values stay under approx. 33 (udma2)? How could i measure the performance difference if there's some?
2. Could be such a big performance defference between two disks running in udma5?
3. AFAIK the -T option doesn't really count, because it measures the speed of the linux memory cache, not hte disk memory cache. It has more likely sg to do with the RAM speed, but the disk speed. Is that right? (My measurements seem to prove this, i think so.)

Jung, Dong Chul · n00b Joined: 30 Aug 2003 Posts: 1 Location: Korea

My System
P4 2.53GHz
IC35L120AVV207-1 [8MB]

zzori nvidia-glx # hdparm -vtT /dev/hda

/dev/hda:
multcount = 16 (on)
IO_support = 0 (default 16-bit)
unmaskirq = 0 (off)
using_dma = 1 (on)
keepsettings = 0 (off)
readonly = 0 (off)
readahead = 256 (on)
geometry = 16383/255/63, sectors = 241254720, start = 0
Timing buffer-cache reads: 1952 MB in 2.00 seconds = 975.17 MB/sec
Timing buffered disk reads: 152 MB in 3.01 seconds = 50.44 MB/sec

zzori nvidia-glx # hdparm -vtT /dev/hda

/dev/hda:
multcount = 16 (on)
IO_support = 0 (default 16-bit)
unmaskirq = 0 (off)
using_dma = 1 (on)
keepsettings = 0 (off)
readonly = 0 (off)
readahead = 256 (on)
geometry = 16383/255/63, sectors = 241254720, start = 0
Timing buffer-cache reads: 1896 MB in 2.00 seconds = 948.14 MB/sec
Timing buffered disk reads: 158 MB in 3.00 seconds = 52.66 MB/sec
zzori nvidia-glx #
_________________
ZZORI

ikaro · Advocate Joined: 14 Jul 2003 Posts: 2527 Location: Denmark

my drive is a Maxtor 120gb ata 133 , 7200 rpm , 2mb cache.

ive done this a long time ago, so i dont remember the old speeds, but this is what i got now:

insanethrees · n00b Joined: 19 Jul 2003 Posts: 10 Location: Seattle, USA

j_c_p · Posted: Tue Sep 09, 2003 1:43 am Post subject:

Hello,

zebbedi · Tux's lil' helper Joined: 05 Sep 2003 Posts: 123

I have a WD 160GB UIDE 8MB Cache and performance seemed rather lacking under linux compared to windows. WD also seem to be rather scared when you ask their tech support about running linux on the drive, but heres my results.

With just "hdparm -d /dev/hda" i get:

/dev/hda:
Timing buffer-cache reads: 1772 MB in 2.00 seconds = 886.00 MB/sec
Timing buffered disk reads: 10 MB in 3.40 seconds = 2.94 MB/sec

Then with "hdparm -m16 -c1 -d1 -u1 -k1 -X69 /dev/hda" i get:

/dev/hda:
Timing buffer-cache reads: 1804 MB in 2.00 seconds = 902.00 MB/sec
Timing buffered disk reads: 160 MB in 3.02 seconds = 52.98 MB/sec

Needless to say i added those lines to my hdparm config file and ran rc-update

Spida · Posted: Mon Oct 06, 2003 6:37 pm Post subject:

Why are you all using hdparm to benchmark your drives? bonnie++ is said to deliver much more reliable results.
-> HDD Benchmarking Thread

grzewho · l33t Joined: 31 Dec 2002 Posts: 626 Location: /home/g

can someone explain me the usage of -A1 , -a64 and -c3 options?
_________________

MrDooM · n00b Joined: 22 Oct 2003 Posts: 69

hdparm -i /dev/hde

/dev/hde:

Model=WDC WD600BB-00CAA1, FwRev=17.07W17, SerialNo=WD-WMA8F1124538
Config={ HardSect NotMFM HdSw>15uSec SpinMotCtl Fixed DTR>5Mbs FmtGapReq }
RawCHS=16383/16/63, TrkSize=57600, SectSize=600, ECCbytes=40
BuffType=DualPortCache, BuffSize=2048kB, MaxMultSect=16, MultSect=4
CurCHS=16383/16/63, CurSects=16514064, LBA=yes, LBAsects=117231408
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
AdvancedPM=no WriteCache=enabled
Drive conforms to: device does not report version:

* signifies the current active mode

hdparm -i /dev/hdg

/dev/hdg:

Model=WDC WD600BB-00CAA1, FwRev=17.07W17, SerialNo=WD-WMA8F1123163
Config={ HardSect NotMFM HdSw>15uSec SpinMotCtl Fixed DTR>5Mbs FmtGapReq }
RawCHS=16383/16/63, TrkSize=57600, SectSize=600, ECCbytes=40
BuffType=DualPortCache, BuffSize=2048kB, MaxMultSect=16, MultSect=4
CurCHS=16383/16/63, CurSects=16514064, LBA=yes, LBAsects=117231408
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
AdvancedPM=no WriteCache=enabled
Drive conforms to: device does not report version:

* signifies the current active mode

hdparm -tT /dev/md0

/dev/md0:
Timing buffer-cache reads: 692 MB in 2.01 seconds = 343.99 MB/sec
Timing buffered disk reads: 116 MB in 3.00 seconds = 38.65 MB/sec

Linux Software Raid 1

MrDooM