Dieter wrote: > In message <[email protected]>, Alexander Motin writes: >>>>> - SiI3124-based - fast and functional. It is actually PCI-X one, but >>>>> there are many boards with built-in PCIe bridges. >>> Do any of these fit in a x1 slot? >> I was surprised, but yes. > > My google-fu fails me. Any make/model, URLs, or keywords?
Syba PCI Express SATA II 4 x Ports RAID Controller Card SY-PEX40008 http://www.amazon.com/Syba-Express-Ports-Controller-SY-PEX40008/dp/B002R0DZWQ/ref=sr_1_22?ie=UTF8&s=electronics&qid=1258452902&sr=1-22 >> And as expected, bus limits it performance >> badly. But it is still works much faster then 3132 at the same bus. > > Hmmm, I must not have tested both disks at once before on the 3132: > > One at a time: > > dd if=/dev/ad18 bs=1m count=1000 of=/dev/null > 1048576000 bytes transferred in 8.807009 secs (119061534 bytes/sec) > > dd if=/dev/ad20 bs=1m count=1000 of=/dev/null > 1048576000 bytes transferred in 8.800526 secs (119149243 bytes/sec) > > Two at once: > > dd if=/dev/ad18 bs=1m count=1000 of=/dev/null & dd if=/dev/ad20 bs=1m > count=1000 of=/dev/null > 1048576000 bytes transferred in 13.766050 secs (76171160 bytes/sec) > 1048576000 bytes transferred in 13.799268 secs (75987799 bytes/sec) > > so 152 MB/s total rather than the expected ~238. PCIe x1 slot is supposed to > be good for 250 MB/s, so it ought to be able to max out both disks at once, > or at least get close. I have close numbers. 250MB/s is actually performance of the physical level. Logical level also creates overhead of somewhere about 30-40 bytes per transfer. As soon as most of desktop chipsets limited with 128bytes transfers, it also shouldn't be forgotten. The interesting fact I have seen yesterday, SiI3132 is able to read 150MB/s, but write 170MB/s. I am not PCIe expert, but looks like transfer capabilities could be asymmetric. Also, and as soon as PCIe is duplex, I've also seen 110MB/s read from one drive, plus 100MB/s write to another, running at the same time. -- Alexander Motin _______________________________________________ [email protected] mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-hardware To unsubscribe, send any mail to "[email protected]"
