Using an old dual CPU server as a gaming PC?

On FedEx vehicle for delivery!

Should be great for flight simulation, just pretend those little fans or
propellers... BZZZZZZZZ!

http://www.ebay.com/itm/371543511693?

"Supermicro 1U Server X8DTU-F 2x Xeon L5630 2.13ghz Quad Core 48gb 4x
Trays DVD"

So far, very cool. Server motherboard manuals are much more thorough
than end-user motherboard manuals. I don't know how, but the thing looks
extremely clean. I've heard of washing components, but it might have an
ultra thin film of grease on some parts, maybe suggesting that the
environment was dust free but not clean-room clean. I don't understand
how they could take the time to clean everything and sell it so cheap. I
will check for dust inside the power supply (assume it's perfectly clean
unless I say otherwise). There are some spare cables in there. Hopefully
including a six pin video card connector.

The memory is Micron MT36JSZF51272PZ-1G4F1AB PC3-10600R 1333 MHz ECC.
Cheap on eBay.

The drive bays/slots! I'm just about to test, but apparently they fit
standard size (3 1/2"?) hard drives with SATA connectors even though
it's labeled an SAS backplane (storage interface). Apparently you just
pop the tray out, fasten the drive to it, and slide the tray into the
case where your SATA drive engages with the SAS power and data
connectors. Looks like my SSD will slide into one of the sockets while
resting on the case.

Hopefully the case will fit most server motherboards in case I continue
on this route. The case is kind of heavy and flexes when picked up by
one corner, so I don't do that.

The product includes two riser cards with x8 output slots.

Significant problem/mistake. The motherboard will not accept a simple
x16 riser card. The manual mentions a power connector. But a power
supply cable does not plug into to that connector. Instead, specially
designed riser cards have a downward facing fin that plugs into that
connector. In other words. That connector doesn't receive power from the
power supply, it provides power to PCI-e riser cards.

So, apparently I will have to do some soldering work (no problem, just
not neat as hoped for). Wires will go from that slot to the riser card
power pins. The manual shows a nice pin out, all 36 pins. But looks like
only six of them are distinct, lots of redundancy, Yea! It's just power
and a couple of switches.

I wonder if the riser card power paths should be severed since the
auxiliary power connector wires will go to the same x16 slot pins.

I haven't looked at the x16 connector pin out, I'm assuming the power
pins are on the left side.

John Doe wrote:
http://www.ebay.com/itm/371543511693?

"Supermicro 1U Server X8DTU-F 2x Xeon L5630 2.13ghz Quad Core 48gb 4x
Trays DVD"

So far, very cool. Server motherboard manuals are much more thorough
than end-user motherboard manuals. I don't know how, but the thing looks
extremely clean. I've heard of washing components, but it might have an
ultra thin film of grease on some parts, maybe suggesting that the
environment was dust free but not clean-room clean. I don't understand
how they could take the time to clean everything and sell it so cheap. I
will check for dust inside the power supply (assume it's perfectly clean
unless I say otherwise). There are some spare cables in there. Hopefully
including a six pin video card connector.

The memory is Micron MT36JSZF51272PZ-1G4F1AB PC3-10600R 1333 MHz ECC.
Cheap on eBay.

The drive bays/slots! I'm just about to test, but apparently they fit
standard size (3 1/2"?) hard drives with SATA connectors even though
it's labeled an SAS backplane (storage interface). Apparently you just
pop the tray out, fasten the drive to it, and slide the tray into the
case where your SATA drive engages with the SAS power and data
connectors. Looks like my SSD will slide into one of the sockets while
resting on the case.

Hopefully the case will fit most server motherboards in case I continue
on this route. The case is kind of heavy and flexes when picked up by
one corner, so I don't do that.

The product includes two riser cards with x8 output slots.

Significant problem/mistake. The motherboard will not accept a simple
x16 riser card. The manual mentions a power connector. But a power
supply cable does not plug into to that connector. Instead, specially
designed riser cards have a downward facing fin that plugs into that
connector. In other words. That connector doesn't receive power from the
power supply, it provides power to PCI-e riser cards.

So, apparently I will have to do some soldering work (no problem, just
not neat as hoped for). Wires will go from that slot to the riser card
power pins. The manual shows a nice pin out, all 36 pins. But looks like
only six of them are distinct, lots of redundancy, Yea! It's just power
and a couple of switches.

I wonder if the riser card power paths should be severed since the
auxiliary power connector wires will go to the same x16 slot pins.

I haven't looked at the x16 connector pin out, I'm assuming the power
pins are on the left side.

I thought the user manual for the X8DTU-F was indicating J10 was
an input ? But I don't see room for routing power to it, if that
is the case. And I don't have a high-res picture of J10, to guess
as to whether it's a Molex, or a "pin and block" or "blade" connector.

As a start, you could check for continuity from J10 P5V to
ATX +5V pins, and see if J10 is in parallel with the
main connector. At least this would prove that J10 is not
completely isolated and guaranteed to be needed. If J10 is
in parallel with the 20 pin ATX, then perhaps the add-in cards
will run without J10 used.

*******

There is an animation here, of how the UIO planar area works.
The planar area, is the "quarter" of your motherboard that is missing.
It leaves room for cards.

http://www.supermicro.com.tr/UIO.cfm.htm

Paul

Paul wrote:

(see the prior message)

I thought the user manual for the X8DTU-F was indicating J10 was an
input ?

The writer might have thought so, too...

"In addition to the Primary 20-pin power connector and the 8-pin CPU PWR
connectors, the Universal IO Power Connector (UIOP) located at J10 is
also required for the riser cards installed on the motherboard. This
power connector is used to provide power to the riser card."

I guess the last sentence there is supposed to be a clue that it's an
output, but it's not much of a clue.

But I don't see room for routing power to it, if that is the case. And
I don't have a high-res picture of J10,

Lots of stuff is blocked in pictures of these server motherboards.

Really tough was trying to figure out whether Intel's legacy S5520xx
motherboards have x16 connectors. Intel's specifications say they have
the option, but it seems that none of those (available) motherboards
actually have x16 connectors.

And I don't have a high-res picture of J10, to guess as to whether
it's a Molex, or a "pin and block" or "blade" connector.

It's a PCI/PCI-e slot style connector. That's how a riser card with a
tail fin fits into it.

As a start, you could check for continuity from J10 P5V to ATX +5V
pins, and see if J10 is in parallel with the main connector. At least
this would prove that J10 is not completely isolated and guaranteed to
be needed. If J10 is in parallel with the 20 pin ATX, then perhaps the
add-in cards will run without J10 used.

Yes, it should be easy to tell using a continuity tester. There are some
(decoupling?) capacitors around that connector. That's probably all it
adds to powering the card. Certainly they had a reason for putting the
capacitors in a separate area instead of right under/at the x16
connector.

I plan to adapt a Molex connector to the video card's auxiliary power
input.

There is an animation here, of how the UIO planar area works. The
planar area, is the "quarter" of your motherboard that is missing. It
leaves room for cards.

http://www.supermicro.com.tr/UIO.cfm.htm

Thanks. Those pictures explain why the riser's right hand side connector
is unusable on mine, because the riser card is designed for a different
motherboard, the location of my motherboard's CPU blocks that side's
connector.

Seems to work, installed Windows 10 with no hiccups.

https://www.flickr.com/photos/275322...n/photostream/

https://www.flickr.com/photos/275322...n/photostream/

No need to click on both, just use the arrows on that page.

John Doe wrote:
Paul wrote:

(see the prior message)

I thought the user manual for the X8DTU-F was indicating J10 was an
input ?

The writer might have thought so, too...

"In addition to the Primary 20-pin power connector and the 8-pin CPU PWR
connectors, the Universal IO Power Connector (UIOP) located at J10 is
also required for the riser cards installed on the motherboard. This
power connector is used to provide power to the riser card."

I guess the last sentence there is supposed to be a clue that it's an
output, but it's not much of a clue.

But I don't see room for routing power to it, if that is the case. And
I don't have a high-res picture of J10,

Lots of stuff is blocked in pictures of these server motherboards.

Really tough was trying to figure out whether Intel's legacy S5520xx
motherboards have x16 connectors. Intel's specifications say they have
the option, but it seems that none of those (available) motherboards
actually have x16 connectors.

And I don't have a high-res picture of J10, to guess as to whether
it's a Molex, or a "pin and block" or "blade" connector.

It's a PCI/PCI-e slot style connector. That's how a riser card with a
tail fin fits into it.

As a start, you could check for continuity from J10 P5V to ATX +5V
pins, and see if J10 is in parallel with the main connector. At least
this would prove that J10 is not completely isolated and guaranteed to
be needed. If J10 is in parallel with the 20 pin ATX, then perhaps the
add-in cards will run without J10 used.

Yes, it should be easy to tell using a continuity tester. There are some
(decoupling?) capacitors around that connector. That's probably all it
adds to powering the card. Certainly they had a reason for putting the
capacitors in a separate area instead of right under/at the x16
connector.

I plan to adapt a Molex connector to the video card's auxiliary power
input.

There is an animation here, of how the UIO planar area works. The
planar area, is the "quarter" of your motherboard that is missing. It
leaves room for cards.

http://www.supermicro.com.tr/UIO.cfm.htm

Thanks. Those pictures explain why the riser's right hand side connector
is unusable on mine, because the riser card is designed for a different
motherboard, the location of my motherboard's CPU blocks that side's
connector.

Seems to work, installed Windows 10 with no hiccups.

So have you figured out how to get your video card in there ?

*******

There is the X58 Northbridge (for desktops) and several
Tylersburg ones as well. This shows some options for
Tylersburg.

http://pc.watch.impress.co.jp/docs/2...gai386_04l.gif

This doesn't go into how small the x16 slots can be chopped up.
You need resources (a hub) for each lane set, so that's a penalty
on the size of the silicon die for the Northbridge. But a thing
with that much I/O, might well be defined by the package for
the Northbridge (many electrical signals, a big BGA package
needed). The other resource needed, is enough high quality
clock signals (PCI Express Rev2 slots need low jitter
clocks). When they make servers, one reason for selecting
x8 slots (if they have enough clocks), is because a $1000
Areca RAID card has an x8 edge card.

http://www.bit-tech.net/hardware/cpu...tecture-dive/7

Now while thinking about that, I'm guessing fitting
a video card isn't going to be that easy. Because the
planar area is designed for their riser - and if their
riser doesn't have open-ended x8 slots, you've got
a problem.

And I don't see any gap in the end of the x8 connectors
on the riser, so I don't see how an x16 video card
can be fitted in there. And while it might be tempting
to "saw a gap in it", that would be pretty hard to do without
damaging something. One of those connectors with a purpose-built
gap in the end, might use thicker side walls to keep the
connector together. So I don't know how feasible it is
to saw a gap so the x16 card can be fitted.

There was a doc on the SuperMicro site with all sorts
of riser products on it. And the twits used fuzzy photos
of each one, so you can't see much in the way of details.
You'd probably want one of the UIO parts. And since
you've got the server sitting right in front of you,
you can see which one is the right one easier than I can.

http://www.supermicro.com/manuals/br...ds_socketR.pdf

Paul

Paul wrote:

John Doe wrote:
Paul wrote:

(see the prior message)

I thought the user manual for the X8DTU-F was indicating J10 was an
input ?

The writer might have thought so, too...

"In addition to the Primary 20-pin power connector and the 8-pin CPU
PWR connectors, the Universal IO Power Connector (UIOP) located at
J10 is also required for the riser cards installed on the
motherboard. This power connector is used to provide power to the
riser card."

I guess the last sentence there is supposed to be a clue that it's an
output, but it's not much of a clue.

But I don't see room for routing power to it, if that is the case.
And I don't have a high-res picture of J10,

Lots of stuff is blocked in pictures of these server motherboards.

Really tough was trying to figure out whether Intel's legacy S5520xx
motherboards have x16 connectors. Intel's specifications say they
have the option, but it seems that none of those (available)
motherboards actually have x16 connectors.

And I don't have a high-res picture of J10, to guess as to whether
it's a Molex, or a "pin and block" or "blade" connector.

It's a PCI/PCI-e slot style connector. That's how a riser card with a
tail fin fits into it.

As a start, you could check for continuity from J10 P5V to ATX +5V
pins, and see if J10 is in parallel with the main connector. At
least this would prove that J10 is not completely isolated and
guaranteed to be needed. If J10 is in parallel with the 20 pin ATX,
then perhaps the add-in cards will run without J10 used.

Yes, it should be easy to tell using a continuity tester. There are
some (decoupling?) capacitors around that connector. That's probably
all it adds to powering the card. Certainly they had a reason for
putting the capacitors in a separate area instead of right under/at
the x16 connector.

I plan to adapt a Molex connector to the video card's auxiliary power
input.

There is an animation here, of how the UIO planar area works. The
planar area, is the "quarter" of your motherboard that is missing.
It leaves room for cards.

http://www.supermicro.com.tr/UIO.cfm.htm

Thanks. Those pictures explain why the riser's right hand side
connector is unusable on mine, because the riser card is designed for
a different motherboard, the location of my motherboard's CPU blocks
that side's connector.

Seems to work, installed Windows 10 with no hiccups.

So have you figured out how to get your video card in there ?

It has an x16 slot. That's what the riser is plugged into, with its tail
fin plugged into J10. I'm not going to use that riser, it's going to be
pulled out of there. It will be used just to determine where J10's
outputs go.

The riser card arrives tomorrow. Between now and then, the continuity
tester will tell which pins J10 goes to. As far as I know, the left side
of an x4/8/16 are all the same. So connecting J10 outputs to the
backside of the riser should be easy enough except for the physical
work.

Also, you might be right about the slot not really needing J10. And
another possibility is to put power supply leads directly to the riser,
like so many handmade powered risers do.

At the moment, I'm trying to dampen the noise of the power supply fan.
I'm going to ask in the electronics design group about sensor pins and
stuff.




--

There is the X58 Northbridge (for desktops) and several Tylersburg
ones as well. This shows some options for Tylersburg.

http://pc.watch.impress.co.jp/docs/2...gai386_04l.gif

This doesn't go into how small the x16 slots can be chopped up. You
need resources (a hub) for each lane set, so that's a penalty on the
size of the silicon die for the Northbridge. But a thing with that
much I/O, might well be defined by the package for the Northbridge
(many electrical signals, a big BGA package needed). The other
resource needed, is enough high quality clock signals (PCI Express
Rev2 slots need low jitter clocks). When they make servers, one reason
for selecting x8 slots (if they have enough clocks), is because a
$1000 Areca RAID card has an x8 edge card.

http://www.bit-tech.net/hardware/cpu...re-i7-nehalem-
architecture-dive/7

Now while thinking about that, I'm guessing fitting a video card isn't
going to be that easy. Because the planar area is designed for their
riser - and if their riser doesn't have open-ended x8 slots, you've
got a problem.

And I don't see any gap in the end of the x8 connectors on the riser,
so I don't see how an x16 video card can be fitted in there. And while
it might be tempting to "saw a gap in it", that would be pretty hard
to do without damaging something. One of those connectors with a
purpose-built gap in the end, might use thicker side walls to keep the
connector together. So I don't know how feasible it is to saw a gap so
the x16 card can be fitted.

There was a doc on the SuperMicro site with all sorts of riser
products on it. And the twits used fuzzy photos

Yep, got to love those fuzzy photos.





of each one, so you can't see much in the way of details. You'd
probably want one of the UIO parts. And since you've got the server
sitting right in front of you, you can see which one is the right one
easier than I can.

http://www.supermicro.com/manuals/br...ds_socketR.pdf

Paul

John Doe wrote:

It has an x16 slot. That's what the riser is plugged into, with its tail
fin plugged into J10. I'm not going to use that riser, it's going to be
pulled out of there. It will be used just to determine where J10's
outputs go.

Well, what I'm worried about, is they did something like this....

| | | |
----- ----------
^ x8 riser out
|
/ ^
--------------------- | 8 of 16 lanes
/ |
| |
| J10 | | | | |
| ----- -------------
Mobo power from ^ x16 mobo (planar)
20-pin ATX connector |
Power pins not connected

They might also make the x16 mobo slot a "regular" one,
and the power section at the end could be connected
in parallel. But a guess would be, they avoided having
to run power routing layers over to the x16 power
section, relying on J10 and riser resources to route the
power.

Perhaps it's because there are potentially more
riser slots than input power busses, that they added
J10 to give more parallel paths. The idea is to get
enough ampacity, have enough pins on the input side of
the riser, to power all the slot power pins on the output
side.

So that's where my ohmmeter would go next. Buzzing the
3.3V and 12V pins of the x16 slot and seeing whether there
is continuity to the 20-pin connector or not. It doesn't
answer the question of copper track ampacity, just proves
they didn't completely cut that path as shown above.

It would make more sense to run all the powers in parallel,
in which case, you'd have no trouble running your card.
If it was like this, it improves the odds of not
actually needing J10. The power section on the
x16 then carries 3.3V @ 3A, 12V @ 2A, and the PCI Express
AUX carries 12V @ X amps.

| | | |
----- ----------
^ x8 riser out
|
| ^
----------------------+ | 8 of 16 lanes
/ | |
| | | --
| J10 | | | | | \__ GTX goes here
| ----- ------------- /
Mobo power from ^ x16 mobo (planar) --
20-pin ATX connector |
Power pins connected

HTH,
Paul

Paul wrote:

John Doe wrote:

It has an x16 slot. That's what the riser is plugged into, with its
tail fin plugged into J10. I'm not going to use that riser, it's
going to be pulled out of there. It will be used just to determine
where J10's outputs go.

Well, what I'm worried about, is they did something like this....

| | | |
----- ----------
^ x8 riser out
| / ^
--------------------- | 8 of 16 lanes
/ |
| |
| J10 | | | | |
| ----- -------------
Mobo power from ^ x16 mobo (planar) 20-pin ATX
connector |
Power pins not connected

They might also make the x16 mobo slot a "regular" one, and the power
section at the end could be connected in parallel. But a guess would
be, they avoided having to run power routing layers over to the x16
power section, relying on J10 and riser resources to route the power.

Perhaps it's because there are potentially more riser slots than input
power busses, that they added J10 to give more parallel paths.

I think it's because the riser can support several cards.

The idea is to get enough ampacity, have enough pins on the input side
of the riser, to power all the slot power pins on the output side.

So that's where my ohmmeter would go next. Buzzing the 3.3V and 12V
pins of the x16 slot and seeing whether there is continuity to the
20-pin connector or not.

Okay.

Since I know the pin out of J10, it should lead me to the only pins of
interest on the PCI-Express slot. And I will test from their to the 20
pin connector, too. But unless there are diodes along the way, I don't
see how they could not be connected from the card to the 20 pin
connector. But I realize you might be talking over my head.

It's extremely dry here (about 27% humidity) lately. Trying very hard to
keep in touch with metal things.



--




It doesn't answer the question of copper track ampacity, just proves
they didn't completely cut that path as shown above.

It would make more sense to run all the powers in parallel, in which
case, you'd have no trouble running your card. If it was like this, it
improves the odds of not actually needing J10. The power section on
the x16 then carries 3.3V @ 3A, 12V @ 2A, and the PCI Express AUX
carries 12V @ X amps.

| | | |
----- ----------
^ x8 riser out
| | ^
----------------------+ | 8 of 16 lanes
/ | |
| | | --
| J10 | | | | | \__ GTX
| goes here
| ----- ------------- /
Mobo power from ^ x16 mobo (planar) -- 20-pin ATX
connector |
Power pins connected

HTH,
Paul

John Doe wrote:


It's extremely dry here (about 27% humidity) lately. Trying very hard to
keep in touch with metal things.

Get yourself an ESD strap, and strap in.

Paul