Which Xeon motherboard? with 800Mhz FSB and AGP slot.

  • Thread starter Thread starter William.R.Reisen
  • Start date Start date
W

William.R.Reisen

Hi,

I'm putting together a Xeon system. I have the requirement that I want
to use an existing expensive AGP card that I have and that I don't
want to make it too expensive.

A lot of the Socket 604 Xeon 800Mhz FSB motherboards have only PCI-X
slots and no AGP slot.

The board that I have found that closest matches my requirements is
this:
Iwill DH800 Dual Socket 604 Motherboard $335 (220 UK pounds)
http://www.amplicon.co.uk/data/iwilldh800.html
http://www.outpost.com/product/4089832

Are there other cheaper ones out there that have the AGP slot with the
socket 604 800Mhz?

Thank you.
 
William.R.Reisen said:
Hi,

I'm putting together a Xeon system.
Click to expand...

Why would you want to do that? Why not build an Opteron system instead?
I have the requirement that I want
to use an existing expensive AGP card that I have and that I don't
want to make it too expensive.

A lot of the Socket 604 Xeon 800Mhz FSB motherboards have only PCI-X
slots and no AGP slot.
Click to expand...

Another reason to build an Opteron machine.
The board that I have found that closest matches my requirements is
this:
Iwill DH800 Dual Socket 604 Motherboard $335 (220 UK pounds)
http://www.amplicon.co.uk/data/iwilldh800.html
http://www.outpost.com/product/4089832

Are there other cheaper ones out there that have the AGP slot with the
socket 604 800Mhz?
Click to expand...

$700 each for Nocona Xeon 3.4 ghz chips? You can get Opteron 246 chips
for only around $440 each.
 
William.R.Reisen said:
Hi,

I'm putting together a Xeon system. I have the requirement that I want
to use an existing expensive AGP card that I have and that I don't
want to make it too expensive.

A lot of the Socket 604 Xeon 800Mhz FSB motherboards have only PCI-X
slots and no AGP slot.
Click to expand...

Many server motherboards aren't concerned about graphics (for obvious
reasons). If you're looking for workstation, there are several with AGP.
Do a google search and you should come up with some quick matches.
 
Hi,

I'm putting together a Xeon system. I have the requirement that I want
to use an existing expensive AGP card that I have and that I don't
want to make it too expensive.

A lot of the Socket 604 Xeon 800Mhz FSB motherboards have only PCI-X
slots and no AGP slot.

The board that I have found that closest matches my requirements is
this:
Iwill DH800 Dual Socket 604 Motherboard $335 (220 UK pounds)
http://www.amplicon.co.uk/data/iwilldh800.html
http://www.outpost.com/product/4089832

Are there other cheaper ones out there that have the AGP slot with the
socket 604 800Mhz?
Click to expand...

Such designs are rather rare beasts, you aren't likely to find many
motherboards with both support for a pair of 800MHz bus speed Xeons
AND an AGP slot. Doing so requires the use of the i875 chipset
(generally targeted at home user/enthusiast boards) as a
dual-processor workstation board, a very rare combination. The only
other board that I'm aware of that would fit this bill is Asus' new
NCCH-DL board:

http://usa.asus.com/products/server/srv-mb/ncch-dl/overview.htm

I suspect that this board will be in the same basic price range as the
IWill board though, perhaps even more expensive.


The real problem here is that the only chipsets that have been
designed with these new 800MT/s bus speed Xeons in mind all use
PCI-Express for graphics instead of AGP. Eventually PCI-Express will
replace AGP as the standard for graphics controllers; fine for going
forward but it doesn't help people in your situation much.


Another option that you might want to look into is a dual AMD Opteron
system instead. I don't know if the Opteron would fit your bill or
not, but it does offer more choice in terms of motherboards supporting
AGP slots.
 
Most meaning AMD shillhouse benchmarks.
Click to expand...

Like Infoworld? A typical corporate IT publication that in the past was
pretty sceptical about anything AMD.

http://www.infoworld.com/article/04/09/17/38FE64shootout2_1.html

A Dual 2.4GHz Opteron roughly 20% faster than a Dual 3.6GHz Xeon on a
bunch of database and web serving tasks.

Nearly every other comparison I've seen indicates similar numbers, but I'd
be more than happy to see some conflicting results if you have them...

Cheers
Anton
 
AD. said:
Like Infoworld? A typical corporate IT publication that in the past was
pretty sceptical about anything AMD.

http://www.infoworld.com/article/04/09/17/38FE64shootout2_1.html

A Dual 2.4GHz Opteron roughly 20% faster than a Dual 3.6GHz Xeon on a
bunch of database and web serving tasks.
Click to expand...

Your own link says Opteron wins on 2 tests and Xeon on the 3rd. One problem
is that gcc currently produces poorly optimized code for Intel platforms. I
have icc 8.1. I'd love to see what that does on both systems as it
generally optimizes AMD platforms nearly double what gcc does as well.
 
Your own link says Opteron wins on 2 tests and Xeon on the 3rd.
Click to expand...

So is it better or not? Quote:

"The test results were conclusive. In every real-world test, the Opteron
250-based Newisys server bested the EM64T Xeon server, despite the fact
that the latter had a faster clock speed."

And of the one artificial benchmark that the Xeon won, it says this:

"But that's not the end of the story. As Goto says, "High scores on the
HPL benchmark do not mean 'high performance computing.'" The
routines in the HPL DGEMM routines can hide long cache latency, which is a
problem on EM64T processors."

Kazushige Goto is one of the writers of that benchmark.
One
problem is that gcc currently produces poorly optimized code for Intel
platforms.
Click to expand...

Evidence? So what? gcc is what will be used by everyone running that
particular software in the 'real world' unless they want to give up their
support.

Even if you are correct and even if it is a valid concern, it's not like
those kind of compiler advantages haven't helped Intel more often than AMD.
I have icc 8.1. I'd love to see what that does on both
systems as it generally optimizes AMD platforms nearly double what gcc
does as well.
Click to expand...

That would be interesting. Be sure to let us know when you test it or find
someone else's test.

Don't get me wrong, I'd be interested in hearing about how the Xeon is
faster or better than the Opteron but I've yet to see anything that shows
that. After all, it was you that claimed it was and haven't offered any
evidence to back it up.

Cheers
Anton
 
AD. said:
So is it better or not? Quote:

"The test results were conclusive. In every real-world test, the Opteron
250-based Newisys server bested the EM64T Xeon server, despite the fact
that the latter had a faster clock speed."
Click to expand...

Every, meaning 2.
And of the one artificial benchmark that the Xeon won, it says this:

"But that's not the end of the story. As Goto says, "High scores on the
HPL benchmark do not mean 'high performance computing.'" The
routines in the HPL DGEMM routines can hide long cache latency, which is a
problem on EM64T processors."

Kazushige Goto is one of the writers of that benchmark.
Click to expand...

A well written program can hide cache latency.
Evidence? So what? gcc is what will be used by everyone running that
particular software in the 'real world' unless they want to give up their
support.
Click to expand...

Unless your company chooses to not use gcc due to poor optimizations.
Even if you are correct and even if it is a valid concern, it's not like
those kind of compiler advantages haven't helped Intel more often than
Click to expand...
AMD.

I have no idea what you're saying.
That would be interesting. Be sure to let us know when you test it or find
someone else's test.
Click to expand...

If I have an opportunity, I'll be sure and test.
Don't get me wrong, I'd be interested in hearing about how the Xeon is
faster or better than the Opteron but I've yet to see anything that shows
that. After all, it was you that claimed it was and haven't offered any
evidence to back it up.
Click to expand...

I claimed it was better. Better meaning more reliable, better support, and
yes, using our compilers and source code knowledge, faster. I just don't
write magazine articles for "evidence".

cheers
 
Every, meaning 2.
Click to expand...

There were 2 real world tests and one artificial one.

So you accept those two were 'real world' while the third one wasn't?

It is a small sample admittedly, but they correspond with other tests
I've seen. As I said, feel free to add counter examples.
A well written program can hide cache latency.
Click to expand...

So what? You are criticising the software not the hardware. Most servers
don't get installed by programmers to run their own code, they get
installed by admins who have to deal with real world apps that are chosen
for plenty of other reasons than their ability to hide cache latency.

If the Opterons lower cache latency helps it run software better without
requiring programmers to 'hide cache latency', isn't that a real world
benefit?
Unless your company chooses to not use gcc due to poor optimizations.
Click to expand...

Which company is that? The ones that recompile their expensive Red Hat and
Suse Enterprise editions from scratch with Intels compiler and probably
jeopardise the support that they're paying for in the first place?

You seem to be overly concerned with custom high performance computing
scenarios which is fine, but most server installations would be more
concerned with how the hardware runs their 'off the shelf' 'real world'
applications.
AMD.

I have no idea what you're saying.
Click to expand...

I'm just saying that argument has at different times and with different
compilers cut both ways and is effectively moot. Most users don't
get much say in which compiler is used to compile the software they use.
I claimed it was better. Better meaning more reliable, better support,
and yes, using our compilers and source code knowledge, faster. I just
don't write magazine articles for "evidence".
Click to expand...

So someone on the internet should just take your word for it without
evidence, when there is plenty of evidence to the contrary?

I'm not discounting your experience, but it must either be limited to a
specialised niche or you have optimised your code so much for Intel that
the Opteron is at a disadvantage. Either is a perfectly valid reason for
preferring Intel, but doesn't necessarily translate into useful
data points for everyday server use.

Cheers
Anton
 
So someone on the internet should just take your word for it without
evidence, when there is plenty of evidence to the contrary?

I'm not discounting your experience, but it must either be limited to a
specialised niche or you have optimised your code so much for Intel that
the Opteron is at a disadvantage. Either is a perfectly valid reason for
preferring Intel, but doesn't necessarily translate into useful
data points for everyday server use.
Click to expand...

Problem is, the 64-bit x86 is very, very, young. It's not useful for any
data points due to it's newness. The software and optimizations WILL get
better as it becomes more mainstream. There aren't a whole lot of tests
between Opteron and Xeon that are 64-bit. Most are 32-bit tests which I
suppose are valid since there are few 64-bit applications out there.

http://www.tomshardware.com/cpu/20040927/opteron_vs_xeon-29.html

AMD - IIIIIIIIIIIIIIII
INTEL - IIIIIIIIIIIIIIIII

Looking at Tom's site, I counted each graphic that had Intel ahead and each
that had AMD ahead. It's a virtual deadlock. I don't get anything out of
it though since it doesn't tell me about anything I do.
 
AMD shillhouse = Anandtech
Click to expand...

Ok, would you prefer SPEC? In CFP2000 Rate (base) a pair of Opteron
250s, in a Sun w2100, manages a score of 35.2 vs. the best Xeon
3.6GHz/800MT/s bus speed score of 28.2 in an IBM e336.

http://www.spec.org/cpu2000/results/rfp2000.html

The Xeon is at least closer in CINT2000 Rate (base) with a score of
33.1 (HP ML370 G4) vs 35.2 for the Opteron (AMD's own setup, Tyan
board).

http://www.spec.org/cpu2000/results/rint2000.html


It is quite close now in SPEC JBB2000, with the dual Opteron (AMD's
own test box) scoring 79544 vs. the dual Xeon 3.6GHz (Dell PowerEdge
2850) score of 78339. The two machines in this case where running
nearly identical software.

http://www.spec.org/jbb2000/results/jbb2000.html


The story continues with SPEC web99_ssl with a pair of Opterons (Sun
Fire v20z) managing a score of 2500 vs. a pair of Xeons (IBM e346)
managing a score of 2400

http://www.spec.org/web99ssl/results/web99ssl.html


Note that I wasn't picking and choosing results here, that's basically
every benchmark on SPECs site in which both chips were compared, using
the fastest result available for each chip. The only ones I didn't
check where the SPEC "Graphics" tests, which are kind of a PITA to
compare one chip to another in since they are so heavily dependant on
the video card used.


Really, there's nothing AMD fanboy-esque about saying that the Opteron
is faster than the Xeon. It *IS* a faster processor in 2P solutions
and a MUCH faster chip in 4P setups. The Xeons shared bus design,
even with the new 800MT/s bus speeds, really hinders it's performance
relative to the integrated memory controller + hypertransport setup of
the Opteron. With the combination of AMD's on-die memory controller
and their high-quality 8000 series chipsets you also get all the same
reliability features.


Of course, in the end it depends largely on what your specific
application requires. There are some apps that are MUCH faster with
the Opteron and some that are MUCH faster with the Xeon. Most though,
are pretty close.
 
Looking at Tom's site, I counted each graphic that had Intel ahead and
each that had AMD ahead. It's a virtual deadlock. I don't get anything
out of it though since it doesn't tell me about anything I do.
Click to expand...

When you ripped into AnandTech, I thought no problem... he just has high
standards for that kind of testing.

But then, you bring up Toms Hardware - whoops!

So your beef with Anand doesn't seem to be with the quality of his
analysis or methodologies anymore (or else you wouldn't have cited Toms
Hardware), just that he doesn't come to the conclusions you want.

I wouldn't quibble with anyone that thought Anand's stuff left a bit to be
desired, but Toms site has historically left a lot more to be desired and
been much closer to being a shill (your word) for certain vendors. Maybe
Toms Hardware has improved lately?

Don't take this the wrong way, but you're just as one-eyed as JK - it's
just that you've got the other eye.
 
AD. said:
When you ripped into AnandTech, I thought no problem... he just has high
standards for that kind of testing.

But then, you bring up Toms Hardware - whoops!

So your beef with Anand doesn't seem to be with the quality of his
analysis or methodologies anymore (or else you wouldn't have cited Toms
Hardware), just that he doesn't come to the conclusions you want.

I wouldn't quibble with anyone that thought Anand's stuff left a bit to be
desired, but Toms site has historically left a lot more to be desired and
been much closer to being a shill (your word) for certain vendors. Maybe
Toms Hardware has improved lately?

Don't take this the wrong way, but you're just as one-eyed as JK - it's
just that you've got the other eye.
Click to expand...

Not that I give a damn, but this is comp.sys.intel that this has been
crossposted to. No doubt, the people on there use Intel CPUs. If I were
JK, I would be on comp.amd.whatever chirping every 2 seconds about Intel.
Anand is definitely an AMD shill. Nothing you say changes that, but I'm
sure in your opinionated way, you will find something to say about that, but
I'm done with this conversation with you. Ciao!
 
Back
Top