A7V133 Rev 1.05 (no dot) and Athlon XP 2100+ _ Can I? How?

[Spiggy Topes]

Hi all,

I've roamed the web endlessly in search of a definitive answer, and I
begin to suspect there's no such animal.

Reading indicates that the 1.05 (no dot) should support Athlon XP's up
to 2100+ in jumpered mode, even though the ASUS website indicates
otherwise. Looks like the German website USED to say it was good, but
now no longer does so.

So, I come to the folks that know, with questions:

1 - What's the likelihood that it will work? Is there anything else
that should be looked at to determine this? Is the ICS 94215AF clock
generator really a factor? I looked, and the chip that should be so
identified has no markings whatsoever. Is that a problem?

2 - What's the best bet? Most of what I read says go Palomino, but
there are obviously some out there using Thoroughbreds and even
Bartons with mixed degrees of success.

3 - Assuming a Palomno XP xx00+ is the way to go, what jumper settings
need changing? What BIOS settings?

4 - Is there a significant risk of making a dog's breakfast of my
current XP Pro SP1 install? Assuming, that is, that I'm a relative
innocent who knows which end of a soldering iron NOT to hold, but
would never put that end willingly inside his case?

Thanks muchly for any constructive advice.

 

[Max]

I've roamed the web endlessly in search of a definitive answer, and I
begin to suspect there's no such animal.

There is. The couple works with no problem, exactly as if it was 1.05 dot.

Reading indicates that the 1.05 (no dot) should support Athlon XP's up
to 2100+ in jumpered mode, even though the ASUS website indicates
otherwise. Looks like the German website USED to say it was good, but
now no longer does so.

So, I come to the folks that know, with questions:

1 - What's the likelihood that it will work? Is there anything else
that should be looked at to determine this? Is the ICS 94215AF clock
generator really a factor? I looked, and the chip that should be so
identified has no markings whatsoever. Is that a problem?

Not that I know. just put it in and rock'n'roll.

2 - What's the best bet? Most of what I read says go Palomino, but
there are obviously some out there using Thoroughbreds and even
Bartons with mixed degrees of success.

Xp T is ok. take a look here.

http://www.a7vtroubleshooting.com/info/cpu/index.htm

3 - Assuming a Palomno XP xx00+ is the way to go, what jumper settings
need changing? What BIOS settings?

I run it in jumperless mode, perfectly. if you want to reach higher speeds,
you need to do the "wire trick" :

http://www.ocforums.com/showthread.php?s=c8f573747b2926eb4a5c6fa494249b02&threadid=279392

assuming the cpu will be stable, that's ok with voltage doesn't warm, etc.

In my humble opinion, standard is perfect, put 1 gb of ram and kick at
2,1Mhz.

4 - Is there a significant risk of making a dog's breakfast of my
current XP Pro SP1 install?

I don't think so, anyway you could perform a safe partition backup,
and if it doesn't work, just go back.

Assuming, that is, that I'm a relative innocent who knows which end
of a soldering iron NOT to hold, but would never put that end
willingly inside his case?

You don't really need. You could kick some 2500+m at higher speed,
but there is no convenience, always in my humble opinion, to push
a 133 bus board to run with a 3000+ cpu instead of trowing it away
and go over an a7v600 with a faster bus. what do you think.

anyway the socket trick just allows you to use lower mapped multipliers
to run at their higher settings, eg, 5x would value 14x and so on.

with a 2100+ you can go jumperless.

Thanks muchly for any constructive advice.

de rien.

 

[Spiggy Topes]

Thanks enormously. There's so much conflicting info out on usenet, I
was worried I was going to follow the wrong path and screw something
up horribly. I'm happy with standard speeds, no real interest in
overclocking for the sake of a few extra cycles, and overclocking for
the sake of overclocking, in my opinion anyway, isn't worth the hertz
OR the hurts.

So, the one remaining question is, what's the limit? I was looking at
the XP2100+ as the fastest availble, but if the Thoroughbreds will
work, how far can I go?

Vielen danke!

 

[Paul]

Thanks enormously. There's so much conflicting info out on usenet, I
was worried I was going to follow the wrong path and screw something
up horribly. I'm happy with standard speeds, no real interest in
overclocking for the sake of a few extra cycles, and overclocking for
the sake of overclocking, in my opinion anyway, isn't worth the hertz
OR the hurts.

So, the one remaining question is, what's the limit? I was looking at
the XP2100+ as the fastest availble, but if the Thoroughbreds will
work, how far can I go?

Vielen danke!

I've read a comment from one poster, that says there is no difference
between all the processors, but the difference is in how the power
is filtered for a PLL in the processor. When you put a more modern
Athlon on a board, it has different filtering requirements than
previous ones, and it is the "luck of the draw", if a new processor
will work in an old board. So, it could work for you, or it might
not. That is why there is so much conflicting info - some people
have good experiences, and others don't.

I'd like to confirm this, but I cannot find document #24363 from
the AMD site. It is supposed to have some of the needed info.

Paul

 

[Spiggy Topes]

I've read a comment from one poster, that says there is no difference
between all the processors, but the difference is in how the power
is filtered for a PLL in the processor. When you put a more modern
Athlon on a board, it has different filtering requirements than
previous ones, and it is the "luck of the draw", if a new processor
will work in an old board. So, it could work for you, or it might
not. That is why there is so much conflicting info - some people
have good experiences, and others don't.

I'd like to confirm this, but I cannot find document #24363 from
the AMD site. It is supposed to have some of the needed info.

Paul

And greater sophistication typically demands greater precision, right?
So there's a lesser chance of success with the Thoroughbreds than with
the Palominos? And all Palominos are created equal? So my best bet
would probably still be the Palomino-based 2100+. Sounds fair. It's a
step up from the current 1200, and not much expenditure on eBay if it
comes to naught.

You mention a "PLL".Phase-locked loop? There's a couple of references
in the archives to an ICS 94215AF clock generator being an important
factor. Found it on the mobo, but where others' illustrations are
quite clearly marked, mine is blank. Unrelieved black. I've put up a
shot of the general vicinity at
http://members.shaw.ca/ubik/a7v133_mobo_pll.jpg, if you want to verify
I'm looking at the correct spot. Is that good, bad, or
inconclusive/indifferent/irrelevant? An ancient copy of SiSoft Sandra
tells me:

Memory Module(s)
Memory Module 1 : SpecTek 23AF8701 512MB 16x(32Mx8) SDRAM
PC133U-333-542 (CL3 upto 133MHz) (CL2 upto 83MHz)

System Clock Generator
Model : ICS 94215
Software Programmable : Yes
Read-Back Support : Yes
Speed Selected By : Software
Spread Spectrum Modulated : Yes
No FSB Speeds : 32
FSB Speed Range : 90 - 166MHz

Voltage Sensor(s)
CPU Core Voltage : 1.79V
Aux Voltage : 0.08V
+3.3V Voltage : 3.54V
+5V Voltage : 4.78V
+12V Voltage : 12.59V
-12V Voltage : -12.25V
-5V Voltage : -5.54V

Does any of that sound problematic at all?



<snipped for brevity>

 

[Tom Joyce]

Hi all,

I've roamed the web endlessly in search of a definitive answer, and I
begin to suspect there's no such animal.

Reading indicates that the 1.05 (no dot) should support Athlon XP's up
to 2100+ in jumpered mode, even though the ASUS website indicates
otherwise. Looks like the German website USED to say it was good, but
now no longer does so.

I used a 2100+ Tbred 'B' in my 1.05

In jumper-free mode it ran at 20*100 with default vcore. With jumpers set
to a 9* multiplier (probably the only viable option with this CPU and the
KT133A chipset) it ran rock solid at 17*133.3 (2266MHz) but needed 1.9
vcore.

There are no guarantees.


Regards,
Tom

 

[Paul]

And greater sophistication typically demands greater precision, right?
So there's a lesser chance of success with the Thoroughbreds than with
the Palominos? And all Palominos are created equal? So my best bet
would probably still be the Palomino-based 2100+. Sounds fair. It's a
step up from the current 1200, and not much expenditure on eBay if it
comes to naught.

You mention a "PLL".Phase-locked loop? There's a couple of references
in the archives to an ICS 94215AF clock generator being an important
factor. Found it on the mobo, but where others' illustrations are
quite clearly marked, mine is blank. Unrelieved black. I've put up a
shot of the general vicinity at
http://members.shaw.ca/ubik/a7v133_mobo_pll.jpg, if you want to verify
I'm looking at the correct spot. Is that good, bad, or
inconclusive/indifferent/irrelevant? An ancient copy of SiSoft Sandra
tells me:

Memory Module(s)
Memory Module 1 : SpecTek 23AF8701 512MB 16x(32Mx8) SDRAM
PC133U-333-542 (CL3 upto 133MHz) (CL2 upto 83MHz)

System Clock Generator
Model : ICS 94215
Software Programmable : Yes
Read-Back Support : Yes
Speed Selected By : Software
Spread Spectrum Modulated : Yes
No FSB Speeds : 32
FSB Speed Range : 90 - 166MHz

Voltage Sensor(s)
CPU Core Voltage : 1.79V
Aux Voltage : 0.08V
+3.3V Voltage : 3.54V
+5V Voltage : 4.78V
+12V Voltage : 12.59V
-12V Voltage : -12.25V
-5V Voltage : -5.54V

Does any of that sound problematic at all?

Well, you are looking in the wrong area. This is an issue with
components used around the processor socket. Since the motherboard
design guide I need cannot be found, my comments will be generic
in nature.

There are two kinds of supply pins on the processor. Most of them
will be labelled VCC and power digital circuits in the processor.
Now, the PLL clock synthesis, takes the low frequency clock signal
generated by the 94215, and multiplies it up by the multiplier
value. The oscillator circuit at the output of the phase
locked loop inside the processor, is effectively an analog
circuit. AMD has a pin on all the processors labelled VCCA, which
acknowledges the different nature of the analog VCC supply, and
that it is a "clean" supply intended just for the PLL (and maybe
a phase detector too).

Normal digital supplies will have perhaps 100mV of switching
noise riding on the rails. If the digital supply is connected
directly to the processor, without a filter, it looks like this.

-------------
100mV |
Vcore -----------|VCCA --->PLL
noise |

A simple RC filter can be added, and the one depicted here,
is a low pass filter, that removes high frequency noise.

-------------
100mV 5mV |
Vcore -----------Resistor-+---------------|VCCA --->PLL
noise | noise |
Capacitor |
|
GND

The second circuit feeds a cleaner version of voltage
to just the PLL circuit. The PLL senses AC signals and if
there is enough noise on its power supply, it will lock
onto noise from the outside world, rather than the clock from
the 94215. (Note: The method depicted above, is only suitable
for circuits which draw tiny DC currents. A PI filter is used
for circuits with heavier requirements.)

So, maybe the difference between a 1.05 and a 1.05. motherboard,
is how that filter is designed. It could be they used different
values for R and C, or they used a single stage or dual stage
PI filter (using L and C). The #24363 document from AMD would
have some of the necessary details, to identify exactly what
is needed.

I can tell you, that some chips have conservative specs for their
filter requirements. A colleague and I were using the same chip
once, and he forgot to add the Capacitor in the above type
circuit. When we got in the lab, his circuit still worked, which
means that his chip could "eat" 100mV of noise without a problem.

And, that is what I mean by the "luck of the draw". If the filter
is missing or wasn't designed to meet the needs of a particular
processor family, it might still work, but there is no way to
guarantee it. I've run into circuits that _none_ of them could
possibly work in a situation like this, and other conservatively
specified circuits where _all_ of them work in a situation like
this, and in this case, based on user comments in Google, this
one is a "roll of the dice".

I don't know of an easy way to trace what is connected to VCCA,
so I don't think reverse engineering this is too practical. And
there might not be any filter on there at all, so nothing to see,
and nothing possible to mod.

HTH,
Paul

 

[Spiggy Topes]

I think I just fried something... We have - had - another box, with an
XP 1800+ chip in it, which we thought we'd use to test out the A7V133
1.05-no-dot. Took out the Athlon 1200, dropped in the XP 1800+ -
should probably have set the BIOS to vanilla first, but didn't. Turned
power back on, hit the start button. There's life, briefly. Display
starts, usual way, and identifies the chip as XP 1800+. Says we're now
cruising at 100MHz FSB / memory, rather than the expected 133, then
.... nothing. Blank screen. Power down, repeat. This time, nothing at
all. No POST beep, nada, rien, nichts.

So, back goes the 1200. Power on, drops into BIOS. Five seconds in,
freezes. Power off, wait ten seconds, repeat. This time, not as far as
the BIOS. Power off, wait longer, stays up a little longer. Freezes.
Repeat. Leave off for 30 mins this time. Alles in ordnung, and it's
been in ordnung ever since. Phew, had me worried there.

Meanwhile, at the other box, the one that had previosuly had the
1800+... Return the processor to its rightful place, power on.
Nothing. No POST beep, rien, nichts, nada. Check to see if we unseated
something, reset BIOS by the approved jumper method, still nothing.
Remove 1800+, drop in ancient Athlon 800 (Duron? - whatever was in the
box before). Nothing. Remove memory, drives, video, network, firewire
card, Nothing.

Take dead box, still with 800 in place, to friend with diagnostic
doohickey wot sits in an ISA slot and makes pretty lights. He says,
your mobo's a dodo. Oh snit. Well, while I'm here, let's give the
1800+ another go in the A7V133 1.05-no-dot. This time, no POST beep,
nothing. Put the 1200 back, life resumes its tranquil course.

So, is there anything I can have done in the process of testing out
the 1800+ that would have fried the processor? Is it possible that
putting that processor back in the other motherboard could have passed
the contagion on and fried the mobo too? Is it possible that the 1800+
is still an option for the A7V133 1.05-no-dot, jumpered or jumperless?
Is it possible that the mobo's not dead and that it's the 800 and
1800+ processors that are stopping box #2 from working?

Help... <:-(((

 

[John Smith]

Spiggy , it is time

Now I know you love you a7v133 , but it is time you moved on
you must let go of your security blanket

I ran mine with an athlon 2000 no problem, but with 3400 out there and P4
3.4
it is time to embrace the new memory and technology

 

[afzan]

I've got two A7V133 motherboards both 1.05 (no dot) and have tried a
palomino XP1700+ in both. Both work ok but may crash with BSOD maybe
once a week, and sometimes as often as once a day. After rebooting
its fine again. I bought a new MSI motherboard for the XP1700+
and it does not crash. So I beleive the websites are correct when
they say only use an Athlon Thunderbird or Duron.
I originally had a Thunderbird 1.4 Ghz and it never crashed.
By the way I will not buy another Asus because I believe Asus could
have distributed an easy fix.
Just think of the stupidity of looking for a dot. It doesn't even rate
a full letter maybe because the difference between the two is minor.

 

[Paul]

I think I just fried something... We have - had - another box, with an
XP 1800+ chip in it, which we thought we'd use to test out the A7V133
1.05-no-dot. Took out the Athlon 1200, dropped in the XP 1800+ -
should probably have set the BIOS to vanilla first, but didn't. Turned
power back on, hit the start button. There's life, briefly. Display
starts, usual way, and identifies the chip as XP 1800+. Says we're now
cruising at 100MHz FSB / memory, rather than the expected 133, then
... nothing. Blank screen. Power down, repeat. This time, nothing at
all. No POST beep, nada, rien, nichts.

So, back goes the 1200. Power on, drops into BIOS. Five seconds in,
freezes. Power off, wait ten seconds, repeat. This time, not as far as
the BIOS. Power off, wait longer, stays up a little longer. Freezes.
Repeat. Leave off for 30 mins this time. Alles in ordnung, and it's
been in ordnung ever since. Phew, had me worried there.

Meanwhile, at the other box, the one that had previosuly had the
1800+... Return the processor to its rightful place, power on.
Nothing. No POST beep, rien, nichts, nada. Check to see if we unseated
something, reset BIOS by the approved jumper method, still nothing.
Remove 1800+, drop in ancient Athlon 800 (Duron? - whatever was in the
box before). Nothing. Remove memory, drives, video, network, firewire
card, Nothing.

Take dead box, still with 800 in place, to friend with diagnostic
doohickey wot sits in an ISA slot and makes pretty lights. He says,
your mobo's a dodo. Oh snit. Well, while I'm here, let's give the
1800+ another go in the A7V133 1.05-no-dot. This time, no POST beep,
nothing. Put the 1200 back, life resumes its tranquil course.

So, is there anything I can have done in the process of testing out
the 1800+ that would have fried the processor? Is it possible that
putting that processor back in the other motherboard could have passed
the contagion on and fried the mobo too? Is it possible that the 1800+
is still an option for the A7V133 1.05-no-dot, jumpered or jumperless?
Is it possible that the mobo's not dead and that it's the 800 and
1800+ processors that are stopping box #2 from working?

Help... <:-(((

http://www.a7vtroubleshooting.com/info/cpu/index.htm

In a table near the bottom of the page (translated from FAQ085 of
the Asus Germany site,) it says for A7V133 when using AthlonXP, note "M"

"Set the multiplier ( FID ) for this CPU manually via DIP-Switch
( JUMPERMODE ), else the system might not boot. Also see "FID Info"
below. Jumperless Mode shouldn't / can't be used in this case. This
is true even if the CPU is not unlocked (has a hardwired multiplier)"

I don't know if that implies the BIOS will get the wrong FID value,
if running in jumperless mode ? Maybe your processor fried because
it was running too fast ? (By running fast, the processor draws
a lot of Vcore current, overheating the Vcore circuit - maybe that is
why the motherboard had to sit for 30 minutes - to cool off.)

I would visually inspect the 1800+ and see if it is burned or discolored.
It could have had a meltdown.

Otherwise, I don't remember hearing of "destruction" awaiting those
who experiment with A7V133 non-1.05. boards. I guess the next A7V133
owner will have to be warned about your experience :-(

And, yes, a dead processor can "spread its poison" to another mother
board. That happened to a poster here not too long ago, who did a
friend a favor, by testing his friend's processor. In that case, I
think the CPU being tested had been involved in a lightning incident.
The thing is, if a processor fails "short', it will drag down the
Vcore circuit. A properly designed circuit is supposed to detect
overcurrent and shut down, but that only happens for a very good
short. A partial short, could just cause the MOSFETs to exceed their
SOA (safe operating area). Depending on the overcurrent detection
method in the Vcore regulator, it is hard for the motherboard designer
to set the current limit, and to avoid a lot of RMAed boards, the
current limit must necessarily be set very high, so healthy processors
don't trigger the protection.

So, have a look at the physical condition of the processor and
see if it looks like it has been heat damaged. With that note
on A7Vtroubleshooting, maybe a multiplier value was used which
caused the core frequency to go really high.

BTW - I'm sorry I didn't see this post of yours back in June.
I only ran into your post again in Google, while searching for
the 24363 design document from AMD. I still want to read the damn
thing, to see if all the Athlon processors are compatible with the
exception of the clock synthesis PLL VCCA noise requirement. My
hypothesis would be, that 1.05 and 1.05. differ in the values
of the filter components used, but the hypothesis is worthless
without a copy of 24363 to verify the details.

Paul