[SI-LIST] Re: PCI at 66 Mhz
- From: Michael Nudelman <mnudelman@xxxxxxxxxxx>
- To: "'david_instone@xxxxxxxxxxxxxx'" <david_instone@xxxxxxxxxxxxxx>,SI <si-list@xxxxxxxxxxxxx>
- Date: Fri, 31 Aug 2001 09:03:22 -0400
David,
Hate to disappoint you, but:
No SCSI harddrive today even the fastest ones, like Quantum (maxtor) 10KII ,
or even 10KIII, or newest 15K RPM Cheetahs, have internal data transfer rate
higher than 80Mb/s. Most have 25-45MB/s, some 35-50.
But this is not the worst news - the "formatted sustained" speed is so far
no more than 45MB/s.
So: for short bursts and with the large on-disc cache like 8Mb you may
temporarily achieve maximum Disc-to-card thruput of 160Mb/s, which than will
be limited somewhat by PCI bottleneck. Which is not quite the bottleneck,
because as you understand, the real bottleneck is the drive media speed
itself. Which is about half of the PCI throughput.
However, in the real system, the PCI bandwidth is even less.
I remember one of the first QLogic Fibre PCI cards; at the time I was
designing RAID5 controller using their 2100 chip and we bumped into a bug.
As the QLogic investigated our complaint, they found out that they never
used the length of burst we used, and the bug wsa seen only at large bursts.
AT least at the time the PCI devices did not use large bursts, and it is
possible (you do not want a PCI device to hog the bus, and so you put a
limited value at the largest burst length register and in the latency
counter) they do not use them today.
And the smaller the burst, the higher the overhead and the lower the thruput
you have. (playinf the devil's advocate, though it does not really matter).
So, at least for a couple of years, while the disc speed is roughly the half
of the PCI I would not worry about the PCI interface for 160 cards anyway.
Mike N.
-----Original Message-----
From: david_instone@xxxxxxxxxxxxxx [mailto:david_instone@xxxxxxxxxxxxxx]
Sent: Friday, August 31, 2001 4:14 AM
To: SI
Subject: [SI-LIST] Re: PCI at 66 Mhz
This is somewhat of the original topic, but if I have a PCI ultra160
SCSI adapter card and plug it into a 33MHz 32 bit PCI bus I think the
max throughput would be 4 * 33 = 132 MBytes per sec, so I'm never going
to be able to get 160 MB/s continuous out of the drive. It would
appear that I would need to use a 66MHz PCI bus to achieve this, which
limits me to only one other card in the PC.
Am I right?
> The PCI Spec does hint that it may be impossible to
> have more than one PCI card per 66 MHz PCI bus, but I am aware that people
> have made 66 MHz PCI work with two card slots. How well they work, I
can't
> say. "It ain't easy."
--
Regards
Dave Instone.
------------------------------------------------------------------
To unsubscribe from si-list:
si-list-request@xxxxxxxxxxxxx with 'unsubscribe' in the Subject field
For help:
si-list-request@xxxxxxxxxxxxx with 'help' in the Subject field
List archives are viewable at:
//www.freelists.org/archives/si-list
or at our remote archives:
http://groups.yahoo.com/group/si-list/messages
Old (prior to June 6, 2001) list archives are viewable at:
http://www.qsl.net/wb6tpu
------------------------------------------------------------------
To unsubscribe from si-list:
si-list-request@xxxxxxxxxxxxx with 'unsubscribe' in the Subject field
For help:
si-list-request@xxxxxxxxxxxxx with 'help' in the Subject field
List archives are viewable at:
//www.freelists.org/archives/si-list
or at our remote archives:
http://groups.yahoo.com/group/si-list/messages
Old (prior to June 6, 2001) list archives are viewable at:
http://www.qsl.net/wb6tpu
Other related posts:
