That's quite the curious find!
It's rather unlikely that the SX835 is damaged.

spacesaver wrote on 2025-02-06, 10:54:

I scored a rare motherboard with both socket 4 and socket 5! That will be great for saving space while being able to test both k […]
Show full quote

I scored a rare motherboard with both socket 4 and socket 5! That will be great for saving space while being able to test both kinds of CPUs.

Socket 5 CPUs run fine, but can't get any socket 4 CPU to post. The manual on the retroweb page says,

A regulator card or jumper card is also required and is installed over the second ZIF socket.

So that strongly suggests the pins marked "P5 POWER CONN" should be connected some how. When using socket 5, a 5V to 3.3V regulator board is plugged into "P54C POWER CONN". If I follow the same pin pattern, then I deduce for the "P5 POWER CONN", that the 2 5V pins should be connected to the 3 pins on the other end. I did that, but it still doesn't POST. I also unset the jumper for bus speed to 60 MHz. I've tried only 1 CPU so far, SX835. It's possible the CPU is broken since it was listed as untested on eBay.

Does anyone have a similar board and can show what the "P5 POWER CONN" looks like? Or know of any other jumpers that should be changed?

Will probably be a stretch to find the 5V regulator as the 560 was the only model that needed it out of the 5xxx series. Guess DEC never envisaged anyone wanting to downgrade their system as the 5V version isn't even in their spare parts catalogue, although the 3.3V version is. There's not even an image of it in the manual or the Celebris 560/590 Field Service/Field Training help app.

Hurray, got it working! The trick was to plug in both the 3.3V regulator and the 5V jumpers at the same time.

I am a little concerned that there's no room for a heatsink due to the 3.3V regulator on top. After running it for a while, I touched the Pentium 60 MHz and it was definitely hot as expected. But I guess the builtin heat spreader helps.

the 560 was the only model that needed it out of the 5xxx series

I guess that means this motherboard is pretty rare and valuable.

Would you mind taking a picture of the cache stick? I have a DEC P5 cache module but I have never been able to confirm which machine it belongs to.

Here you go. Strange you say P5. My stick says P54C and it works for both P5 and P54C

Also, I forgot to mention an experiment I did earlier

1                       needs 5V jumper?          needs 3.3V regulator?
2socket4                       1                         1
3socket5                       0                         1

This is suggesting socket 4 CPUs use 3.3V, which is strange.

Awesome - if you find the time, maybe take a few more shots of the system (running a few benches).

Here is mine, says 5V on it

spacesaver wrote on 2025-02-09, 02:06:

Hurray, got it working! The trick was to plug in both the 3.3V regulator and the 5V jumpers at the same time. https://www.vogon […]
Show full quote

Hurray, got it working! The trick was to plug in both the 3.3V regulator and the 5V jumpers at the same time.

I am a little concerned that there's no room for a heatsink due to the 3.3V regulator on top. After running it for a while, I touched the Pentium 60 MHz and it was definitely hot as expected. But I guess the builtin heat spreader helps.

Are you sure that this is correct configuration for power feeding when s. 4 CPU is used? There's a header for VRM add-in card near the socket 5 too - J26 "P5 POWER CONN".

Yeah I can't bring myself to actually believe this is correct/intended even if it is POSTing. The socket 4 chips ran very hot and a heatsink is absolutely required

maxtherabbit wrote on 2025-02-09, 23:42:

Yeah I can't bring myself to actually believe this is correct/intended even if it is POSTing. The socket 4 chips ran very hot and a heatsink is absolutely required

Exactly. In my previous comment I spared it, that now I'm expecting a new thread with title like "Suddenly my socket 4 Pentium stopped working, what could be the problem?".

Are you sure that this is correct configuration for power feeding when s. 4 CPU is used?

Fairly sure, because the P5 uses 5V, so no need for a regulator. Also, the manual reinforces this, "A regulator card or jumper card is also required and is installed over the second ZIF socket." Strange, why bother with a jumper instead of hard wiring it. There are 2 holes for plastic pins around the P5 connector, just like there are 3 holes for the 3.3V regulator. So maybe the jumper was a board that covers the entire socket 5.

Or maybe the cache module needs to be 5V when using socket 4, thus eliminating the need for the 3.3V regulator. As you saw above, mine is 3.3V. But that doesn't make sense because the L2 cache is connected to the memory controller, not CPU, so then it shouldn't change.

Maybe the lack of heatsink is compensated by the chassis fan. But I don't see any at the front or rear of the chassis Celebris 590 (not mine)

spacesaver wrote on 2025-02-10, 01:18:

Fairly sure, because the P5 uses 5V, so no need for a regulator. Also, the manual reinforces this, "A regulator card or jumper card is also required and is installed over the second ZIF socket." Strange, why bother with a jumper instead of hard wiring it.

Maybe the lack of heatsink is compensated by the chassis fan. But I don't see any at the front or rear of the chassis Celebris 590 (not mine)

Well, you're reading the manual like the devil reads the bible. This "A regulator card or jumper card is also required and is installed over the second ZIF socket." (then which one of the two sockets is "the first ZIF socket" if s. 4 is older than s. 5? 😁 ) or "If you install a higher performance CPU, a regulator card might be required.
Connect the regulator card to the power connector, align the holes in the card with
the plastic inserts, and then snap into place." (again, seems like this is only corresponding for P54C, but I wouldn't rely on something written like that by an obviously technically incompetent personality - I mean the one of the manual's authors) may also refer to other VRM header J26 "P5 POWER CONN" as this J26 VRM header is not there just for decoration 😀

I'll wait for the "what happened to my s. 4 Pentium...?" thread...

What happens if you remove the cache and the 3v3 regulator?

What if these pins also match, when VRM module is turned by 180 degrees and connected to J26?

maxtherabbit wrote on 2025-02-09, 23:42:

Yeah I can't bring myself to actually believe this is correct/intended even if it is POSTing. The socket 4 chips ran very hot and a heatsink is absolutely required

I agree. The socket 4 needs headsink.

I also found this on Google Groups, a post from 1999:
https://groups.google.com/g/comp.sys.dec/c/Xj … /m/yI_k-KUCfzsJ

This Celebris with Socket 4 and Socket 5 is a weird duck. […]
Show full quote

This Celebris with Socket 4 and Socket 5 is a weird duck.

To run a Pentium 60, you need another little voltage regulator part
that attaches so that it hangs over the Socket 5.

To run a Pentium in Socket 5, you need another little voltage
regulator part that attaches so that it hangs over the Socket 4.
Limit on CPU speed is 100 MHz for Socket 5.

AFAIK, you do not need the cache module installed, because the system
comes up testing for presence of the cache, and runs OK with cache
absent (or disabled in setup).

Also found this on the cache module in the service manual:

EDIT, might be worth getting in touch with Alvaro84. He seems to have the socket 4 voltage board.
Re: Any Love For Socket 5?
Re: I recently found this hardware, AKA the Dumpster find thread.

Thanks for all the tips. This is quite a mystery. Earlier, I thought the cache voltage would always be the same since it would be attached to the chipset. But apparently, the chipset voltage itself changes depending on whether using socket 4 or 5. So I decided to measure some cache and chipset voltages.

Finding the chipset was a little challenging given the uncommon VLSI chips and needing to strip the suffix. This board uses VL82C591 https://theretroweb.com/chipsets/578
In all cases, the L2 cache was disabled in the BIOS. chipset_v1 is the voltage @ pin 1, and chipset_v105 is voltage @ pin 105, both at the end of each row of pins. Pin 1 isn't power, so I find it's not stable until reaching the DOS prompt. Also, I didn't measure the exact voltage in many cases and rounded it to 5.

1P54C regulator connected?   P5 jumpers connected?   cache installed?    posts?      chipset_v105    chipset_v1       cache_volts
2             1                         0                   1              1               5             3.3             3.3
3             1                         0                   0              0               5             3.3             N/A
4             0                         1                   1              0               5          not stable         4.8
5             0                         1                   0              0               5          not stable         N/A
6             1                         1                   1              1               5              5              4.8

So, clearly the chipset voltage (pin 105) changes depending on whether socket 4 or 5 is used

Does the chipset or some other component still use 3.3V when socket 4 is used? That could support why the 3.3V regulator is still needed. Inconclusive.

Taking out the cache always makes it fail to POST.

What if these pins also match, when VRM module is turned by 180 degrees and connected to J26?

If you try to plug the 3.3V regulator board into the P5 power connector, the holes in the board for the plastic push pins, don't align

spacesaver wrote on 2025-02-10, 10:50:

If you try to plug the 3.3V regulator board into the P5 power connector, the holes in the board for the plastic push pins, don't align

Ok, I thought that the VRM card has two pairs of holes for the plastic support, so it cold match the other mounting position over the socket 5.

So, you'll need another VRM add-in card for the 5 V regulator header - the one that Alvaro84 mentions in his comments.

Just don't use s. 4 Pentium without a heatsink and a fan as it is now in your somehow-workinng current power feed configuration.

Hi everyone, I've arrived and I'll take some pics and think this config over again - tbh I haven't played with it for a good while but IIRC I used either the VRM _or_ the 5V connection, both in ways that blocked the other socket, and of course it needed the appropriate cache module too. I should have both in the same cupboard and it must work, I'll have to give it a try. Sorry, I'm definitely not fast at _anything_ but I do listen.

Edit. (+ a few hours):

Well, here's complete the Socket4 setup plus the accessories...

The jumper bridge is in position, I rarely use this board with anything else than socket 4 CPUs.

I don't think that regulator should be there at all for 5 volt operation, I find it really unlikely for it to be intended to be over the a cpu with no heatsink. Like mentioned there should be a way to just jumper that connector for it to work without the regulator card, unless it's a little more complicated then that and that's why the manufacturer had some other jumper card or something to plug in to the connector.