JTAG Strange overheating problem with Falcon JTAG

[amesie2]

I have a JTAG Falcon that is on 16747 and running FSD 3 775. I think the sensor for the ram and GPU is messed up since the ram stays at 33.8 F and the GPU runs about 143 F then all of a sudden the temp will rise on the ram well over 200 F and the GPU will go up to 180 F or so then it shuts down with 2 flashing lights.

This happens at various times but will generally run between 10-20 minutes before it happens. I have cleaned both CPU and GPU and replaced with AS5 and it has had the X-clamps removed.

I am attaching a picture to show also that there is no colored bar for the GPU or ram, the GPU temp will go up and down but the ram will stay at 33.8 F until the rapid climb in temp and then the shutdown (at this point the colored bars do work). This happens pretty quick where I can't snap a picture of it.



I figured that maybe I had a bad NAND so I rebuilt it and flashed it again, I also reinstalled FSD with the same results. I JATAGGED this a while ago and put it on the shelf and didn't mess with it much until now and discovered this problem.

Does anyone have any thoughts on this? CRAZY

 

[AllyNerd]

Well before the gpu shoots up is the CPU temp high also? Have u edited dash launch temps?

 

[amesie2]

No the CPU stays pretty close to the 120 F mark plus or minus a couple of degrees. I never changed the temps in dash launch. I think it is strange that there is no bar within the GPU or Ram reading. The temps on the GPU and Ram just shoot up in a matter of a couple of seconds.

 

[AllyNerd]

You sure your heat sync is on right? Also put the xclamp back on they r perfectly fine to use. Maybe take the top off and actually feel how hot it's getting. I def have never seen that happen before. If it's overly hot to the touch then something's off... But I would def edit target temps anyways

 

[amesie2]

I do have the top off and when it spikes and shuts down the temp is warm to the touch but not overly hot. I think there is something wrong with the sensor itself as the temp stays at 33.8 F on the ram then at random times it spikes to over 200 so does GPU but that one usually runs at about 140-145 F the whole time until the ram starts going up so does the GPU then shut down.
I have pulled the heatsink off the GPU and it looks to be making good contact, I also have the fans running on high speed.

What would you recommend for target temps ?

 

[Taijigamer2]

U need to replace heatsink paste with quality paste if u haven't done so already. Then make sure u'r heatsinks are on properly as has been suggested, sounds like classic overheat due to inefficient heat transfer.

 

[amesie2]

I have replaced thermal paste with AS5 already and made sure the heat sink is in good contact. I really think this is an issue with the temp sensors since the ram stays at 33.8 F for between 10-20 minutes or sometimes longer then it will spike very quickly to over 200 F as will the GPU which runs at around 140-145 F then will spike at the same time the ram does.
You can also see from the picture there is no green, yellow or red line in the GPU or Ram temp reading.

 

[Krafter]

U need to replace heatsink paste with quality paste if u haven't done so already. Then make sure u'r heatsinks are on properly as has been suggested, sounds like classic overheat due to inefficient heat transfer.

^^^^^^^^
What he said. It's really easy to disturb the thermal compound when pulling out the mobo. The compound becomes brital over time so any flexing of the board can break the seal between the heatsync and the chip.

 

[amesie2]

I have replaced the thermal paste and checked that the heat sink is seating properly as stated a couple of times in previous responses. There seems to be an issue with the sensor not picking up properly, where is it located and can it be disabled or can I change the SMC_CONFIG and change thresh holds?

 

[jligg]

I have had a jasper do this exact thing. I would run it for a while and it would spike and shutdown, then I checked heatsink and it was warm and so was both chips but not hot by any means. I ended up cutting the trace to the sensor and its been fine ever since..heres a picture of the trace. But you better be sure its definitely not overheating before cutting it. Do at your own risk

 

[amesie2]

I have checked the heatsink after it shutdown and it was not all that warm to the touch so I'm guessing it is something with the sensor being messed up as the ram or mem reading stays at 33.8 F then at random it will spike to over 200 F as will the GPU and shut down.
The spot where you show to cut the trace is that for the CPU or does it handle both CPU and GPU? Thanks

 

[COMPLETE_TOOL]

Try a reflow or reball if not replace GPU if that don't help cut the F ing temp trace lol

 

[amesie2]

Guess I'll try to cut the trace.

Thanks for the input.

 

[AllyNerd]

I don't know if it's safe but maybe u can even edit the shutdown temps of the ram and gpu and put them up high so even when it spikes it still won't shutoff maybe just a guess

 

[amesie2]

Allynerd, maybe that would be a better option since I know it isn't actually hitting those temps. Like I said it runs fine from anywhere 10-20 minutes (sometimes longer) and the Ram doesn't even show higher than 33.8 F then will randomly shoot to over 200 F in less than 5 seconds same as GPU which hovers around 140-145 F until ram spikes.

 

[amesie2]

Well I decided to try a reflow as suggested by complete_tool and the box was on running for over 4 hours without shutting down. I did set the SMC_Config threshold temps to 60c across the board, the fans do spin up to around 70% I would say and the temps run around 148 F for GPU and 130 for Ram.

Thanks for the input guys!

 

[Krafter]

Well I decided to try a reflow as suggested by complete_tool and the box was on running for over 4 hours without shutting down. I did set the SMC_Config threshold temps to 60c across the board, the fans do spin up to around 70% I would say and the temps run around 148 F for GPU and 130 for Ram.

Thanks for the input guys!

So it's fixed? Much safer solution than cutting the trace or adjusting the temps in the SMC config.