[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: Regarding Kirkwood cpu support



On Tue, Mar 17, 2009 at 12:17:39AM +0100, Laurent GUERBY wrote:

> > > we have also ordered a few SheevaPlugs but our retailer said that it
> > > will take at least two more weeks to deliver. We are very interested
> > > in data throughput, could you please post some very basic ethernet
> > > performace numbers?
> > 
> > My main concern so far has been to get a working kernel image into the
> > archive as soon as possible, so I haven't had a chance to look at
> > performance yet.
> > 
> > What kind of test would you like to see?
> 
> Basic test from the plug "gcc55" to my PC "pc2":
> 
> gcc55 ~ # iperf -c pc2  -N -n 1000M
> ------------------------------------------------------------
> Client connecting to pc2, TCP port 5001
> TCP window size: 16.0 KByte (default)
> ------------------------------------------------------------
> [  3] local 192.168.1.118 port 45282 connected with 192.168.1.101 port 5001
> [  3]  0.0-51.5 sec  1000 MBytes    163 Mbits/sec
> 
> My PC has:
> 02:00.0 Ethernet controller [0200]: Realtek Semiconductor Co., Ltd. RTL8111/8168B PCI Express Gigabit Ethernet controller [10ec:8168] (rev 02)
> 
> Note: the plug was running two compilations by farm users during the
> network test and the kernel has lock checking enabled so it's
> unlikely an upper bound :).

On the 6281 at 1.2 GHz, I get wire speed TCP transmit when GSO is
enabled, and wire speed TCP receive when LRO is enabled (which
mv643xx_eth supports since recently -- it's currently in net-next).

Note that the onboard (PCI) gige network card in my desktop machine
was the bottleneck in most of the tests I tried initially -- I couldn't
saturate the KW board until I put a relatively beefy PCIe NIC in my
desktop.  (The desktop machine itself is a Core 2 Quad, so that isn't
the issue, it's purely the onboard PCI NIC being crap.)


Reply to: