Skip site navigation (1)Skip section navigation (2)
Date:      Thu, 27 Aug 2009 00:06:18 +0300
From:      Alexander Motin <mav@FreeBSD.org>
To:        freebsd-hardware@freebsd.org
Subject:   Small SATA performance contest
Message-ID:  <4A95A3CA.3060306@FreeBSD.org>

next in thread | raw e-mail | index | archive | help
I've just made several tests, comparing older SATA 1.x SiI3114 
(32bit/66MHz PCIX), newer SATA 2.x JMicron JMB363 and SiI3132 (PCIe x1) 
cards and on-board ICH8 AHCI with two fast drives connected to each:

Linear read with 1MB block from two drives, summary:
SiI3114 in PCI 32/33 slot: 120MB/s.
SiI3114 in PCIX 32/66 slot: 190MB/s.  <-- near to SATA 1.x limit
SiI3132 in PCIe x1 slot: 150MB/s
JMB363 in PCIe x1 slot: 188MB/s
ICH8 AHCI on-board: 310MB/s           <-- drives limit

So, faster bus can be a significant bonus even to old card. Inefficient 
design can kill even new bus. Same time fastest internal connections of 
ICH8, combined with effective design can give unique results.

Linear read with 512B block:
SiI3114 in PCIX 32/66 slot: 16Ktps.
SiI3132 in PCIe x1 slot: 17Ktps
ICH8 AHCI on-board: 17Ktps

10 streams linear reads with 512B block:
SiI3114 in PCIX 32/66 slot: 10Ktps.
SiI3132 in PCIe x1 slot: 25Ktps
ICH8 AHCI on-board: 25Ktps

But even slower newer card can win in some situations, just because of 
additional functionality, such as command queuing in this case.

-- 
Alexander Motin



Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?4A95A3CA.3060306>