Should I have underclocked.

[Donald McTrevor]

I recently purchased a 300Mhz AMD K6/2 as 300 if the
max the motherboard could manage, however I did notice
a 400Mhz chip which I could have probably got for the same
price.
I am wondering whether I should have got the 400Mhz
model an underclocked it at 300Mhz?
I presume there would have been a lot less heat.

Also there may be even faster chips which I could have
underclocked, producing even less heat, maybe even
enough to run without a fan, which would be great
noise wise.

I note the K6/2 goes up to 550Mhz.


Also I just read else where
http://www.computing.net/cpus/wwwboard/forum/11920.html
"an AMD k6-2 will interpret a 2x multiplier as a 6x multipilier"
So I could have got a 450Mhz K6/2?
Actually I think someone here, possibly Kony had already
told me this (or something similar)

 

[DaveW]

You made the right choice.

 

[Donald McTrevor]

You made the right choice.

Care to explain why?

 

[kony]

I recently purchased a 300Mhz AMD K6/2 as 300 if the
max the motherboard could manage, however I did notice
a 400Mhz chip which I could have probably got for the same
price.
I am wondering whether I should have got the 400Mhz
model an underclocked it at 300Mhz?

Yes, you definitely should have.
K6-2 are not mulitplier locked, their spec'd frequency is
merely the guaranteed upper limit of stability. In other
words, buying a 400MHz version you had a guarantee it would
run at 300MHz if you chose to do so for any reason, as it
would be quite significantly (relatively speaking)
underclocked. With the 300MHz version you have far lesser
chance of overclocking it to 400MHz. Unlike the Intel chips
of that era, none of the AMD chips were mulitplier locked
and they all depend on the motherboard setting the
multiplier, or at least manipulating the mulitplier pins so
the final multiplier is determined in that method of
external signaling.

I presume there would have been a lot less heat.

?
Do you know what you're doing?
There is no reason to believe it makes much difference,
they're both low heat by modern standards but even so,
running one at 400MHz is higher heat than 300MHz.

Also there may be even faster chips which I could have
underclocked, producing even less heat, maybe even
enough to run without a fan, which would be great
noise wise.

You would have to undervolt them too, are you able to test
stability in such scenarios?

If you underclock a 400MHz CPU to 300MHz but leave the
voltage the same it will produce exactly same amount of heat
as a 300MHz CPU will, because they're actually the same CPU.
Sometimes there is a very minor deviation to this as very
slight core changes are made to facilitate higher clock
speeds in later models per CPU family, but generally
speaking the difference isn't worth much consideration.

Also, the whole point was getting more performance, yes?
If you wanted to underclock to reduce heat then this has all
been an exercise in madness.

I note the K6/2 goes up to 550Mhz.


Also I just read else where
http://www.computing.net/cpus/wwwboard/forum/11920.html
"an AMD k6-2 will interpret a 2x multiplier as a 6x multipilier"
So I could have got a 450Mhz K6/2?
Actually I think someone here, possibly Kony had already
told me this (or something similar)

yes, the best CPU would've been at least 450MHz, one spec'd
for 2.2V default voltage (some used 2.4V).

 

[Donald McTrevor]

Yes, you definitely should have.

Well thats a shame because there was on going for 99p
on the same day. However I guess there will be more in the
future, here is K6/2 450 with mobo and ram for £1.21
(looks a bit like mine but with a bettter chipset) £5 postage though.

Anyway I should be able to pick one up cheap in the future, because,
as people point out, its obsolecent stuff, like my four sticks of 16meg EDO
ram, I could try and sell them on ebay but I would be lucky to find a buyer.

Anyway I have 3 blobs of heat sink compouond ( I should have bought
a big tube).

K6-2 are not mulitplier locked, their spec'd frequency is
merely the guaranteed upper limit of stability. In other
words, buying a 400MHz version you had a guarantee it would
run at 300MHz if you chose to do so for any reason, as it
would be quite significantly (relatively speaking)
underclocked. With the 300MHz version you have far lesser
chance of overclocking it to 400MHz. Unlike the Intel chips
of that era, none of the AMD chips were mulitplier locked
and they all depend on the motherboard setting the
multiplier, or at least manipulating the mulitplier pins so
the final multiplier is determined in that method of
external signaling.



?
Do you know what you're doing?
There is no reason to believe it makes much difference,
they're both low heat by modern standards but even so,
running one at 400MHz is higher heat than 300MHz.


You would have to undervolt them too, are you able to test
stability in such scenarios?

No idea!! If it work it is stable I guess.

If you underclock a 400MHz CPU to 300MHz but leave the
voltage the same it will produce exactly same amount of heat
as a 300MHz CPU will, because they're actually the same CPU.
Sometimes there is a very minor deviation to this as very
slight core changes are made to facilitate higher clock
speeds in later models per CPU family, but generally
speaking the difference isn't worth much consideration.

Also, the whole point was getting more performance, yes?
If you wanted to underclock to reduce heat then this has all
been an exercise in madness.

I was thinking of maybe buying a faster this if possibe as the max I can
go is 6 X 75 so maybe I could use 600mhz chip underclocked to 450.

yes, the best CPU would've been at least 450MHz, one spec'd
for 2.2V default voltage (some used 2.4V).

Well I have got my chip now and I will be attempting to install it
soon, but I want to try and get some benchmarks so I can
measure any increase in performance.
I have never replaced the CPU before so if you don't hear from
me for a while it will be because I have fryed my computer!!

 

[kony]

No idea!! If it work it is stable I guess.

No, it is guaranteed, certain that if the CPU voltge is too
low the system could run but be instable. Test it with
Prime95 Torture Test (Google will find it). That will also
produce near the maximum (reasonably possible) heat from the
CPU so it is also a good test of the adequacy of the CPU
heatsink. As soon as (if) the test produces errors, you can
stop it immediately as the system is not stable.

I was thinking of maybe buying a faster this if possibe as the max I can
go is 6 X 75 so maybe I could use 600mhz chip underclocked to 450.

There are no 600MHz K6-2 CPUs. A 450MHz spec'd K6-2 would
be fine for that purpose of running at 6 x 75, so long as
your motherboard is able to adjust the PCI divider so 75MHz
isn't out of spec on the PCI bus. Previously (in another
thread?) I pointed out the jumper you need to change for
that.

Well I have got my chip now and I will be attempting to install it
soon, but I want to try and get some benchmarks so I can
measure any increase in performance.
I have never replaced the CPU before so if you don't hear from
me for a while it will be because I have fryed my computer!!

Just make sure the voltage is set correctly with the jumpers
before trying to turn on the system, and print out the
jumper settings beforehand.

 

[Donald McTrevor]

No, it is guaranteed, certain that if the CPU voltge is too
low the system could run but be instable. Test it with
Prime95 Torture Test (Google will find it). That will also
produce near the maximum (reasonably possible) heat from the
CPU so it is also a good test of the adequacy of the CPU
heatsink. As soon as (if) the test produces errors, you can
stop it immediately as the system is not stable.



There are no 600MHz K6-2 CPUs. A 450MHz spec'd K6-2 would
be fine for that purpose of running at 6 x 75, so long as
your motherboard is able to adjust the PCI divider so 75MHz
isn't out of spec on the PCI bus. Previously (in another
thread?) I pointed out the jumper you need to change for
that.



Just make sure the voltage is set correctly with the jumpers
before trying to turn on the system, and print out the
jumper settings beforehand.

Well I did get the setting wrong at first ( more on that
later, however the up shot is that my computer is now twice
as slow as it used to be!!!!!!!!!!! aaaaaaaaaarrrrrrrrhhhhhhhhhhh!!

I will add some info a little later, but its seems very slow
takes an age to boot, there may be something else wrong
or the K6/2 is a useless processor!! (worth £1 though, just).

 

[Donald McTrevor]

No, it is guaranteed, certain that if the CPU voltge is too
low the system could run but be instable. Test it with
Prime95 Torture Test (Google will find it). That will also
produce near the maximum (reasonably possible) heat from the
CPU so it is also a good test of the adequacy of the CPU
heatsink. As soon as (if) the test produces errors, you can
stop it immediately as the system is not stable.



There are no 600MHz K6-2 CPUs. A 450MHz spec'd K6-2 would
be fine for that purpose of running at 6 x 75, so long as
your motherboard is able to adjust the PCI divider so 75MHz
isn't out of spec on the PCI bus. Previously (in another
thread?) I pointed out the jumper you need to change for
that.



Just make sure the voltage is set correctly with the jumpers
before trying to turn on the system, and print out the
jumper settings beforehand.

How do I print then out/see thm??
Also in the manual only the K6 is quoted, not the K6-2
could that be a problem?

Would any BIOS changes help?

 

[kony]

How do I print then out/see thm??

They will either be in the manual or the link you supplied
days ago. Use the appropriate application to print that.

Also in the manual only the K6 is quoted, not the K6-2
could that be a problem?

I vaguely recall the link you supplied, showed K6-2 among
those supported. Double-check that link you gave. Even so,
"usually" if a board supports a K6 (1), it also will run a
K6-2. An early post I made had mentioned the power supply
subcircuit on the motherboard, that it is sufficient so that
is not at issue. Also the link showed the necessary 2.2V
CPU voltage setting.

Would any BIOS changes help?

I don't know what's wrong yet. What EXACTLY have you done?
Set the voltage to 2.2V. Set multiplier to 2X. Set FSB to
66MHz. Test that first, and only if it is stable, THEN
proceed to try 75MHz FSB and also changing that other PCI
jumper I previously mentioned- only changing it for 75MHz
attempts, not for 66MHz FSB attempts. If your original CPU
used 75MHz FSB, that means you would have to change that
jumper to the 66MHz FSB setting, as shown on that link you
provided previously.

 

[kony]

They will either be in the manual or the link you supplied
days ago. Use the appropriate application to print that.



I vaguely recall the link you supplied, showed K6-2 among
those supported. Double-check that link you gave. Even so,
"usually" if a board supports a K6 (1), it also will run a
K6-2. An early post I made had mentioned the power supply
subcircuit on the motherboard, that it is sufficient so that
is not at issue. Also the link showed the necessary 2.2V
CPU voltage setting.


I don't know what's wrong yet. What EXACTLY have you done?
Set the voltage to 2.2V. Set multiplier to 2X. Set FSB to
66MHz. Test that first, and only if it is stable, THEN
proceed to try 75MHz FSB and also changing that other PCI
jumper I previously mentioned- only changing it for 75MHz
attempts, not for 66MHz FSB attempts. If your original CPU
used 75MHz FSB, that means you would have to change that
jumper to the 66MHz FSB setting, as shown on that link you
provided previously.

Also, Google for "CPU-Z" and run it to determine the actual
operating frequency of your CPU. It is quite possible that
because you bought an old 300MHz K6-2, when you set it to 2X
it does not interpret it as 6X. That (re)interpretation is
not applicable to the earliest K6-2 models but was to the
later ones having higher spec'd speed. Essentially, my
recommendation was to buy at least a 400 or 450MHz CPU. If
CPU-Z shows it isn't running at 6X but rather 2X, then set
your multiplier to the highest setting the board does list
which of course is lower than 6X.

 

[Blackthorn]

Also, Google for "CPU-Z"

Feck's sake...google for Donald McTrevor +troll

 

[CBFalconer]

.... snip ...

If you underclock a 400MHz CPU to 300MHz but leave the
voltage the same it will produce exactly same amount of heat
as a 300MHz CPU will, because they're actually the same CPU.
Sometimes there is a very minor deviation to this as very
slight core changes are made to facilitate higher clock
speeds in later models per CPU family, but generally
speaking the difference isn't worth much consideration.

Not so, because the primary current flow mechanism is to charge and
discharge many itty-bitty capacitors once per clock cycle. This
results in a certain Q (coulombs) per clock period, and when you
increase the period you decrease the net Q per unit time, which is
current. Thus the 400 to 300 frequency ratio should result in 3/4
the current draw, and thus 3/4 the power level.

If you are comparing to a CPU rated at only 300, then it is
obviously not the identical CPU for some reason or another. If it
is built from the same masks using the same doping etc. it may very
well act the same as the derated 400, but there is no guarantee of
that. It may be derated because of some mechanism that prevents
charging the capacitors fully in the shorter period, for example.

 

[Donald McTrevor]

They will either be in the manual or the link you supplied
days ago. Use the appropriate application to print that.



I vaguely recall the link you supplied, showed K6-2 among
those supported. Double-check that link you gave. Even so,
"usually" if a board supports a K6 (1), it also will run a
K6-2. An early post I made had mentioned the power supply
subcircuit on the motherboard, that it is sufficient so that
is not at issue. Also the link showed the necessary 2.2V
CPU voltage setting.


I don't know what's wrong yet. What EXACTLY have you done?
Set the voltage to 2.2V. Set multiplier to 2X. Set FSB to
66MHz. Test that first, and only if it is stable, THEN
proceed to try 75MHz FSB and also changing that other PCI
jumper I previously mentioned- only changing it for 75MHz
attempts, not for 66MHz FSB attempts. If your original CPU
used 75MHz FSB, that means you would have to change that
jumper to the 66MHz FSB setting, as shown on that link you
provided previously.

SOrry about the late reply but my computer reset whilst I was
writing the last reply!!
So I decided to set it to 66mhz not the 75 it was however
that also had a boot problem (reset half way throuogh) so
I gave up and I am now now back on the Cyrix, which
booted up as Cryix MII 200, when it should have been 300!!
Also a prog called aida32 said it was running at 150mhz!!

So I have some problems, one thing I am thinking is that the
jumpers are rusty and not connected properly, actually I might have it
set at 75X2 not the 75 X 3, I will check.

More shortly.

 

[Donald McTrevor]

Also, Google for "CPU-Z" and run it to determine the actual
operating frequency of your CPU. It is quite possible that
because you bought an old 300MHz K6-2, when you set it to 2X
it does not interpret it as 6X. That (re)interpretation is
not applicable to the earliest K6-2 models but was to the
later ones having higher spec'd speed. Essentially, my
recommendation was to buy at least a 400 or 450MHz CPU. If
CPU-Z shows it isn't running at 6X but rather 2X, then set
your multiplier to the highest setting the board does list
which of course is lower than 6X.

What I did was to set it at 75 X 4 because 75 X 6 is two high
for a 300mhz processor so I never used the X2 'trick'.

I did thenk the problem might be due to a 75 bus being to
high, however I sstill had problems on 66 (X4) so I gave
up and went back to the Cyrix which I may have
mistakenly set to 75 X 2 when I meant 75X3 the
setting for 2 = 0 0 1
and for 3 = 1 0 0 and I think I may have read the
jumpers the wrong way around, manual not that clear, and
circuit board marking hard to see, I will check later.

Going by that logic as X4 = 1 1 0 then I *might* have set it
to 0 1 1, which is undefined and may well have defaulted to
X1.5, which would give 75 X 1.5= 112.5 Mhz, which is actually
what it felt like, 1/3 of my normal speed!!
I hope that is the solution, I will try later!!

 

[Donald McTrevor]

Feck's sake...google for Donald McTrevor +troll

Don't listen to him, I am not trolling. (Obviously). More likely he is a
troll.

 

[Donald McTrevor]

Not so, because the primary current flow mechanism is to charge and
discharge many itty-bitty capacitors once per clock cycle. This
results in a certain Q (coulombs) per clock period, and when you
increase the period you decrease the net Q per unit time, which is
current. Thus the 400 to 300 frequency ratio should result in 3/4
the current draw, and thus 3/4 the power level.

Yes I would agree with that, posibbly or even better, however is not
power current squared X resistance in which case it wuld be
3/4 X 3/4 = 9/16 = about half the power?
If true thats quite a saving on heat and you might get away
without a heatsink even better to drop from 600 to 300
which would be only 25% of normal power.

 

[Donald McTrevor]

What I did was to set it at 75 X 4 because 75 X 6 is two high
for a 300mhz processor so I never used the X2 'trick'.

I did thenk the problem might be due to a 75 bus being to
high, however I sstill had problems on 66 (X4) so I gave
up and went back to the Cyrix which I may have
mistakenly set to 75 X 2 when I meant 75X3 the
setting for 2 = 0 0 1
and for 3 = 1 0 0 and I think I may have read the
jumpers the wrong way around, manual not that clear, and
circuit board marking hard to see, I will check later.

Going by that logic as X4 = 1 1 0 then I *might* have set it
to 0 1 1, which is undefined and may well have defaulted to
X1.5, which would give 75 X 1.5= 112.5 Mhz, which is actually
what it felt like, 1/3 of my normal speed!!
I hope that is the solution, I will try later!!

Indeed my Cyrix was set at 2X75, I have now put it
right and it is a 3X75= MII 300.
I don't know if I had a similar problem with the K6-2, it's
a bigger job changing CPU again, also more jumpers to get
wrong!!
I may try again later, or more likely tomorrow.

 

[Donald McTrevor]

SOrry about the late reply but my computer reset whilst I was
writing the last reply!!
So I decided to set it to 66mhz not the 75 it was however
that also had a boot problem (reset half way throuogh) so
I gave up and I am now now back on the Cyrix, which
booted up as Cryix MII 200, when it should have been 300!!
Also a prog called aida32 said it was running at 150mhz!!

So I have some problems, one thing I am thinking is that the
jumpers are rusty and not connected properly, actually I might have it
set at 75X2 not the 75 X 3, I will check.

Yes it was 2X75 and now fixed back to 3X75 on the Cryix,
not sure about the K6/2 yet though.

 

[Donald McTrevor]

Yes it was 2X75 and now fixed back to 3X75 on the Cryix,
not sure about the K6/2 yet though.

Well I have retried the K6 and it booted up in good time and
ran well, not sureifit was any faster, boot took a little longer I think.
Inconclusive test on a .wmv, may have been a little faster.
I was about to report some success when the computer
rebooted it self :O(

Getting a bit fed up with it now. I was running at 75X4 so I may
try 66X4 tomorrow, but I am not very optimistic.

I am back on the cyrix now, almost thought I had fried that but
I forgot to push the CPU insertion leaver down!!

I would be better going to PC world and buying that lovely
Athlon64 methinks.

I am lucky I still have a working machine after all the fiddling.

 

[kony]

Not so, because the primary current flow mechanism is to charge and
discharge many itty-bitty capacitors once per clock cycle. This
results in a certain Q (coulombs) per clock period, and when you
increase the period you decrease the net Q per unit time, which is
current. Thus the 400 to 300 frequency ratio should result in 3/4
the current draw, and thus 3/4 the power level.

Yes, perhaps I wasn't clear. I meant that running a 400MHz
(spec'd) CPU at 300MHz, would produce same heat as running a
300MHz (spec'd) CPU at 300MHz, providing both had same
voltage.

If you are comparing to a CPU rated at only 300, then it is
obviously not the identical CPU for some reason or another.

Depends a bit on what's being compared, but towards the end
of the K6-2 product cycle, yes the 300MHz parts were the
same as 400MHz parts. I don't recall the earlier core
refinements that allowed them to ramp the speeds up but it
seems unlikely that made any significant difference in heat
generation, except that after a certain point they did drop
the default voltage from 2.4V to 2.2V across-the-board, for
the same speeds.

If it
is built from the same masks using the same doping etc. it may very
well act the same as the derated 400, but there is no guarantee of
that. It may be derated because of some mechanism that prevents
charging the capacitors fully in the shorter period, for example.

Perhaps that it true in certain situations but I don't think
it applies to the K6-2 to any significant extent. It did
appear to have a fairly linear heat increase with frequency.
If there were other mitigating factors I would consider them
slight, not a very significant difference in power (heat)
per MHz.