I am looking at the Asus A8V Mbd (Via K8T800 chipset, ~$62.00 on pricewatch) for a hypothetical box, lightweight LAN server. How is Linux support for this chipset ? I have an older (S370 PIII) box, this one as it happens, w/ a Via 694 chipset & love it. Can I look forward to the same great performance & reliability under Linux (FC5 most likely) w/ the K8T800 ? TIA

Tue, 2006-07-11 at 14:27 -0500, William A. Mahaffey III wrote:
I am looking at the Asus A8V Mbd (Via K8T800 chipset,
~$62.00 on pricewatch) for a hypothetical box, lightweight
LAN server.

You don't want to buy a desktop chipset for a server. People do this
all-the-time, even for Intel (e.g., i845/i865), and pay the price in I/
limitations and stability.

Paying $200 for an entry-level, but true _server_ mainboard instead of
$100 for a desktop mainboard is well worth the added $100.

The entry-level ServerWorks HT1000 is a nice solution with a single
PCI-X channel for Socket-939 for under $200. Adding a HT2000 typically
brings up the cost over $300 to what the nVidia nForce Pro with 40 PCIe
channels plus the AMD8131/8132 dual-PCI-X typically costs.

Intel's entry-level 72xx series (based on a license of the ServerWorks
IV/Grand Champion) starts with dual PCIe x8 I/ in the 7230 for under
$200 (sometimes under $150) and you can add an ESB6x00 series for PCI-X
for under $300.

Again, _never_ use a desktop chipset for server duties. And especially
_not_ a "consumer" rated/tested (e.g., i865 instead of i875).

How is Linux support for this chipset ?

ViA typically has ATA controller issues (long story).
is recommended in many cases.

I have an older (S370 PIII) box, this one as it happens, w/
a Via 694 chipset & love it.

The 694 had many performance and stability issues. Especially early
revisions. Latter 694X versions fared better, but were still _not_ a
_server_ chipset.

E.g., the 694 still had only 1/4-1/8 the I/ of a ServerSet IIILE which
started at $250 when they were new (and under $200 later on). I
replaced many i440BX/GX and 693A/694X dual-processor systems with the
IIILE and file server performance was instantly 3x (literally, +200%).

Can I look forward to the same great performance & reliability under
Linux (FC5 most likely) w/ the K8T800 ? TIA

I guess if you're idea of performance and reliability is low, yes. ;->
-- Bryan

P.S. William -- you're not going to believe this, but I wrote this and
was just about to send and just realized you posted it. I honestly
wrote it _before_ I noticed it was you.

P.P.S. Lay off the desktop chipsets -- seriously. I'll loan you an
Intel 72xx (or even a 75xx) series -- you'll see what I mean! ;->

Bryan J. Smith wrote:

Tue, 2006-07-11 at 14:27 -0500, William A. Mahaffey III wrote:

>
>I am looking at the Asus A8V Mbd (Via K8T800 chipset,
$62.00 on pricewatch) for a hypothetical box, lightweight
>>LAN server.
>
>>
>
>You don't want to buy a desktop chipset for a server. People do this
>all-the-time, even for Intel (e.g., i845/i865), and pay the price in I/
>limitations and stability.
>
>Paying $200 for an entry-level, but true _server_ mainboard instead of
>$100 for a desktop mainboard is well worth the added $100.
>
>The entry-level ServerWorks HT1000 is a nice solution with a single
>PCI-X channel for Socket-939 for under $200. Adding a HT2000 typically
>brings up the cost over $300 to what the nVidia nForce Pro with 40 PCIe
>channels plus the AMD8131/8132 dual-PCI-X typically costs.
>
>Intel's entry-level 72xx series (based on a license of the ServerWorks
>IV/Grand Champion) starts with dual PCIe x8 I/ in the 7230 for under
>$200 (sometimes under $150) and you can add an ESB6x00 series for PCI-X
>for under $300.
>
>Again, _never_ use a desktop chipset for server duties. And especially
>_not_ a "consumer" rated/tested (e.g., i865 instead of i875).
>

>
>>How is Linux support for this chipset ?
>
>>
>
>ViA typically has ATA controller issues (long story).
is recommended in many cases.


>
>>I have an older (S370 PIII) box, this one as it happens, w/
>>a Via 694 chipset & love it.
>
>>
>
>The 694 had many performance and stability issues. Especially early
>revisions. Latter 694X versions fared better, but were still _not_ a
>_server_ chipset.
>
>E.g., the 694 still had only 1/4-1/8 the I/ of a ServerSet IIILE which
>started at $250 when they were new (and under $200 later on). I
>replaced many i440BX/GX and 693A/694X dual-processor systems with the
>IIILE and file server performance was instantly 3x (literally, +200%).
>

>
>>Can I look forward to the same great performance & reliability under
>>Linux (FC5 most likely) w/ the K8T800 ? TIA
>
>>
>
>I guess if you're idea of performance and reliability is low, yes. ;->
>
Bryan
>
>P.S. William -- you're not going to believe this, but I wrote this and
>was just about to send and just realized you posted it. I honestly
>wrote it _before_ I noticed it was you.
>
>P.P.S. Lay off the desktop chipsets -- seriously. I'll loan you an
>Intel 72xx (or even a 75xx) series -- you'll see what I mean! ;->
>

, you're just too kind :-). When I say lightweight, I *mean*
lightweight. N public access, N apache, N DNS, N several dozen users
at any 1 time, just CPU, RAM, & code to run ad nauseum. It *will* live
at runlevel 3, but the server traits end there, for the most part. I was
mostly asking about reasonably complete/stable Linux support for that
chipset, decent %-age of full speed for RAM-intensive calculations, etc.
e.g. good S/chipset interaction for my somewhat truncated requirements.
Someone else (another list) indicated all was well, care to dispute that ?

Wed, 2006-07-12 at 00:24 -0400, Bryan J. Smith wrote:
Paying $200 for an entry-level, but true _server_ mainboard instead of
$100 for a desktop mainboard is well worth the added $100.

P.S. William -- you're not going to believe this, but I wrote this and
was just about to send and just realized you posted it. I honestly
wrote it _before_ I noticed it was you.
P.P.S. Lay off the desktop chipsets -- seriously. I'll loan you an
Intel 72xx (or even a 75xx) series -- you'll see what I mean! ;->

FYI, NewEgg _does_ have a server mainboard category:

An Intel E7210 with two (2) PCI-X 66MHz slots starts at $180:

A ServerWorks HT1000 with one (1) PCI-X 133MHz slot starts at $235:

There are also various options with PCIe x4 and x8 slots (with or
without PCI-X) for storage (in case your storage is a PCIe x4 or x8
card) for both single Socket-478/LGA-775 as well as Socket-939.

Bryan J. Smith wrote:

Wed, 2006-07-12 at 00:24 -0400, Bryan J. Smith wrote:

>
>>Paying $200 for an entry-level, but true _server_ mainboard instead of
>>$100 for a desktop mainboard is well worth the added $100.
>
>>P.S. William -- you're not going to believe this, but I wrote this and
>>was just about to send and just realized you posted it. I honestly
>>wrote it _before_ I noticed it was you.
>>P.P.S. Lay off the desktop chipsets -- seriously. I'll loan you an
>>Intel 72xx (or even a 75xx) series -- you'll see what I mean! ;->
>
>>
>
>FYI, NewEgg _does_ have a server mainboard category:

>
>An Intel E7210 with two (2) PCI-X 66MHz slots starts at $180:

>
>A ServerWorks HT1000 with one (1) PCI-X 133MHz slot starts at $235:

>
>There are also various options with PCIe x4 and x8 slots (with or
>without PCI-X) for storage (in case your storage is a PCIe x4 or x8
>card) for both single Socket-478/LGA-775 as well as Socket-939.
>
>

1st link got me the generic server board page, nothing specific, my
misfire ? I *am* looking at , & found the K8T800 board on a price
basis alone. Tyan makes a S939 board, NewEgg got 'em for ~$195.00 last I
looked. I *don't* need the I/ just now, just good RAM/CPU/reliability

Bryan J. Smith wrote:

Wed, 2006-07-12 at 00:24 -0400, Bryan J. Smith wrote:

>
>>Paying $200 for an entry-level, but true _server_ mainboard instead of
>>$100 for a desktop mainboard is well worth the added $100.
>
>>P.S. William -- you're not going to believe this, but I wrote this and
>>was just about to send and just realized you posted it. I honestly
>>wrote it _before_ I noticed it was you.
>>P.P.S. Lay off the desktop chipsets -- seriously. I'll loan you an
>>Intel 72xx (or even a 75xx) series -- you'll see what I mean! ;->
>
>>
>
>FYI, NewEgg _does_ have a server mainboard category:

>
>An Intel E7210 with two (2) PCI-X 66MHz slots starts at $180:

>
>A ServerWorks HT1000 with one (1) PCI-X 133MHz slot starts at $235:

>
>There are also various options with PCIe x4 and x8 slots (with or
>without PCI-X) for storage (in case your storage is a PCIe x4 or x8
>card) for both single Socket-478/LGA-775 as well as Socket-939.
>
>

Just tried that last link, sweet. NewEgg had the Foxconn S940 board for
$195.00 before they (Foxconn) tanked that board, too bad I waited

Tue, 2006-07-11 at 23:40 -0500, William A. Mahaffey III wrote:
, you're just too kind :-).

Seriously William. We had several threads on your issues with the i865
(as well as i845) month after month after month. That wasn't even a
"professional/corporate desktop" chipsets, that's the i875.

I'm just trying to help you friend.

When I say lightweight, I *mean* lightweight. N public access,

What does "public" have _anything_ to do with it?
In fact, LAN I/ is typically far more intense.

N apache, N DNS,

Apache and DNS typically have more to do with _computational_
performance. If you're throwing files around on a LAN, that's I/!

We've had this discussion before, when you've had compatibility and
performance issues. Moving from an Intel i845/i865 (again, not even a
i875 that is tested for professional, desktop use -- and totally
ignoring the E72xx/75xx) to a ViA KT series is a _lateral_ move.

Again, referring back to your experience with the 694 -- back in the
late '90s, I used to replace i440BX/GX and 693/694 chipsets with
ServerSet IIIs -- often for ~$250. A measly $100 more in the mainboard
makes _all_ the difference when you're spending $1K!

N several dozen users at any 1 time, just CPU, RAM, & code to run ad
nauseum.

What is its function?

Again, why save $100 on a system that costs $500-1,000 anyway, when
another $100 will give you server performance and quality?

It *will* live at runlevel 3,

When does "graphics" have _anything_ to do with "storage/communcation"
I/? William, I think this is where you keep missing the point, and why
you keep running into issue after issue on your systems.

I can't believe you've now taken all those issues and discussions we had
months ago with your issues and just thrown them out-the-window. Spend
$175+ on an E72xx series mainboard for P4, or $235 for a HT1000 for

but the server traits end there, for the most part. I was mostly
asking about reasonably complete/stable Linux support for that
chipset, decent %-age of full speed for RAM-intensive calculations,

Can you elaborate on the application?

etc. e.g. good S/chipset interaction for my somewhat truncated
requirements.

Intel 72xx series hsa pretty much been the "Tier-1 PC EM Gold Standard"
for P4 (as well as Xeon) for workstation/entry-level server for 3+ years
(and that's not looking at the new 5000 series which just came out a few
months ago).

AMD, it's been either the aging AMD8000 series, or the 18 month
proven nForce Professional with the AMD8131/8132 added for PCI-X.
the low-end (single Socket-939), the ServerWorks HT1000 does the job for
kernel 2.6 just fine.

Someone else (another list)

Yes, LUNA.

indicated all was well, care to dispute that ?

Yes, I do! And I responded to him as well! It's clear he's used to
desktop uses as well. Desktop I/ is _different_ than server I/, even
just for a few clients.

William, you've had a _lot_ of issues that I've helped you with. I
didn't mind, especially since you said didn't know about the E72xx/75xx
series prior. And we discussed this at length.

But now you're going right back into the rut. And you're going to run
into the same issues. I honestly thought you really appreciated it when
we went into the i8xx/9xx v. E72xx/75xx series differences months back.

I guess not! @-ppp

This isn't a "my way dammit" discussion. This is me saying, "Hey,
remember all those issues you had? I thought you weren't going to dot
hem again?"

Spend an extra $100. It's damn worth it.

Tue, 2006-07-11 at 23:54 -0500, William A. Mahaffey III wrote:
1st link got me the generic server board page, nothing specific, my
misfire ?

It's a _different_ page than the normal mainboard page.

I *am* looking at ,

Just because it's an does _not_ mean any mainboard that takes it
is server-quality! We've had that discussion before too! You'd be
better off going with an older P4 with a cheap E7210 mainboard (although
they are becoming more scarce as the E7221/7230 are more popular now).

& found the K8T800 board on a price basis alone.

Yes! I know!

Tyan makes a S939 board,

Tyan makes server, workstation _and_ desktop mainboards too.
Socket-939 is a desktop and workstation/entry-level server platform.

NewEgg got 'em for ~$195.00 last I looked.

What chipset is on that board?

I *don't* need the I/ just now, just good RAM/CPU/reliability

And what's your application?

Again, an extra $100 _before_ you put the $1K+ system makes
all-the-difference. It's not just server I/

Bryan J. Smith wrote:

Tue, 2006-07-11 at 23:40 -0500, William A. Mahaffey III wrote:

>
>>, you're just too kind :-).
>
>>
>
>Seriously William. We had several threads on your issues with the i865
>(as well as i845) month after month after month. That wasn't even a
>"professional/corporate desktop" chipsets, that's the i875.
>
>I'm just trying to help you friend.
>

>
>>When I say lightweight, I *mean* lightweight. N public access,
>
>>
>
>What does "public" have _anything_ to do with it?
>In fact, LAN I/ is typically far more intense.
>

>
>>N apache, N DNS,
>
>>
>
>Apache and DNS typically have more to do with _computational_
>performance. If you're throwing files around on a LAN, that's I/!
>
>We've had this discussion before, when you've had compatibility and
>performance issues. Moving from an Intel i845/i865 (again, not even a
>i875 that is tested for professional, desktop use -- and totally
>ignoring the E72xx/75xx) to a ViA KT series is a _lateral_ move.
>
>Again, referring back to your experience with the 694 -- back in the
>late '90s, I used to replace i440BX/GX and 693/694 chipsets with
>ServerSet IIIs -- often for ~$250. A measly $100 more in the mainboard
>makes _all_ the difference when you're spending $1K!
>

>
>>N several dozen users at any 1 time, just CPU, RAM, & code to run ad
>>nauseum.
>
>>
>
>What is its function?

See below.


>Again, why save $100 on a system that costs $500-1,000 anyway, when
>another $100 will give you server performance and quality?

Probably more like $700.00 total, apparently. $100.00+ does matter, ~15%


>
>>It *will* live at runlevel 3,
>
>>
>
>When does "graphics" have _anything_ to do with "storage/communcation"
>I/? William, I think this is where you keep missing the point, and why
>you keep running into issue after issue on your systems.
>
>I can't believe you've now taken all those issues and discussions we had
>months ago with your issues and just thrown them out-the-window. Spend
>$175+ on an E72xx series mainboard for P4, or $235 for a HT1000 for
>
>

>
>>but the server traits end there, for the most part. I was mostly
>>asking about reasonably complete/stable Linux support for that
>>chipset, decent %-age of full speed for RAM-intensive calculations,
>
>>
>
>Can you elaborate on the application?

Sure. Finite Element grid/mesh generation & subsequent analysis, often
coupled w/ CFD grid generation/analysis. Fluid-structure interaction
problems (deforming SRM propellant grains under stress from the flow
resulting from their combustion). Reasonably well resolved 2-D (i.e.,
not 3-D, thus I *can* afford pretty good 2-D resolution, borderline
large-eddy type resolution on the fluid side, similar resolution on the
solid side since I spend most of my CPU time on the fluid side & don't
want/need to answer needling questions about skimping on resolution
*anywhere* in the analysis). Turbulent, sub-sonic/transonic, complex
internal geometries on the fluid side, largish deformations on the solid
side. Definitely CPU/RAM intensive, not quite so I/ intensive. Local
drives large enough to catch the results, then processed by another box
on the LAN (where the Tecplot license lives), but still by me, i.e. only
one user at any 1 time. These boxen are like droids, w/ only 1 master,
not dozens like most public servers. I/ is less important,
CPU/RAM/stability are paramount.


>
>>etc. e.g. good S/chipset interaction for my somewhat truncated
>>requirements.
>
>>

Wed, 2006-07-12 at 00:23 -0500, William A. Mahaffey III wrote:
Probably more like $700.00 total, apparently. $100.00+ does matter,
~15%

Consider your time swapping out a board/chipset that was specifically
designated as "does not test to 24x7 tolerances" (e.g., i865 v. i875) in
that cost estimate.

Sure. Finite Element grid/mesh generation & subsequent analysis, often
coupled w/ CFD grid generation/analysis.
Fluid-structure interaction problems (deforming SRM propellant grains
under stress from the flow resulting from their combustion).
Reasonably well resolved 2-D (i.e., not 3-D, thus I *can* afford
pretty good 2-D resolution, borderline large-eddy type resolution on
the fluid side, similar resolution on the solid side since I spend
most of my CPU time on the fluid side & don't want/need to answer
needling questions about skimping on resolution *anywhere* in the
analysis). Turbulent, sub-sonic/transonic, complex internal geometries
on the fluid side, largish deformations on the solid side. Definitely
CPU/RAM intensive, not quite so I/ intensive. Local drives large
enough to catch the results, then processed by another box on the LAN
(where the Tecplot license lives), but still by me, i.e. only one user
at any 1 time.
These boxen are like droids, w/ only 1 master, not dozens like most
public servers. I/ is less important,

But what's your data rate from the distributed clients to the master(s)?

At what point will the combination of possible local storage and network
communication saturate each other?

That all affects how linearly the application scales.

If you're pushing back just 10MBps to the master per client with at
least 10 clients, you've already saturated a 32-bit PCI bus for just
network -- not including the mass inefficiency of Ethernet or the added
throughout to put to disk (if required).

That's why even 8 years ago**, we used 64-bit PCI NICs for GbE,
typically with all other I/ (storage, etc) on a separate PCI bus
when we were looking at anything close to a dozen nodes

CPU/RAM/stability are paramount.

They why do you buy i865 solutions when they _fail_ Intel's tolerances
for i875? We've been through this before William. I really don't think
you understood anything I said before, and I really wish the best for
you.

Your time easily pays for the premium. Saving $100/node isn't worth
your time -- before we even look at the performance/bottleneck aspects.
-- Bryan

**NTE: I (among others on this list) rolled out such grid computing
for CFD and other, distributed applications with Linux in the late '90s.
Even back then, a single 32-bit PCI bus didn't cut it. The options are
a little better now with PCIe x1 channels being more standard -- but the
data rate has always increased as well.

Some are using Infiniband right on the HTX (HyperTransport eXtension),
but that's overkill for you. All I'm advocating is that you don't
bottleneck yourself to the point where you are sacrificing any
scalability beyond 4-5 nodes so you can buy a few more.

Wed, 2006-07-12 at 01:58 -0400, Bryan J. Smith wrote:
Consider your time swapping out a board/chipset that was specifically
designated as "does not test to 24x7 tolerances" (e.g., i865 v. i875) in
that cost estimate.

the cost of your engineer's time and/or deadlines for a failed run.

**NTE: I (among others on this list) rolled out such grid computing
for CFD and other, distributed applications with Linux in the late '90s.
Even back then, a single 32-bit PCI bus didn't cut it. The options are
a little better now with PCIe x1 channels being more standard -- but the
data rate has always increased as well.

If cost is really paramount, then get the cheapest _refurb_ mainboard
and Celeron or Sempron you can find, run diskless (if possible -- it
does make management easier) keep the cost under $200/unit. The only
issue is if and when the I/ becomes a bottleneck for the DTR --
especially when you have masters in the distributed system (which is why
you should _never_ skimp there).

You also need to ensure your applications -- especially if engineering
runs take days at a time -- can handle failover and restart an
interation/process if a node fails. That right there is typically the
biggest issue and cost (far, far more than hardware and IT time) -- code
that is non-interruptable/restartable when runs take days.