VOGONS


Voltage problem on Socket 7 board

Topic actions

First post, by naujoks

User metadata
Rank Newbie
Rank
Newbie

I have a FIC PA-2007 board, which has a power problem.
https://theretroweb.com/motherboards/s/fic-pa-2007#docs

Funnily, when I put in a 100Mhz Pentium, everything works just fine, but when putting in a Pentium 166MHz CPU (2.8V), the board doesn't boot at all.
I'm measuring just 2V at the Vss pins on the socket.
I think I got the jumpers right (166MHz 66MHz 2.5x Clk 1 & 2, Clk 2 1 & 2, Clk 3 2 & 3, Freq 1 2 & 3, Freq 2, 2 & 3 , Freq 3 1 & 2, VR 2 & 3 (3.3v/2.8V).

What could be the reason for not getting the right kind of power?

Reply 1 of 22, by majestyk

User metadata
Rank Oldbie
Rank
Oldbie

Mainboards from this period, like the PA-2007, 2010-2013, 503+..., often suffer from the capacitor plague. The larger electrolytics have to be replaced to restore the proper function of the step down voltage regulators.
You find these electrolytics left of and below the CPU socket in the picture.

Reply 3 of 22, by lti

User metadata
Rank Member
Rank
Member

What voltage do you get with the 100MHz Pentium? The socket 7 boards I've seen only used the 3.3V regulator with a single-rail CPU, so check the I/O voltage (VCC3) as well.

Vss is usually ground, so that's a little confusing. The core voltage on socket 7 is VCC2.

Reply 4 of 22, by Horun

User metadata
Rank l33t++
Rank
l33t++

Yes the p-100 is single core V cpu and sounds like the low V core VRM is having an issue. The board has two VRM: a 3.3v and the one for dual V cpu's (aka 2.8v).
Check the caps specially around the inductor. If the P-166 is not an MMX it should be single V core 3.3v

Hate posting a reply and then have to edit it because it made no sense 😁 First computer was an IBM 3270 workstation with CGA monitor. Stuff: https://archive.org/details/@horun

Reply 6 of 22, by majestyk

User metadata
Rank Oldbie
Rank
Oldbie
kingcake wrote on 2024-06-10, 04:44:

Motherboard appears to have adjustable linear regulators, not switch mode VRMs. A bad electrolytic shouldn't make it go low. As asked above, why are you measuring Vss to ground?

This is _not_ a linear regulator, it´s a classic buck / step-down regulator:

pa2007buck.jpg
Filename
pa2007buck.jpg
File size
57.74 KiB
Views
442 views
File license
Public domain

Reply 7 of 22, by kingcake

User metadata
Rank Oldbie
Rank
Oldbie
majestyk wrote on 2024-06-10, 05:19:
kingcake wrote on 2024-06-10, 04:44:

Motherboard appears to have adjustable linear regulators, not switch mode VRMs. A bad electrolytic shouldn't make it go low. As asked above, why are you measuring Vss to ground?

This is _not_ a linear regulator, it´s a classic buck / step-down regulator:
pa2007buck.jpg

Oh yeah, looks like it. I did a quick google and got LT1085s. Obviously I was looking at the wrong board.

Reply 8 of 22, by dionb

User metadata
Rank l33t++
Rank
l33t++
naujoks wrote on 2024-06-08, 07:13:

[...]

but when putting in a Pentium 166MHz CPU (2.8V), the board doesn't boot at all.

Exactly which CPU do you have there? If in doubt, post the s-Spec for the CPU in question.

The regular P166 is a 3.3V CPU, so almost certainly won't boot with 2.8V set.

Reply 9 of 22, by naujoks

User metadata
Rank Newbie
Rank
Newbie

My apologies, I got all my Vccs Vsss in a twist.
I've replaced all caps around the CPU socket.
I'm measuring 2.02V on Vcc2 and 2.79V on Vcc3. No boot codes whatsoever. The CPU remains stone cold.
I'm attaching a picture of the P166 MMX CPU I'm using. It's the one which has the contacts on the other side, which makes is very convenient to measure voltages.

Attachments

  • IMG_3295.JPG
    Filename
    IMG_3295.JPG
    File size
    1.62 MiB
    Views
    320 views
    File license
    Public domain

Reply 10 of 22, by Sphere478

User metadata
Rank l33t++
Rank
l33t++

Yeah those voltages are all wonky.

Should have 3.3-3.51 on vcc/vcc3
Should have 2.45-2.8 for that cpu on vcc2 (there were p55c models with that voltage range, most should work anywhere in that range at 233mhz and lower)

Do you have a p54 ppga pentium? (Looks the same, but no mmx)

You may just have your voltage jumpers set wrong.

If the cpu is set to 2x and fsb is at 50mhz the cpu won’t make much heat. It could all just be settings.

You don’t need a cpu to check voltages.

Use a small resistor leg and some jumper wires to test the cpu socket and a bunch of jumper configs.

You should power down between changes. As the fet driver can glitch between settings on some mobos and feed 5v direct.

Sphere's PCB projects.
-
Sphere’s socket 5/7 cpu collection.
-
SUCCESSFUL K6-2+ to K6-3+ Full Cache Enable Mod
-
Tyan S1564S to S1564D single to dual processor conversion (also s1563 and s1562)

Reply 11 of 22, by naujoks

User metadata
Rank Newbie
Rank
Newbie

I have a non MMX P166 CPU, it works just fine. VR setting is the same as for the 100MHz CPU (ie VR pin 3 & 4).

Can someone check that I've got the jumpers set right, please?

For the 166MHz Intel P166 MMX CPU I have set:
CLK 1 1-2
CLK2 1-2
CLK3 2-3

FRQ1 2-3
FRQ2 2-3
FRQ3 1-2

VR 9-10

Only one jumper needs to be set on the VR, right? So 9-10 for the 166 MMX?

Reply 12 of 22, by majestyk

User metadata
Rank Oldbie
Rank
Oldbie

Jumpers are correct.
You can also try VR -> 7-8 for 2.9V vore voltage.

Strange that V I/O is too low also. It has it´s own linear VRM.

...Your high resolution pics of this part of the mainboard are still a bit "insufficient".

Reply 13 of 22, by Sphere478

User metadata
Rank l33t++
Rank
l33t++
majestyk wrote on 2024-06-11, 13:52:
Jumpers are correct. You can also try VR -> 7-8 for 2.9V vore voltage. […]
Show full quote

Jumpers are correct.
You can also try VR -> 7-8 for 2.9V vore voltage.

Strange that V I/O is too low also. It has it´s own linear VRM.

...Your high resolution pics of this part of the mainboard are still a bit "insufficient".

I wonder if vcc3 is broken and the board is powering up from vcc2 (or vice versa) when jumpered by a p54 cpu (those cpus bridge vcc2/3)

Probably not, as both read low.

Here’s a thought…

What if your volt meter isn’t reading correctly? Both readings are equally low…

What are you using for ground reference? Which pins are you measuring vcc2 and vcc3 from?

What if the problem is in fact the cpu.

Try random multiplier jumper settings I’ve seen that hang boards before. Don’t ask me why. Also reset cmos after cpu change. And be sure you have the latest bios. (Check Jans website)

If the board has a p55/p54/dual/single jumper I have seen it where setting it incorrectly actually is the way that makes it work.

Sphere's PCB projects.
-
Sphere’s socket 5/7 cpu collection.
-
SUCCESSFUL K6-2+ to K6-3+ Full Cache Enable Mod
-
Tyan S1564S to S1564D single to dual processor conversion (also s1563 and s1562)

Reply 14 of 22, by naujoks

User metadata
Rank Newbie
Rank
Newbie

I'm getting my ground from the tinned screw holes.
I've actually got two P166 MMX CPUs and I've tried both, neither works (the CPUs are working fine in another Socket 7 board I own).
I've tried all of the 10 available BIOSes.
There doesn't seem to be a P55/P54/etc jumper.

Reply 15 of 22, by majestyk

User metadata
Rank Oldbie
Rank
Oldbie

Never trust a manual, unless you are 100% sure it´s for the very revision of your mainboard that you have.
Since you didn´t take any pictures, we don´t even know the revision of your PA-2007. Please have a look at the bottom left corner and let us know.

I have a revision 1.2 on the table right now and for 66 MHz FSB the clock jumpers CLK 1-3 need to be set "1-2, 1-2, 1-2" - NOT "1-2, 1-2, 2-3" like you did.

This mainboard detects the "split voltage CPU present" pin at the CPU socket to switch between the modes, no need to set jumpers.

If a single voltage CPU is inserted, Vcc2 and Vcc3 pins are shorted inside the CPU. This shorts the outputs of both VRMs.

That´s why the manual demands you set the Vcore VRM to 3.5V for 3.5V (P54) CPUs, the CPU gets 3.5 V for both core and I/O from the Vi/o regulator, the second (Vi/o) VRM gets latched.
In case of a 3.3V P54 according to the manual you set the Vcore VRM to 3.3V. Both VRMs share the load in this case and produce one (high current) 3.3 V supply.

If a P55 CPU is inserted, there´s no inside bridging, the output of the 3.3V VRM supplies Vi/o directly to the Vcc3 pins of the CPU. The Vcore VRM is set to any voltage below 3.3V and supplies Vcore to the Vcc2 pins of the CPU.

As you can see, the smaller Vi/o VRM with it´s tiny heatsink is never forced to deliver the full I/O plus core power to any CPU. It either supplies I/O only (P55), supplies a part of the power (3.3V / P54) or is inactive / latched (3.5V / P54 cpu case).

Most older mainboards don´t have _all_ mounting holes grounded, but only some of them. Better use the center pins of the PSU connector for ground when masuring voltages.

Last edited by majestyk on 2024-06-14, 08:17. Edited 6 times in total.

Reply 16 of 22, by Sphere478

User metadata
Rank l33t++
Rank
l33t++

Normally I would’t recommend trying this, but in theory it wouldn’t do anything that installing a p54 wouldn’t already be doing. And the cpu should survive.

You would normally not do this for over volting reasons of the cpu, but as many others have pointed out, mmx chips always seem to survive it. (Though I don’t like it)

Anyway, disclaimers out of the way, at your own risk, you could bridge vcc2 and vcc3 on the fets with a wire if you are able to locate the proper fet legs/tabs. This will effectively be as if you installed a p54 chip. As far as the mobo is concerned.

You need to do it at a high current junction like the fets, as doing it at a single processor pin may burn the pin.

As far as the chip is conserned it will be as if you installed a pmmx in a socket 5 mobo.

The higher voltage reg will take the load and the lower voltage reg will just free wheel with the higher voltage on the output.

Anyway, it should boot up, the mmx chip will get hot as shit. (So heat sink it with fan and paste) But in theory it should work

many have run it this way (3.3v) for years, but I suggest you don’t keep it this way.

Many have also ran p54 chips on dual plane boards which make this same bridge I am suggesting.

If this does not boot up, then your problem isn’t voltage related. If it does, then your problem is vrm related. Vcc2 or vcc3

I still think you should confirm that your DVM is reading voltage correctly

Sphere's PCB projects.
-
Sphere’s socket 5/7 cpu collection.
-
SUCCESSFUL K6-2+ to K6-3+ Full Cache Enable Mod
-
Tyan S1564S to S1564D single to dual processor conversion (also s1563 and s1562)

Reply 17 of 22, by majestyk

User metadata
Rank Oldbie
Rank
Oldbie

The linear regulator for 3.3V I/O with it´s tiny heatsink (right bottom corner) is designed for providing Vi/o only. It will probably fail early when it is forced to supply both Vcore and Vi/o.
I know many designs with decent regulators and heatsinks for both voltages where this might be a workaround for a broken Vcore regulator.

pa_2007_vrm.JPG
Filename
pa_2007_vrm.JPG
File size
1.88 MiB
Views
109 views
File license
Public domain

There isn´t much left to examine here: The integrated regulator / MOSFET driver circuit U20 (ADP1148) or one of the two MOSFETS (NDP6030L) could be the culprits.
On the Vi/o side it´s one NDP6020L (less power) and a 3310A-3.3 for driving it.

The FIC PA-2007 is a very reliable, versatile high performance Socket 7 mainboard and it´s definitely worth repairing it.

Reply 18 of 22, by Sphere478

User metadata
Rank l33t++
Rank
l33t++

My understanding is that is basically how installing a p54 typically works, the vio reg powers both. And vcc2 just kinda sits there because it is only producing a lower voltage so it carries none of the load.

On some s7 mobos you can set both regs to the same voltage and this shares the load. It’s a good idea. I like that. Typically one reg will be taking a little more load though because of slight voltage mismatch.

I’m curious what the p54/p55 jumper on some mobos accomplishes. Does it set both regs the same? Seems like a logical feature. Or does it bridge them? I haven’t looked into that jumper. I guess it was reported that this mobo doesn’t have that. Many dont.

In any case going from 5v to 3.51-3.3 v doesn’t make as much heat as 5v to 1.6-3.2v And many s5 mobos have a single tiny reg. The hottest most power hungry s5/s7/ss7 chips actually were single plane cpus. The cyrix 6x86 133mhz got up to about 22w

IMG_3623.jpeg
Filename
IMG_3623.jpeg
File size
304.6 KiB
Views
98 views
File license
Public domain

That’s not to say that if you are rocking a s5 and a 22w cpu a nice dual reg with big heatsink wouldn’t be a good idea, just saying that many mobos ran the cpu on one reg.

Attachments

  • IMG_3624.jpeg
    Filename
    IMG_3624.jpeg
    File size
    156.27 KiB
    Views
    83 views
    File license
    Public domain
Last edited by Sphere478 on 2024-06-13, 17:48. Edited 1 time in total.

Sphere's PCB projects.
-
Sphere’s socket 5/7 cpu collection.
-
SUCCESSFUL K6-2+ to K6-3+ Full Cache Enable Mod
-
Tyan S1564S to S1564D single to dual processor conversion (also s1563 and s1562)

Reply 19 of 22, by majestyk

User metadata
Rank Oldbie
Rank
Oldbie

The heatsink beside the Cyrix is _far_ larger than the one on the PA-2007.

On the PA-2007 for a P54 cpu you have to jumper the Vcore regulator for 3.3V. (The I/O reg has a fixed driver circuit for 3.3V).

There are also designs where all power passes the Vi/o reg. first, then the Vcore VRM is fed 3.3V and reduces it to 2.8V or whatever. This splits the heat dissipation, since both transistors are producing a voltage drop.

TP54/55 jumpers are often 3-5 jumpers in parallel and they just short the outputs of both regulators.