16 Nov
1999
16 Nov
'99
4:40 p.m.
For laughs I've been running a 486dx/33 (intel) at 66mhz in a dell box.
The only mod if the heat sink is a moose I had and its attached to the
chip using a BEo filled thermal epoxy. Runs cool and it's solid.
I've tried that in a socket5 Mb I have using a 486dx2/50 at 100mhz, seems
fine after three months. The cooling is not big a problem.
I consider both abuse, but hey I have a few. Besides I've done that before
with 8085s and Z80s with no ill effects.
Allison
17 Nov
17 Nov
2:13 a.m.
For laughs I've been running a 486dx/33 (intel) at
66mhz in a dell box.
The only mod if the heat sink is a moose I had and its attached to the
chip using a BEo filled thermal epoxy. Runs cool and it's solid.
486dx33 at 66 MHz ? Well, I don't want to be picky, but wouldn't
this result in 66 MHz CPU bus ? Sounds a bit off for 486 Boards.
I've tried that in a socket5 Mb I have using a
486dx2/50 at 100mhz, seems
fine after three months. The cooling is not big a problem.
Jep, the most notable CPU is still the AMD 5x86 (486dx4/133 with
extended cache) - I have them running at 4x40 MHz == 160 MHz without
any problem - only for high load apps the cooling must be presen,
a Linux (web) server can even run without a fan.
I consider both abuse, but hey I have a few. Besides
I've done that before
with 8085s and Z80s with no ill effects.
I found it hard to run NMOS 6502 on more than 1.3 times their nominal speed.
Gruss
H.
P.S.: When designing a new cooler element/radiator, keep in mind that
silver is a way better thermal conductor than Al or Cu.
--
Der Kopf ist auch nur ein Auswuchs wie der kleine Zeh.
H.Achternbusch
4:53 a.m.
On Wed, 17 Nov 1999, Hans Franke wrote:
P.S.: When designing a new cooler element/radiator,
keep in mind that
silver is a way better thermal conductor than Al or Cu.
Considering that the aluminum heat-exchanger/cold-plates from my last
water-cooled design weigh more than 7lbs total, I don't think I'm going to
be re-implementing my design in silver any time too soon...
Aaron
8:02 a.m.
Rumor has it that Aaron Christopher Finney may have mentioned these words:
On Wed, 17 Nov 1999, Hans Franke wrote:
Aw, c'mon! By my (admittedly rough) calculations, that's only about $510USD
worth of silver (at $5/troy oz spot price). Isn't your hobby worth that? ;-)
See ya,
Roger "Merch" Merchberger
--
Roger "Merch" Merchberger --- sysadmin, Iceberg Computers
Recycling is good, right??? Ok, so I'll recycle an old .sig.
If at first you don't succeed, nuclear warhead
disarmament should *not* be your first career choice.
P.S.: When designing a new cooler
element/radiator, keep in mind that
silver is a way better thermal conductor than Al or Cu.
Considering that the aluminum heat-exchanger/cold-plates from my last
water-cooled design weigh more than 7lbs total, I don't think I'm going to
be re-implementing my design in silver any time too soon...
18 Nov
18 Nov
6:19 a.m.
New subject: E.U.N.U.C.H. (Cost)
>> P.S.: When designing a new cooler
element/radiator, keep in mind that
>> silver is a way better thermal conductor than Al or Cu.
>Considering that the aluminum
heat-exchanger/cold-plates from my last
>water-cooled design weigh more than 7lbs total, I don't think I'm going to
>be re-implementing my design in silver any time too soon...
Aw, c'mon! By my (admittedly rough) calculations,
that's only about $510USD
worth of silver (at $5/troy oz spot price). Isn't your hobby worth that? ;-)
Well, true, Ag is cheap - and you don't need a real lot.
Just if Aaron is now using SEVEN pound of Al, he might
need now 7 *(10,5/2,7) or 27,2 lbs - quite some more stuff.
But on the other hand I used on my silver K6-2 cooler just
600g (1.5lbs), worth ~290 Mark (~150 USD).
Gruss
H.
It's a simple design, operating just on water, and always around
room temperature - and I pushed an K6-2 300 up to 500 MHz...
(Of course copper would also be ok ... I just compared the price,
and for less than 50 bucks difference I got a cool silver thingy :)
--
Stimm gegen SPAM: http://www.politik-digital.de/spam/de/
Vote against SPAM: http://www.politik-digital.de/spam/en/
Votez contre le SPAM: http://www.politik-digital.de/spam/fr/
Ich denke, also bin ich, also gut
HRK
8:29 a.m.
New subject: E.U.N.U.C.H. (Cost)
> >> P.S.: When designing a new cooler
element/radiator, keep in mind that
> >> silver is a way better thermal conductor than Al or Cu.
It's a simple design, operating just on water, and
always around
room temperature - and I pushed an K6-2 300 up to 500 MHz...
(Of course copper would also be ok ... I just compared the price,
and for less than 50 bucks difference I got a cool silver thingy :)
With a cooler that is chip/plate/water don't you want to keep the "plate"
as thin as possible?
19 Nov
19 Nov
5:33 a.m.
New subject: E.U.N.U.C.H. (Materialthickness)
>> >> P.S.: When designing a new cooler
element/radiator, keep in mind that
>> >> silver is a way better thermal conductor than Al or Cu.
>It's a simple design, operating just on water,
and always around
>room temperature - and I pushed an K6-2 300 up to 500 MHz...
>(Of course copper would also be ok ... I just compared the price,
>and for less than 50 bucks difference I got a cool silver thingy :)
With a cooler that is chip/plate/water don't you
want to keep the "plate"
as thin as possible?
Jep, since phi = lambda*S*delta-T / delta . Just, I didn't use a
simple plate and only one waterlayer, but rather a system of pipes
(drilled holes) to have a) a better controll for the water flow and
b) enlarge the plate / water contact by three. As larger the
surface (S) as higher the possible thermal current (vulgo amount
of transported heat). I can't enlarge the chips surface, so S is
fixed between chip and 'plate', but I can enlarge S between 'plate'
and water.
The minimal thickness is defined by your tools and skills - I used
0.8 mm (~1/30 inch) as minimal thickness betreen one pipe and the
chip.
Gruss
H.
--
Stimm gegen SPAM: http://www.politik-digital.de/spam/de/
Vote against SPAM: http://www.politik-digital.de/spam/en/
Votez contre le SPAM: http://www.politik-digital.de/spam/fr/
Ich denke, also bin ich, also gut
HRK
8:36 a.m.
New subject: E.U.N.U.C.H. (Materialthickness)
With a cooler
that is chip/plate/water don't you want to keep the "plate"
as thin as possible?
Jep, since phi = lambda*S*delta-T / delta . Just, I didn't use a
simple plate and only one waterlayer, but rather a system of pipes
(drilled holes) to have a) a better controll for the water flow and
b) enlarge the plate / water contact by three. As larger the
surface (S) as higher the possible thermal current (vulgo amount
of transported heat). I can't enlarge the chips surface, so S is
fixed between chip and 'plate', but I can enlarge S between 'plate'
and water.
The minimal thickness is defined by your tools and skills - I used
0.8 mm (~1/30 inch) as minimal thickness betreen one pipe and the
chip.
10:08 a.m.
New subject: E.U.N.U.C.H. (Materialthickness)
>> With a cooler that is chip/plate/water
don't you want to keep the "plate"
>> as thin as possible?
>Jep, since phi = lambda*S*delta-T / delta . Just,
I didn't use a
>simple plate and only one waterlayer, but rather a system of pipes
>(drilled holes) to have a) a better controll for the water flow and
>b) enlarge the plate / water contact by three. As larger the
>surface (S) as higher the possible thermal current (vulgo amount
>of transported heat). I can't enlarge the chips surface, so S is
>fixed between chip and 'plate', but I can enlarge S between 'plate'
>and water.
>The minimal thickness is defined by your tools and
skills - I used
>0.8 mm (~1/30 inch) as minimal thickness betreen one pipe and the
>chip.
Since we are talking fluid cooling, anybody putting a
heatsink on both
sides of the chip?
I would not worry too much about the water side,
pressure and flow can work
wonders there, but would keep the copper plate just thick enough to avoid
hot spots and cover as much of the chip surface as electrically possible.
Well, sounds gut at first sight - just three problems:
First you have only a _very_ small free area on the 'other' side - there
are some nasty pins, and somehor it is no-good to touch them.
Ans second, this is hard to do in any standard board - and even
with custom extenders you still have to connect them and the
wire length becomes a problem - adding 10% corespeed and loosing
20% bus speed isn't exactly a gain.
Gruss
H.
--
Stimm gegen SPAM: http://www.politik-digital.de/spam/de/
Vote against SPAM: http://www.politik-digital.de/spam/en/
Votez contre le SPAM: http://www.politik-digital.de/spam/fr/
Ich denke, also bin ich, also gut
HRK
5:28 p.m.
New subject: E.U.N.U.C.H. (Materialthickness)
Since we are
talking fluid cooling, anybody putting a heatsink on both
sides of the chip?
I would not worry too much about the water side,
pressure and flow can work
wonders there, but would keep the copper plate just thick enough to avoid
hot spots and cover as much of the chip surface as electrically possible.
Well, sounds gut at first sight - just three problems:
First you have only a _very_ small free area on the 'other' side - there
are some nasty pins, and somehor it is no-good to touch them.
Ans second, this is hard to do in any standard board - and even
with custom extenders you still have to connect them and the
wire length becomes a problem - adding 10% corespeed and loosing
20% bus speed isn't exactly a gain.
20 Nov
20 Nov
5:03 a.m.
New subject: E.U.N.U.C.H. (DrillingHoles)
Rumor has it that Mike Ford may have mentioned these words:
What I had in mind was a hole in the motherboard,
perhaps two (water
in/out), with a small plate making contact directly under the "chip". My
guess is that it could be substantly more effective given the internal
construction of most chips.
Unforch, with the density of the traces & datapaths necessary on newer
motherboards, you'll surely ruin the motherboard drilling thru it at the
CPU socket area... not good for the operation of the computer, IMHO. ;-)
There are some Pentium AT motherboards that have a thru-hole under the chip
socket for support... if you found other ways to support the mobo and you
are *very* good with small tubing, you might be able to work this to your
advantage... but it would be tricky work and I'm not sure how much benefit
you'd see from it.
Take care,
Roger "Merch" Merchberger
--
Roger "Merch" Merchberger --- sysadmin, Iceberg Computers
Recycling is good, right??? Ok, so I'll recycle an old .sig.
If at first you don't succeed, nuclear warhead
disarmament should *not* be your first career choice.
22 Nov
22 Nov
1:52 a.m.
New subject: E.U.N.U.C.H. (Materialthickness)
>> Since we are talking fluid cooling, anybody
putting a heatsink on both
>> sides of the chip?
>> I would not worry too much about the water side, pressure and flow can work
>> wonders there, but would keep the copper plate just thick enough to avoid
>> hot spots and cover as much of the chip surface as electrically possible.
>Well, sounds gut at first sight - just three
problems:
>First you have only a _very_ small free area on the 'other' side - there
>are some nasty pins, and somehor it is no-good to touch them.
>Ans second, this is hard to do in any standard board - and even
>with custom extenders you still have to connect them and the
>wire length becomes a problem - adding 10% corespeed and loosing
>20% bus speed isn't exactly a gain.
What I had in mind was a hole in the motherboard,
perhaps two (water
in/out), with a small plate making contact directly under the "chip". My
guess is that it could be substantly more effective given the internal
construction of most chips.
Well, take a look at any ordinary board, and you will notice that
the 'free' space below the CPU isn't free at all - 4 to 6 layers
of connections - and I just asume they may have some sense - also
the reachaple surface is _very small (less than 3cm^2).
I just see I forgot to mention problem number 3:
The bottom side has (AFAIK) never a usable (and thermal
relevant) connection to the chip itself. wouldn't it be
somewhat stupid to rely on thermal transport into a region
where under all normal circumstanses no or only a minimal
heatremovel is possible ?).
Servus
Hans
--
Stimm gegen SPAM: http://www.politik-digital.de/spam/de/
Vote against SPAM: http://www.politik-digital.de/spam/en/
Votez contre le SPAM: http://www.politik-digital.de/spam/fr/
Ich denke, also bin ich, also gut
HRK
7:23 a.m.
New subject: E.U.N.U.C.H. (Materialthickness)
I just see I forgot to mention problem number 3:
The bottom side has (AFAIK) never a usable (and thermal
relevant) connection to the chip itself. wouldn't it be
somewhat stupid to rely on thermal transport into a region
where under all normal circumstanses no or only a minimal
heatremovel is possible ?).
That would be the fundamental problem, and I thought it was just the
opposite, but I suspect you may be right. The bottom does look a lot like a
"cap" now that I think about it.
17 Nov
17 Nov
5:11 a.m.
On Wed, 17 Nov 1999, Hans Franke wrote:
ISA not PCI. The bus is still cpuclock /n and n in that case is 8.
I have about 8 of them running here at work including one as a NT3.51
server all at 133mhz. Very good machines!
For laughs
I've been running a 486dx/33 (intel) at 66mhz in a dell box.
The only mod if the heat sink is a moose I had and its attached to the
chip using a BEo filled thermal epoxy. Runs cool and it's solid.
486dx33 at 66 MHz ? Well, I don't want to be picky, but wouldn't
this result in 66 MHz CPU bus ? Sounds a bit off for 486 Boards. I've tried that in a socket5 Mb I have using
a 486dx2/50 at 100mhz, seems
fine after three months. The cooling is not big a problem.
Jep, the most notable CPU is still the AMD 5x86 (486dx4/133 with
extended cache) - I have them running at 4x40 MHz == 160 MHz without
any problem - only for high load apps the cooling must be presen,
a Linux (web) server can even run without a fan. I found it hard to run NMOS 6502 on more than 1.3
times their nominal speed.
The 8085s and z80s would run that fast but things like propagation times
don't shrink so the access times and all do not scale well. For example
take a 4mhz z80 to 6mhz and you better have memory that is faster than
what would be needed first at glance. Same thing pushing 8085H-2 (6mhz,
to 8mhz).
Allison
18 Nov
18 Nov
6:19 a.m.
New subject: E.U.N.U.C.H. (Overclocking in the old time)
For laughs I've been running a 486dx/33 (intel) at
66mhz in a dell box.
The only mod if the heat sink is a moose I had and its attached to the
chip using a BEo filled thermal epoxy. Runs cool and it's solid.
486dx33 at
66 MHz ? Well, I don't want to be picky, but wouldn't
this result in 66 MHz CPU bus ? Sounds a bit off for 486 Boards. I've tried that in a socket5 Mb I have using a
486dx2/50 at 100mhz, seems
fine after three months. The cooling is not big a problem.
Jep, the most notable
CPU is still the AMD 5x86 (486dx4/133 with
extended cache) - I have them running at 4x40 MHz == 160 MHz without
any problem - only for high load apps the cooling must be presen,
a Linux (web) server can even run without a fan. I found it
hard to run NMOS 6502 on more than 1.3 times their nominal speed.
The 8085s and
z80s would run that fast but things like propagation times
don't shrink so the access times and all do not scale well. For example
take a 4mhz z80 to 6mhz and you better have memory that is faster than
what would be needed first at glance. Same thing pushing 8085H-2 (6mhz,
to 8mhz).
5:46 a.m.
New subject: E.U.N.U.C.H. (Overclocking in the old time)
an (non multiplieable) basic 486DX CPU, the only way
is to raise bus
speed - and 66 MHz is way out of Spec for all 486es I know - regulary
only the AMD Parts are speced for 40 or the old Intel DX50 (was quite
It was abuse but it works.
True. But for the 6502 even more the CPU itself was
often the hardest
part to tweak. Geting fast Memorys wasn't that problem - the CPU itself
starts to malfunction, possibly the propagation times within the CPU
where already stressed.
That is the true limiting speed of most CPUs. The RCA CDP1802 cmos part
is an excellent example of this. At certain voltages and high speeds
the part will not work due to the internal propagation delays.
Allison
9082
days inactive
9088
days old
15 comments
5 participants
participants (5)
-
af-list＠is1.wfi-inc.com -
allisonp＠world.std.com -
Hans.Franke＠mch20.sbs.de -
mikeford＠socal.rr.com -
zmerch＠30below.com