Date: Thu, 14 Aug 1997 12:12:55 +0100 (BST) From: Stephen Roome <steve@visint.co.uk> To: mika ruohotie <bsdhack@shadows.aeon.net> Cc: chat@FreeBSD.ORG Subject: Re: speed test Message-ID: <Pine.BSF.3.95.970814114229.1580M-100000@dylan.visint.co.uk> In-Reply-To: <199708141005.NAA12170@shadows.aeon.net>
next in thread | previous in thread | raw e-mail | index | archive | help
On Thu, 14 Aug 1997, mika ruohotie wrote:
> this one goes to chat... (from hackers)
>
> > > numbers are Mbyt/sec. can anyone beat those tx97-e speeds?
> > root@beast-> dd if=/dev/zero of=/dev/null bs=1m count=1000
> > 1000+0 records in
> > 1000+0 records out
> > 1048576000 bytes transferred in 6 secs (174762666 bytes/sec)
>
> *gasp*
I'm using a plain Intel VX based motherboard with a vanilla Pentium 166
(150 overclocked) and with that test I get:
root@dylan# dd if=/dev/zero of=/dev/null bs=1m count=1000
1000+0 records in
1000+0 records out
1048576000 bytes transferred in 8.404485 secs (124763861 bytes/sec)
I'm just using 60ns EDO on a normal non-overclocked board (66Mhz bus
etc.), but this test is beginning to confuse me..
1) whats the blocksize got to do with the number..
i.e. if you use smaller blocksize does it not go through main
memory and only touch cache ?
2) If I was running this board at 83MHz I would expect (naively perhaps)
that figure to go 83/66 times faster, which *should* give me 156900007
bytes/sec. okay so that's 149Mb/s. (assuming a Megabyte is 1048576 bytes).
3) I've seen stuff on these lists implying that xxx000000 bytes/second
counts as xxxMB/s. It's almost like we're selling hard drives, advertising
size or speed in terms of decimal MB instead of real MB.
> uh, let me re-phrase, can anyone get past that 151,4Mbyt/sec with x86 tech?
> (it was average, highest i saw was somewhere in the 153M class)
[See above (3),] was it really not 153xxxxxx bytes/second, I'd be
interested what the actual maximum theoretical bandwidth is, because I'm
not so sure about this test anymore. Using a PP200 yesterday with
2.1-STABLE I only got about 60Mb/s, and I'm concerned how much of a
reliable test this is.
I was under the impression that at 83Mhz bus speed with good RAM and
assuming absolutely no delays and waits for ram to precharge etc.. blah
blah you should get a top transfer of 166MB/s. Obviously some will be lost
so I wouldn't expect to get much more than that 153M you saw.
> actually, now that i think about it, the alpha is not that much faster at all,
> i have to say i might be even dissapointed to see such a slow stat from
> a machine costing multiple times the tx97-e motherboard...
But the alpha thing is odd, what's the alpha's memory bus speed and what
byte width memory does it use. Is that 166MB/s show above approaching the
alpha's limit (it's one of those limit sounding numbers). If so, x86
platforms will be outperforming this as soon as 100Mhz motherboards come
out.
I wouldn't be surprised if that's not some time in the next few months, or
are they already out.
If so, then that alpha better not have cost more than 2000 UK pounds (or
$2000 I suppose considering the price of hardware here.)
--
Steve Roome - Vision Interactive Ltd.
Tel:+44(0)117 9730597 Home:+44(0)976 241342
WWW: http://dylan.visint.co.uk/
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?Pine.BSF.3.95.970814114229.1580M-100000>