<!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.0 Transitional//EN">
<HTML><HEAD>
<META content="text/html; charset=us-ascii" http-equiv=Content-Type>
<META name=GENERATOR content="MSHTML 8.00.6001.18702"></HEAD>
<BODY>
<DIV><SPAN class=328452909-10032010><FONT color=#0000ff size=2 face=Arial>A few
comments about your question:</FONT></SPAN></DIV>
<DIV><SPAN class=328452909-10032010><FONT color=#0000ff size=2
face=Arial></FONT></SPAN> </DIV>
<DIV><SPAN class=328452909-10032010><FONT color=#0000ff size=2 face=Arial>1) SDR
cards can only reach about 7.7G bit of bw due to the 8:10 encoding and other
overhead.</FONT></SPAN></DIV>
<DIV><SPAN class=328452909-10032010><FONT color=#0000ff size=2 face=Arial>2)
There haven't been a lot of work done on performance on the last years, but
checksum offload and large send offload have been added (assuming that your card
support it. Connectx is an example to a card that supports this
features).</FONT></SPAN></DIV>
<DIV><SPAN class=328452909-10032010><FONT color=#0000ff size=2 face=Arial>3)
Using 4k MTU should allow you to reach higher bw. (again, you will need
Connectx).</FONT></SPAN></DIV>
<DIV><SPAN class=328452909-10032010><FONT color=#0000ff size=2 face=Arial>4)
Since computers are much stronger today than they have been on 2007, you should
be able to reach very close to the theoretical bw (of SDR). </FONT></SPAN></DIV>
<DIV><SPAN class=328452909-10032010><FONT color=#0000ff size=2 face=Arial>5)
When working with full bw cpu of the receiver will be the
bottleneck.</FONT></SPAN></DIV>
<DIV><SPAN class=328452909-10032010><FONT color=#0000ff size=2 face=Arial>6) Our
software is being tested on windows server 2008 R2. This is one of the operating
systems that we do whql on. We believe that it should also work for windows 7 64
bits, but we don't really test there.</FONT></SPAN></DIV>
<DIV><SPAN class=328452909-10032010><FONT color=#0000ff size=2
face=Arial></FONT></SPAN> </DIV>
<DIV><SPAN class=328452909-10032010><FONT color=#0000ff size=2
face=Arial>Thanks</FONT></SPAN></DIV>
<DIV><SPAN class=328452909-10032010><FONT color=#0000ff size=2
face=Arial>Tzachi</FONT></SPAN></DIV>
<DIV><SPAN class=328452909-10032010><FONT color=#0000ff size=2
face=Arial></FONT></SPAN> </DIV><BR>
<BLOCKQUOTE
style="BORDER-LEFT: #0000ff 2px solid; PADDING-LEFT: 5px; MARGIN-LEFT: 5px; MARGIN-RIGHT: 0px">
<DIV dir=ltr lang=en-us class=OutlookMessageHeader align=left>
<HR tabIndex=-1>
<FONT size=2 face=Tahoma><B>From:</B> ofw-bounces@lists.openfabrics.org
[mailto:ofw-bounces@lists.openfabrics.org] <B>On Behalf Of </B>Ishaan
Dalal<BR><B>Sent:</B> Monday, March 08, 2010 2:03 AM<BR><B>To:</B>
ofw@lists.openfabrics.org<BR><B>Subject:</B> [ofw] Current IPoIB performance
with WinOF<BR></FONT><BR></DIV>
<DIV></DIV>Hi,<BR><BR>I'd like to experiment with IB as a cheap 10+ Gbps
point-to-point interconnect (4x SDR adapters are available quite cheap)
between a video editing workstation (Win 7 x64) and a storage server (Server
2008 R2). SMB will be used to maintain compatibility with other existing
clients. The server is a fairly powerful machine (dual-core Core i5) with a
RAID subsystem that can saturate a gigabit link several times
over.<BR><BR>Digging through list archives to determine IPoIB performance in
Windows, the latest I could find were posts from 2006-2007 that claimed
performance was poor because the way IPoIB was implemented was extremely
CPU-intensive. How much has IPoIB performance improved with the newer
WinOF/NDIS versions? Is it still as tightly
CPU-bound?<BR><BR>Thanks,<BR>-Ishaan<BR></BLOCKQUOTE></BODY></HTML>