|
|
|
|
| Author |
Message |
Ulf Samuelsson
Guest
|
 Posted: Mon Jan 10, 2005 4:19 pm Post subject: S2462UNGM reboot problem |
 |
|
Have this strange problem with the board.
The first problem is that the board crashes with
DRIVER_IRQL_NOT_LESS_OR_EQUAL.
The PC has internal USCSI-LVD,Dual Ethernet, VGA and has
* Creative Geforce 4 Ti4400
* Creative Audigy Platinum
* Pinnacle DC1000 framegrabber
* 5 port USB High Speed USB Host.
* Corsair 512 mB Reg ECC SDRAM
* Enhance 460 W P/W
I have conluded that this is due to the interrupt sharing, which is not
acceptable to the Audigy
according to other posts.
Happens after running for a couple of hours, and I learned to live with
this.
Didnt try too much shifting the PCI boards, sharing went from bad to worse.¨
Would appreciate to have some guidelines on how to move around the boards...
======================================================
The second problem is more annoying.
When I want to restart from Windows XP, the PC shuts down OK but refuses to
restart.
Have removed all addin cards from PCI slot and AGP slot and it still behaves
this way.
If I pull the power cord for 5-10 minutes, then the PC starts windows OK.
If I start it after 3-4 minutes, it could start, but would freeze during the
boot process.
I went into system setup, and looked at the CPU temperature, and one of the
CPUs
were running high temperature, going up to 73 'C, (which is above the 70'C
limit)
and then froze. The Other CPU kept to the 60'C where it has operated before.
The Zalman fan is running at max as it should,and has kept the temperature
within limits before.
The system have run at 60'C+ temperature since it was commissioned in mid
2002.
(I know this is high, but what to do, have 9 fans in the machine...)
I removed the heating CPU and the situation seems to improve little, but not
100% sure about this.
Noticed that there was very little silicon paste between the CPU and the
cooler.
Have tried reducing bus speed to 200 MHz, problems remains.
One key thing is the CMOS clear jumper.
This seems to have been forgotten in the 2-3 position after the last BIOS
upgrade.
According to the manual, this should be in the 1-2 position except if you
wan to clear the
battery backed up CMOS SRAM. Then you power off, and shortcircuit [2-3] for
a few seconds.
Anyone that knows what happens if you short circuit this for an extended
period?
I think that it has been there for more than 6 months, and the PC has been
shut
off most of the time for a period of 3 months during that time.
Maybe it is as simple as a bad battery? Will try this next!
As I see it, the problem is either
1) Motherboard
2) CPU
3) Memory
4) Power Supply
The PC is running memtest86 today to check the memory.
Will try to get my hands on a spare CPU to test that.
Any other ideas on how to test?
--
Best Regards
Ulf at atmel dot com
These comments are intended to be my own opinion and they
may, or may not be shared by my employer, Atmel Sweden. |
|
| Back to top |
|
 |
Fix your Windows Problems - FAST.
FREE Safe Scan Registry Check. Locate & Fix Errors in Minutes!
|
|
Floyd L. Davidson
Guest
|
| Posted: Mon Jan 10, 2005 8:12 pm Post subject: Re: S2462UNGM reboot problem |
|
|
"Ulf Samuelsson" <ulf@NOSPAMatmel.com> wrote:
| Quote: | Have this strange problem with the board.
The first problem is that the board crashes with
DRIVER_IRQL_NOT_LESS_OR_EQUAL.
The PC has internal USCSI-LVD,Dual Ethernet, VGA and has
* Creative Geforce 4 Ti4400
* Creative Audigy Platinum
* Pinnacle DC1000 framegrabber
* 5 port USB High Speed USB Host.
* Corsair 512 mB Reg ECC SDRAM
* Enhance 460 W P/W
I have conluded that this is due to the interrupt sharing, which is not
acceptable to the Audigy
according to other posts.
Happens after running for a couple of hours, and I learned to live with
this.
Didnt try too much shifting the PCI boards, sharing went from bad to worse.¨
Would appreciate to have some guidelines on how to move around the boards...
|
I'm not up on that, but I'd just try random changes... :-)
However, given the rest of this description I'd hold off on
doing that until you find out what is going on with heat. You
might find the first problem disappears when the second one
is fixed.
| Quote: | ======================================================
The second problem is more annoying.
When I want to restart from Windows XP, the PC shuts down OK but refuses to
restart.
Have removed all addin cards from PCI slot and AGP slot and it still behaves
this way.
If I pull the power cord for 5-10 minutes, then the PC starts windows OK.
If I start it after 3-4 minutes, it could start, but would freeze during the
boot process.
|
That sounds like what happens whenever the system goes into a
protective mode and shuts itself down. I've had overheating do
that, and I've slipped and shorted out things like the 3.3 SB
volt bus and had it lock up for a couple minutes. (Gives one a
real fright to see sparks and have a shutdown and then it won't
even turn on again! The first time I did that I thought "I've
really done it this time...")
| Quote: | I went into system setup, and looked at the CPU temperature, and one of the
CPUs
were running high temperature, going up to 73 'C, (which is above the 70'C
limit)
and then froze. The Other CPU kept to the 60'C where it has operated before.
The Zalman fan is running at max as it should,and has kept the temperature
within limits before.
The system have run at 60'C+ temperature since it was commissioned in mid
2002.
(I know this is high, but what to do, have 9 fans in the machine...)
|
*NINE* fans???? You need to look at just how they are all
arranged. For example, if seven of them are blowing air out of
the box and the other two are blowing air on the CPU's... you
might just have one big static block of hot air right in the
middle of the box! (AT&T once built a computer that had two
fans, one on each side of the hard disk... and the first
modification made to that design was to remove one of those
fans, because there simply was no air movement *across* the hard
disk, and the machine was eating disk drives for breakfast.)
Look very carefully at air flow, and make sure there *is* flow.
With that many fans you'll need some blowing in and some blowing
out. (Ideally, all fans should blow *in*, and have filters in
front of them. But computers aren't usually engineered that
way.)
Also, some power supplies can shutdown due to heat. I have a
S2462UNG here that, if the room temperature gets too high, will
shutdown due to an over heated power supply. The CPUs are
running only 53-57C, but all the heat from them is being sucked
right into the power supply... it beeps at me four times or so,
and goes bye bye.
| Quote: | I removed the heating CPU and the situation seems to improve little, but not
100% sure about this.
Noticed that there was very little silicon paste between the CPU and the
cooler.
|
There really shouldn't be very much. But I'll tell you a secret
as far as replacing it (which you really should do if you pull
the heat sink off the CPU)... Go to an auto parts store and buy
some anti-seize compound (made by Permatex, for example). It is
copper based, and is intended to be used on things like engine
head bolts or on boiler systems, but it is a great heat transfer
paste! Just stir it up real good, and apply a very thin coat.
I use some made for Napa Auto Parts ("Copper Anti-Seize
Lubricant" by Permatex), and has a part number 765-2569. A four
ounce container cost me about $5, and will easily last the rest
of my life. The only thing better is the silver based stuff that
cost $12 per CPU... and it isn't enough better to measure the
difference.
| Quote: | Have tried reducing bus speed to 200 MHz, problems remains.
|
That would probably only show up under load. So with an idle
box, you won't see much change.
| Quote: | One key thing is the CMOS clear jumper.
This seems to have been forgotten in the 2-3 position after the last BIOS
upgrade.
According to the manual, this should be in the 1-2 position except if you
wan to clear the
battery backed up CMOS SRAM. Then you power off, and shortcircuit [2-3] for
a few seconds.
Anyone that knows what happens if you short circuit this for an extended
period?
|
Nothing. It removes the battery from the circuit, and puts a
short across the filter capacitor, to drain any charge on it.
If the power is turned off, the result is no voltage at all to
the NVRAM. But there is a diode between it and the NVRAM, so
that if AC power is turned on the power supply provides voltage
to the NVRAM, but because of the diode it doesn't see that
shorted capacitor.
1 2 3
o<--->o o diode diode
| | | +| /| |\ |+
| +-----------|< |---+---| >|-----------o +5 volt bus
+ | + | | | \| | |/ |
----- ----- | +---+
--- ----- | | |
| | | +---+
| | | |
+-----+-----+------------+------------------o GND
BAT CAP NVRAM
| Quote: | I think that it has been there for more than 6 months, and the PC has been
shut
off most of the time for a period of 3 months during that time.
Maybe it is as simple as a bad battery? Will try this next!
|
It should be just fine. The effect, however, is that no matter
what anyone did to change setup parameters that are saved in
NVRAM, they lasted only until the power was removed, and then
the defaults were once again set!
| Quote: | As I see it, the problem is either
1) Motherboard
2) CPU
3) Memory
4) Power Supply
The PC is running memtest86 today to check the memory.
Will try to get my hands on a spare CPU to test that.
Any other ideas on how to test?
|
Hmmm... none of the above? :-)
Look for air flow problems. How clean is the power supply
inside? And are any of the cables blocking the air flow either
for the CPUs or the the power supply? And make sure you aren't
trying to pull a big vacuum in the middle of the box! In order
to blow air out, air has to come in from somewhere... And a
number of fans (such as the CPU fans, video card fans, hard disk
coolers, etc.) commonly don't really do anything but re-arrange
the air inside the case, which doesn't really help much if there
isn't enough air flow into and out of the box.
And pay special attention to the one CPU that is over heating.
Is the heat sink clean? Rather than find a spare CPU, I'd just
consider swapping the two of them, heat sink and all. If the
high temp goes with the cpu/cooler, something is wrong with it.
If it doesn't go with it, something is wrong with the air flow
in that area (that motherboard has a problem with air flow there
anyway...) I would expect the high temp to stay with the
cpu/cooler, and I'd also expect that cleaning the heat sink and
applying a thin coat of anti-seize lube would fix it.
--
Floyd L. Davidson <http://web.newsguy.com/floyd_davidson>
Ukpeagvik (Barrow, Alaska) floyd@barrow.com |
|
| Back to top |
|
|
bhoover
GURU
Joined: 25 Aug 2004
Posts: 134
|
| Posted: Tue Jan 11, 2005 2:02 pm Post subject: Re: S2462UNGM reboot problem |
|
|
Ulf Samuelsson wrote:
| Quote: | Have this strange problem with the board.
The first problem is that the board crashes with
DRIVER_IRQL_NOT_LESS_OR_EQUAL.
The PC has internal USCSI-LVD,Dual Ethernet, VGA and has
* Creative Geforce 4 Ti4400
* Creative Audigy Platinum
* Pinnacle DC1000 framegrabber
* 5 port USB High Speed USB Host.
* Corsair 512 mB Reg ECC SDRAM
* Enhance 460 W P/W
I have conluded that this is due to the interrupt sharing, which is not
acceptable to the Audigy
according to other posts.
|
Have you turned on the interrupt table option in the bios? And btw, what
version bios is it? You might want to flash with the latest one.
| Quote: | Happens after running for a couple of hours, and I learned to live with
this.
Didnt try too much shifting the PCI boards, sharing went from bad to worse.¨
Would appreciate to have some guidelines on how to move around the boards...
======================================================
The second problem is more annoying.
When I want to restart from Windows XP, the PC shuts down OK but refuses to
restart.
Have removed all addin cards from PCI slot and AGP slot and it still behaves
this way.
If I pull the power cord for 5-10 minutes, then the PC starts windows OK.
If I start it after 3-4 minutes, it could start, but would freeze during the
boot process.
I went into system setup, and looked at the CPU temperature, and one of the
CPUs
were running high temperature, going up to 73 'C, (which is above the 70'C
limit)
and then froze. The Other CPU kept to the 60'C where it has operated before.
The Zalman fan is running at max as it should,and has kept the temperature
within limits before.
The system have run at 60'C+ temperature since it was commissioned in mid
2002.
(I know this is high, but what to do, have 9 fans in the machine...)
I removed the heating CPU and the situation seems to improve little, but not
100% sure about this.
Noticed that there was very little silicon paste between the CPU and the
cooler.
Have tried reducing bus speed to 200 MHz, problems remains.
|
Since on the subject of temps, thought I'd share a little of what I'm getting
with this board (non-scsi version). I put the system together a little over 4
months ago. It runs more or less continuously.
CPU (1.2Ghz MP's) temps at nearly idle are around 50c, and 47c, and 48c, and 45c
idle. VRM1 runs around 54c, and peaks up to around 57c under a healthy load.
It's at around 51c idle. I've got a front intake fan, two rear exhaust fans,
and an Enermax psu with two fans in it (and of course fans on each cpu  ).
Though I'm running without the side panel of my Chenming tower, until I cut a
hole, and add a fan. CPU thermal compound would be a good place to start.
You may also want to consider what rpm's your fans run. SpeedFan reports mine
max are around 7300 rpm, and I run them at 65%, and 70% respectively -- more for
harmonics, but when the load is heavy, I'll bump cpu1 to 75% -- keeps it at 50c
on cpu1, as cpu2 comes up to match it (again, I don't bring up cpu2
correspondingly because of case harmonics). I can run them as low as 65%, and
50% respectively, with roughly the same temps, at idle.
Bryan
| Quote: | One key thing is the CMOS clear jumper.
This seems to have been forgotten in the 2-3 position after the last BIOS
upgrade.
According to the manual, this should be in the 1-2 position except if you
wan to clear the
battery backed up CMOS SRAM. Then you power off, and shortcircuit [2-3] for
a few seconds.
Anyone that knows what happens if you short circuit this for an extended
period?
I think that it has been there for more than 6 months, and the PC has been
shut
off most of the time for a period of 3 months during that time.
Maybe it is as simple as a bad battery? Will try this next!
As I see it, the problem is either
1) Motherboard
2) CPU
3) Memory
4) Power Supply
The PC is running memtest86 today to check the memory.
Will try to get my hands on a spare CPU to test that.
Any other ideas on how to test?
--
Best Regards
Ulf at atmel dot com
These comments are intended to be my own opinion and they
may, or may not be shared by my employer, Atmel Sweden. |
|
|
| Back to top |
|
|
Floyd L. Davidson
Guest
|
| Posted: Tue Jan 11, 2005 9:08 pm Post subject: Re: S2462UNGM reboot problem |
|
|
Bryan Hoover <bhoover@wecs.com> wrote:
| Quote: |
Since on the subject of temps, thought I'd share a little of what I'm getting
with this board (non-scsi version). I put the system together a little over 4
months ago. It runs more or less continuously.
CPU (1.2Ghz MP's) temps at nearly idle are around 50c, and 47c, and 48c, and 45c
idle. VRM1 runs around 54c, and peaks up to around 57c under a healthy load.
It's at around 51c idle. I've got a front intake fan, two rear exhaust fans,
and an Enermax psu with two fans in it (and of course fans on each cpu ).
Though I'm running without the side panel of my Chenming tower, until I cut a
hole, and add a fan. CPU thermal compound would be a good place to start.
You may also want to consider what rpm's your fans run. SpeedFan reports mine
max are around 7300 rpm, and I run them at 65%, and 70% respectively -- more for
harmonics, but when the load is heavy, I'll bump cpu1 to 75% -- keeps it at 50c
on cpu1, as cpu2 comes up to match it (again, I don't bring up cpu2
correspondingly because of case harmonics). I can run them as low as 65%, and
50% respectively, with roughly the same temps, at idle.
Bryan
|
Interesting comments!
When I started using the first of the two S2462 motherboards I
have, a lot of experimenting with fans convinced me it either
needed a hole cut in side panel with a high volume fan mounted,
or if the hole was large enough it wouldn't need the fan at all.
So they are both being used with the side panel removed. That
seems to be very significant with the one using MP 1600 CPU's,
and not so important on the one running MP 2400's.
Also the first one is an early version of the board, and the
speed of the CPU fans cannot be controlled, while it works just
fine on the later board.
I'm running Linux on both, and the one with MP 1600 CPUs is
plotting various temperatures and voltages. It generates a new
plot every ten minutes, and samples every two minutes. An early
set of plots is available on my web page (along with the various
scripts to do it).
http://web.newsguy.com/floyd_davidson/sensors
--
Floyd L. Davidson <http://web.newsguy.com/floyd_davidson>
Ukpeagvik (Barrow, Alaska) floyd@barrow.com |
|
| Back to top |
|
|
Ulf Samuelsson
Guest
|
| Posted: Tue Jan 11, 2005 10:13 pm Post subject: Re: S2462UNGM reboot problem |
|
|
"Ulf Samuelsson" <ulf@NOSPAMatmel.com> skrev i meddelandet
news:cru7f2$m09$1@public2.atmel-nantes.fr...
| Quote: | Have this strange problem with the board.
The first problem is that the board crashes with
DRIVER_IRQL_NOT_LESS_OR_EQUAL.
The PC has internal USCSI-LVD,Dual Ethernet, VGA and has
* Creative Geforce 4 Ti4400
* Creative Audigy Platinum
* Pinnacle DC1000 framegrabber
* 5 port USB High Speed USB Host.
* Corsair 512 mB Reg ECC SDRAM
* Enhance 460 W P/W
|
Airflow problem found.
It was INSIDE the fan.
3 mm layer of dust between the fan and the heatsink explains a lot.
Cleaned it out and bought some new Arctic Alumina Thermal compound.
Instructions at website: So how read when the CPUs are outside the box?
When I put the CPUs into the box the last time, I only put thermal compound
on the CPU chip protection resulting in 58-65 ' Celsius.
Now I put a thick layer of this compound on the complete heatspreader of the
fan.
Resulting temperature is 46 'Celsius after running for 10 minutes.
THis is not recommended by the store though!
Used the whole tube and got a new one, this time "Arctic Silver 5"
which is 3 x the price and supposedly 5 times bettewr then the Arctic
Alumina.
This one , I only put on the CPU chip protection, as recommended by the
www.arcticsilver.com web site. Resulting temperature is 56 'Celsius!!!
I understand that compound might, if it conducts electricity, shortwire
some of the configuration stuff on top of the MCU, but you could conceive a
small ring with the same height as the cpu protection and fill everyting
inside the ring
with the thermal compound then this seems to be superior to the silver based
componud
mounted as recommended.
Now it boots with the onchip VGA, but seems to hate my Geforce 4.
I had this problem sometime before and then it suddenly started to work.
Tried an old Voodoo 3, which also boots.
Maybe I zapped the board, will test in an old PC.
======some more info on my machine ================
I use the last BIOS (2.14?)
Lian-Li 71 chassis
Fans:
Two front intake fans.
Two CPU fans blowing air on the CPU heat spreaders. Maybe this is
wrong?
Should the fans blow air from the CPU?
Two back exhaust fans close to the CPUs
One back exhaust fan inside the Enhance Power Supply
Two back exhaust fans at the top
Now (with open case)
CPU0: Large blob of compound : 2700 rpm 2" fan and 52'C
CPU1: Thin layer of silver compound, 4680 rpm and 48'C
| Quote: | --
Best Regards
Ulf at atmel dot com
These comments are intended to be my own opinion and they
may, or may not be shared by my employer, Atmel Sweden.
|
|
|
| Back to top |
|
|
Fix your Windows Problems - FAST.
FREE Safe Scan Registry Check. Locate & Fix Errors in Minutes!
|
|
bhoover
GURU
Joined: 25 Aug 2004
Posts: 134
|
| Posted: Wed Jan 12, 2005 1:21 am Post subject: Re: S2462UNGM reboot problem |
|
|
Ulf Samuelsson wrote:
| Quote: | "Ulf Samuelsson" <ulf@NOSPAMatmel.com> skrev i meddelandet
news:cru7f2$m09$1@public2.atmel-nantes.fr...
Have this strange problem with the board.
The first problem is that the board crashes with
DRIVER_IRQL_NOT_LESS_OR_EQUAL.
The PC has internal USCSI-LVD,Dual Ethernet, VGA and has
* Creative Geforce 4 Ti4400
* Creative Audigy Platinum
* Pinnacle DC1000 framegrabber
* 5 port USB High Speed USB Host.
* Corsair 512 mB Reg ECC SDRAM
* Enhance 460 W P/W
Airflow problem found.
It was INSIDE the fan.
3 mm layer of dust between the fan and the heatsink explains a lot.
Cleaned it out and bought some new Arctic Alumina Thermal compound.
Instructions at website: So how read when the CPUs are outside the box?
When I put the CPUs into the box the last time, I only put thermal compound
on the CPU chip protection resulting in 58-65 ' Celsius.
Now I put a thick layer of this compound on the complete heatspreader of the
fan.
Resulting temperature is 46 'Celsius after running for 10 minutes.
THis is not recommended by the store though!
Used the whole tube and got a new one, this time "Arctic Silver 5"
which is 3 x the price and supposedly 5 times bettewr then the Arctic
Alumina.
This one , I only put on the CPU chip protection, as recommended by the
www.arcticsilver.com web site. Resulting temperature is 56 'Celsius!!!
I understand that compound might, if it conducts electricity, shortwire
some of the configuration stuff on top of the MCU, but you could conceive a
small ring with the same height as the cpu protection and fill everyting
inside the ring
with the thermal compound then this seems to be superior to the silver based
componud
mounted as recommended.
Now it boots with the onchip VGA, but seems to hate my Geforce 4.
I had this problem sometime before and then it suddenly started to work.
Tried an old Voodoo 3, which also boots.
Maybe I zapped the board, will test in an old PC.
======some more info on my machine ================
I use the last BIOS (2.14?)
Lian-Li 71 chassis
Fans:
Two front intake fans.
Two CPU fans blowing air on the CPU heat spreaders. Maybe this is
wrong?
Should the fans blow air from the CPU?
Two back exhaust fans close to the CPUs
One back exhaust fan inside the Enhance Power Supply
Two back exhaust fans at the top
Now (with open case)
CPU0: Large blob of compound : 2700 rpm 2" fan and 52'C
CPU1: Thin layer of silver compound, 4680 rpm and 48'C
|
Blowing down onto cpu is correct -- in any event, you should be able to tell
which way the fan goes by looking at it. If there's a label in the middle, it
should be facing up.
I'm only visually familiar with the 1.2Ghz MP. Assuming yours are the same, you
should have compound only on the core -- the up-raised cube in the middle of the
die. And only a thin a layer -- good description I read said the volume of
compound should be about the size of a rice kernal, and spread evenly over the
core -- some just put the rice kernal dab in the middle of the core, and let the
heatsink installation spread it. Or use a credit card to spread it.
Bryan
| Quote: | --
Best Regards
Ulf at atmel dot com
These comments are intended to be my own opinion and they
may, or may not be shared by my employer, Atmel Sweden.
|
|
|
| Back to top |
|
|
Ulf Samuelsson
Guest
|
| Posted: Wed Jan 12, 2005 1:21 am Post subject: Tyan S2462 UNGM (PCI slots and Interrupts) |
|
|
Here is a brief analysis of how the interrupts are allocated.
Tested 6 combinations today. Dont claim it is always correct, but it
might be worth a try.
PCI Interrupts are allocated on interrupts 16-19.
For slot 1,2,3,4,5 (from the AGP slot) the first interrupt per
board (base interrupt) is 16,17,18,19 and 16.
If the board supports multiple interrupts then they
are allocated as base interrupt + 1, base interrupt +2 etc.
with wraparound from 19 to 16.
My USB board has two OHCI and one EHCI so it uses 3 interrupts.
If it goes into slot 1, it uses 16,17 and 18
If it goes into slot 4 it uses 19,16 and 17
If it goes into slot 5, it uses 16,17 and 18
Some interrupts appears to be fixed: SCSI(16,17), Ethernet(18,19) and
USB(19)
A table would look like:
IRQ16: SCSI1, PCI1#1, PCI4#2, PCI3#3 PCI5#1
IRQ17: SCSI2, PCI2#1, PCI1#2, PCI4#3 PCI5#2
IRQ18: ENET#1,PCI3#1, PCI2#2, PCI1#3 PCI5#3
IRQ19: ENET#2,PCI4#1, PCI3#2, PCI2#3 USB
Thisi s not completely true since I have seen some PCI boards
allocate in the lower 16 interrupts as well.
The Pinnacle DC1000 framegrabber does this.
Have no clue how the AGP interrupt will map.
It appears like there is no way to have an addin board with a dedicated
interrupt.
Still, I hope this may help someone. |
|
| Back to top |
|
|
Floyd L. Davidson
Guest
|
| Posted: Wed Jan 12, 2005 1:21 am Post subject: Re: S2462UNGM reboot problem |
|
|
Bryan Hoover <bhoover@wecs.com> wrote:
| Quote: |
I'm only visually familiar with the 1.2Ghz MP. Assuming yours are the same, you
should have compound only on the core -- the up-raised cube in the middle of the
die. And only a thin a layer -- good description I read said the volume of
compound should be about the size of a rice kernal, and spread evenly over the
core -- some just put the rice kernal dab in the middle of the core, and let the
heatsink installation spread it. Or use a credit card to spread it.
|
I wouldn't let the heatsink installation spread it. Particularly as
some of these compounds are fairly thick, and won't really spread well
at all. Even the thinest aren't that thin.
Generally I use a single edge razor blade to spread it, but a credit
card should work just about as well.
The point is to fill in any air gaps between the two metal
surfaces, so there is genuinely 100 surface to surface contact,
rather than just the high points of each surface touching. I
don't know if the difference is 50% contact without it and 98%
with it, or just what; but that is probably a fairly close
figure.
But it is best to realize that the heat tranfer characteristics
of the compound are _not_ as good as the metal itself. The
thickness of needs only be the same as the uneveness between the
two metal surfaces!
Anything thicker than that actually _reduces_ the heat transfer.
--
Floyd L. Davidson <http://web.newsguy.com/floyd_davidson>
Ukpeagvik (Barrow, Alaska) floyd@barrow.com |
|
| Back to top |
|
|
bhoover
GURU
Joined: 25 Aug 2004
Posts: 134
|
| Posted: Wed Jan 12, 2005 1:21 am Post subject: Re: S2462UNGM reboot problem |
|
|
"Floyd L. Davidson" wrote:
| Quote: | Bryan Hoover <bhoover@wecs.com> wrote:
Since on the subject of temps, thought I'd share a little of what I'm getting
with this board (non-scsi version). I put the system together a little over 4
months ago. It runs more or less continuously.
CPU (1.2Ghz MP's) temps at nearly idle are around 50c, and 47c, and 48c, and 45c
idle. VRM1 runs around 54c, and peaks up to around 57c under a healthy load.
It's at around 51c idle. I've got a front intake fan, two rear exhaust fans,
and an Enermax psu with two fans in it (and of course fans on each cpu ).
Though I'm running without the side panel of my Chenming tower, until I cut a
hole, and add a fan. CPU thermal compound would be a good place to start.
You may also want to consider what rpm's your fans run. SpeedFan reports mine
max are around 7300 rpm, and I run them at 65%, and 70% respectively -- more for
harmonics, but when the load is heavy, I'll bump cpu1 to 75% -- keeps it at 50c
on cpu1, as cpu2 comes up to match it (again, I don't bring up cpu2
correspondingly because of case harmonics). I can run them as low as 65%, and
50% respectively, with roughly the same temps, at idle.
Bryan
Interesting comments!
When I started using the first of the two S2462 motherboards I
have, a lot of experimenting with fans convinced me it either
needed a hole cut in side panel with a high volume fan mounted,
or if the hole was large enough it wouldn't need the fan at all.
So they are both being used with the side panel removed. That
seems to be very significant with the one using MP 1600 CPU's,
and not so important on the one running MP 2400's.
|
You might find interesting as a reference -- at least with regards to the temps I
gave -- this amd whitepaper:
http://www.amd.com/us-en/assets/content_type/white_papers_and_tech_docs/25325.pdf
They thermal evaluate several setups with this board, and one of them -- the Chenming
tower -- happens to be almost identical to mine -- I'm using the "6000 rpm" Vantec
hsf's.
I understand earlier versions didn't have heatsink on chipset either. And the VRM's
are hot. Side panel fan seems much recommended.
The whitepaper -- Chenming anyway -- tests are with 1.2Ghz. Recommends additional
thermal solution for anything higher than 1.2Ghz.
| Quote: | Also the first one is an early version of the board, and the
speed of the CPU fans cannot be controlled, while it works just
fine on the later board.
I'm running Linux on both, and the one with MP 1600 CPUs is
plotting various temperatures and voltages. It generates a new
plot every ten minutes, and samples every two minutes. An early
set of plots is available on my web page (along with the various
scripts to do it).
http://web.newsguy.com/floyd_davidson/sensors
|
Floyd Davidson from SA? Ha! How do? I'd recognize those graphs anywhere .
I add, with the side panel on, AGP, and harddrive temps drop to like 38c, and 36
(harddrive as low as 34), because, I assume, better air direction from the front
intake.
Bryan
| Quote: | --
Floyd L. Davidson <http://web.newsguy.com/floyd_davidson
Ukpeagvik (Barrow, Alaska) floyd@barrow.com |
|
|
| Back to top |
|
|
bhoover
GURU
Joined: 25 Aug 2004
Posts: 134
|
| Posted: Wed Jan 12, 2005 5:01 am Post subject: Re: S2462UNGM reboot problem |
|
|
Bryan Hoover wrote:
| Quote: | "Floyd L. Davidson" wrote:
You might find interesting as a reference -- at least with regards to the temps I
gave -- this amd whitepaper:
http://www.amd.com/us-en/assets/content_type/white_papers_and_tech_docs/25325.pdf
They thermal evaluate several setups with this board, and one of them -- the Chenming
tower -- happens to be almost identical to mine -- I'm using the "6000 rpm" Vantec
hsf's.
I understand earlier versions didn't have heatsink on chipset either. And the VRM's
are hot. Side panel fan seems much recommended.
The whitepaper -- Chenming anyway -- tests are with 1.2Ghz. Recommends additional
thermal solution for anything higher than 1.2Ghz.
|
Correction -- the tests are done with 1.4Ghz, but concluding recommendations for
additonal cooling for anything over 1.2Ghz.
| Quote: | Also the first one is an early version of the board, and the
speed of the CPU fans cannot be controlled, while it works just
fine on the later board.
I'm running Linux on both, and the one with MP 1600 CPUs is
plotting various temperatures and voltages. It generates a new
plot every ten minutes, and samples every two minutes. An early
set of plots is available on my web page (along with the various
scripts to do it).
http://web.newsguy.com/floyd_davidson/sensors
Floyd Davidson from SA? Ha! How do? I'd recognize those graphs anywhere .
|
No, that's right -- fan speed, etc. conversation.
| Quote: | I add, with the side panel on, AGP, and harddrive temps drop to like 38c, and 36
(harddrive as low as 34), because, I assume, better air direction from the front
intake.
|
Should add, case off temps for these are around 41, 44.
Bryan
| Quote: | Bryan
--
Floyd L. Davidson <http://web.newsguy.com/floyd_davidson
Ukpeagvik (Barrow, Alaska) floyd@barrow.com |
|
|
| Back to top |
|
|
Fix your Windows Problems - FAST.
FREE Safe Scan Registry Check. Locate & Fix Errors in Minutes!
|
|
Ulf Samuelsson
Guest
|
| Posted: Wed Jan 12, 2005 2:06 pm Post subject: Re: S2462UNGM reboot problem |
|
|
| Quote: | Now it boots with the onchip VGA, but seems to hate my Geforce 4.
I had this problem sometime before and then it suddenly started to work.
Tried an old Voodoo 3, which also boots.
Maybe I zapped the board, will test in an old PC.
|
Yesterday, I configured the machine to use both cards, and setting = PCI.
It booted from the ATI properly, but in the middle of the Windows boot
the screen went black, and It would not work when I connected the screen to
the GeForce 4 card.
I think I know what happened!
The ATI is disabled by Windows, only Geforce Active.
The Geforce 4 is set to run 1600 x 1200 for my nice 20" Sun monitor,
but the test were run on the junky old 17" with max 1280 x 1024 resolution.
In the middle of the boot, the card switched over to the Geforce
but the screen could not handle it.
Just got another to run Linux, and the Sun screen was connected to this one.
This, morning, I connected two monitors and then it started to boot from the
ATI,
and windows came up on the GeForce.
Then I changed to AGP in setup again and then suddenly it started to work
booting from the GeForce 4.
It is very strange that the S2462 did not boot from the AGP
when the ATI was disabled by the Jumper! I think it should...
It would be very nice if there was a setting to choose the boot screen
but I guess, they wont update the BIOS any more.
Anyway, I think my current problem is solved,
Maybe I test changing the compound to make it thinner,
but if the temperature spreads in the package, then
very thin compound on the chip, and a thick compound layer around
it is probably a workable idea, assuming you dont mess with the
configuration.
Probably the package could improve. Why not a larger contact surface with
the fan?
| Quote: | --
Best Regards
Ulf at atmel dot com
These comments are intended to be my own opinion and they
may, or may not be shared by my employer, Atmel Sweden.
|
|
|
| Back to top |
|
|
bhoover
GURU
Joined: 25 Aug 2004
Posts: 134
|
| Posted: Fri Jan 14, 2005 4:32 pm Post subject: Re: S2462UNGM reboot problem |
|
|
Bryan Hoover wrote:
| Quote: | Bryan Hoover wrote:
"Floyd L. Davidson" wrote:
You might find interesting as a reference -- at least with regards to the temps I
gave -- this amd whitepaper:
http://www.amd.com/us-en/assets/content_type/white_papers_and_tech_docs/25325.pdf
They thermal evaluate several setups with this board, and one of them -- the Chenming
tower -- happens to be almost identical to mine -- I'm using the "6000 rpm" Vantec
hsf's.
I understand earlier versions didn't have heatsink on chipset either. And the VRM's
are hot. Side panel fan seems much recommended.
The whitepaper -- Chenming anyway -- tests are with 1.2Ghz. Recommends additional
thermal solution for anything higher than 1.2Ghz.
Correction -- the tests are done with 1.4Ghz, but concluding recommendations for
additonal cooling for anything over 1.2Ghz.
Also the first one is an early version of the board, and the
speed of the CPU fans cannot be controlled, while it works just
fine on the later board.
I'm running Linux on both, and the one with MP 1600 CPUs is
plotting various temperatures and voltages. It generates a new
plot every ten minutes, and samples every two minutes. An early
set of plots is available on my web page (along with the various
scripts to do it).
http://web.newsguy.com/floyd_davidson/sensors
Floyd Davidson from SA? Ha! How do? I'd recognize those graphs anywhere .
No, that's right -- fan speed, etc. conversation.
I add, with the side panel on, AGP, and harddrive temps drop to like 38c, and 36
(harddrive as low as 34), because, I assume, better air direction from the front
intake.
Should add, case off temps for these are around 41, 44.
|
Manutia, but I'll add, I can maintain the above two side panel on temps by moving the front
intake into the 3'' drive bay just above it -- AGP is a little higher -- around 39 -- like
this.
Moving the fan up neglets the board I guess, from right around at the last pci slot. The
full sized scsi card in slot 5 does not feel warm, seems to fair well -- and actually, the
air still makes it across the bottom of the card. So the trade off seems minimal.
This probably increases flow up around the chipset, and even the VRM's I should think --
these temps seem to be a little more responsive to the increased air flow. CPU1 heatsink
is very cool to the touch as are the nearby inductors copper windings. CPU2 heatsink, a
little warm, and the copper inductor windings still rather hot to touch.
Right now, with side panel off, fans speeds for CPU1, CPU2, Front Intake at 65%, 60%, and
70% respectively, temps are:
VRM1 51c
VRM2 54c
CPU1 47c
CPU2 49c
DDR 40c
AGP 39c
HD0 38c
Anyway, this is my solution with the fans I've presently got, until I get a side panel with
fans. Side panel off with the front intake moved up seems a very good compromise.
Bryan
| Quote: | Bryan
Bryan
--
Floyd L. Davidson <http://web.newsguy.com/floyd_davidson
Ukpeagvik (Barrow, Alaska) floyd@barrow.com |
|
|
| Back to top |
|
|
|
|
| |
Watched Topics
Search
Register
Log in to check your private messages
Profile
Log in
Memberlist
Usergroups