Google Groups no longer supports new Usenet posts or subscriptions. Historical content remains viewable.
Dismiss

XP Timings Uncovered (cold boot problem)

1 view
Skip to first unread message

Kaptian Krunch

unread,
Mar 12, 2002, 2:58:34 AM3/12/02
to
Hey Kaptian Krunch Komes Through! Oh Joy, Oh Joy, Oh Joy!
If you have a cold boot problem with the XP processor This is a must read
post.

taken from the AMD white papers #24309
and are the requirements for a cold boot (power on)

<SNIP>
7. The FID[3:0] signals are valid within 100 ns after PWROK is
asserted. The chipset must not sample the FID[3:0] signals
until they become valid. Refer to the AMD AthlonT
Processor-Based Motherboard Design Guide, order# 24363, for
the specific implementation and additional circuitry
required.
8. The FID[3:0] signals become valid within 100 ns after
RESET# is asserted. Refer to the AMD AthlonT Processor-
Based Motherboard Design Guide, order# 24363, for the
specific implementation and additional circuitry required.
<SNIP>

I though I was going to go crazy trying to find proof of what I have been
saying all along in the news groups. So if I read this right, on Pwr up
(cold) timings have to be < 100ns in order to cold boot the system. This
document also talks about the power must stablize within 10ns or no boot
either. So that is why AMD has Approved Power Suppl;ies.

Any comments?

---Hey Hey Hey, It's Kaptian 'K"---


Bill

unread,
Mar 12, 2002, 1:22:42 PM3/12/02
to
In article <3c8d...@news.mhogaming.com>,
kaptiankrunch=remov...@comcast.net says...

Oh, jeez, I thought everybody knew that. ;)

Excellent detective work there. Ya done good.

Bill
--
"Give a man a fish, and you feed him for a day...
Force everybody else to give him fish, and he'll vote
for you forever."
A Taxpayer

Matt

unread,
Mar 12, 2002, 3:48:02 PM3/12/02
to

> Any comments?

ya actually...so what does all this mean for us no so techie types? I
understand a bit that the cpu has to kinda do it's thing first before it can
boot right? So it could be the power supply but it could still just be a
mobo cpu timing issue no? and if it's not the supply how do I fix it? I
e-mailed soltek to see if they had an answer and I did some more research on
my power supply...It has a combined output of 175w. Enhance (the type of
supply I have) "does" have athlon xp approved supplies but not the exact
model I have (mine is actually athlon approve up too xp)...figures eh! I
guess what I'd like to know is should I order the new psu or should I just
wait and see what soltek has to say or is there another fix, such as, bios
settings that can be done? appreciate all the research you've done
Kaptian...what a pain in the rumpess....hehehee...one more thing...This
enhance supply is loud...any recommendations on a quiet one? Thx again

matt


Nitza666

unread,
Mar 12, 2002, 6:05:23 PM3/12/02
to
These are the mobotherboard design guide and how the motherboard should boot
or reset it doesnt have anything to do with the PSU.AMD has the approved PSU
list because AMD CPUs use a lot of power.
"Kaptian Krunch" <kaptiankrunch=remov...@comcast.net> wrote in message
news:3c8d...@news.mhogaming.com...

Kaptian Krunch

unread,
Mar 12, 2002, 9:38:14 PM3/12/02
to
well from what I can acertain (I aint no einstine) from the document the
power supply has to ascetain a power good signal and has to do it within a
certian amount of time (10ns I think) anyways, on a simplistic, I would wait
for the soltec people to get back with you as to their side of it..maybe
they have known issues and fixes for this board, but, on the other hand, if
you are only getting 175w off the 5v and the 3.3v line I would say you are
25w short of a happy meal. Both the power and timings have to be just right
inorder to boot, I would lean to your PS side of the problem because if the
power planes of the board are not engerized within the specified amount of
time and a power good signal (cheepo ps just emit a PS OK signal reguardless
of stable voltages) cold boot wouldnt work because of this, but by the time
you hit the Reset button they had stablized and the processor starts up. as
for a recommendation, get one that has the lowest tolorances Ä…1% for the 5
and 3.3 v rails and at least 50watts more then you think you need.for future
expansionso if you need a 350 watter, then go for at least 400watts and AMD
has a list of them on their site for your selection HTTP://WWW.AMD.COM

good luck

(Myself I have been looking into taking 2 - 250 watters and hooking them
together in parallel to provide a total output of 500w. Circuit board
design is almost finished and will make it available to all who want it
after I test it out on my own system. Circuit board will cost around $35.00
and have very stable current output through high quality voltage regulators
and filtering through a big bank of caps)

"Matt" <katm...@SpamNOhotmail.com> wrote in message
news:6Qtj8.5344$Oo3.200...@newssvr21.news.prodigy.com...

Kaptian Krunch

unread,
Mar 12, 2002, 9:51:06 PM3/12/02
to
read my previous post... I strongly disagree because of all the steps it
takes for a processor of this type to pull itslf up to a running condition
and a lot of factors that didnt exist before the XP now do exist and any
old 350watter isnt gonna do the job... look for your self at the paper I
quoted it is on AMD's web site. I only quotes 2 of the 8 or 9 steps that are
required to boot an XP

"Nitza666" <nitz...@FAhotmail.com> wrote in message
news:a6m1j9$k72$1...@nic.grnet.gr...

Nitza666

unread,
Mar 12, 2002, 11:00:02 PM3/12/02
to
Actually the TBirds need a good PSU not the Athlon XP and both the
ThunderBird and the Athlon XP need the same conditions as almost any modern
CPU from P4,P3 and VIA's CPUs.As for the cold boot i never had any cold boot
problems in all the PCs i have built and i have never used an approved AMD
PSU.

"Kaptian Krunch" <kaptiankrunch=remov...@comcast.net> wrote in message
news:3c8ebeaa$1...@news.mhogaming.com...

Nitza666

unread,
Mar 12, 2002, 11:12:10 PM3/12/02
to
The P4 needs exactly the same timings as the Athlon XP.

"Kaptian Krunch" <kaptiankrunch=remov...@comcast.net> wrote in message
news:3c8d...@news.mhogaming.com...

Matt

unread,
Mar 12, 2002, 11:15:28 PM3/12/02
to
thx again kaptain...I'll be sure to post when I get it fixed and I will get
it fixed because I'm a freak when it comes to things working properly...cya

Matt

"Kaptian Krunch" <kaptiankrunch=remov...@comcast.net> wrote in message

news:3c8ebba7$1...@news.mhogaming.com...

Kaptian Krunch

unread,
Mar 13, 2002, 2:07:57 PM3/13/02
to
Does it? I wouldnt know as I am AMD fan and dont use intel if I can help
it. Do you have any links or excerpts from Intel to back that statement up?

KK

"Nitza666" <nitz...@FAhotmail.com> wrote in message

news:a6mjie$qsn$1...@nic.grnet.gr...

Kaptian Krunch

unread,
Mar 13, 2002, 2:06:23 PM3/13/02
to
Yes T-birds DO need a good PS but are not as pickey as to when the Power OK
line is asserted, The XP's ARE. Have you read the article I quoted from
AMD?
They are certinally power hungry processors (t-birds) and the XP's dont have
that need for power, but they are finickey for when that power is applied to
the power planes of the mobo and how stable it is at the time of boot.
Guess that is why AMD only reccomends certian PS's for the XP's, or might it
be they recommend certian mobos and ps because they get kick backs from only
those companies?

KK


"Nitza666" <nitz...@FAhotmail.com> wrote in message

news:a6mirm$qjo$1...@nic.grnet.gr...

Kaptian Krunch

unread,
Mar 13, 2002, 2:16:48 PM3/13/02
to
I hope this discussion has shed some lite on this seemingly mysterious
workings of the XP processor. I know I have gleaned some from it through my
research. My interest in it is only to gain as much info I can so I can
build my XP system without the pitfalls of others. I started reading in the
NG IWILL, cause I liked the design of the board, but wanted to see what the
glitches were before I jumped in with my hard earned money. I am glad I
waited because IWILL came out with a new revision of the KK266 +R board
(chanded from the AMI controller to the HiPoint) and all users of that board
have no problems with XP so far ;-) This group is full of knowledge and
info also and I hope I have given as much as I have gotten. Ya I am a freak
too when things dont quite work the way I want them to also.

Thanx all
KK


"Matt" <katm...@SpamNOhotmail.com> wrote in message

news:AnAj8.5498$Ds5.211...@newssvr21.news.prodigy.com...

Nitza666

unread,
Mar 13, 2002, 5:44:27 PM3/13/02
to
Go to Intel site and download the .pdf for the P4 as i did its not very
difficult.

"Kaptian Krunch" <kaptiankrunch=remov...@comcast.net> wrote in message
news:3c8fa3a0$1...@news.mhogaming.com...

Kaptian Krunch

unread,
Mar 14, 2002, 3:14:11 AM3/14/02
to
Dont want intel garbage on my computer, but if you did d/l the pdf. then
just quote the section about the timings and I will as others would be happy
to read them.

thank you for your time to reply.
KK

"Nitza666" <nitz...@FAhotmail.com> wrote in message

news:a6okno$204$1...@nic.grnet.gr...

Nitza666

unread,
Mar 14, 2002, 8:10:21 AM3/14/02
to
You are the CPU detective...........

"Kaptian Krunch" <kaptiankrunch=remov...@comcast.net> wrote in message
news:3c905bf4$1...@news.mhogaming.com...

Kaptian Krunch

unread,
Mar 14, 2002, 12:47:04 PM3/14/02
to
then I guess either you back up your statment. subject closed.

"Nitza666" <nitz...@FAhotmail.com> wrote in message

news:a6q7f6$plj$1...@nic.grnet.gr...

Kaptian Krunch

unread,
Mar 14, 2002, 2:27:47 PM3/14/02
to
ment to say CANT back up your statement... subject close again
kk

"Kaptian Krunch" <kaptiankrunch=remov...@comcast.net> wrote in message
news:3c90...@news.mhogaming.com...

Nitza666

unread,
Mar 14, 2002, 9:02:38 PM3/14/02
to
I dont need to back mine statment,if you want to know go and get the .pdf
from Intel its not very difficult and having an Intel .pdf will not damage
your PC.The information is there at the Intel site available to everyone as
AMD does go and see.

"Kaptian Krunch" <kaptiankrunch=remov...@comcast.net> wrote in message
news:3c90...@news.mhogaming.com...

Skid

unread,
Mar 15, 2002, 11:38:05 AM3/15/02
to
Thanks, Ed. That is my conclusion as well. These specs are really AMD's
design advice for motherboard manufacturers, and they explain -- among other
things -- why the Athlon XP works fine in some older KT133A mobos and not
others. This whole issue was hotly discussed in regards to the Abit KT7A,
and this explanation is lifted from the FAQ at
http://www.viahardware.com/faq/kt7/kt7faq.htm
Why does only the KT7A(-RAID) version 1.3 support the AthlonXP processor?
There has been much heated discussion about the failure of early KT7A
motherboard revisions to support the AthlonXP and Duron (Morgan core)
processors. These discussions have also been confused by incorrect
statements on the ABIT (China) website, suggesting that the 64 BIOS release
provided support for these processors. This is not true, and the issue is
more complicated. The difference between the earlier Athlon processors and
the new AthlonXP processor lies in the processor boot process. You can see
the relevant boot timing diagrams on page 35, figure 10 of the Athlon Model
4 Data Sheet (for the older Athlons) and page 43, figure 12 of the Athlon
Model 6 Data Sheet (for the Athlon XP). During the boot process, the
motherboard must sample three data lines called FID(0:3) in order to
determine the CPU multiplier setting. In the new AthlonXP processor, the
timing requirements for this sampling has changed, and these lines only
become valid 100ns after the Vcore power OK (PWROK) signal is asserted.

This means that if the motherboard does not apply an extra circuit to delay
the NB_RESET# signal after the FIDs are fully valid, the motherboard may
potentially get the wrong values from the FID lines. This is the reason why
the earlier motherboard versions cannot support the AthlonXP processor.
Getting the wrong FIDs may not always result in the system being unable to
boot up - but it may result in the motherboard using the wrong parameters to
initialise the processor, causing system instabilities under certain
circumstances. On the earlier Athlon processors the FIDs are always valid
after the Vcore of the processor is valid and this is why no motherboard has
a problem supporting the highest frequency of old Athlon.

In practice, this change is very subtle. Many AthlonXP processors will work
fine on earlier motherboard revisions. Others will work fine for some
percentage of the time, but will require a hard reset during boot if this
timing assertion is violated. However, only revision 1.3 of the motherboard
contains the necessary hardware modifications to fully support AMD's revised
timing specification. This is why ABIT cannot officially support the
AthlonXP processor on earlier motherboard revisions - although if you can
live with the occasional hard reset there is no problem. Furthermore, ABIT
have made attempts to improve the timing compatibility with the 64 BIOS
release, but still cannot officially support the processor on these
revisions.

No BIOS release will allow ABIT to properly support the AthlonXP on
motherboard versions prior to v1.3. However, in practice, the timing
difference is sufficiently subtle that many processors will work fine, and
others will simply require the occasional reset during boot. Once the
processor has booted, the stability should usually be the same on all
motherboards.

In some respects ABIT are simply being more honest than other motherboard
manufacturers. In most cases the AthlonXP will appear to work OK - or have
occasional instabilities. However, rather than pretending that a simple BIOS
update can resolve the problem, ABIT are only officially claiming that the
board with the proper additional circuitry can support these new processors.

I respect KK's research, but do not agree with his conclusions. The FID
specs tell you how long the chipset has to recognize the multiplier on the
cpu, they have nothing to do with power supplies. The FID specs changed
between the Thunderbird and the XP, and that meant that some motherboards
could handle the new chips without mods, some needed a bios update to work
reliably, and some wouldn't work at all. Newer motherboards were designed
with the XP in mind and have no such problems.

"Ed" <n...@home.com> wrote in message
news:evf39uotg2bfuuhg3...@4ax.com...

> That is for Motherboard Design Guides, really has nothing to do with a
> PSU. The Mobo has to do the timing tricks.

Kaptian Krunch

unread,
Mar 15, 2002, 1:13:46 PM3/15/02
to
44 Signal and Power-Up Requirements Chapter 8
Power-Up Timing Requirements. The signal timing requirements are
as follows:

1. RESET# must be asserted before PWROK is asserted.
The AMD Athlon XP processor does not set the correct
clock multiplier if PWROK is asserted prior to a RESET# < relies on
PWR OK from power supply
assertion. It is recommended that RESET# be asserted at
least 10 nanoseconds prior to the assertion of PWROK.
In practice, a Southbridge asserts RESET# milliseconds
before PWROK is deasserted.

2. All motherboard voltage planes must be within
specification before PWROK is asserted.
PWROK is an output of the voltage regulation circuit on the
motherboard. PWROK indicates that VCC_CORE and all
other voltage planes in the system are within specification.
The motherboard is required to delay PWROK assertion for
a minimum of three milliseconds from the 3.3 V supply
being within specification. This delay ensures that the
< same here
system clock (SYSCLK/SYSCLK#) is operating within
specification when PWROK is asserted.
The processor core voltage, VCC_CORE, must be within
specification as dictated by the VID[4:0] pins driven by the
processor before PWROK is asserted. Before PWROK
assertion, the AMD Athlon processor is clocked by a ring
oscillator.
The processor PLL is powered by VCCA. The processor PLL
does not lock if VCCA is not high enough for the processor
logic to switch for some period before PWROK is asserted. <
power ok
VCCA must be within specification at least five
microseconds before PWROK is asserted.
In practice VCCA, VCC_CORE, and all other voltage planes
must be within specification for several milliseconds before
PWROK is asserted.
After PWROK is asserted, the processor PLL locks to its
operational frequency.

3. The system clock (SYSCLK/SYSCLK#) must be running
before PWROK is asserted.
When PWROK is asserted, the processor switches from
driving the internal processor clock grid from the ring
oscillator to driving from the PLL. The reference system
24309D-January 2002 AMD AthlonT XP Processor Model 6 Data Sheet <
same here
Preliminary Information
clock must be valid at this time. The system clocks are
designed to be running after 3.3 V has been within
specification for three milliseconds.

4. PWROK assertion to deassertion of RESET#
The duration of RESET# assertion during cold boots is
intended to satisfy the time it takes for the PLL to lock with
a less than 1 ns phase error. The processor PLL begins to
run after PWROK is asserted and the internal clock grid is
switched from the ring oscillator to the PLL. The PLL lock
time may take from hundreds of nanoseconds to tens of
microseconds. It is recommended that the minimum time
< hmmm more power ok
between PWROK assertion to the deassertion of RESET# be
at least 1.0 milliseconds. Southbridges enforce a delay of
1.5 to 2.0 milliseconds between PWRGD (Southbridge
version of PWROK) assertion and NB_RESET# deassertion.

5. PWROK must be monotonic and meet the timing
< power ok here too
requirements as defined in Table 13, "General AC and DC
Characteristics," on page 36. The processor should not
switch between the ring oscillator and the PLL after the
initial assertion of PWROK.

6. NB_RESET# must be asserted (causing CONNECT to also
assert) before RESET# is deasserted. In practice all
Southbridges enforce this requirement.
If NB_RESET# does not assert until after RESET# has
deasserted, the processor misinterprets the CONNECT
assertion (due to NB_RESET# being asserted) as the
beginning of the SIP transfer. There must be sufficient
overlap in the resets to ensure that CONNECT is sampled
asserted by the processor before RESET# is deasserted.

7. The FID[3:0] signals are valid within 100 ns after PWROK is

< power ok here again


asserted. The chipset must not sample the FID[3:0] signals
until they become valid. Refer to the AMD AthlonT
Processor-Based Motherboard Design Guide, order# 24363, for
the specific implementation and additional circuitry
required.

8. The FID[3:0] signals become valid within 100 ns after
RESET# is asserted. Refer to the AMD AthlonT Processor-
Based Motherboard Design Guide, order# 24363, for the
specific implementation and additional circuitry required.

Ok Ed, the Mother board design guide is a seprate document, but has info for
the additional circuitry
that the mobo manufacturers must impliment inorder to get the above timings
correct.
You say it has nothing to do with a power supply, however it does mention in
7 places about the PWROK signal
which comes from the power supply as well as the voltages associated with
it.
sorry group if this is so long, but some people need some education, and
some just cant be educated.

KK

"Ed" <n...@home.com> wrote in message
news:evf39uotg2bfuuhg3...@4ax.com...
> That is for Motherboard Design Guides, really has nothing to do with a
> PSU. The Mobo has to do the timing tricks.
>
> On Tue, 12 Mar 2002 02:58:34 -0500, "Kaptian Krunch"
> <kaptiankrunch=remov...@comcast.net> wrote:
>

> >Hey Kaptian Krunch Komes Through! Oh Joy, Oh Joy, Oh Joy!
> >If you have a cold boot problem with the XP processor This is a must read
> >post.

<snip to shorten>


Kaptian Krunch

unread,
Mar 15, 2002, 1:33:59 PM3/15/02
to
Hi Skid, Good reply with good solid facts, not just what you think. I know
you will read the steps I put in a previous post and realise not only the
mobo circuitry must be right, BUT also the power supply timings(PWROK) must
also be right for the system to be completly stable. As you can see in the
8 steps it takes for the processor to boot, 7 of those 8 relies heavly on
the PWROK signal from the powersupply (PIN 8 of the PS or gray wire)

"Skid" <sk...@attbi.com> wrote in message
news:Nrpk8.31836$WP2.8...@typhoon.atl.ipsvc.net...


> Thanks, Ed. That is my conclusion as well. These specs are really AMD's
> design advice for motherboard manufacturers, and they explain -- among
other
> things -- why the Athlon XP works fine in some older KT133A mobos and not
> others. This whole issue was hotly discussed in regards to the Abit KT7A,
> and this explanation is lifted from the FAQ at

<snip>

Nitza666

unread,
Mar 15, 2002, 4:18:00 PM3/15/02
to
Oh my god KK,PWROK signal has nothing to do with the PSU,its applied from
the mobo to the CPU when the on board power supply of the VCore is ready to
give stable voltage to the CPU.Mobo starts to use the timings AMD says when
it gets stable Voltages from the PSU and not the from the time you press the
Power Button.

"Kaptian Krunch" <kaptiankrunch=remov...@comcast.net> wrote in message
news:3c923eaf$1...@news.mhogaming.com...

Kaptian Krunch

unread,
Mar 15, 2002, 4:42:21 PM3/15/02
to
Oh my gawd Nitz666, you mean that I dont need a power supply because the
mobo has its own? NOT! guess the PWROK line (the little gray wire that
comes from the PSU TO the mobo on pin 8 of the power connector) is totally
useless? NOT... just for kick cut it and see if you can boot.

"Nitza666" <nitz...@FAhotmail.com> wrote in message

news:a6todb$hlf$1...@nic.grnet.gr...

Nitza666

unread,
Mar 15, 2002, 7:58:54 PM3/15/02
to
Are you saying that the PWROK of the PSU goes to the Athlon are you serious
KK this is no the same signal????????
Of course the mobo hasa power supply(Can you see on your mobo 6 our 4
MOSFETs?? these are the on board PSU of the CPU) it has a DC to DC converter
that uses the 5V of the PSU to convert it and supply with it the CPU or do
you think that the CPU gets 5V and the in the die it converts them to
1.75V???

"Kaptian Krunch" <kaptiankrunch=remov...@comcast.net> wrote in message
news:3c926ae0$1...@news.mhogaming.com...

Kaptian Krunch

unread,
Mar 15, 2002, 9:53:33 PM3/15/02
to
yes I am saying that the PWROK goes to the cpu pin PWROK how do you think
the cpu knows when the power supply is ready to supply it with the power it
needs? Before you speak again I would suggest you take a course in
Electronics engineering like I did for 2 years and have an associates degree
in... sheesh!

"Nitza666" <nitz...@FAhotmail.com> wrote in message

news:a6u5c2$mh1$1...@nic.grnet.gr...

Kaptian Krunch

unread,
Mar 15, 2002, 11:03:49 PM3/15/02
to
Let me re reply to this in a more intelligent mannor then the writer of the
below message and what I wrote before to explain to him what the PWROK
signal does in the powersupply (he doesnt know diddly about electronics
evedemtly)
The PWROK signal is a low voltage signal that comes off of pin 1 of an LM393
compairator. It will not assert a power good signal till the power supply
has sufficent voltage and amps. basically a timing circuit (yes right in
that little PSU). when the ps is first turned on (we are talking
milliseconds here voltages are unstable and there are spikes and dips.and
the LM393 compairs the signals off the switching voltage regular TL494 chip
for the right voltage. not till the voltagages have settled down will the
LM393 give the ok for the processor to boot up.
the signal from the LM393 is in the area of 2.2v and is compatable with most
logic circuits and cmos. Just type in LM393 in google to see the circuitry.

KK
-It is a proven fact that people can not hear with their mouths wide open -

"Nitza666" <nitz...@FAhotmail.com> wrote in message

news:a6u5c2$mh1$1...@nic.grnet.gr...

0 new messages