AMD K6/2 faster than a 933MHz Pentium II?

[Pen]

In my recollection and experience, the K6-2 300 would only
overclock
about 10%, or to 66x5.

 

[Donald McTrevor]

In my recollection and experience, the K6-2 300 would only
overclock
about 10%, or to 66x5.

I maybe able to X4.5 and X5 although not offically in the manual.
I think I need to overclock it by about 300% unfortunately!!

 

[kony]

I see it has no L2 cache but I think this ia feature rather than a problem.

But.....
http://www2.geek.com/discus/messages/27/5678.html?1013720343

SO it seems you can turn it on in the BIOS but I don't know how to do that!!
Any assistance appreciated :O)

Lack of L2 cache on the CPU is a problem, in that it makes
the CPU significantly slower. However, earlier socket 7
cpus didn't have L2 cache integral either, so at least your
motherboard does have L2 cache on it. You do not need to
change any board bios settings for the L2 cache, certainly
it was enabled all these years with you Cyrix CPU (else it
would've ran so slow it wasn't useable at all) and remains
enabled unless you manually disable it.

 

[kony]

In my recollection and experience, the K6-2 300 would only
overclock
about 10%, or to 66x5.

His variant is a weak K6-2, too, it probably has less
headroom than the avg (later version of) 300MHz chip. The
last ones sold did sometimes make it past 400MHz but it
usually not to 500MHz or more, and at that point the 400MHz
versions weren't even much more expensive.

 

[Donald McTrevor]

Lack of L2 cache on the CPU is a problem, in that it makes
the CPU significantly slower. However, earlier socket 7
cpus didn't have L2 cache integral either, so at least your
motherboard does have L2 cache on it. You do not need to
change any board bios settings for the L2 cache, certainly
it was enabled all these years with you Cyrix CPU (else it
would've ran so slow it wasn't useable at all) and remains
enabled unless you manually disable it.

There is a jumper which allows you to run L2 cache at
'linear burst mode' for the Cyrix, so maybe I can only
have L2 cache with a Cyrix? I don't know.
I am trying to see if I can update my BIOS.
(not got very far).

 

[kony]

There is a jumper which allows you to run L2 cache at
'linear burst mode' for the Cyrix, so maybe I can only
have L2 cache with a Cyrix? I don't know.
I am trying to see if I can update my BIOS.
(not got very far).

Read your motherboard manual. You can use simple benchmark
tests for memory, cache performance in Sisoft Sandra. It
will also provide a crude CPU benchmark too. Linear burst
mode was Cyrix specific, IF I remember correctly. K6-2
should use the default cache setting shown in your manual,
the same as a Pentium 1/1-MMX did.

 

[Donald McTrevor]

Read your motherboard manual. You can use simple benchmark
tests for memory, cache performance in Sisoft Sandra. It
will also provide a crude CPU benchmark too. Linear burst
mode was Cyrix specific, IF I remember correctly. K6-2
should use the default cache setting shown in your manual,
the same as a Pentium 1/1-MMX did.

The manual is not too helpful:-

"Cache 0/512KB, Write Back Direct Mapped organisation, Pipeline Burst
Cache soldered onto the motherboard" is what it says.

I can't find anything in the manual to enable cache for the K6, I get the
impression
it cannot be done. There are no more jumper setting mentioned and nothing
I could do in the BIOS when I booted with the K6.
I will try looking at the BIOS with the Cyrix.
Maybe a BIOS upgrade might help.

 

[Donald McTrevor]

The manual is not too helpful:-

"Cache 0/512KB, Write Back Direct Mapped organisation, Pipeline Burst
Cache soldered onto the motherboard" is what it says.

I can't find anything in the manual to enable cache for the K6, I get the
impression
it cannot be done. There are no more jumper setting mentioned and nothing
I could do in the BIOS when I booted with the K6.
I will try looking at the BIOS with the Cyrix.
Maybe a BIOS upgrade might help.

The BIOS menu is the same with the Cyrix.
There are more BIOS menu options in the manual than
in the actual BIOS setup.
"BIOS features set up" seems to be missing, or
may it was "Standard Cmos set up" I forget, but it is one of the
two.
There is also a an option to load the default settings but I am
wary of doing that.

In the link I suspected was my mobo
http://www.uktsupport.co.uk/pb/mb/850.htm

There is another link.(BIOS upgrade)
http://support.packardbell.com/uk/mypc/?PibItemNr=REFBIOS00300200#show
File Title: PB850 (FR500) BIOS

Note it says FR500 which is what my manual says.

Also another version
http://support.packardbell.com/uk/mypc/?PibItemNr=REFBIOS00300100#show

But I would probably do more damage than good .
BEst wait untill I have another computer to fall back on.

 

[kony]

The manual is not too helpful:-

"Cache 0/512KB, Write Back Direct Mapped organisation, Pipeline Burst
Cache soldered onto the motherboard" is what it says.

OK, what's wrong with that?

I can't find anything in the manual to enable cache for the K6, I get the
impression
it cannot be done.

You don't seem to be grasping something- you do NOT need to
"enable the cache for the K6". The cache was already
enabled the whole time, there is no "enabling" to do.

There are no more jumper setting mentioned and nothing
I could do in the BIOS when I booted with the K6.
I will try looking at the BIOS with the Cyrix.
Maybe a BIOS upgrade might help.

Help with what?
There are ONLY five things you need(ed) to do to go from
(having done nothing, right after you made your very first
post) to being finished:

1) Buy a 450MHz+ K6-2
2) Install it and the heatsink/fan.
3) Set jumpers for voltage, multiplier, and FSB. Any other
jumpers should remain the same, the jumpers should look
identical to how they would if you had a Pentium 133 MMX
installed, except voltage would be 2.2V or 2.4V (as stamped
on the K6-2).
4) Run CPU-Z to confirm speed and Sisoft Sandra to check
performance.

The entire process takes a little under 30 minutes, I can't
understand why you're making something easy, so difficult.

 

[kony]

The BIOS menu is the same with the Cyrix.
There are more BIOS menu options in the manual than
in the actual BIOS setup.
"BIOS features set up" seems to be missing, or
may it was "Standard Cmos set up" I forget, but it is one of the
two.
There is also a an option to load the default settings but I am
wary of doing that.

In the link I suspected was my mobo
http://www.uktsupport.co.uk/pb/mb/850.htm

There is another link.(BIOS upgrade)
http://support.packardbell.com/uk/mypc/?PibItemNr=REFBIOS00300200#show
File Title: PB850 (FR500) BIOS

When you put in any CPU except a Cyrix, you need to open
JP17 as linked above. There is no other cache setting or
bios menu setting you should need to change.

Note it says FR500 which is what my manual says.

Also another version
http://support.packardbell.com/uk/mypc/?PibItemNr=REFBIOS00300100#show

But I would probably do more damage than good .
BEst wait untill I have another computer to fall back on.

What do you hope to gain?
Your L2 cache is enabled.
Did you ever figure out what speed to run the K6-2?

This is madness, if you had simply bought, installed, and
used the new CPU it might've been reasonable to upgrade for
a buck, but it is not worth the time you're spending.

My last recommendation- just set it and forget it, get on
with using the system.

 

[Donald McTrevor]

When you put in any CPU except a Cyrix, you need to open
JP17 as linked above. There is no other cache setting or
bios menu setting you should need to change.

I always do, it won't even boot up otherwise as I found when I forgot
to do it once.

What do you hope to gain?
Your L2 cache is enabled.
Did you ever figure out what speed to run the K6-2?

It was fine at 50 X 6 and 66X4 (posibblly faster)
also ran at 4.5 X 66 but did have a crash on second boot
so probably not stable.

This is madness, if you had simply bought, installed, and
used the new CPU it might've been reasonable to upgrade for
a buck, but it is not worth the time you're spending.

Well I have learn't a bit, used to take meover an hour to
swop CPU, I can probably do it in a 5 minutes now.

 

[Donald McTrevor]

OK, what's wrong with that?

I think only the Cyrix can use that cache.

You don't seem to be grasping something- you do NOT need to
"enable the cache for the K6". The cache was already
enabled the whole time, there is no "enabling" to do.

But I have to disable the pipeline burst bit which seems to disable the
cache altogeather as AIDA32 shows no L2 cache for the K6-2

Help with what?
There are ONLY five things you need(ed) to do to go from
(having done nothing, right after you made your very first
post) to being finished:

1) Buy a 450MHz+ K6-2
2) Install it and the heatsink/fan.
3) Set jumpers for voltage, multiplier, and FSB. Any other
jumpers should remain the same, the jumpers should look
identical to how they would if you had a Pentium 133 MMX
installed, except voltage would be 2.2V or 2.4V (as stamped
on the K6-2).
4) Run CPU-Z to confirm speed and Sisoft Sandra to check
performance.

The entire process takes a little under 30 minutes, I can't
understand why you're making something easy, so difficult.

It didn't take that long it just was not stable at many settings.and
when it crashes it takes a long time to reboot/scandisk on the wrong
speed/slow speed.
As for benchmarks it is whether it will do what I want it to do
in the real world which matters most.
I doubt the 450 CPU would be stable in my system anyway, it has
enough trouble with the 300.

 

[kony]

I think only the Cyrix can use that cache.

Then you have jumped to a conclusion without any reason to
do so. The cache will be enabled, no matter what CPU,
unless you take specific steps to "disable" it. Since you
haven't disabled it, it is still being used.

As I'd mentioned already, you need to test the cache by
benchmarking. Do not guess about the potential for a
problem, benchmark and focus on any problems factually
shown.

But I have to disable the pipeline burst bit which seems to disable the
cache altogeather as AIDA32 shows no L2 cache for the K6-2

There is no L2 on a K6-2, but that does not mean there is no
L2 on the board itself. AIDA will not report an L2 on a
board is an L2 on a CPU, apparently, and rightly so because
it ISN'T on the CPU. The jumper does NOT appear to disable
the cache.

Read slowly and carefully, this is the last time I"m going
to tell you:

The board always uses L2 cache unless you see a VERY
specific setting that "DISABLES L2 CACHE". If there is any
other description of Cyrix or Write-Back, Write-through,
etc, etc, it is NOT disabling the L2 cache.

With cache benchmarks it is quite easy to see not only
whether cache is enabled but how much. There is a plateau
that drops off on a graph beyond the cache size. Sisoft
Sandra.

It didn't take that long it just was not stable at many settings.and
when it crashes it takes a long time to reboot/scandisk on the wrong
speed/slow speed.

It did take that long, including the newsgroup posts. I
really meant "total" time, including ordering, receiving,
heatsink compound, reading the manual to get jumper
settings, benchmarking, etc, etc.

As for benchmarks it is whether it will do what I want it to do
in the real world which matters most.

Yes but the benchmark is the indicator of proper function,
including the L2 cache. As for whether it will "do what you
want", that depends on the performance potential of the
part, and was why the original suggestion included mention
that it may not be worthwhile, contrasted with a new(er)
system.

I doubt the 450 CPU would be stable in my system anyway, it has
enough trouble with the 300.

It usually is stable. I've upgraded dozens if not over 100
old pre-"super" socket 7 boards to run K6-2 CPUs. You had
it easy, sometimes it requires guessing and checking voltage
levels with a multimeter to come up with the correct jumpers
for ~ 2.2-2.4V because many boards came close (enough) to
that target but didn't list any settings not (then used at
the time of creating the manual) applicable to any earlier
socket 7 CPUs.

 

[Donald McTrevor]

Then you have jumped to a conclusion without any reason to
do so.

Apart from the AIDA reporting no L2 cache, if there was L2
cache why not says so?
Actually I have just looked again and under 'chipset' (not
a great title it does say 512 cache for the cryix, I will have to
put the K6 back in and see what it says for that. I now
expect it will say I have 512 L2 cache too, but not pipeline
burst.

The cache will be enabled, no matter what CPU,
unless you take specific steps to "disable" it. Since you
haven't disabled it, it is still being used.

As I'd mentioned already, you need to test the cache by
benchmarking. Do not guess about the potential for a
problem, benchmark and focus on any problems factually
shown.



There is no L2 on a K6-2, but that does not mean there is no
L2 on the board itself. AIDA will not report an L2 on a
board is an L2 on a CPU, apparently, and rightly so because
it ISN'T on the CPU. The jumper does NOT appear to disable
the cache.

OK the L2 cache is listed under 'Chipset', not a great place to put
it.

Read slowly and carefully, this is the last time I"m going
to tell you:

The board always uses L2 cache unless you see a VERY
specific setting that "DISABLES L2 CACHE". If there is any
other description of Cyrix or Write-Back, Write-through,
etc, etc, it is NOT disabling the L2 cache.

With cache benchmarks it is quite easy to see not only
whether cache is enabled but how much. There is a plateau
that drops off on a graph beyond the cache size. Sisoft
Sandra.

I have that Software now, had to do another download to get it
to run. (MDAC whatever that is)
I am not sure I can save reports, I cant find the ones I made,
that option might be disabled, maybe I can cut and paste them (nope).

It did take that long, including the newsgroup posts. I
really meant "total" time, including ordering, receiving,
heatsink compound, reading the manual to get jumper
settings, benchmarking, etc, etc.



Yes but the benchmark is the indicator of proper function,
including the L2 cache. As for whether it will "do what you
want", that depends on the performance potential of the
part, and was why the original suggestion included mention
that it may not be worthwhile, contrasted with a new(er)
system.



It usually is stable. I've upgraded dozens if not over 100
old pre-"super" socket 7 boards to run K6-2 CPUs. You had
it easy, sometimes it requires guessing and checking voltage
levels with a multimeter to come up with the correct jumpers
for ~ 2.2-2.4V because many boards came close (enough) to
that target but didn't list any settings not (then used at
the time of creating the manual) applicable to any earlier
socket 7 CPUs.

Main problem was I was trying to run at 75mhz I think, didn't
seem that much higher than 66 (to me). Then all the crashes and
scandisks at about 100 Mhz, very slow.
I will try it again at 4X66 which seemed stable.

 

[Donald McTrevor]

Apart from the AIDA reporting no L2 cache, if there was L2
cache why not says so?
Actually I have just looked again and under 'chipset' (not
a great title it does say 512 cache for the cryix, I will have to
put the K6 back in and see what it says for that. I now
expect it will say I have 512 L2 cache too, but not pipeline
burst.


OK the L2 cache is listed under 'Chipset', not a great place to put
it.



I have that Software now, had to do another download to get it
to run. (MDAC whatever that is)
I am not sure I can save reports, I cant find the ones I made,
that option might be disabled, maybe I can cut and paste them (nope).

Main problem was I was trying to run at 75mhz I think, didn't
seem that much higher than 66 (to me). Then all the crashes and
scandisks at about 100 Mhz, very slow.
I will try it again at 4X66 which seemed stable.

Well where should I start, I ran a load of benchmaks on the cyrix
from Sandra.
Switched to K6, fail to boot (reboooted itself), check jumpers, all OK,
checked CPU, cleaned new thermal blob. Retried gain still at 4X66,
booted OK **BUT** it took 4:12 minutes to boot almost twice as long
as cyrix at 2:10.
So no point in benchmarking that it's obviously useless!!
AIDA reports 512 external cache enabled, sync pipeline burst.

One other thing from
http://www.uktsupport.co.uk/pb/mb/850.htm
PCI/ CPU Bus Synch.
JP18 1-2 Asynch. (CPU Bus Speed at 75 MHz)
2-3 Synch. (CPU Bus Speed 66.6 MHz or less)

So I am going to try the 2-3 setting as I am at 66.6Mhz

Anyhowits not looking good, something is seriously wrong the
boot was just too slow.


FRom sandra multimedia bench

int float

cyr 646 124
k6 1414 1729

So K6 faster here

 

[Donald McTrevor]

Well where should I start, I ran a load of benchmaks on the cyrix

from Sandra.
Switched to K6, fail to boot (reboooted itself), check jumpers, all OK,
checked CPU, cleaned new thermal blob. Retried gain still at 4X66,
booted OK **BUT** it took 4:12 minutes to boot almost twice as long
as cyrix at 2:10.
So no point in benchmarking that it's obviously useless!!
AIDA reports 512 external cache enabled, sync pipeline burst.

One other thing from
http://www.uktsupport.co.uk/pb/mb/850.htm
PCI/ CPU Bus Synch.
JP18 1-2 Asynch. (CPU Bus Speed at 75 MHz)
2-3 Synch. (CPU Bus Speed 66.6 MHz or less)

So I am going to try the 2-3 setting as I am at 66.6Mhz

Anyhowits not looking good, something is seriously wrong the
boot was just too slow.


FRom sandra multimedia bench

int float

cyr 646 124
k6 1414 1729

So K6 faster here

OK... I tried the new J18 setting (sync/async) and rebooted, it rebooted in
2:23 which is OK but a little slower than the 2:11 for the cyrix, but 'not
bad',
and fairlt proportional to the bus speed ratio 66/75 =0.88 (131s/143s)
=0.91.
However I think it has booted at a similar speed before without the J18
change
so I am not sure of the reason why.

Anyway I did the Sandra cache benchemark.
cryix k6 k6 with new J18 (or the quick boot)
Combo 128 87 94
speed 22.8 3.6 4.6
2kb 751 176 237 MB/s
4 776 176 206
8 772 172 186
16 772 166 178
32 769 157 161
64 464 135 132
128 151 103 103
256 72 87 103
512kb 54 74 82
1 meg 41 61 64
4 35 50 52
16 35 50 52
64 meg 35 49 54 MB/s


So....it seems the cyrix is much faster on the small data sizes but
slower on the larger stuff.

Something to do with the caches I expect??

I also did a test playing a smallish (4meg) .wmv file and the results
were basically identical.

So overall similar perforamce I think.

However I am not sure if I would trust this K6 to play poker with,
it might cost me more than I paid fot the chip.

 

[~misfit~]

Really? Well I struggled to find them with google, I even posted here
earlier, requesting benchmarks (without any replies).
If any one can post benchmarks to contradict the site please do!!

http://www.tomshardware.com/cpu/20030217/index.html

"Benchmark Marathon: 65 CPUs from 100MHz to 3066MHz".

While they don't actually benchmark your particular CPU they do include the
K6-2 500 (CPU#1). Also they don't include the PIII 933 so I'll use the PII
800 (CPU#2) for comparative purposes:

On the Multimedia PC Mark 2002 benchmark CPU#1 got 725, CPU#2 got 2194

On the PC Mark 2002 memory access benchmark CPU#1 got 683 and CPU#2 got 2405

(Keep in mind that CPU#1 is in fact a lot faster than yours and CPU#2 is
slower than the one you're crowing about yours being better than).

Etcetera. Read it yourself.

Tom's Hardware is a pretty reputable site

If they were incorrect I imagine it would have been spotted now,
after several years!!

Probably been laughed at for years. I see you've used it to make yourself
feel happy about your purchase. Therefore the site has it's uses.

If you still doubt them why not drop them an email?
mailto:[email protected]

And spoil it for people who want to feel good about buying junk? Why would I
do that? ;-)

See? You're happy! That's the main thing. It's a shame to spoil that with
hard evidence. Sorry.

Well yes but you cannot compare CPU's running in boxes with
vastly different configurations.
Well I might get my K6 tomorrow, I have already forked out
£2.50 on some thermal compound, (more than it cost for the CPU!!).

On slighty unhappier note it seems that a an AMD K6-2 will
interpretate a X2 clock as a X6 clock, so I might be able to run a
450MHz AMD.
So maybe I should have bought a faster processor!!
Will I have to spend another £1 to get one?!!

Maybe the graphics card was a bottle neck but it still showed the
PIII running slower.

I don't think the benchmarks are wrong because if they are they
must have got a whole series of them wrong, here the PIII 700
is also slower.
I think the truth is the K6-2 is a great processor which punches
above it weight in some applications.

Then, sorry to say, you are deluded. The _K6-3_ was a good CPU but even
that, at 450MHz, was "slower" than Intel's lowly Celeron 400 Mendicino.

I never thought I'd see the day! Me, a dyed-in-the-wool AMD man posting
evidence that certain Intel CPUs were better than certain AMD ones.

 

[~misfit~]

OK, what's wrong with that?


You don't seem to be grasping something- you do NOT need to
"enable the cache for the K6". The cache was already
enabled the whole time, there is no "enabling" to do.



Help with what?
There are ONLY five things you need(ed) to do to go from
(having done nothing, right after you made your very first
post) to being finished:

1) Buy a 450MHz+ K6-2
2) Install it and the heatsink/fan.
3) Set jumpers for voltage, multiplier, and FSB. Any other
jumpers should remain the same, the jumpers should look
identical to how they would if you had a Pentium 133 MMX
installed, except voltage would be 2.2V or 2.4V (as stamped
on the K6-2).
4) Run CPU-Z to confirm speed and Sisoft Sandra to check
performance.

The entire process takes a little under 30 minutes, I can't
understand why you're making something easy, so difficult.

LOL. I remember the days of messing about with systems like this. If I
hadn't done that I doubt I would be at the stage where I can build a new PC
in a few hours with the latest components etc.

Don't forget Dave, it wasn't always easy for us (or at least for me) when we
were learning. I'll always remember the sense of achievement I got when I
got my Pentium 200 MMX running stably at 250Mhz on an 83MHz FSB. That was
one fast machine. <g>.

If I was Donald, playing with old PCs, I'd get a socket 370 board that
supports Coppermines (Maybe a Slot 1 with slocket) and have a play with some
of the Celerons in the ~600Mhz range. They should be cheap enough, several
times faster than what he's currently playing with, and eminently
overclockable. I've had the only two 600s I've owned running stably at 900.
Quite a usable machine for very little money.

Even cheaper, get a Mendicino 400. Not a bad little work-horse. Hell, if he
wasn't on the other side of the world from me I'd give him one, and probably
a mobo too. I have at least 3 Celly 400's in my CPU drawer. One or two 500's
as well, *and* my two Coppermine 600's. They're not worth selling and I
can't bring myself to throw them out. :-(

It seems we're a dying breed mate. Just about everyone I know is buying Dell
etc. these days. So cheap! I used to build quite a few machines for friends,
save them some money and they'd get a good, solid machine. Now, there are
Dells around at less than I can build a machine for. Granted they're crap
but everything is made to be thrown away in a couple years these days
anyway. I've heard people say "So what if it craps out afer the warranty is
up? I'll have got my money's worth, I'll just get another, faster one".

Very sad. When I started playing with hardware the average PC cost 4 times
what it does now and the average wage was probably half. It was worth
knowing how to upgrade and build machines then, I had good systems and could
help friends. Practical, you could save money. Now it's just a hobby that
actually costs me money truth-be-told.

 

[kony]

Apart from the AIDA reporting no L2 cache, if there was L2
cache why not says so?

It is reporting CPU specs, apparently.

There is no point in wondering, benchmark the cache.

OK the L2 cache is listed under 'Chipset', not a great place to put
it.

It is a great place to put it. The cache is part of the
chipset and motherboard, not the CPU.

I have that Software now, had to do another download to get it
to run. (MDAC whatever that is)
I am not sure I can save reports, I cant find the ones I made,
that option might be disabled, maybe I can cut and paste them (nope).

Why would you need to cut & paste it?
Just look at the graph and note where it drops. There
should be a sharp decline in memory throughput right after
512K.

Main problem was I was trying to run at 75mhz I think, didn't
seem that much higher than 66 (to me). Then all the crashes and
scandisks at about 100 Mhz, very slow.
I will try it again at 4X66 which seemed stable.

If you run it at 75MHz, you'd need two things:

1) Set the PCI divider jumper as shown on the previously
linked charts.

2) A CPU spec'd for that FSB or higher- the CPU you bought
is a rare K6-2 in that it's only spec'd to 66MHz FSB. Most
are rated for 100MHz.

 

[kony]

On Thu, 11 Aug 2005 23:41:50 GMT, "Donald McTrevor"

<snip>

booted OK **BUT** it took 4:12 minutes to boot almost twice as long
as cyrix at 2:10.
So no point in benchmarking that it's obviously useless!!

Hmmmmm.

You seem to have forgotten to mention that it wasn't
application performance that you were trying to improve but
rather boot time.

Frankly, at this point I think the experiment is over, you
have all the data you need and I have nothing more to add,
except that now all you have left to do is see if it will
help with those 2 card table game screens you wanted.