Date: Sun, 23 Jun 2002 14:01:10 -0400 (EDT) From: Andrew Gallatin <gallatin@cs.duke.edu> To: Yoriaki FUJIMORI <fujimori@prinz.fujimori.cache.waseda.ac.jp> Cc: obrien@FreeBSD.ORG, freebsd-alpha@FreeBSD.ORG Subject: Re: What are my options PC164 and Symbios 53C895 Message-ID: <15638.3302.913567.234842@grasshopper.cs.duke.edu> In-Reply-To: <200206230755.g5N7t2N53755@prinz.fujimori.cache.waseda.ac.jp> References: <20020622231921.A22789@dragon.nuxi.com> <200206230755.g5N7t2N53755@prinz.fujimori.cache.waseda.ac.jp>
next in thread | previous in thread | raw e-mail | index | archive | help
Yoriaki FUJIMORI writes:
> One more comment. You cannot exploit the performance of LVD on 21164
> systems. (Not on entry level 21264 systems, either.) I did not measure
> it precisely, but my feeling is the top speed of pci bus on 164LX is
> still below 30MB/sec. Naturally, much worse on PC164.
Its not quite that bad.
For large tranfers, a 32-bit card should be able to do ~70MB/sec for
DMA reads (write to disk/ send on network) and over 100MB/sec for DMA
writes (read from disk, recv on network). Here is actual data from a
164SX, which is somewhat similar to the LX:
DMA rate for 8192 Byte pages (32bit / 33MHz bus)
Timing 32 pages.
bus_read (send) = 72 MBytes/s
bus_write (recv) = 126 MBytes/s
I measured my UP1000 and it was also somewhere in the this
neighborhood.
64-bit cards in tsunami based machines are quite a bit better, but
still are nowhere near as good as most modern x86 (or even Apple)
chipsets.
Drew
To Unsubscribe: send mail to majordomo@FreeBSD.org
with "unsubscribe freebsd-alpha" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?15638.3302.913567.234842>