Uneven CPU thermals!

Did a Cinebench R23 to compare to others in this thread.

13582 out of a 7940HS. Seriously disappointing.

2 Likes

Uneven is true, lol, just find out mine is the same.
Mine is like 45w max and 15515 R23 score (7940)

1 Like

Well, that doesn’t look too bad… I mean in the delta’s. Your cores all get a bit steamy though.

1 Like

During the benchmark, all cores should use as much power as they can until thermal throttling starts at 100°C. So some cores sitting at that temperature is good and expected, but some showing the delta of 20°C or the like is bad, as this means the cooling plate sits at 70° or some such (less than the coolest core), but the hot cores can’t transfer their heat to the cooler.
If all cores are below 100°C and the power usage is below 50W, there may be some power restrictions at play. Testing with the 180W charger is recommended.

Well I was hoping I didn’t have this issue but I’m not so sure. Cinebench R23 score is rather low.

Not sure about the temperatures and power readings. The spread is not great but the package power being at 55.8 max and hovering around 53 when on full load with Cinebench sounds correct…?

Power mode is set to performance. I toyed for a few days with undervolting but I couldn’t lower it in any significant way without causing instability and I couldn’t find a proper way to apply it permanently or only once at startup.

Is it worth it to contact support about it? Reading the thread makes it sound like it’s a gamble whether a replacement would fix it or I’d be asked to send the whole thing in.

1 Like

For the 7940hs this is definitly to low. The Package Power looks okay, but but not the Score. But its alot better than many of us had. The 7940 should sit above 16k in this Benchmark. It’s either on you to opt for a Replacement and if Framework will perform the Replacement.

1 Like

Mmm. I guess I’ll wait to see if Framework comes up with an official response. I’m not currently in dire need of extra performance but if it’s faulty it’d be nice to have a permanent fix. Maybe I’ll do the PTM7950 switch once the LM replacement guide is out, I’m not super confident I can do a good job without a guide. Thank you and everyone else here for sharing your findings.

1 Like

I don’t think a replacement really solves the core problem, you basically have to receive a lottery winning thermal solution or not.
In my opinion there are 2 issues; one, the plate that’s between the die and vapor chamber is sometimes installed with too much thermal resistance (can only speculate on the cause). Two, insufficient liquid metal to account for “out flow” degrading the thermal performance over time.
One can be solved with your own paste or ptm7950, the other needs to be fixed by frameworks supplier, coolermaster.
That is an expensive fix as it would entail some type of replacement or repair.

2 Likes

I wonder if the core deltas are partly an AMD problem? It seems like even after a lot of effort, the overall performance scores and package wattage can be brought up, but the chips still show large core-to-core temp variation. I wonder if it’s partly a silicon lottery type of thing as well with these chips from AMD?

1 Like

It is a good question, I wanted to look into it more but seems the people with more knowledge are not on this forum, as my question only appealed to crickets.

Looking at the layout, I would say, it depends…

see: amd ryzen 9 7940hs cpu core diagram

1 Like

Got my replacement mainboard. I’ll install it this weekend and do some tests. Should be fun. Gives me an excuse to tinker.

2 Likes

The worst part of all this is that I already got a replacement mainboard for an unrelated USB port issue (which wasn’t fixed by the replacement) so now I’m going to have replaced the mainboard twice in the same laptop through warranty. I’ll be surprised if I’m not outright blacklisted as a customer for this.

It is not your Fault! They let it slip. Sorry to say so. I know it’s outrageous.

I just installed my new heatsink with the possibly, almost sure fake PTM7950.

First test, 10 minutes run:

Second test, Single run:

I installed this heatsink without any modifications.
I’ll keep an eye on the performance.

3 Likes

Well as you did Record and also had HWInfo Running i suspect you lost about 200-300 Points. Great Result nevertheless. Maybe you scored a Goal on the Heatsink Lottery :wink: (Also another Core is hotter than Standard :stuck_out_tongue_winking_eye: )

Where did you get it from? From what I’ve heard (via LTT), it can actually be quite difficult to find the legitimate product, which is why they added it to their store after doing the leg work.

This is the one I bought.
# Thermal Conductive 8.5W Pad -Honeywell PTM7950 Phase Change Silicone Pad Sheet Laptop CPU GPU Silicone Grease Pad

I’m not sure if this is the real stuff, but, so far it works quite nicely.

1 Like

What’s strange, is that now code #5 is hot, instead of #4.

Re:

@BigT with the new heatsink it changed from #4 to #5.
Heatsink 1 and 2 resulted in core #4 being the hot one, while heatsink #3 moved from core 4 to core 5.

2 Likes

Ah, well then in your case at least, it sure doesn’t seem to be the processor. Thanks for that detail.

It can not be AMD the Troiblemaker at that Point. Any other Review/Benchmark i saw from the 7840/7940 had pretty even Temps with Deviation of 2-5K and all of them did sit well below the 100C even with 60w. Also the Heatsink of the Framework is pretty beefy it should handle it much better. Vapor Chamber + 3 pretty Big Heatpipes for the CPU Alone. The Air coming out of the Heatpipe is also not hot enough for the TDP and Temps its running. Thats an Indicator of the High Temperature Resistance.