Subject: Re: individual hme sbus cards vs quad hme sbus cards
To: Bruce O'Neel <edoneel@sdf.lonestar.org>
From: Greg A. Woods <woods@weird.com>
List: port-sparc
Date: 08/12/2006 17:06:47
--pgp-sign-Multipart_Sat_Aug_12_17:06:39_2006-1
Content-Type: text/plain; charset=US-ASCII
Content-Transfer-Encoding: quoted-printable

At Sat, 12 Aug 2006 18:51:56 +0000,
Bruce O'Neel wrote:
>=20
> On Thu, Aug 10, 2006 at 08:00:53PM +0200, doomwarrior wrote:
> <snip>
> >=20
> > so i start my apache on the 'sparky' and wget from my  pc, first with=20
> > le0, second with the hme0
> >=20
> > 96.419.840   876.52K/s    ETA 00:00
> > 96.419.840     2.13M/s    ETA 00:00
> >=20
> > only 2Megs per seconds.  I thourghed I would get at least a boost to=20
> > 5Megs or so.  What can I do to speed up the system to (at least)
> > 3Megs, because I want to upgrade to a faster internet connection,=20
> > without changing my lovely 'sparky'.
>=20
> I've never been able to get more than 2 some meg/secs through a hme on
> a ss20 when the data is going into user space (say via ftp, wget,
> etc).

On my relatively slow ss20 I get a consistent rate of just over 4 MB/s
with FTP from the fastest machine on my network:

ftp> get cd3.iso /dev/null
local: /dev/null remote: cd3.iso
229 Entering Extended Passive Mode (|||58545|)
150 Opening BINARY mode data connection for 'cd3.iso' (375980032 bytes).
226 Transfer complete.
375980032 bytes received in 01:27 (4.10 MB/s)

cpu0 at mainbus0: TMS390Z50 v0 or TMS390Z55 @ 75 MHz, on-chip FPU
cpu0: physical 20K instruction (64 b/l), 16K data (32 b/l), 1024K external =
(32 b/l): cache enabled

hme0 at sbus0 slot 3 offset 0x8c00000 level 4 (ipl 7): Sun Happy Meal Ether=
net (SUNW,hme)
hme0: Ethernet address 08:00:20:71:6a:18
nsphy0 at hme0 phy 1: NS DP83840 10/100 media interface, rev. 0
nsphy0: 10baseT, 10baseT-FDX, 100baseTX, 100baseTX-FDX, auto

(and of course it is running at 100baseTX-FDX through a good switch to
the fast server)

(My 85 MHz sparc-5 is still easily fast enough for a 4 Mbit aDSL
connection, even when running quite a few IPF rules and altq.  I have a
quad-HME card in it and only recently moved the LAN connection over from
le0 to an HME port when I got an even newer 100baseTX switch.)


> With my collection of SS20s and Mbus cards I've found that:
>=20
> - 180mhz Hypersparcs are the fastest in raw CPU.  That said they don't
> do such a good job with i/o and things go downhill fast i/o wise when
> you have 2 of them.  My guess here is that the mbus gets saturated
> talking to memory.

hmmm.... dunno much about the mbus specs myself, but that makes sense.

> - 85mhz SuperSparc IIs are quite fast i/o wise.  The big cache might
> help.

That would be 1MB cache on those ones too?

> - 133 and 142mhz Hypersparcs are also fast, and, i/o wise, seem faster
> than the 180mhz ones.  The 142s with the 1meg of cache are very
> pleasing to work with, but, they also put out a bit of heat.

perhaps they're just slow enough to not often saturate the mbus?

--=20
						Greg A. Woods

H:+1 416 218-0098 W:+1 416 489-5852 x122 VE3TCP RoboHack <woods@robohack.ca>
Planix, Inc. <woods@planix.com>       Secrets of the Weird <woods@weird.com>

--pgp-sign-Multipart_Sat_Aug_12_17:06:39_2006-1
Content-Type: application/pgp-signature
Content-Transfer-Encoding: 7bit

-----BEGIN PGP SIGNATURE-----
Version: PGPfreeware 5.0i for non-commercial use
MessageID: jWWMI/cn9v3qTb3gJ9uWFiaqet8+TIJN

iQA/AwUBRN5C4WJ7XxTCWceFEQJ1SwCg0Y+7fjRZiauLKWalCsH2Xy5vq2IAnjxG
MF9o+4KGDBt8OvXvFaoFOIrM
=GjIJ
-----END PGP SIGNATURE-----

--pgp-sign-Multipart_Sat_Aug_12_17:06:39_2006-1--