Dell Precision 7710/7720 weird CPU temperature readings. Doing 40c to 99c with fresh paste

[SuperMG]

Hello.

 

I bought 2 Dell precision 7720.

 

One was with a Quadro P3000, working fine, until I put the RTX 4070 MXM. I modded the heatsink so it has a good contact.

 

The CPU temperature reading are WTF. When the CPU or GPU is doing something, the CPU is pegged at 99-100C. With PTM 7950 and cleaned vents...

 

When I insert back the P3000, same issue.

 

So I bought another 7720 with M1200.

 

M1200 and CPU works fine.

 

Then I reuse the modified heatsink with the RTX 4070 and I get the same problem, CPU temperature readings from 40c to 99c in an instant... Does random spikes at 70-80c in idle too...

 

Windows 10 and 11 fresh copy.

 

When I remove the MXM GPU, the CPU no longer reaches 99c.

 

Is this a heatsink problem or is this an RTX 4070 MXM problem?

 

The RTX 4070 stays cool at idle and no random spikes.

 

Did anyone get this issue before?

Edited yesterday at 03:23 PM by SuperMG

[Chalybion]

Since the issue persists with the P3000 reinstalled, the problem is with your modified heat sink - either the CPU is no longer making contact, or the fluid flow/wicking cycle in the heatpipes is blocked.  But all three heatpipes would have to be damaged to get 99C immediately, and even then there is some heat transfer by conduction.  So my guess is that the CPU is not making good contact -  does the PTM 7950 show contact w/ the heatsink after the heatsink is removed?

Can you describe the modifications that you made?

[SuperMG]

35 minutes ago, Chalybion said:

Since the issue persists with the P3000 reinstalled, the problem is with your modified heat sink - either the CPU is no longer making contact, or the fluid flow/wicking cycle in the heatpipes is blocked.  But all three heatpipes would have to be damaged to get 99C immediately, and even then there is some heat transfer by conduction.  So my guess is that the CPU is not making good contact -  does the PTM 7950 show contact w/ the heatsink after the heatsink is removed?

Can you describe the modifications that you made?

Yes it does show contact.

 

My modification is just to add one piece of 4cmx4cmx1.2mm copper plate on the GPU's section.

 

I did the same modification on the newer heatsink, same issue I think?

 

There's good amount of pressure on the CPU, I can't do that far.

 

This happens on two different 7720 laptops. 6920HQ and 7820HQ

[Chalybion]

That still sounds like a contact problem.  The heatsink isn't mounted with much pressure, and all of the pressure comes from the springs around each of the four mounting screws.  So if the GPU side is raised by 1.2 mm, then those springs have to push the heatpipes downward by 1.2 mm, and that much less force is applied to the CPU.

Is it possible to do away with the GPU copper plate by using thinner thermal pads?  Or shim the CPU heatsink an equal amount?

[SuperMG]

1 hour ago, Chalybion said:

That still sounds like a contact problem.  The heatsink isn't mounted with much pressure, and all of the pressure comes from the springs around each of the four mounting screws.  So if the GPU side is raised by 1.2 mm, then those springs have to push the heatpipes downward by 1.2 mm, and that much less force is applied to the CPU.

Is it possible to do away with the GPU copper plate by using thinner thermal pads?  Or shim the CPU heatsink an equal amount?

I don't know, the CPU doesn't seem hot. I feel the wind on my hand, it's cold. If it doesn't do good contact, how is it possible for the CPU to cool down? 

 

Visually with the other heatsink with the copper plate, I see that they're aligned good, like on line between the CPU and GPU parts. Like there's some sort of an elevation gap

 

Let me remove the copper plate on the other heatsink and test it with P3000

[SuperMG]

I removed the copper plate and used the previous Dell Precision 7720 where the Quadro P3000 and 4070 with the custom heatsink had weird CPU temp spikes issues.

 

I removed the copper and added the M1200 because the P3000 was somewhere else.

 

And now the CPU doesn't act up! It's not doing 40c to 99c. Or doing 99c at 35W! I'll test further more.

 

Edit: no problems with unmodified heatsink (removed copper plate) on both Dell Precisions with P3000 and M1200.

Edited yesterday at 06:56 PM by SuperMG

[SuperMG]

I still get the same issue with the copper plate installed. I installed it correctly before it wasn't installed correctly. I had to cut the plate so it can fit flat. VRAM pads are ok now. Still peaks at 99c when being at 35W

 

I touch the CPU side, it's not even hot, the vents have cold air.

 

The GPU temps don't even go at 99c. Max is like 80c but when at 115W 100%. The GPU temps are like 27-30c idle. No weird spikes on the GPU side.

 

When I insert an USB key, CPU spikes at 99c...

 

I'll try 0.8mm thick and 1mm thick copper plates in 2 weeks. 1.2mm is probably too much. Because I have the NV-A heatsink and it doesn't even touch the "tall" R22 inductor, like there's a 0.2-0.3mm gap "visually". On the NV-B heatsink (old 7720), 0mm gap. Need to cut impossible corner.

 

I want to try undervolting at -100mV but it's locked of course and I have the latest bios.

 

 

FurMark 1.37: RTX 4070 MXM, 116W max, 79C max, 205 fps AVG 1080p

[darnok44]

Hi. Let me get this straight, because maybe I don't get something. It's OK with stock heatsink? But when you add 1.2mm plate on the heatsink in the place of gpu die and you install it cpu temps are mess? 

I quickly checked how this heatsink looks like and it's two plates basically with shared heatpipes. So if you add at one side/gpu 1.2mm height... I don't see how could you clamp it down on the cpu side? Logically thinking cpu side/ die plate/ has to be raised that same amount - 1.2mm to maintain that same level a flatness of the whole heatsink. If ou added height at gpu side and screwed down heatsink everywhere/ including cpu side/ it had to bend the heatsink (I guess heatpipes itself between gpu and cpu). There is no way that cpu side is sitting flat on the cpu itself. Maybe is touching in some place but surely it's not sitting on the whole die. That's just my guess. Check everything closely again. There has to be something bent. Maybe photo would share some light and give clues.

[Chalybion]

3 hours ago, darnok44 said:

Logically thinking cpu side/ die plate/ has to be raised that same amount - 1.2mm to maintain that same level a flatness of the whole heatsink.

Exactly this, and need to add thicker thermal pads for the CPU VRM to compensate.

 

4 hours ago, SuperMG said:

I touch the CPU side, it's not even hot, the vents have cold air.

The vents are blowing cold air because the heatsink is not contacting the CPU well, so heat is not being transferred to the heatsink

 

4 hours ago, SuperMG said:

FurMark 1.37: RTX 4070 MXM, 116W max, 79C max, 205 fps AVG 1080p

Really cool that the 4070 is working!

 

4 hours ago, SuperMG said:

I want to try undervolting at -100mV but it's locked of course and I have the latest bios.

The last BIOS which allows undervolting on the 7720 is 1.15.1

The procedure to downgrade BIOS is here, and moderately complicated:  https://github.com/vuquangtrong/Dell-PFS-BIOS-Assembler

The basic steps are:

- Download 1.15.1 and the newest BIOS 1.40.0

- Extract the contents of both BIOSes (this will include packages for the dell BIOS, as well as Intel Management Engine, etc.

- Replace the 1.40.0 package with the 1.15.1 package

- Recompile the 1.40.0 BIOS

- The version check will pass, and the 1.15.1 BIOS will install

 

Also, keep an original version of BIOS 1.40.0 on a USB drive, and learn how to recover from USB just in case the BIOS update goes wrong