Re: P2B-DS and generic slotkets



Frank Galphin wrote:
"P2B" <p2b@xxxxxxxxxxxx> wrote in message news:Dxj4f.11066$vD4.618719@xxxxxxxxxxxxxxxxxxxxxxxx


Frank Galphin wrote:

"P2B" <p2b@xxxxxxxxxxxx> wrote in message news:xJ%3f.6840$S43.1114160@xxxxxxxxxxxxxxxxxxxxxxxx


Frank Galphin wrote:



"Paul" <nospam@xxxxxxxxxx> wrote in message news:nospam-2809050108000001@xxxxxxxxxxxxxxxx



In article <rK6dnTKsOMzsKKTeRVn-ug@xxxxxxxxxxxx>, "Frank Galphin"
<fgalphin@xxxxxxxxxxx> wrote:




"Paul" <nospam@xxxxxxxxxx> wrote in message
news:nospam-2009050258060001@xxxxxxxxxxxxxxxx



In article <YIedndI8cNDpoLLeRVn-gQ@xxxxxxxxxxxx>, "Frank Galphin"
<fgalphin@xxxxxxxxxxx> wrote:




"Stephan Grossklass" <sgrokla-nospam04q2@xxxxxxxx> wrote in message
news:dgkejf$qsv$04$1@xxxxxxxxxxxxxxxxxxxx



Frank Galphin schrieb:




Hay all,
Well subject says it all. I have 2 PIII 850's which are not the
same
stepping
value. I have 2 generic slotkets the are supposedly dual capable. The
system
will boot with one of the 850's in slot 1 or slot 2. However when both
are
installed
the system will not boot. The slockets are labeled "Cpu Card Rev.1.0"
any
ideas?

"Dual capable" could also mean dual *Celeron* (Mendocino) capable, and
many slot adapters are. This does not imply that dual CuMines work as
well. Sometimes this also requires changing jumper settings, like on
the
MSI 6905 Master. Both the dual capable 6905 Master and the S370-DL have
a 4-pin CPU selection jumper, which makes me suspect that this is a
requirement for unmodded dual CuMine operation.

Stephan
--
Home: http://stephan.win31.de/
PC#6: i440BX, 2xP3-500E, 704 MiB, 250+80 GB, R9k AGP 64 MiB, 110W

Hi Stephan,
Well I tried the PIII 450 slot 1 and a celeron 500 "Mendocino core"
together using
the generic sloket and same as before would not boot in dual mode. Any
other
ideas?
I know i need to find a Asus 370-dl slocket for really cheap. Then it
might
work.

Frank 

Upgradeware Slot-T Conversion Kit  $30 each
http://www.bytewizecomputers.com/products/7/9/104/1048

The Upgradeware web page:
http://www.upgradeware.com/english/product/slott/slott.htm

Modification procedure for making Slot-T (Tualatin) slocket
dual processor ready (for PIII-S Pentium):

http://www.tipperlinne.com/slot-t.htm

I would try the adapter without the modification first.
With your processors, the Slot-T may work without any mod.

HTH,
 Paul


Well I kinda got it working. The secret is the N33 pin as described here.
http://www.hardwarecentral.com/hardwarecentral/reviews/1673/4/
Both processors are working fine in Dual mode however,
the voltages are flaky. According to the hardware monitor the
processors are requesting 2.0 volts instead of 1.65 volts. I really
don't want to run these processors at 2.0 volts for to long. I don't
think they will last. What I don't understand about that article
is why is R3 being removed? What was it's purpose?
Again any hints suggestions welcome.

Frank 

"440bx design guide" - PDF pg.30
http://www.intel.com/design/chipsets/designex/29063401.pdf

"GTL+ is a low output swing, incident wave switching, open-drain
bus with external pull-up resistors that provide both the high
logic level and termination at the end of the bus."

One purpose of the resistors, is to function as the load
resistor and also as a transmission line termination, to
reduce signal reflections. If R3 is a 56 ohm resistor, perhaps
that is its purpose.

A good technical article on modding, should explain what each
of the modifications is doing, and what the old connectivity
achieved before the modder cut it up. Such information makes
it possible for a third party to check the workmanship of
the modder, and interpret what he/she is doing. Since Mr. Sassen
chose not to help us in that way, it is hard to say what he
has done.

For example, I don't understand why Mr. Sassen has connected
a 56 ohm resistor to ground. It could be that this grounding
resistor is physically underneath where R3 is located on
the other side of the board. In other words, a signal used
to have a 56 ohm resistor to Vtt, and now it has a leaded
resistor connected to ground, on the other side of the board,
as a modification. A 56 ohm resistor might make a good high
speed termination for some kind of logic, but if the signal
on that line is GTL+, a pullup resistor to Vtt makes a lot
more sense.

Without a schematic diagram or a block diagram of what he has
done, I would need the module in my hands to figure out what
has been done.

With regard to the voltage being requested, use an ohmmeter
to see if there is continuity from the VID pins in the socket,
to the SC242 edge connector. It could be the slocket doesn't
have them wired, or there is a DIP switch or something ? Are
you sure the slocket doesn't have a way to modify the VID
value ?

This article describes how to do a "wire trick", in a situation
where there is no DIP switch to set the Vcore voltage. When
I needed to do a mod like this, I just cut the appropriate
VID tracks near the edge connector, and wired the signals as
required. (I modded a slocket that had no VID jumpers.)

http://www.tipperlinne.com/p2b-ds.htm

Other info:

http://developer.intel.com/design/chipsets/440/documentation.htm

(A dual processor 440BX motherboard schematic. This will not
teach you much, as the real action is on the slocket. Still,
this is one of the few reference schematics available for
download.)

http://www.intel.com/design/chipsets/designex/BXDPDG10.PDF

Paul


Well I got it working after a little effort. I ordered two of the Upgradeware Slot-T adapters from Strattton computers. They were the newer Version 1.1 adapters so
I did not need to relocate resistor R2 to R3 Position as described here:
http://tipperlinne.com/slot-t.htm

I haven't seen a v1.1 Slot-T yet. Did they add a jumper so you can select X2 or N33 for the bus request signal, or just omit R2 and install R3 to make N33 the factory default?

I suppose I should order a pair, check them out, and update my site...



But let me tell you what I had to do to make those two PIII 850's cooperate with each other. It turns out that one of the PIII 850's was set to request 1.70 volts instead of 1.65 volts. Difference in stepping I assume. So I set the Slot-T's to 1.65 volts for CPU 1 and 1.70 volts for CPU 2. System booted and recognized two processors. I think YEAH!!!! well that was short lived. I checked the hardware monitor and guess what? They were requesting 2.0 volts. Boy was I confused. So just for the hell of it I decided to set both Slot-T's to 1.65 volts. Well that fixed the problem. The hardware monitor was reading 1.6 volts. YEAH happiness. I guess the minor voltage difference was throwing off the onboard voltage regulators. Well anyway like I said it is now working. 

Interesting!

The two voltage regulators are quite independent, each controlled by the VID signals from their respective processors (or the VID jumper settings when running Slot-Ts), so there's no reason the processors cannot run at quite different voltages - and software like MBM will correctly report differing CPU voltages. The Slot-Ts in my main system have been set to 1.35v and 1.40v for some time - the lowest stable settings - and MBM reports 1.37v and 1.41v respectively. Last time I checked with a meter, the actual voltages were about 100mV above what MBM reports. The BIOS displays 1.5v - it's not very accurate, and appears to round up in 0.5v increments.

The BIOS only displays one CPU voltage. I would have expected that to be taken from one or other of the regulators - and your post prompted me to find out which one.

I was wrong - testing proves the BIOS reports it's idea of the higher of the two Vcore voltages, regardless of which regulator is delivering it. I suppose reading both and displaying the higher one only took a few lines of code, but I was surprised. A few more lines and it could display both :-)

I can think of three explanations for your observations when you had the Slot-Ts set differently:

1. Upgradeware screwed up the VID jumpers on the v1.1 adapters (unlikely)
2. The BIOS reported a 1.7v setting as 2.0v because the regulators always run a little higher than the VID request, and the BIOS rounds up (less unlikely)
3. You accidentally set the jumpers on one adapter to a higher voltage than you intended (no offense intended, I've done it)

It would be most interesting if you could reproduce the 2.0v BIOS reading when the adapters are set to 1.7 and 1.65, then see what MBM reports. I would be perplexed if it wasn't within 250mV of the jumper settings.

BTW, if you aren't overclocking, those PIII 850's are likely to be perfectly happy (and a little cooler) running at the 1.55v setting.

P2B 


Hey P2B thanks for the response.

Regarding the Ver 1.1 Slot T's there is no jumper setting to select either X2 or N33. The layout
seems to be similiar to the picture posted on your site. Where you show R2 there is a 56 Ohm resistor
and R3 is allready populated with a 0(zero) Ohm resistor. Wish I had a digital camera to show you. 

No need, I get the picture :-)


Looks like N33 is selected by default!

Indeed. The 56 Ohm X2 pullup is not required according to Intel datasheets, but it can't hurt.

I don't suppose I'll be selling any more modded Slot-T adapters :-( It was a nice sideline for the 2+ years it took Upgradeware to adopt my fix.


Well I did verify that I did have the jumpers set right and they were (No offense taken. I have done it too!)

Well when I had the processors set to 1.65 volts for CPU 1 and 1.70 volts for CPU 2 I went ahead
and booted into XP pro and used Everest at Home to look at the hardware sensors. It was reporting
2.00 volts and 2.05 volts respectfully for each CPU. So it seems as though it's not the BIOS reporting
incorrect voltages unless Everest is pulling down the info from the BIOS??. I thought it would poll the
Hardware sensor chip to get the info. Very confusing.

Everest definitely gets it's data from the sensor chip, so this is very mysterious. When you set 1.65v on both adapters, Everest reports close to that, right?



Yes that is what everest is reporting around 1.63 vots.


Do you have a meter to confirm the CPUs are really getting 2.0v with the 1.65/1.70 setting? Maybe the sensor chip is the problem - which one is it? It's between the end of the ISA slot and the BIOS chip - I expect to see a Winbond 83781D there. What revision is your P2B-DS? It must be at least 1.05 - and if it's one of the rare oddball 1.05s with mismatched regulators, perhaps the answer lies there...



*I do have a meter but am leary of using it. I don't want to short out
the test lead to an adjacent pin. Besides I am not sure whcih pin
to check.

I hear you on that - I've done it and crashed the system, luckily did not permanently damage anything. Had me worried for a bit, though - it wouldn't POST after I cycled power, had to remove +5vsb for a few minutes to wake it up again.

I built test points onto a pair of Slot-Ts so I could easily measure various voltages without risk. If you care to go to the trouble, the easiest way to measure Vcore is to solder a short length of wire-wrap wire onto the back of pin AA37 (Top row, 7th pin from the left as viewed from the back of the Slot-T), and connect the meter to the other end.

I think this is worth tracking down, because at this point I'm not sure I'd trust anything but a meter to tell me what voltage the CPUs are really getting, regardless of jumper settings.


I was going to overclock these to 952Mhz by setting 112 Mhz FSB and up the voltages on the CPU's
to 1.75Volts. Should be stable I think.

Most Coppermines will overclock at least 12% on default voltages. Don't give 'em more juice until they prove their need :-)



Left at 1.65 volts system would lock up with 112Mhz selected through SoftFSB
So I think I need a voltage boost.

Quite possibly - but I'd still select 112Mhz via jumpers to confirm the instability at default voltage. I've experienced odd behavior using SoftFSB on the P2B-DS, and don't really trust it.

Thanks for the input P2B. Much appreciated.
Any further questions please ask.

Your problems seem to be solved, so perhaps you'd prefer to move on - but if you're interested in understanding anomalies encountered along the way, the comprehensive list of useful information would be:

- Motherboard revision and PCBA level (the latter is a D0x number, sometimes printed somewhere on the component side, but often only on a sticker on the side of the last ISA connector - which requires a mirror to read after the board is installed)
- Regulator chip IDs (should be HIP6004xx[x] and HIP6019xx[x])
- Sensor chip ID (see above)
- CPU sSpecs (Sxxxx, printed on the CPUs)
- Meter readings for Vcore when BIOS shows 2.0v, and again when it reads 1.6v


Motherboard is a Rev 1.06 PCBA Level D03 latest and greatest I heard!

Definitely the latest and greatest from Asus - only a Tipperlinne P2B-DS is better :-)

Regulator chips: HIP6004BCB and HIP6019BCB 

As expected on 1.06 D03

Sensor chip: ASUS AS97127F

Not sure what that one does (Asus doesn't make their chip datasheets available), but it isn't the sensor. The Winbond is at the other end of the ISA slots, between the inner one and the BIOS chip - pretty much guaranteed to be a W83781D.

CPU sSpecs: CPU 1 SL43H CPU 2 SL4CC

Different steppings, but according to Intel "Currently no known issues associated with mixing these steppings". I very much doubt you've found an issue, too - my gut says it's something else.

*See answer above


I'm seriously considering ordering a pair of v1.1 Slot-Ts to play with, and I have a few spare P3 933s I haven't gotten around to selling - maybe I can reproduce the anomaly.

P2B 


Again thanks for the input P2B. Hope that answers your questions. As befor
any further questions please ask.

Thanks. No more questions. At this point I suspect an undocumented erratum in the Winbond sensor - erroneously reports 2.0v Vcore under your specific configuration. We'll leave it at that unless you decide to take meter readings, or I reproduce the anomaly.

P2B
.