tehsiggi wrote on 2025-11-03, 09:44:

No, all right. You've measured directly at the ISL, so that's good.. Let my grind my gears for a little while and I'll come back to you 😁

No worries 😁 Do you think there would be any point in replacing the ISL? My replacement chips are arriving any day now.

If you have them at hand, just give it a shot. I'm a bit out of ideas on this one right now..

Here a couple things just to make sure:

• FB (pin 5) is 0V as well?
• PHASE (pin 8 ) is connected to MID_FET12 (the one side of the inductor)?
• Can you check that the COMP network C1622, C1623 and R1265 look okay and are connected how they should?
• Voltage at COMP (ping 4) when the card is running?

tehsiggi wrote on 2025-11-03, 17:21:

If you have them at hand, just give it a shot. I'm a bit out of ideas on this one right now.. […]
Show full quote

If you have them at hand, just give it a shot. I'm a bit out of ideas on this one right now..

Here a couple things just to make sure:

• FB (pin 5) is 0V as well?
• PHASE (pin 8 ) is connected to MID_FET12 (the one side of the inductor)?
• Can you check that the COMP network C1622, C1623 and R1265 look okay and are connected how they should?
• Voltage at COMP (ping 4) when the card is running?

You're on the right track tehsiggi. The ISL6525 replacement chip arrived today. WHAT AN IMPROVEMENT!!

I powered on the card and as soon as I saw the core starting warming up I jumped 😉 I then saw this on the screen....

Here are the new voltage readings.

What are your thoughts?

Hey, things start moving again!

So good to know, the ISL6525 was a goner. Alright. NVVDD is 1.6V, should be 1.66V, so good enough I'd say. We're clearly missing the memory voltages.
FBVDD (main memory voltage) should be 2.8V.
VTT (derived from main memory voltage) should by around 1.2V.

Let's check a couple of things (perhaps even in that order).

• State of NVVDD_PWRGD - pin 7 of the NVVDD regulator ISL6525 (during operation) - if that one is not high, the memory regulator will stay offline.
• Resistance of FBVDD against ground (not installed into a system)
• Resistance of VTT against ground (not installed into a system)
• Voltage on R1285 (both sides) should be around 5V - (during operation)
• Signals on R1291 (low FET gate for FBVDD) as well as for R1292 (high FET gate for FBVDD) - (during operation)

We're looking at the ISL6225 now, which is a dual output buck converter built for memory power supply applications:

We'll figure it out.

I've noticed that your riser cable introduces a significant drop on the 3.3V apparently, if only 2.4V remain at the card. 😉

Awesome ! Will go through the list tomorrow. It gets a little tricky now…with the core online, it heats up pretty quick without a heatsink so I only have a very short time to do any testing. I’m not too keen on trying to balance a heatsink on the core in case it slips off and shorts something 😗

Got no cooler handy? I take those 40x40mm coolers, put a nice load of paste onto them and then firmly push them onto the core.. The paste will keep it in place most of the time.. for everything else during testing there are cable ties 😁

Looking forward to your results.. my ISL6525 arrived as well.. let's see if I find some time to get my eval boards done..

tehsiggi wrote on 2025-11-10, 10:46:

Got no cooler handy? I take those 40x40mm coolers, put a nice load of paste onto them and then firmly push them onto the core.. The paste will keep it in place most of the time.. for everything else during testing there are cable ties 😁

Looking forward to your results.. my ISL6525 arrived as well.. let's see if I find some time to get my eval boards done..

I’ve got some little coolers sitting around. I have some Noctua paste that is pretty sticky. I’ll give it a try.

I'll cross my fingers for you 😁

tehsiggi wrote on 2025-11-10, 10:29:

Let's check a couple of things (perhaps even in that order). […]
Show full quote

Let's check a couple of things (perhaps even in that order).

• State of NVVDD_PWRGD - pin 7 of the NVVDD regulator ISL6525 (during operation) - if that one is not high, the memory regulator will stay offline.
• Resistance of FBVDD against ground (not installed into a system)
• Resistance of VTT against ground (not installed into a system)
• Voltage on R1285 (both sides) should be around 5V - (during operation)
• Signals on R1291 (low FET gate for FBVDD) as well as for R1292 (high FET gate for FBVDD) - (during operation)

We're looking at the ISL6225 now, which is a dual output buck converter built for memory power supply applications:

The attachment isl6225.pdf is no longer available

We'll figure it out.

I've noticed that your riser cable introduces a significant drop on the 3.3V apparently, if only 2.4V remain at the card. 😉

Here we go! Apologies for the delay.

• State of NVVDD_PWRGD - pin 7 of the NVVDD regulator ISL6525 (during operation) - if that one is not high, the memory regulator will stay offline. - 4.6v so high
• Resistance of FBVDD against ground (not installed into a system) - 1.6 ohms - used TP30
• Resistance of VTT against ground (not installed into a system) - 6.8k omhs - used TP29
• Voltage on R1285 (both sides) should be around 5V - (during operation) - 4.7v both sides
• Signals on R1291 (low FET gate for FBVDD) as well as for R1292 (high FET gate for FBVDD) - (during operation)

Hmm..

Could you measure R1291 and R1292 against ground (no operation)? The high side signal looks weird. Almost like stuck to the 0.8V you measured at FBVDD.
The resistance of FBVDD to ground concerns me. 1.6 Ohms is not a lot. Not sure I've measured anything like that on any mem rail of BGA memory cards.
If it's true, it would mean 1.75A of current at 2.8V, nothing too crazy, the regulator should be able to provide that (even into a resistive load). So not bothering too much about it - for now. The FET singal looks more interesting to me right now.

tehsiggi wrote on 2025-11-12, 12:05:

Hmm.. […]
Show full quote

Hmm..

Could you measure R1291 and R1292 against ground (no operation)? The high side signal looks weird. Almost like stuck to the 0.8V you measured at FBVDD.
The resistance of FBVDD to ground concerns me. 1.6 Ohms is not a lot. Not sure I've measured anything like that on any mem rail of BGA memory cards.
If it's true, it would mean 1.75A of current at 2.8V, nothing too crazy, the regulator should be able to provide that (even into a resistive load). So not bothering too much about it - for now. The FET singal looks more interesting to me right now.

Yeah it does look a bit low. I was surprised.

R1291 = 7.6 mega ohms
R1292 = 8.3 mega ohms

Should I order a new ISL6225 as a precaution? (2 weeks delivery usually)

zuldan wrote on 2025-11-12, 21:23:

Yeah it does look a bit low. I was surprised. […]
Show full quote

Yeah it does look a bit low. I was surprised.

R1291 = 7.6 mega ohms
R1292 = 8.3 mega ohms

Should I order a new ISL6225 as a precaution? (2 weeks delivery usually)

Ha, sounds you are stacking up 2006 finest voltage regulators. Not sure what killed the ISL6525, perhaps it also took down the ISL6225. I'll leave it to you, don't want you to spend money you might not want to 😉

I have a couple of things to look at:

• Voltage over C1641 (10nF) during operation - it should be above 1.5V to start (see datasheet threshold on page 3)
• Resistance of R1277 - should be 49.9kOhms

I just read that you measured the rail resistance of FBVDD on TP30, but that one is the feedback network testpoint for VTT, the correct one would be TP31. Can you re-check TP31 to ground?
For VTT, you measured using TP29, but that is the feedback network testpoint for FBVDD. I haven't found a proper testpoint for it, you might just wanna measure any pin of L814 against ground for that.

However, 1.6Ohms of TP30 against ground is concerning. How do R1276 and R1279 look?

tehsiggi wrote on 2025-11-13, 10:19:

Ha, sounds you are stacking up 2006 finest voltage regulators. Not sure what killed the ISL6525, perhaps it also took down the ISL6225. I'll leave it to you, don't want you to spend money you might not want to 😉

I have a big collection now from other cards 😉 New ISL6225 is on it's way!

• Voltage over C1641 (10nF) during operation - it should be above 1.5V to start (see datasheet threshold on page 3) = 0.5v
• Resistance of R1277 - should be 49.9kOhms = 50k ohms
• Re-check TP31 to ground = 340 ohms
• Any pin of L814 against ground = 280k ohms
• R1276 = 490 ohms (490 ohms to ground)
• R1279 = 535 ohms (490 ohms to ground)

zuldan wrote on 2025-11-14, 09:15:

I have a big collection now from other cards ;-) New ISL6225 is on it's way! […]
Show full quote

I have a big collection now from other cards 😉 New ISL6225 is on it's way!

• Voltage over C1641 (10nF) during operation - it should be above 1.5V to start (see datasheet threshold on page 3) = 0.5v
• Resistance of R1277 - should be 49.9kOhms = 50k ohms
• Re-check TP31 to ground = 340 ohms
• Any pin of L814 against ground = 280k ohms
• R1276 = 490 ohms (490 ohms to ground)
• R1279 = 535 ohms (490 ohms to ground)

The attachment C1641.JPG is no longer available

Alright.. so that TP31 measurement gives me hope. 340Ohms for the FBVDD rail is good. And the 280kOhms for the VTT rail is good as well.
So C1641 is the soft start. I wonder, can you get a scope reading of it (continuous trigger, trigger at 0.3V) against ground? I would like to see if the signal is ramping or not.
Perhaps this card starts setting a track record for randomly dead switching regulators..

tehsiggi wrote on 2025-11-14, 14:22:

Alright.. so that TP31 measurement gives me hope. 340Ohms for the FBVDD rail is good. And the 280kOhms for the VTT rail is good as well.
So C1641 is the soft start. I wonder, can you get a scope reading of it (continuous trigger, trigger at 0.3V) against ground? I would like to see if the signal is ramping or not.
Perhaps this card starts setting a track record for randomly dead switching regulators..

Ok here you go. It didn't really capture anything on C1641 but I think that's because I don't have the scope configured correctly? or SS isn't going at all.