nV News Forums

 
 

nV News Forums (http://www.nvnews.net/vbulletin/index.php)
-   General Linux (http://www.nvnews.net/vbulletin/forumdisplay.php?f=27)
-   -   nforce4, SATA and a low HD performance (http://www.nvnews.net/vbulletin/showthread.php?t=85432)

bugger 01-31-07 06:51 PM

nforce4, SATA and a low HD performance
 
1 Attachment(s)
Configuration:
board A8N-SLI, nforce4, Debian Etch AMD64, Athlon64x2, 2GB, 2 SATA and 1 IDE disk.
uname -a
Linux kirk 2.6.20-rc6 #5 SMP PREEMPT Wed Jan 31 13:34:55 NZDT 2007 x86_64 GNU/Linux

frustrating times: although I thought that the sata_nv driver in the new kernel branch 2.6.20-rcX supports ADMA/UCQ the disk performance hasn't changed at all.

hdparm -Tt /dev/sda
/dev/sda:
Timing cached reads: 1370 MB in 2.00 seconds = 684.76 MB/sec
Timing buffered disk reads: 162 MB in 3.01 seconds = 53.76 MB/sec

hdparm -Tt /dev/sdb
/dev/sdb:
Timing cached reads: 1434 MB in 2.00 seconds = 716.94 MB/sec
Timing buffered disk reads: 224 MB in 3.00 seconds = 74.56 MB/sec


see also the dmesg log attached. Has anyone a clue why the performance of the SATA disks is slightly below the the old IDE disk?

mlau2 02-01-07 02:40 AM

Re: nforce4, SATA and a low HD performance
 
hdparm does linear reads, in which case NCQ won't help one iota.
Your test should be running multiple apps which access lots of data
on the disk simultaneously. And then you should only notice a slight
decrease in seek times and no miracle increase in throughput...

bugger 02-06-07 07:40 PM

Re: nforce4, SATA and a low HD performance
 
Yes, perhaps. hdparm may or may not be the appropriate tool for SATA benchmarks or a simple comparison. On the other hand, if you google through the internet for some other SATA hdparm results you'll find significantly higher transfer rates. Apart from that my feeling is that someone pulled the handbrake on the disks and the transfer rate can be much better than it is right now.
What is your experience with the nforce4/SATA combination?

chunkey 02-07-07 03:34 PM

Re: nforce4, SATA and a low HD performance
 
No,

ADMA only "offloads" some interrupt & DMA work, but it doesn't make your
HDD to spin faster. And NCQ helps only for random disc access (It even slows down linear reads because of the higher latency!)

if you really want a better results, then get 15.000 RPM Ultra320 SCSI HDDs OR build a RAID0!

e.g:
RAID0 with two 58MB/s hdds:
Code:

/dev/md4:
 Timing buffered disk reads:  340 MB in  3.01 seconds = 113.01 MB/sec

SATA or SCSI shootout

robhancock 02-19-07 09:49 PM

Re: nforce4, SATA and a low HD performance
 
Those performance numbers seem fine for a single disk. Where are you seeing people getting better numbers than that?


All times are GMT -5. The time now is 07:03 AM.

Powered by vBulletin® Version 3.7.1
Copyright ©2000 - 2014, Jelsoft Enterprises Ltd.
Copyright 1998 - 2014, nV News.