No temperauture increase with processing?!?

On Mon, 28 Jun 2004 19:54:41 +0100, rstlne <..DeleteThis@text.news.virgin.net> wrote:

 >
 > "Peter Hucker" <hucker.DeleteThis@clara.co.uk> wrote in message
 > news:opsaa6qwfoaiowgp@blue...
  >> I have a dual Athlon 2800+, the CPUs are hovering around 50C (within a few
 > degrees - assuming error margins in motherboard monitors). They are at this
 > temperature after a few minutes of turning the PC on (when sat in the BIOS),
 > and also when idling in windows XP. I started a couple of programs to use
 > both CPUs 100%, and the temperatures remain the same! Is something not
 > quite right here???
  >>
  >> --
 >
 > 100% cpu usage is not a direct relation to cpu strain.
 > Loading the cpu might make no difference if the cooling is pretty good.. You
 > might only see 1-2c change.. Maybee your only seeing the socket temp also,
 > that would give a fairly constant result.

Could be the sensor in this board is not positioned as well as what I'm used to - I thought they all used an internal sensor built into the athlon nowadays? TJunc I've seen it called. But every other board I've used, the reading rockets up when using the CPU. I used to get about 30C at idle and 60C at full load on some machines. If the cooling is pretty ood - why am I getting as high as 50C?



--
13 parrots and rising <a style='text-decoration: underline;' href="http://www.petersparrots.com" target="_blank">http://www.petersparrots.com</a>
93 silly video clips <a style='text-decoration: underline;' href="http://www.insanevideoclips.com" target="_blank">http://www.insanevideoclips.com</a>
1259 digital photos <a style='text-decoration: underline;' href="http://www.petersphotos.com" target="_blank">http://www.petersphotos.com</a>
Served from a watercooled dual 3.2GHz silent Athlon with 512GB of watercooled Raid.

Can you be a closet claustrophobic?<!-- ~MESSAGE_AFTER~ -->
 >> Stay informed about: No temperauture increase with processing?!? 

Peter Hucker wrote:
 > I have a dual Athlon 2800+, the CPUs are hovering around 50C (within
 > a few degrees - assuming error margins in motherboard monitors).
 > They are at this temperature after a few minutes of turning the PC on
 > (when sat in the BIOS), and also when idling in windows XP. I
 > started a couple of programs to use both CPUs 100%, and the
 > temperatures remain the same! Is something not quite right here???

Most (all?) dual-Athlon boards do not support halt-disconnect, so there is
very little fluctuation in temperatures at all. With a K7D (measures
internal diode temps, though seems to read higher than some other boards
with the same CPU and HSF) and two 2500's@2.1GHz with Silent Boosts, I
usually have ~50 deg C idle, and ~60 deg C load with burnk7. When the "load"
is something like prime95 torture test, then the delta is only around 6-7
deg C, and is even lower with things like 3dmark and games. Also, if you
don't affinity-lock the processes with task manager, the switching overhead
grows and the temperature drops a degree or two.

I'd be surprised if it didn't move at all though, regardless of the cooling
you're using

--
Michael Brown
<a style='text-decoration: underline;' href="http://www.emboss.co.nz" target="_blank">www.emboss.co.nz</a> : OOS/RSI software and more
Add michael@ to emboss.co.nz - My inbox is always open<!-- ~MESSAGE_AFTER~ -->
 >> Stay informed about: No temperauture increase with processing?!? 

On Tue, 29 Jun 2004 12:30:38 +1200, Michael Brown <see DeleteThis @signature.below> wrote:

 > Peter Hucker wrote:
  >> I have a dual Athlon 2800+, the CPUs are hovering around 50C (within
  >> a few degrees - assuming error margins in motherboard monitors).
  >> They are at this temperature after a few minutes of turning the PC on
  >> (when sat in the BIOS), and also when idling in windows XP. I
  >> started a couple of programs to use both CPUs 100%, and the
  >> temperatures remain the same! Is something not quite right here???
 >
 > Most (all?) dual-Athlon boards do not support halt-disconnect, so there is
 > very little fluctuation in temperatures at all.

!!!! Doesn't that mean I am running them unnecessarily hot?!?!? I wonder, does that mean I wouldn't actually use any more electricity by running tham at 100%? Time to start up those distributed computing tasks again!

 > With a K7D (measures
 > internal diode temps, though seems to read higher than some other boards
 > with the same CPU and HSF)

Well the water temp is 35C, and the diode temp is 50C - does this seem reasonable? I don't really like this silicon grease, I've ordered some pads - they always seem to transfer heat better than the grease (the pads that come with boxed athlons).

 > and two 2500's@2.1GHz with Silent Boosts, I
 > usually have ~50 deg C idle, and ~60 deg C load with burnk7. When the "load"
 > is something like prime95 torture test, then the delta is only around 6-7
 > deg C, and is even lower with things like 3dmark and games. Also, if you
 > don't affinity-lock the processes with task manager, the switching overhead
 > grows and the temperature drops a degree or two.
 >
 > I'd be surprised if it didn't move at all though, regardless of the cooling
 > you're using

Well it might by a couple of degrees - but the fluctuation in the readings is a few degrees so I wouldn't notice.



--
13 parrots and rising <a style='text-decoration: underline;' href="http://www.petersparrots.com" target="_blank">http://www.petersparrots.com</a>
93 silly video clips <a style='text-decoration: underline;' href="http://www.insanevideoclips.com" target="_blank">http://www.insanevideoclips.com</a>
1259 digital photos <a style='text-decoration: underline;' href="http://www.petersphotos.com" target="_blank">http://www.petersphotos.com</a>
Served from a watercooled dual 3.2GHz silent Athlon with 512GB of watercooled Raid.

There's no need to worry, I have the memory of a.. a.. what's that thing called?<!-- ~MESSAGE_AFTER~ -->
 >> Stay informed about: No temperauture increase with processing?!? 

Peter Hucker wrote:
 > On Tue, 29 Jun 2004 12:30:38 +1200, Michael Brown
 > <see DeleteThis @signature.below> wrote:
 >
  >> Peter Hucker wrote:
   >>> I have a dual Athlon 2800+, the CPUs are hovering around 50C (within
   >>> a few degrees - assuming error margins in motherboard monitors).
   >>> They are at this temperature after a few minutes of turning the PC
   >>> on (when sat in the BIOS), and also when idling in windows XP. I
   >>> started a couple of programs to use both CPUs 100%, and the
   >>> temperatures remain the same! Is something not quite right here???
  >>
  >> Most (all?) dual-Athlon boards do not support halt-disconnect, so
  >> there is very little fluctuation in temperatures at all.
 >
 > !!!! Doesn't that mean I am running them unnecessarily hot?!?!?

It means the chip is wasting power due to a limitation of the OS, so yes,
sorta If you run Linux (kernel 2.6 and later IIRC), you'll find the idle
temperatures should drop significantly as this does do halt-disconnect on
idle.

 > I
 > wonder, does that mean I wouldn't actually use any more electricity
 > by running tham at 100%?

The power usage of the CPUs would be roughly proportional to the delta
between the water temp and the die temp. So if you went from 50 deg C to 52
dec C with a water temp of 35, then the power usage of the CPUs would have
increased about 13%, give or take. This would equate to (again, roughtly
speaking, basing on a 80W full load power and 60% efficiency through the
system) an increase from the wall socket of about 35W (taking into account
both CPUs). It's not zero, but it's also not that big.

 > Time to start up those distributed
 > computing tasks again!



  >> With a K7D (measures
  >> internal diode temps, though seems to read higher than some other
  >> boards with the same CPU and HSF)
 >
 > Well the water temp is 35C, and the diode temp is 50C - does this
 > seem reasonable?

Water temp sounds a little high unless you've got a high room temperature,
but a delta of 15 deg C sounds "about right" for low-end water cooling.

 > I don't really like this silicon grease, I've
 > ordered some pads - they always seem to transfer heat better than the
 > grease (the pads that come with boxed athlons).

I'd recommend against going with the pads purely for ease-of-removal
reasons, especially as there doesn't seem to be much of a difference in
temperatures regardless of the type of compound used (Arctic silver,
bubblegum pads, toothpaste, etc etc).

[...]

--
Michael Brown
<a style='text-decoration: underline;' href="http://www.emboss.co.nz" target="_blank">www.emboss.co.nz</a> : OOS/RSI software and more
Add michael@ to emboss.co.nz - My inbox is always open<!-- ~MESSAGE_AFTER~ -->
 >> Stay informed about: No temperauture increase with processing?!? 

On Tue, 29 Jun 2004 17:58:28 +1200, Michael Brown <see.DeleteThis@signature.below> wrote:

 > Peter Hucker wrote:
  >> On Tue, 29 Jun 2004 12:30:38 +1200, Michael Brown
  >> <see.DeleteThis@signature.below> wrote:
  >>
   >>> Peter Hucker wrote:
   >>>> I have a dual Athlon 2800+, the CPUs are hovering around 50C (within
   >>>> a few degrees - assuming error margins in motherboard monitors).
   >>>> They are at this temperature after a few minutes of turning the PC
   >>>> on (when sat in the BIOS), and also when idling in windows XP. I
   >>>> started a couple of programs to use both CPUs 100%, and the
   >>>> temperatures remain the same! Is something not quite right here???
   >>>
   >>> Most (all?) dual-Athlon boards do not support halt-disconnect, so
   >>> there is very little fluctuation in temperatures at all.
  >>
  >> !!!! Doesn't that mean I am running them unnecessarily hot?!?!?
 >
 > It means the chip is wasting power due to a limitation of the OS, so yes,
 > sorta If you run Linux (kernel 2.6 and later IIRC), you'll find the idle
 > temperatures should drop significantly as this does do halt-disconnect on
 > idle.

But this never happened on a dual PIII. Can't be the OS.

  >> I wonder, does that mean I wouldn't actually use any more electricity
  >> by running tham at 100%?
 >
 > The power usage of the CPUs would be roughly proportional to the delta
 > between the water temp and the die temp. So if you went from 50 deg C to 52
 > dec C with a water temp of 35, then the power usage of the CPUs would have
 > increased about 13%, give or take. This would equate to (again, roughtly
 > speaking, basing on a 80W full load power and 60% efficiency through the
 > system) an increase from the wall socket of about 35W (taking into account
 > both CPUs). It's not zero, but it's also not that big.
 >
  >> Time to start up those distributed
  >> computing tasks again!
 >
 >

I switched them off before when the amount of heat coming frm the PCs was alarming (as in a lot of electricity bill!) Even in winter, it heats the house but nowhere near as cheaply as gas heating.

   >>> With a K7D (measures
   >>> internal diode temps, though seems to read higher than some other
   >>> boards with the same CPU and HSF)
  >>
  >> Well the water temp is 35C, and the diode temp is 50C - does this
  >> seem reasonable?
 >
 > Water temp sounds a little high unless you've got a high room temperature,
 > but a delta of 15 deg C sounds "about right" for low-end water cooling.

I was surprised at how LOW it was. The room is about 23 to 24C (warm weather just now, and this room has three 19" monitors.

  >> I don't really like this silicon grease, I've
  >> ordered some pads - they always seem to transfer heat better than the
  >> grease (the pads that come with boxed athlons).
 >
 > I'd recommend against going with the pads purely for ease-of-removal
 > reasons, especially as there doesn't seem to be much of a difference in
 > temperatures regardless of the type of compound used (Arctic silver,
 > bubblegum pads, toothpaste, etc etc).

TOOTHPASTE!

My theory was that the gunk stuff oozes away. For example I noticed one CPU getting 10C extra because the gunk had not stayed where I put it.



--
13 parrots and rising <a style='text-decoration: underline;' href="http://www.petersparrots.com" target="_blank">http://www.petersparrots.com</a>
93 silly video clips <a style='text-decoration: underline;' href="http://www.insanevideoclips.com" target="_blank">http://www.insanevideoclips.com</a>
1259 digital photos <a style='text-decoration: underline;' href="http://www.petersphotos.com" target="_blank">http://www.petersphotos.com</a>
Served from a watercooled dual 3.2GHz silent Athlon with 512GB of watercooled Raid.

"Americans will always do the right thing when they have exhausted all other alternatives." -- Winston Churchill<!-- ~MESSAGE_AFTER~ -->
 >> Stay informed about: No temperauture increase with processing?!? 

Michael Brown wrote:

 > Peter Hucker wrote:
 >
  >>On Tue, 29 Jun 2004 12:30:38 +1200, Michael Brown
  >><see.RemoveThis@signature.below> wrote:
  >>
  >>
   >>>Peter Hucker wrote:
   >>>
   >>>>I have a dual Athlon 2800+, the CPUs are hovering around 50C (within
   >>>>a few degrees - assuming error margins in motherboard monitors).
   >>>>They are at this temperature after a few minutes of turning the PC
   >>>>on (when sat in the BIOS), and also when idling in windows XP. I
   >>>>started a couple of programs to use both CPUs 100%, and the
   >>>>temperatures remain the same! Is something not quite right here???
   >>>
   >>>Most (all?) dual-Athlon boards do not support halt-disconnect, so
   >>>there is very little fluctuation in temperatures at all.
  >>
  >>!!!! Doesn't that mean I am running them unnecessarily hot?!?!?
 >
 >
 > It means the chip is wasting power due to a limitation of the OS, so yes,
 > sorta If you run Linux (kernel 2.6 and later IIRC), you'll find the idle
 > temperatures should drop significantly as this does do halt-disconnect on
 > idle.

It has nothing to do with 'the OS'. It has to do with the nature of the
Athlon bus and the relatively new s2k bus disconnect, which is required for
it to go into low power mode, and which chipsets support it.

<a style='text-decoration: underline;' href="http://www.xbitlabs.com/articles/cpu/display/athlonxp-3000.html" target="_blank">http://www.xbitlabs.com/articles/cpu/display/athlonxp-3000.html</a>

 >
 >
  >>I
  >>wonder, does that mean I wouldn't actually use any more electricity
  >>by running tham at 100%?
 >
 >
 > The power usage of the CPUs would be roughly proportional to the delta
 > between the water temp and the die temp. So if you went from 50 deg C to 52
 > dec C with a water temp of 35, then the power usage of the CPUs would have
 > increased about 13%, give or take. This would equate to (again, roughtly
 > speaking, basing on a 80W full load power and 60% efficiency through the
 > system) an increase from the wall socket of about 35W (taking into account
 > both CPUs). It's not zero, but it's also not that big.
 >
 >
  >>Time to start up those distributed
  >>computing tasks again!
 >
 >
 >
 >
 >
   >>>With a K7D (measures
   >>>internal diode temps, though seems to read higher than some other
   >>>boards with the same CPU and HSF)
  >>
  >>Well the water temp is 35C, and the diode temp is 50C - does this
  >>seem reasonable?
 >
 >
 > Water temp sounds a little high unless you've got a high room temperature,
 > but a delta of 15 deg C sounds "about right" for low-end water cooling.
 >
 >
  >>I don't really like this silicon grease, I've
  >>ordered some pads - they always seem to transfer heat better than the
  >>grease (the pads that come with boxed athlons).
 >
 >
 > I'd recommend against going with the pads purely for ease-of-removal
 > reasons, especially as there doesn't seem to be much of a difference in
 > temperatures regardless of the type of compound used (Arctic silver,
 > bubblegum pads, toothpaste, etc etc).
 >
 > [...]
 >
 > --
 > Michael Brown
 > <a style='text-decoration: underline;' href="http://www.emboss.co.nz" target="_blank">www.emboss.co.nz</a> : OOS/RSI software and more
 > Add michael@ to emboss.co.nz - My inbox is always open
 >
 ><!-- ~MESSAGE_AFTER~ -->
 >> Stay informed about: No temperauture increase with processing?!? 

On Tue, 29 Jun 2004 02:00:05 -0500, David Maynard <dNOTmayn.RemoveThis@ev1.net> wrote:

 > Michael Brown wrote:
 >
  >> Peter Hucker wrote:
  >>
   >>> On Tue, 29 Jun 2004 12:30:38 +1200, Michael Brown
   >>> <see.RemoveThis@signature.below> wrote:
   >>>
   >>>
   >>>> Peter Hucker wrote:
   >>>>
   >>>>> I have a dual Athlon 2800+, the CPUs are hovering around 50C (within
   >>>>> a few degrees - assuming error margins in motherboard monitors).
   >>>>> They are at this temperature after a few minutes of turning the PC
   >>>>> on (when sat in the BIOS), and also when idling in windows XP. I
   >>>>> started a couple of programs to use both CPUs 100%, and the
   >>>>> temperatures remain the same! Is something not quite right here???
   >>>>
   >>>> Most (all?) dual-Athlon boards do not support halt-disconnect, so
   >>>> there is very little fluctuation in temperatures at all.
   >>>
   >>> !!!! Doesn't that mean I am running them unnecessarily hot?!?!?
  >>
  >>
  >> It means the chip is wasting power due to a limitation of the OS, so yes,
  >> sorta If you run Linux (kernel 2.6 and later IIRC), you'll find the idle
  >> temperatures should drop significantly as this does do halt-disconnect on
  >> idle.
 >
 > It has nothing to do with 'the OS'. It has to do with the nature of the
 > Athlon bus and the relatively new s2k bus disconnect, which is required for
 > it to go into low power mode, and which chipsets support it.

It's a bloody AMD chipset!

<font color=purple> > <a style='text-decoration: underline;' href="http://www.xbitlabs.com/articles/cpu/display/athlonxp-3000.html</font" target="_blank">http://www.xbitlabs.com/articles/cpu/display/athlonxp-3000.html</font</a>>

Intersting. WHOAH! 72C?? Surely that's a little high? Mine lock up at around 70C. In fact it's the highest setting there is in the BIOS for the CPU overheat alarm!


--
13 parrots and rising <a style='text-decoration: underline;' href="http://www.petersparrots.com" target="_blank">http://www.petersparrots.com</a>
93 silly video clips <a style='text-decoration: underline;' href="http://www.insanevideoclips.com" target="_blank">http://www.insanevideoclips.com</a>
1259 digital photos <a style='text-decoration: underline;' href="http://www.petersphotos.com" target="_blank">http://www.petersphotos.com</a>
Served from a watercooled dual 3.2GHz silent Athlon with 512GB of watercooled Raid.

New here? Pull up a chair and we'll plug you in.<!-- ~MESSAGE_AFTER~ -->
 >> Stay informed about: No temperauture increase with processing?!? 

Peter Hucker wrote:
 > On Tue, 29 Jun 2004 02:00:05 -0500, David Maynard <dNOTmayn RemoveThis @ev1.net> wrote:
 >
  >> Michael Brown wrote:
  >>
   >>> Peter Hucker wrote:
   >>>
   >>>> On Tue, 29 Jun 2004 12:30:38 +1200, Michael Brown
   >>>> <see RemoveThis @signature.below> wrote:
   >>>>
   >>>>
   >>>>> Peter Hucker wrote:
   >>>>>
   >>>>>> I have a dual Athlon 2800+, the CPUs are hovering around 50C (within
   >>>>>> a few degrees - assuming error margins in motherboard monitors).
   >>>>>> They are at this temperature after a few minutes of turning the PC
   >>>>>> on (when sat in the BIOS), and also when idling in windows XP. I
   >>>>>> started a couple of programs to use both CPUs 100%, and the
   >>>>>> temperatures remain the same! Is something not quite right here???
   >>>>>
   >>>>>
   >>>>> Most (all?) dual-Athlon boards do not support halt-disconnect, so
   >>>>> there is very little fluctuation in temperatures at all.
   >>>>
   >>>>
   >>>> !!!! Doesn't that mean I am running them unnecessarily hot?!?!?
   >>>
   >>>
   >>>
   >>> It means the chip is wasting power due to a limitation of the OS, so
   >>> yes,
   >>> sorta If you run Linux (kernel 2.6 and later IIRC), you'll find
   >>> the idle
   >>> temperatures should drop significantly as this does do
   >>> halt-disconnect on
   >>> idle.
  >>
  >>
  >> It has nothing to do with 'the OS'. It has to do with the nature of the
  >> Athlon bus and the relatively new s2k bus disconnect, which is
  >> required for
  >> it to go into low power mode, and which chipsets support it.
 >
 >
 > It's a bloody AMD chipset!

I presume your exclamation is meant to imply if it's an AMD chipset then it
must, of necessity, do bus disconnect (properly) but that isn't true.

 >
<font color=green>  >> <a style='text-decoration: underline;' href="http://www.xbitlabs.com/articles/cpu/display/athlonxp-3000.html</font" target="_blank">http://www.xbitlabs.com/articles/cpu/display/athlonxp-3000.html</font</a>>
 >
 >
 > Intersting. WHOAH! 72C?? Surely that's a little high? Mine lock up
 > at around 70C. In fact it's the highest setting there is in the BIOS
 > for the CPU overheat alarm!
 >
 ><!-- ~MESSAGE_AFTER~ -->
 >> Stay informed about: No temperauture increase with processing?!? 

David Maynard wrote:
 > Michael Brown wrote:
 >
  >> Peter Hucker wrote:
  >>
   >>> Michael Brown wrote:
   >>>
   >>>> Peter Hucker wrote:
   >>>>
   >>>>> I have a dual Athlon 2800+, the CPUs are hovering around 50C
   >>>>> (within a few degrees - assuming error margins in motherboard
   >>>>> monitors). They are at this temperature after a few minutes of
   >>>>> turning the PC on (when sat in the BIOS), and also when idling in
   >>>>> windows XP. I started a couple of programs to use both CPUs
   >>>>> 100%, and the temperatures remain the same! Is something not
   >>>>> quite right here???
   >>>>
   >>>> Most (all?) dual-Athlon boards do not support halt-disconnect, so
   >>>> there is very little fluctuation in temperatures at all.
   >>>
   >>> !!!! Doesn't that mean I am running them unnecessarily hot?!?!?
  >>
  >> It means the chip is wasting power due to a limitation of the OS, so
  >> yes, sorta If you run Linux (kernel 2.6 and later IIRC), you'll
  >> find the idle temperatures should drop significantly as this does do
  >> halt-disconnect on idle.
 >
 > It has nothing to do with 'the OS'. It has to do with the nature of
 > the Athlon bus and the relatively new s2k bus disconnect, which is
 > required for it to go into low power mode, and which chipsets support
 > it.
 >
<font color=purple> > <a style='text-decoration: underline;' href="http://www.xbitlabs.com/articles/cpu/display/athlonxp-3000.html</font" target="_blank">http://www.xbitlabs.com/articles/cpu/display/athlonxp-3000.html</font</a>>

The 760MP/MPX chipsets do not do halt-disconnect properly in SMP mode,
despite what it says in the docs (though it should work fine in uniprocessor
mode). Trying to do a halt-disconnect equavalent in such a system results in
a lockup most of the time (due to various bus contention issues). Hence it's
not enabled by the BIOS/vcool etc. However, it's still possible to make the
CPUs enter C2 (stop-grant, what halt-disconnect results in) and C3 (clock
control, runs the CPU at 1/[x] it's normal speed, [x] being 64 or 128
normally) through the southbridge, but only when both CPUs are idle. This is
what Linux does. However, since this requires modification of the idle loop
(among other problems), such a modification is very non-trivial to do in
Windows.

Incidentally, bus disconnect et al have been around since at least the
Thunderbird. Just hasn't been supported by many chipsets, and only was made
a requirement for validation relatively recently (as per your quoted
article).

[...]

--
Michael Brown
<a style='text-decoration: underline;' href="http://www.emboss.co.nz" target="_blank">www.emboss.co.nz</a> : OOS/RSI software and more
Add michael@ to emboss.co.nz - My inbox is always open<!-- ~MESSAGE_AFTER~ -->
 >> Stay informed about: No temperauture increase with processing?!? 

On Tue, 29 Jun 2004 20:39:20 +1200, Michael Brown <see.DeleteThis@signature.below> wrote:

 > David Maynard wrote:
  >> Michael Brown wrote:
  >>
   >>> Peter Hucker wrote:
   >>>
   >>>> Michael Brown wrote:
   >>>>
   >>>>> Peter Hucker wrote:
   >>>>>
<snip>
   >>>>>
   >>>>> Most (all?) dual-Athlon boards do not support halt-disconnect, so
   >>>>> there is very little fluctuation in temperatures at all.
   >>>>
   >>>> !!!! Doesn't that mean I am running them unnecessarily hot?!?!?
   >>>
   >>> It means the chip is wasting power due to a limitation of the OS, so
   >>> yes, sorta If you run Linux (kernel 2.6 and later IIRC), you'll
   >>> find the idle temperatures should drop significantly as this does do
   >>> halt-disconnect on idle.
  >>
  >> It has nothing to do with 'the OS'. It has to do with the nature of
  >> the Athlon bus and the relatively new s2k bus disconnect, which is
  >> required for it to go into low power mode, and which chipsets support
  >> it.
  >>
<font color=green>  >> <a style='text-decoration: underline;' href="http://www.xbitlabs.com/articles/cpu/display/athlonxp-3000.html</font" target="_blank">http://www.xbitlabs.com/articles/cpu/display/athlonxp-3000.html</font</a>>
 >
 > The 760MP/MPX chipsets do not do halt-disconnect properly in SMP mode,
 > despite what it says in the docs (though it should work fine in uniprocessor
 > mode). Trying to do a halt-disconnect equavalent in such a system results in
 > a lockup most of the time (due to various bus contention issues). Hence it's
 > not enabled by the BIOS/vcool etc. However, it's still possible to make the
 > CPUs enter C2 (stop-grant, what halt-disconnect results in) and C3 (clock
 > control, runs the CPU at 1/[x] it's normal speed, [x] being 64 or 128
 > normally) through the southbridge, but only when both CPUs are idle. This is
 > what Linux does. However, since this requires modification of the idle loop
 > (among other problems), such a modification is very non-trivial to do in
 > Windows.

There is an optin for stopgrant in the bios - might have been during suspend not idel, but I'll go check!

 > Incidentally, bus disconnect et al have been around since at least the
 > Thunderbird. Just hasn't been supported by many chipsets, and only was made
 > a requirement for validation relatively recently (as per your quoted
 > article).


--
13 parrots and rising <a style='text-decoration: underline;' href="http://www.petersparrots.com" target="_blank">http://www.petersparrots.com</a>
93 silly video clips <a style='text-decoration: underline;' href="http://www.insanevideoclips.com" target="_blank">http://www.insanevideoclips.com</a>
1259 digital photos <a style='text-decoration: underline;' href="http://www.petersphotos.com" target="_blank">http://www.petersphotos.com</a>
Served from a watercooled dual 3.2GHz silent Athlon with 512GB of watercooled Raid.

Why is there no Disneyland China?
No one's tall enough to go on the good rides.<!-- ~MESSAGE_AFTER~ -->
 >> Stay informed about: No temperauture increase with processing?!? 

On Tue, 29 Jun 2004 20:39:20 +1200, Michael Brown <see.TakeThisOut@signature.below> wrote:

 > David Maynard wrote:
  >> Michael Brown wrote:
  >>
   >>> Peter Hucker wrote:
   >>>
   >>>> Michael Brown wrote:
   >>>>
   >>>>> Peter Hucker wrote:
   >>>>>
<snip>
   >>>>>
   >>>>> Most (all?) dual-Athlon boards do not support halt-disconnect, so
   >>>>> there is very little fluctuation in temperatures at all.
   >>>>
   >>>> !!!! Doesn't that mean I am running them unnecessarily hot?!?!?
   >>>
   >>> It means the chip is wasting power due to a limitation of the OS, so
   >>> yes, sorta If you run Linux (kernel 2.6 and later IIRC), you'll
   >>> find the idle temperatures should drop significantly as this does do
   >>> halt-disconnect on idle.
  >>
  >> It has nothing to do with 'the OS'. It has to do with the nature of
  >> the Athlon bus and the relatively new s2k bus disconnect, which is
  >> required for it to go into low power mode, and which chipsets support
  >> it.
  >>
<font color=green>  >> <a style='text-decoration: underline;' href="http://www.xbitlabs.com/articles/cpu/display/athlonxp-3000.html</font" target="_blank">http://www.xbitlabs.com/articles/cpu/display/athlonxp-3000.html</font</a>>
 >
 > The 760MP/MPX chipsets do not do halt-disconnect properly in SMP mode,
 > despite what it says in the docs (though it should work fine in uniprocessor
 > mode). Trying to do a halt-disconnect equavalent in such a system results in
 > a lockup most of the time (due to various bus contention issues). Hence it's
 > not enabled by the BIOS/vcool etc. However, it's still possible to make the
 > CPUs enter C2 (stop-grant, what halt-disconnect results in) and C3 (clock
 > control, runs the CPU at 1/[x] it's normal speed, [x] being 64 or 128
 > normally) through the southbridge, but only when both CPUs are idle. This is
 > what Linux does. However, since this requires modification of the idle loop
 > (among other problems), such a modification is very non-trivial to do in
 > Windows.

There is an optin for stopgrant in the bios - might have been during suspend not idle, but I'll go check!

DRAT. It's a suspend mode - "Stop Grant (saves the state of the entire system to disk and then powers off the system)"

 > Incidentally, bus disconnect et al have been around since at least the
 > Thunderbird. Just hasn't been supported by many chipsets, and only was made
 > a requirement for validation relatively recently (as per your quoted
 > article).


--
13 parrots and rising <a style='text-decoration: underline;' href="http://www.petersparrots.com" target="_blank">http://www.petersparrots.com</a>
93 silly video clips <a style='text-decoration: underline;' href="http://www.insanevideoclips.com" target="_blank">http://www.insanevideoclips.com</a>
1259 digital photos <a style='text-decoration: underline;' href="http://www.petersphotos.com" target="_blank">http://www.petersphotos.com</a>
Served from a watercooled dual 3.2GHz silent Athlon with 512GB of watercooled Raid.

NEWSFLASH!!! Bouncing elephantiasis woman destroys central Portsmouth<!-- ~MESSAGE_AFTER~ -->
 >> Stay informed about: No temperauture increase with processing?!? 

Michael Brown wrote:

 > David Maynard wrote:
 >
  >>Michael Brown wrote:
  >>
  >>
   >>>Peter Hucker wrote:
   >>>
   >>>
   >>>>Michael Brown wrote:
   >>>>
   >>>>
   >>>>>Peter Hucker wrote:
   >>>>>
   >>>>>
   >>>>>>I have a dual Athlon 2800+, the CPUs are hovering around 50C
   >>>>>>(within a few degrees - assuming error margins in motherboard
   >>>>>>monitors). They are at this temperature after a few minutes of
   >>>>>>turning the PC on (when sat in the BIOS), and also when idling in
   >>>>>>windows XP. I started a couple of programs to use both CPUs
   >>>>>>100%, and the temperatures remain the same! Is something not
   >>>>>>quite right here???
   >>>>>
   >>>>>Most (all?) dual-Athlon boards do not support halt-disconnect, so
   >>>>>there is very little fluctuation in temperatures at all.
   >>>>
   >>>>!!!! Doesn't that mean I am running them unnecessarily hot?!?!?
   >>>
   >>>It means the chip is wasting power due to a limitation of the OS, so
   >>>yes, sorta If you run Linux (kernel 2.6 and later IIRC), you'll
   >>>find the idle temperatures should drop significantly as this does do
   >>>halt-disconnect on idle.
  >>
  >>It has nothing to do with 'the OS'. It has to do with the nature of
  >>the Athlon bus and the relatively new s2k bus disconnect, which is
  >>required for it to go into low power mode, and which chipsets support
  >>it.
  >>
  >>http://www.xbitlabs.com/articles/cpu/display/athlonxp-3000.html
 >
 >
 > The 760MP/MPX chipsets do not do halt-disconnect properly in SMP mode,
 > despite what it says in the docs (though it should work fine in uniprocessor
 > mode). Trying to do a halt-disconnect equavalent in such a system results in
 > a lockup most of the time (due to various bus contention issues). Hence it's
 > not enabled by the BIOS/vcool etc. However, it's still possible to make the
 > CPUs enter C2 (stop-grant, what halt-disconnect results in) and C3 (clock
 > control, runs the CPU at 1/[x] it's normal speed, [x] being 64 or 128
 > normally) through the southbridge, but only when both CPUs are idle. This is
 > what Linux does. However, since this requires modification of the idle loop
 > (among other problems), such a modification is very non-trivial to do in
 > Windows.
 >
 > Incidentally, bus disconnect et al have been around since at least the
 > Thunderbird. Just hasn't been supported by many chipsets, and only was made
 > a requirement for validation relatively recently (as per your quoted
 > article).

Yes, it's been around for a long time but it also hasn't always worked
properly for a long time, which is the reason for the recent certification
push, and many early motherboards which 'had it' suddenly disabled it in
later BIOS updates because of lock up problems.


 > [...]
 >
 > --
 > Michael Brown
 > <a style='text-decoration: underline;' href="http://www.emboss.co.nz" target="_blank">www.emboss.co.nz</a> : OOS/RSI software and more
 > Add michael@ to emboss.co.nz - My inbox is always open
 >
 ><!-- ~MESSAGE_AFTER~ -->
 >> Stay informed about: No temperauture increase with processing?!? 

Peter Hucker wrote:

 > Michael Brown wrote:
 >
  >> Peter Hucker wrote:
  >>
   >>> Michael Brown wrote:
   >>>
   >>>> Peter Hucker wrote:
   >>>>
   >>>>> I have a dual Athlon 2800+, the CPUs are hovering around 50C
   >>>>> (within a few degrees - assuming error margins in motherboard
   >>>>> monitors). They are at this temperature after a few minutes of
   >>>>> turning the PC on (when sat in the BIOS), and also when idling in
   >>>>> windows XP. I started a couple of programs to use both CPUs
   >>>>> 100%, and the temperatures remain the same! Is something not
   >>>>> quite right here???
   >>>>
   >>>> Most (all?) dual-Athlon boards do not support halt-disconnect, so
   >>>> there is very little fluctuation in temperatures at all.
   >>>
   >>> !!!! Doesn't that mean I am running them unnecessarily hot?!?!?
  >>
  >> It means the chip is wasting power due to a limitation of the OS, so
  >> yes, sorta If you run Linux (kernel 2.6 and later IIRC), you'll
  >> find the idle temperatures should drop significantly as this does do
  >> halt-disconnect on idle.
 >
 > But this never happened on a dual PIII. Can't be the OS.

See other post for a bit more elaboration on this point.

[...]
   >>>> With a K7D (measures
   >>>> internal diode temps, though seems to read higher than some other
   >>>> boards with the same CPU and HSF)
   >>>
   >>> Well the water temp is 35C, and the diode temp is 50C - does this
   >>> seem reasonable?
  >>
  >> Water temp sounds a little high unless you've got a high room
  >> temperature, but a delta of 15 deg C sounds "about right" for
  >> low-end water cooling.
 >
 > I was surprised at how LOW it was. The room is about 23 to 24C (warm
 > weather just now, and this room has three 19" monitors.

With a good radiator and fan setup, you should be able to get the water/air
delta to somewhere in the 3 deg C to 5 deg C range. Of course, cooling a
dually and keeping the delta this small requires a fairly hefty radiator

   >>> I don't really like this silicon grease, I've
   >>> ordered some pads - they always seem to transfer heat better than
   >>> the grease (the pads that come with boxed athlons).
  >>
  >> I'd recommend against going with the pads purely for ease-of-removal
  >> reasons, especially as there doesn't seem to be much of a difference
  >> in temperatures regardless of the type of compound used (Arctic
  >> silver, bubblegum pads, toothpaste, etc etc).
 >
 > TOOTHPASTE!

Heh heh, you haven't heard of the infamous DansData comparison?

<a style='text-decoration: underline;' href="http://www.dansdata.com/goop.htm" target="_blank">http://www.dansdata.com/goop.htm</a>

 > My theory was that the gunk stuff oozes away. For example I noticed
 > one CPU getting 10C extra because the gunk had not stayed where I put
 > it.

Correct, though it doesn't happen very quickly. You should (IMO) redo paste
every year or so if your computer is in a tower case, but it's probably not
so urgent in a desktop case. Low-pressure heatsinks (and indirectly, some
shims) also make the problems worse. However, you need to clean out thin-fin
HSFs about once a year anyhow, so I just do a general spring cleaning of my
computers every so often.

--
Michael Brown
<a style='text-decoration: underline;' href="http://www.emboss.co.nz" target="_blank">www.emboss.co.nz</a> : OOS/RSI software and more
Add michael@ to emboss.co.nz - My inbox is always open<!-- ~MESSAGE_AFTER~ -->
 >> Stay informed about: No temperauture increase with processing?!?