~OT: Tyan's new Opteron PCIe-SLI mobo

[gimp]

here's the link to tyan's PCIe express dual opteron board with SLI:

http://www.tyan.com/products/html/thunderk8we.html

however this board includes PCI-X as well - which makes it ultra
expensive. wasn't half the point of PCIe to get rid of PCI-X and end up
with cost effective workstation boards...??? is anyone making a dual
opteron board that has SLI other than tyan...?

 

[Ed]

here's the link to tyan's PCIe express dual opteron board with SLI:

http://www.tyan.com/products/html/thunderk8we.html

however this board includes PCI-X as well - which makes it ultra
expensive. wasn't half the point of PCIe to get rid of PCI-X and end up
with cost effective workstation boards...??? is anyone making a dual
opteron board that has SLI other than tyan...?

ABIT WN-2S+
Dual AMD Opteron, 8GB DDR memory, 2 Gigabit Ethernet, 4-channel SATA II,
Dual PCI-E x16 Slots
http://www2.abit.com.tw/page/en/ser...&fMTYPE=Workstation Boards&pPRODINFO=Overview

 

[Rob Stow]

here's the link to tyan's PCIe express dual opteron board with SLI:

http://www.tyan.com/products/html/thunderk8we.html

however this board includes PCI-X as well - which makes it ultra
expensive. wasn't half the point of PCIe to get rid of PCI-X and end up
with cost effective workstation boards...??? is anyone making a dual
opteron board that has SLI other than tyan...?

The problem with getting rid of PCI-X is that there are so far
almost no PCI-E products to replace the kinds of things that
normally get put into a PCI-X slot, such as RAID controllers and
gigabit NICs. 32 bit PCI slots will also be needed for a while
for things like sound cards, for which there are also few - if
any - PCI-E options yet.

Someone I'm going to build a few systems for has ordered 6 of
those boards, plus processors, RAM, video cards, etc. For each
system, the RAID cards and the HDTV capture/encoding cards each
need a PCI-X slot and the sound cards need a PCI slot.


Personally, I'd like to see a board like the S2895 but without
the SLI crap. For most systems I'd rather have one of the 16
lane PCI-E slots replaced with a few 4 lane slots so that when
PCI-E versions of RAID cards, NICs, etc start to show up later
this year I will have slots to put them in.

I've heard rumours of a mezzanine card that will convert a 16
lane slot into a few 4 lane slots, but my crude understanding of
how PCI-E works makes me skeptical about that notion. I'll
believe it when I see one working.

 

[daytripper]

I've heard rumours of a mezzanine card that will convert a 16
lane slot into a few 4 lane slots, but my crude understanding of
how PCI-E works makes me skeptical about that notion. I'll
believe it when I see one working.

Make room in your crude understanding for PCIe bridges. NEC has a nice one...

 

[Rob Stow]

Make room in your crude understanding for PCIe bridges. NEC has a nice one...

I've seen PCI-E to (PCI or PCI-X) bridges, but so far none for
(single PCI-E) to (multiple PCI-E). I'll go looking for that NEC
one now but if you already have an URL saved somewhere it would
be nice if you could post it.

 

[pixel]

Personally, I'd like to see a board like the S2895 but without the SLI
crap.

i'll soon be running 1920 x 1200 resolution so SLI is all good in my
books for the average 1024 or even 1280 user SLI might not make much
of a difference. in my last few workstations i havn't had the need for
any 32bit PCI cards.... modems are usually external routers and modern
onboard sound is good quality, but yeah i see why its important to keep
it around for a bit.

 

[daytripper]

I've seen PCI-E to (PCI or PCI-X) bridges, but so far none for
(single PCI-E) to (multiple PCI-E). I'll go looking for that NEC
one now but if you already have an URL saved somewhere it would
be nice if you could post it.

Google µPD720401

Likely to find its way into my next design....

Many ways to use this device...even as the basis of an "add-in PCIe riser" for
terminally nerdy case modders ;-)

/daytripper

 

[pixel]

ABIT WN-2S+
Dual AMD Opteron, 8GB DDR memory, 2 Gigabit Ethernet, 4-channel SATA II,
Dual PCI-E x16 Slots
http://www2.abit.com.tw/page/en/ser...&fMTYPE=Workstation Boards&pPRODINFO=Overview

thanks for the link, pretty impressive... it all fits in an ATX form
factor too do the PCIe-16X slots run at full speed in SLI mode like
the tyan...?

 

[George Macdonald]

here's the link to tyan's PCIe express dual opteron board with SLI:

http://www.tyan.com/products/html/thunderk8we.html

however this board includes PCI-X as well - which makes it ultra
expensive. wasn't half the point of PCIe to get rid of PCI-X and end up
with cost effective workstation boards...??? is anyone making a dual
opteron board that has SLI other than tyan...?

Yep, I believe PCIe is supposed to replace PCI-X but I'm sure there are
people who have shelled out big $$ for PCI-X cards who don't want to junk
them.

 

[brundlefly76]

I wish Tyan would just ship a decent Opteron serverboard.

They are the only ones who are really working that market, but I have
had terrible experiences with these boards - of the two I have in
production now, one is running on a secret beta bios I got from support
just so it will post correctly, and the other one I had to put red tape
over one of the ethernet ports caused it shit the bed.

 

[Rob Stow]

I wish Tyan would just ship a decent Opteron serverboard.

They are the only ones who are really working that market, but I have
had terrible experiences with these boards - of the two I have in
production now, one is running on a secret beta bios I got from support
just so it will post correctly, and the other one I had to put red tape
over one of the ethernet ports caused it shit the bed.

I have built about a dozen dualie servers using the S2881 and
S2882, plus one quad using the S4882.

Server building doesn't get much easier than that.


Tyan is not the only one working that market - Arima also makes
some nice boards.
Haven't personally used one but they sound good on paper ;-)

 

[brundlefly76]

I wish I had the same experience as you.

Its just impossible to ask my sysadmin to install another one after the
hardware failures on the first 3 boards.

Similarly, I built my business on 7 homebuilt servers using the
cheapest ECS consumer AMD boards, and they ran like a top for 2 years
under *heavy* load. Then I read that people were returning some absurd
number of the same model because of failure - oh well, didnt happen to
me!

All a numbers game and ymmv.

 

[Rob Stow]

I wish I had the same experience as you.

Its just impossible to ask my sysadmin to install another one after the
hardware failures on the first 3 boards.

I sympathize with your reluctance to tell him to install another
one, but I also wonder if he might be the problem ? Have you
watched over his shoulder while he works to see if he really
knows what he is doing ? I have seen a few sysadmins that are
geniuses when it comes to arranging computers, cables, switches,
routers, etc to set up and maintain a complicated network - yet
should be shot before they are allowed to touch anything /inside/
a computer.

And what about the vendor you bought the boards from ? Do you
trust them enough to send you new boards instead of something
that was returned, in perhaps less than pristine condition, by
someone else ?

If you buy your processors and mobos from the same vendor, do you
ask them to test the combo before shipping ?

Similarly, I built my business on 7 homebuilt servers using the
cheapest ECS consumer AMD boards, and they ran like a top for 2 years
under *heavy* load. Then I read that people were returning some absurd
number of the same model because of failure - oh well, didnt happen to
me!

I've also had good success with ECS boards, albeit for desktops
and not servers.

I have also watched a lot of DIY-ers building systems and the
number that just don't have a clue is appalling - and their cpu
and mobo failure rates are correspondingly high. So many people
can't fathom that static electricity kills. You tell them that
standing on a carpet in 25% relative humidity is risky enough
when you *do* use elementary precautions like wearing a wrist
strap and then they go ahead and handle parts *without* a wrist
strap.

 

[keith]

I sympathize with your reluctance to tell him to install another
one, but I also wonder if he might be the problem ? Have you
watched over his shoulder while he works to see if he really
knows what he is doing ? I have seen a few sysadmins that are
geniuses when it comes to arranging computers, cables, switches,
routers, etc to set up and maintain a complicated network - yet
should be shot before they are allowed to touch anything /inside/
a computer.

And what about the vendor you bought the boards from ? Do you
trust them enough to send you new boards instead of something
that was returned, in perhaps less than pristine condition, by
someone else ?

All good points. Given that the hardware is zapped (they all use the same
chips) it seems that the supply chain is "interrupted" here somewhere.

If you buy your processors and mobos from the same vendor, do you ask
them to test the combo before shipping ?

Onesy-twosy? It doesn't seem there is enough volume to get any leverage
here.

When I was in the x86 biz, I didn't like ECS at all. That's half a decade
ago, so... I have been impressed with Tyan though. Sure, I've not had
a perfect ride with this motherboard, but I'm not perfect either. ;-)
The fact that the do have support means a lot to me.

I've also had good success with ECS boards, albeit for desktops and not
servers.

I have also watched a lot of DIY-ers building systems and the number
that just don't have a clue is appalling - and their cpu and mobo
failure rates are correspondingly high. So many people can't fathom
that static electricity kills. You tell them that standing on a carpet
in 25% relative humidity is risky enough when you *do* use elementary
precautions like wearing a wrist strap and then they go ahead and handle
parts *without* a wrist strap.

OTOH, I've never had such a failure and don't protect things as I should.
At work we have protection for customer product, but the rest isn't in a
perfect ESD environment. ESD is a problem, but it's not *that* big of a
problem.


Exactly. A few units aren't a useful sample. The fact that the units had
I/O ports destroyed )particularly something like the Ethernet port) is
interesting though.

 

[Rob Stow]

All good points. Given that the hardware is zapped (they all use the same
chips) it seems that the supply chain is "interrupted" here somewhere.




Onesy-twosy? It doesn't seem there is enough volume to get any leverage
here.

Depends on how many you are buying. One mobo and two Opty 2xx
procs will generally set you back $50 for a pre-test before it is
shipped. Spend another $800 by adding RAM, case, PSU and the fee
for assembly and testing of your bare bones system will often by
waived if you are a good customer. Buy six boards, six cases, 12
procs, 48 GB of DDR and many vendors will happily throw in the
pre-shipping testing even if you are a first time buyer.

When I was in the x86 biz, I didn't like ECS at all. That's half a decade
ago, so... I have been impressed with Tyan though. Sure, I've not had
a perfect ride with this motherboard, but I'm not perfect either. ;-)
The fact that the do have support means a lot to me.




OTOH, I've never had such a failure and don't protect things as I should.
At work we have protection for customer product, but the rest isn't in a
perfect ESD environment. ESD is a problem, but it's not *that* big of a
problem.

ESD is just one example of how many DIYers are careless. Other
common errors I see are
- poor cpu/heatsink assembly - such as no goop or waaaay
too much.
- using excessive force to insert cards/DIMMs/CPUs
- powering up motherboards, hard drives, etc while they are
lying on or in contact with a conductor - including
a sweaty palm. Duh.
- not properly securing things into place.
- using something other than the manufacturer's proper utility
for flashing the BIOS.

 

[Henry Nettles]

When I was in the x86 biz, I didn't like ECS at all. That's half a decade
ago, so... I have been impressed with Tyan though. Sure, I've not had
a perfect ride with this motherboard, but I'm not perfect either. ;-)
The fact that the do have support means a lot to me.

I have probably built 50 systems using the ECS K7S5A motherboard. Of
those, about 10 were flakey. In my experience, this is a very high
percentage. The ones that were good never gave any trouble.

More recently, I have built maybe 25 or 30 systems using the ECS
k7VTA3 motherboard. I have not had a single problem with these
motherboards.

Tyan motherboards tend to be fairly good, with the exception of the
S2460 (the original dual AMD motherboard). Over the years, I have
built maybe 10 systems with this motherboard, all of which caused me
grief. I believe most of the problems were associated with the AMD
chipset on the motherboard, but not all....

 

[chris.schoenfeld]

More bad news for Opteron is that IBM has released its Hurricane
chipset to help Xeon memory performance match Opteron's, and combined
with them stopping mentioning Opteron systems in press releases,
analysts say the love with AMD is over at IBM:

http://www.theregister.co.uk/2005/03/03/ibm_amd_storm/

So, that leaves pretty much HP and Sun, both of which manufacture their
own systems (vs HP, which used third-party).

I suspect I cant expect to see much more innovation and competition in
the third-party Opteron serverboard market moving forward with the
second-largest Opteron vendor pushing it to the back burner.

 

[George Macdonald]

OTOH, I've never had such a failure and don't protect things as I should.
At work we have protection for customer product, but the rest isn't in a
perfect ESD environment. ESD is a problem, but it's not *that* big of a
problem.

Are the grounds in most buildings even worth the bother of a wrist strap?
I'm sure I'm like many - just touch some metal first before messing with
something and then grab/touch the chassis while working. BTW I noticed
(eventually) that, with our hard disks in Promise SuperSwap enclosures,
touching something metal just before inserting the key to power down the
drive to be swapped, prevented the hangs I'd been experiencing prior to
that. This was even in Summer with only mild A/C so there's small amounts
of ESD present all the time.

Exactly. A few units aren't a useful sample. The fact that the units had
I/O ports destroyed )particularly something like the Ethernet port) is
interesting though.

I'm getting suspicious of Ethernet "compatibilty" after recent experiences.
I had to junk a couple of Linksys NICs because, on shutdown of the computer
with the Linksys, they would hang our D-Link 10/100 hub, i.e. the whole
network would stop and a power cycle on the hub was the only way to get it
going again... which also brings up the point that paying ~$500. or so for
a network switch which doesn't even have a power switch on it is slightly
annoying. More recently I'm finding several 10/100 devices which won't
work with an old 10Mbit hub we've had for several years - the hub still
works fine with several other NICs of the same mfr (Intel as it happens) as
those which don't work.

 

[nobody]

I have built about a dozen dualie servers using the S2881 and
S2882, plus one quad using the S4882.

Server building doesn't get much easier than that.


Tyan is not the only one working that market - Arima also makes
some nice boards.
Haven't personally used one but they sound good on paper ;-)

And also MSI - they have at least one server board (PCI-X, but no AGP)
and one workstation board (K8T Master2-FAR, has AGP, but no PCI-X). I
have K8T since last May, and have not a single bad thing to say about
it.

 

[Rob Stow]

And also MSI - they have at least one server board (PCI-X, but no AGP)
and one workstation board (K8T Master2-FAR, has AGP, but no PCI-X). I
have K8T since last May, and have not a single bad thing to say about
it.

I've used two of the 2-FAR boards but I didn't mention them
because they don't have PCI-E, which appears to be important
to the OP.