lkml.org 
[lkml]   [1999]   [Sep]   [2]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
Date
From
SubjectRe: NT IO Bandwidth limitations?
On Thu, 2 Sep 1999, Jesus David Diaz Leal wrote:

> Anybody knows why the IO bandwidth on a NT Intel box is that much
> limited? Why it is that hard to get even close to 100 Mbyte/sec, if the
> theoretical bandwidth of PCI can be much higher? like: 66Mhz x 32bit bus
> = 66Mhz x4 byte = 266 Mbyte/sec?

None of the common Intel chipsets support 66MHz PCI busses. To the best
of my recollection, the only way to get a 66MHz PCI bus with Intel CPUs is
to buy one of the [ridiculously] expensive N way Xeon boxes.

> What cause this difference between theoretical and practical
> bandwidth? is it CPU, chipset, or cache coherence mechanism?

It's PCI, plain and simple. Remember that PCI runs a seperate cycle to
transfer the adress before using the bus to transfer data. If you're
transfering data sequentially and the burst transfers are long enough,
then you can get closer to PCI's max bandwidth. From the PCI spec, take
the following chart (for 66 MHz, double the numbers):

Data Phases Bandwidth (MB/s)
8 60
16 80
32 96
64 107

> Is it just the ugly Operating System? Is possible to get more on a Linux
> on Intel machine? On an Sgi/Irix, with THE SAME PCI bus you can easily
> get over 200 Mbyte/sec...

It sounds like the SGI is using a 66 MHz PCI bus.

-ben


-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/

\
 
 \ /
  Last update: 2005-03-22 13:53    [W:0.037 / U:20.324 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site