[AG-TECH] New hardware specification for AG 2.0?

Osland, CD (Chris) C.D.Osland at rl.ac.uk
Wed Feb 19 05:34:44 CST 2003


I know it's really bad form to reply to your own mailing, but I
left out a couple of caveats ...

Running a single machine for all three major functions means
running them all under one OS - probably Windows, but I have no
preference.  I don't know what, if anything, this compromises,
as I haven't explored the PIG discussions yet.

If the basic software (vic, rat) doesn't multi-thread, I guess
there may either be problems, or there could be a problem of
capture vic and display vic running on the same processor - I
don't know how to avoid that, if it is a problem.

I'm assuming that the audio side of the Osprey 220 cards can
be used by rat - although I have no proof of that.

Also (not a caveat but a postscript)  ...

As I've mentioned in another thread, we're interested in higher
resolution video streams - PAL for starters, XGA would also be
nice - and these will presumably require CPU power, presumably
in proportion to the area (in pixels).  Therefore, with the
current video streams being about 110K pixels and requiring
about 1400/20 = 70 MHz, PAL streams (768x576 = 440K pixels
will probably require no less than 280 MHz per stream, and 
XGA (1024x768 = 800K pixels) will require about 500 MHz per
stream.  This assumes that the encoding algorithms are
equally [in]efficient, which is probably a false assumption!
Tests of video streaming with a non-Osprey card showed that
it was not difficult to use all the power on an AMD 1600,
so it looks like the higher-res image streams will be able to
mop up CPU speed upgrades for a few more years!  Mercifully
it looks as though our usage is likely to be either a large
number of sites are current resolution, or a small number of
sites with a few higher res streams.

Cheers

Chris



____________________________________________________________________
Chris Osland                         Office tel: +44 (0) 1235 446565
Digital Media and Access Grid      Medialab tel: +44 (0) 1235 446459
BIT Department             Access Grid room tel: +44 (0) 1235 445666
e-mail:   C.D.Osland at rl.ac.uk               Fax: +44 (0) 1235 445597

CLRC Rutherford Appleton Laboratory (Bldg. R18)
Chilton, DIDCOT, Oxon OX11 0QX, UK

[The contents of this email are confidential and 
are for the use of the intended recipient only.
If you are not the intended recipient do not take 
any action on it or show it to anyone else,
but return this email to the sender and delete your copy of it.]





-----Original Message-----
From: Osland, CD (Chris) [mailto:C.D.Osland at rl.ac.uk]
Sent: 19 February 2003 10:39
To: ag-tech at mcs.anl.gov
Subject: RE: [AG-TECH] New hardware specification for AG 2.0?


[Warning - this reply has grown a bit long...!]

Can I chip in with some thoughts on this.  I don't think AG 2.0
is actually the agent of change, so much as the passing of time.
When the 'recommended' hardware (which has previously been noted
is more of a 'reference known to work' hardware set) was written
down, 500 MHz was about tops for CPU power.  If you wanted more,
you went for dual 500 MHz.

Now AMD 2200+ MPs is equally affordable on a per-chip basis, as are
2.xx GHz Pentium 4s.  Dual processor motherboards are not as
expensive, compared to single processor ones, as they were, and they
appear to be much more reliable than dual mobos a few years ago.

As a result, it appears to me that a single machine with dual
processors (AMD or Intel to personal choice) can provide the
power to run what I'm currently running on 3 1.4 GHz AMD single
proc machines:

	display machine peaking at 100% utilization with about
	20-25 video windows (and about half that when encrypted) (**)

	capture machine running at about 15% with 4 Hauppage cards
	active

	audio machine barely idling

(**) The fly in the ointment is the Gentner control software which
hogs all remaining cycles as a result of polling over a serial
line (ugh!!!).  Setting its Windows priority to Low (the lowest)
improves things, but in critical meetings I still kill it off
to avoid 'sticky windows' problems.

I like things rack-mounted, so am considering a single 4U machine
with, say, twin 2400 MP processors (or whatever is affordable the
week I place the order), Osprey twin capture cards (2 off 220s -
which have the advantage of balanced audio I/O - saves a card
and a bal-unbal box), and possible a 1U pensioned off server box
just for the Gentner software!!!  For display I like the idea of
the Parhelia 3-output card (AGP) for the wall and something
OK but not earth-shattering for the monitor - I might use a dual
output card for that, as the combined windows from display, audio
and video capture functions might be a bit crowded on a single
monitor - my eyes aren't up to 1600x1200 displays unless they're
about 32" wide!

In summary, the power needed these days depends on the most complex
meetings you aim to handle effortlessly.  If you're regularly
running 15 site meetings encrypted (we are headed that way) the
display machine will probably need as much power as you can cram
in one case all to itself.  If 8 sites without encryption is more
your style, you can probably put everything in one box, insofar
as CPU power is concerned.

Just my 0.02's worth.

Chris

____________________________________________________________________
Chris Osland                         Office tel: +44 (0) 1235 446565
Digital Media and Access Grid      Medialab tel: +44 (0) 1235 446459
BIT Department             Access Grid room tel: +44 (0) 1235 445666
e-mail:   C.D.Osland at rl.ac.uk               Fax: +44 (0) 1235 445597

CLRC Rutherford Appleton Laboratory (Bldg. R18)
Chilton, DIDCOT, Oxon OX11 0QX, UK

[The contents of this email are confidential and 
are for the use of the intended recipient only.
If you are not the intended recipient do not take 
any action on it or show it to anyone else,
but return this email to the sender and delete your copy of it.]





-----Original Message-----
From: Jon Johansson [mailto:jon.johansson at ualberta.ca]
Sent: 18 February 2003 22:45
To: ag-tech at mcs.anl.gov
Subject: [AG-TECH] New hardware specification for AG 2.0?


Hi, 

We have finally got some money to buy hardware for an AG Room
and I am now wondering whether the Access Grid Hardware Specification
is going to change for AG 2.0. I know that several groups have 
got nodes working with 2 computers (one for display and control, and 
the other for both audio and video capture), is this going to become
the new specification? This question is motivated by my interest in
buying pieces, installing software and, as much as possible, having
it work through the change from AG1 to AG2.

We are considering the following hardware: 

   Display computer:	dual Xeon - 1 GB Ram
                        Matrox G450 Dual Head AGP Graphics Adapter
                        Matrox G200 Quad PCI Graphics Adapter
   Video Cap computer:	dual Xeon - 1 GB Ram
                        Hauppauge video capture adapters x4
   Audio Cap computer:	single Xeon - 512 MB Ram
                        Creative Ensoniq AudioPCI Audio Adapter x4
   Control computer:	single Xeon - 512 MB Ram

All computers have Ethernet and ATI RAGE XL 8MB video on the motherboards
and IDE drives. 

I am particularly interested in audio and video capture cards with multiple
inputs, are there any that people are finding usable? These seem to be 
necessary if you are going to reduce the number of PCI slots available
by reducing the number of computers.
The CPU usage on the capture computers must depend on how much processing
is done in the capture cards, are there multi-input capture cards that
will leave little for the CPU to do in terms of encoding?

Are there alternatives to the Matrox graphics cards that are cheaper/
better?

The AG specification gives a price for the Matrox G450 Dual Head AGP 
Graphics Adapter of $200. When I go to the Matrox website the G450 cards
seem to be PCI now and the prices are $600US for the dual head and $800 US
for the quad head. Matrox doesn't seem to be selling the AGP cards anymore.
Are these PCI cards acceptable for driving the displays? 
Are there alternative AGP graphics cards that would be suitable?

Thank you,
Jon Johansson.

-- 
Jon I Johansson             *  Research Computing Support
jon.johansson at ualberta.ca   *  Computing and Network Services
Tel.: (780) 492-9304        *  University of Alberta
Fax.: (780) 492-1729        *  Edmonton, Alberta, CANADA



More information about the ag-tech mailing list