linux-kernel - Slow PCI-E speed

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [day] [month] [year] [list]

Message-ID: <4DA50102.8050900@keff.org>
Date:	Wed, 13 Apr 2011 03:48:50 +0200
From:	sh@...f.org
To:	linux-kernel@...r.kernel.org
Subject: Slow PCI-E speed

Hi

Even if this is probably,even if I really think it's not, hardware
related I'm putting this question out there cause I have no idea on what
can be causing my problems.

Recently I've had the following cards in a HP DL145 G2

1x Myrinet 10G PCI-E 4x
2x Intel Quad 82546EB 1GE NIC
1x OCZ Revodrive 120GB

--

None of the cards above can deliver more than 375MB/s. The tests for the
Myrinet-card is done using multiple wgets. For the Intel Quads using
LACP and multiple wgets. The server I'm wget:ing from has shown
performance >=500Mbit/s in the past. And for the Revodrive using dd/cat.
Everything to /dev/null.

Because the DL145 G2 is quite old I thought it would be the chipsets etc
being the bottleneck.

I then replaced the server with a Supermicro X7DBE (2x E5405 CPUs with
13 GB Ram) which is based on a Intel 5000P Chipset. The Revodrive card
now sits in a dedicated PCI-E 4x slot connected to the 5000P chipset.
The other cards sits on an Intel ESB2-chipset connected with a 10.7GB/s
pipe to the 5000P (where the memory,cpu etc is).

The Revodrive should be able to handle ~500MB/s and well the Myrinet
~1000MB/s (atleast way more than 375MB/s). And the Intel Quads atleast
more than 375MB/s when bonded..

But everything tops out at 375MB/s. And yes..375 = 3Gbit/s :)

--

Like I said this probably isn't a Linux kernel problem but I want to
know if someone has come across something like this before and maybe
maybe...there's some kernel setting that fixes it cause i'm clueless.

Thanks!

// Sebastian H

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