PC Graphics

[doanhanam]

I have used water cooling system for the three zones as i circled in red color . When the temperature of the metal back plate approached 34.5 degrees , the system shut  down ! . I know this  vga problem because other parts in my computer works flawlessly . So anyone can identify any zone that need to be cooled ! . I bought a strong fan for extra cooling for the fear that the the temperature sensor . It works little but everything remains the same ! . I attached image here so that any one can comment or help . Thank ahead everybody ! . 

[elstaci]

Need more information like Make & Model of your GPU Card and Make & Model of Water Cooling for your GPU Card.

Seems like the Water Cooling is not installed correctly or something got damaged when you installed the GPU Water cooler on the GPU card.

Just guessing.

[doanhanam]

My vga is PowerColor PCS+ R9 380X MYST Edition as cited in the picture . I makes the water cooler for this vga by myself . And I have identified the three zones and use all water coolers for them . The vga still work when temperature is under 34 degrees . I use the temperature circuit to check the iron plate . i have used thermal pad with 12 W/km for the zone near the ports of vga and this zone was   heat before and now it is coolest place . Everything goes better but the vga still halt the pc ! . I will try to use the best thermal pad with gpu , last night i touched and feel that is extremely hot . But gpu still work to 100 degrees as i have tried but now the thermal tube is out , i will use thermal pad for better reusage ! . No thing damaged at all because the vga still work until temperature is under 34 degrees ! .

[elstaci]

The only thing that comes to my mind is a Thermal sensor in the GPU card is defective and giving out the wrong temperature.

For instance, at 34C when it shuts down, the defective Thermal Sensor could be giving out temps of over 104C thus the GPU Chip shuts down before damage occurs.

But I imagine if it has a defective Thermal Sensor the GPU card should have automatically throttled before reaching 34C.

Or it could be an electronic part that once it starts to heat up it causes the GPU card to shut down the computer.

There is one User here at AMD Forums, Colesdav, that could probably help you find out what is wrong with your GPU card since he is a technician and fixes his own GPU cards if possible.

You could try as a troubleshooting procedure, re-install the GPU original Cowl with Heatsink and see if it continues to shut down at 34C. That would eliminate the CPU water Cooler if it does the same thing.

[doanhanam]

Thank you .

I have just replace the thermal conductive liquid by thermal pad 12 W/km and everything goes happy . I have not tested the temperature that this vga would shut all down pc . But i have experienced a lot that the temperature that makes sensors stop working is quite problematic ! . I have identified the zone near port of vga and gotten them cooled and every thing went better until now . And I have just used another thermal pad 12 W/km for gpu and now thing changes dramatically ! . It works but i stop for the lunch ! . This afternoon , I will install debian for it and check again for sure . Thank for your precious guidance . Have good day ! .

[doanhanam]

Now when everything has been cooled all three zones and with the sensor circuit outside . All the temperature is at room temperature 31 degrees in ho chi minh city . The vga still shutdown the computer after couple minutes ! . I think there is the problem with the thermal sensor and i am sure that replacement must be happened in order to make this vga work correctly . Or i must buy other vga to use ! . I will try Colesdav in order to know more about this ! .

[doanhanam]

Thing changes a lot when i use thermal pad 12 W/Km . I have even used it at the back side of PCB vga so that the temperature could be conducted into the iron plate . And one thing curious happened . When the temperature of the back side of iron plate of vga cpu zone only reach at 42 or 43 degrees , the vga shut down the computer ! . This is the weird thing . If the back side of pcb of gpu zone only needed to be cooled , why producer did not make a cool for it ! . This such weird thing makes my water cooling system more tedious l. Anyone have met this thing before ? . Now , only one zone need to be cooled ! . In foreign countries , 42 or 43 degrees is remarkable in comparison with 25 degrees in room . But nowadays , 34 or 35 degrees in room in vietnam in sai gon is so familiar ! . Thank ahead ! .

[colesdav]

I think this is a review of your GPU - Is this correct?
https://www.overclockersclub.com/reviews/powercolor_pcs_r9_380x_myst/ 

The first thing I would do would be to reassemble the GPU with it's original air cooler and see if I could get it to work properly with that.

One thing that can catch you out is the thickness of thermal pads on the VRAM Memory Chips when you reassemble your GPU.

Those thermal pads often degrade and you can purchase new thermal pads that are usually meant for use with AIO GPU cooling solutions.

In other words you remove the original air cooler. Use an AIO closed loop cooler directly on the GPU.
Then you use small thermal pads, heatsink, and small aircooled heatsinks on the VRAM, cooled with an additional VRAM fan. 

They can be thicker than the original VRAM thermal pads which can mean insufficient contact pressure with GPU heatsink when the cooler is reassembled.

That would cause exact symptoms you describe - the GPU will power up and then cut out because the GPU Temp is too high since cooler is not contacting with the GPU properly.

Interestingly you have not identified the VRAM chips on your GPU in need of cooling.
Additionally in the teardown of the GPU in that review, it looks like they are not cooled by anything other than airflow from the fans on the GPU cooler.

Thanks.

[doanhanam]

Thank for your reply ,

The problem is that I use water cooling system to make all the corner of

the vga to work with temperature at the room temperature in hcm city 32

degrees . But even then , the shutdown of the pc by the vga still happen !

.

In order to make more cool for the vram and other not smooth surface , i

use strong fan . But thing is still exact the same and everything shutdown

at the same temperature . This means , fan contributes no effect at all !

. I think that the thermal sensors has failure . And with the basic

electric equipment tool but i am software not hardware expert , can you

show me how to check if thermal sensors condition ? . Thank a lot .

[colesdav]

There will be a thermal sensor on the GPU.
If that is tripping out the PC then it means that there is insufficient contact from the heatsink-thermal paste-GPU die surface.
If there is another thermal sensor on the PCB it is likely a Negative Temperature Coefficient Thermistor (Resistor). 
http://www.resistorguide.com/ntc-thermistor/ 
Failure Mechanisms of Thermistors 
You will have to find that component on the PCB and test it.
Once identified you will have to source a replacement part.

[doanhanam]

I am sure that the contact between the heatsink thermal gpu surface will

never happen . I use the water cooling and the water cooling device

surface covers completely and totally the surface of gpu die .I use thermal

conductive pad 12 W m2/m , which is the highest conductive .

The rest is the thermal sensor as you mention . I think that the

sensors are like

https://industrial.panasonic.com/ww/products/thermal-solutions/ntc-thermistor-chip-type/ntc-thermistor

because

there is no other thermal sensor matches . There are a lot NTC

themistors . But the failure NTC themistor is only for the old age because

the vga board still work and after couple minutes in OS , ít shut down . I

read it from https://www.sensorsci.com/failure-mechanisms-of-thermistors

. So how can i find out the failure NTC themistor ? . But whât we find

here is the faster declination ò NTC themistor than normal so that circuit

is short and shut down the pc . It sound very dificult to happen ! . With

the forest of NTC , it is not easy to check ! .

[doanhanam]

You mean the thermal sensors are in fact is only NTC thermistor only ? .

And so there is no thing so called "thermal sensor " as it appears in the

software view ? . Thank your for your guidance . I am totally new to this

field . Happy day .

[colesdav]

Download and run TechPowerUp GPUz.

It should give you an idea of how many temperature sensors are on your GPU.

Here are some screenshots from an R9 390X.

GPU Temperature is very likely a Temp Sensor on the GPU die.

VRM Temperature 1 and Temperature 2 are likely placed somewhere near the VRM on the board. 

There may be some clues on any text on the motherboard as well.

[doanhanam]

In the psensor in debian 10 , there are 3 temperature . one is 42 degrees , -127 , 0 degree . I know that 42 degrees C is the core . Other two sensors are not correct and i have never seen they indicated any other degree in temperature . I can not copy the screen the because it shut down . 

The mainboard is GA-EP45T-UD3LR  https://www.gigabyte.com/Motherboard/GA-EP45T-UD3LR-rev-13#ov   and i am sure that nothing wrong with the mainboard . My cpu is E8400 OC to 4.2 ghz and ram 1600 corsair to 2000 mhz  with 16 MB capacity . 

The temperature is 32 degrees in hcm city and that is the room temperature . About 10 higher degrees C in comparison with your country .

Thanks and have nice day .

[colesdav]

RE: In the psensor in debian 10 , there are 3 temperature . one is 42 degrees , -127 , 0 degree

I would not guarantee anything w.r.t. AMD GPU hardware getting reported correctly in Linux.
Try Windows 10.

[doanhanam]

I windows 10 , it takes more time to boot completely into OS than in debian 10 . So i have the time to view the sensor temperature . In windows 10 , i can not view gpuz  before it shuts down the machine  before you click gpuz ! . For the copyright , I only use linux and windows 10 is unregistered . I know that there is only effective driver of this vga in windows only . I have spent a lot of time to install driver for many linux branches and all stopped hopeless ! . Only windows , the driver is recognized and i can use it to render image in blender . I follow the copyright as the truth and serious attitude towards the career . No cracked version for commercial purpose . 

[colesdav]

Can you get hold of another GPU just to make sure there is nothing else wrong on your PC System?

[doanhanam]

This mainboard is tested carefully . This is sure because i test it carefully before buy it . So if now i use another vga , this is the same as the test made with the seller with his vga . So i am sure that the mobo E8400 + ga-ep45t-ud3lr + 16 gb corsairs 1600 mhz ddr3 works flawlessly . The only problem is from the vga only ! . I am sure and the seller expert on pc hardware is sure too . Moreover that is so simple test ! . I follow the way you suggest . That is direct confront with the problem from vga ! . Good day ! .

[colesdav]

Have you checked all VRM Mosfets are OK and not blown?

[doanhanam]

For the vrm mosfets and vram , I use very strong fan for it . But when use the fan and no use ,very thing is the same .  degree to shutdown the pc are the same ! . and even the time before shut down is the same ! . So with this vga , no problem at all with vrm mosfets and vram .

[colesdav]

Yes I know you use a strong fan, and you say the shutdown temperature fan or no fan is the same. 

But if the Thermistors are broken/degraded such that they have an incorrect offset resistance, the output of protection circuits connected to them will interpret a higher temperature and could shut the GPU down.

You need to identify where any thermistors are located on the GPU PCB.

Quick way to find out is using TechPowerUP GPUz.
If you get any temperatures reported apart from GPU Temperature, that means there are additional thermistors on the PCB.

[doanhanam]

So , you mean the two zones with -127 degrees C and 0 degrees . They rests always at that degree . -127 or 0 degree this indicate the malfunction of the thermistors . Yes , now i know your meaning . And use VOM to check the  thermistors in the forest of thermistors ! . And i will use the hair drying machine to test . That is really the bright spark at the bottom of the tunnel ! . Thank your guidance . Luckily i learn a lot from this guidance . I still do not know what is thermistor before i get guidance from you ! . Thank a lot . I will try to check for the forest of thermistors . 

[colesdav]

RE: And use VOM to check the  thermistors in the forest of thermistors !

By VOM you mean Voltage-Ohm-Meter?

There should be only two thermistors on the PCB based on what you say. 
The R9 390x example I show you above indicates only two thermistors on the PCB located near the VRMs.

The rest should be normal surface mount resistors, or surface mount capacitors.

[doanhanam]

Can you guide more specificly in askimage+2.png  ? . I can only see the forest of thermistors in front and back pcb ! . thank you so much .

[colesdav]

I am sorry I cannot do more.
I would need to have the GPU to find the 2 thermistors.

[doanhanam]

Yes , thank you for your precious guidance . I find your guidance  helpful . I will try to find .

[doanhanam]

I am learning electric component with online material . I heard you say about the mainboard . This makes me to remember that with the another machine that i have , when i install this r9 380 into the mainboard . It works flawlessly and i even use benchmark software to view the change of of the temperature of gpu upto 100 degrees C but it does not shut down then . Maybe the present way is reinstall to that machine and use it . I change because that machine use ddr2 and just with 4 GB ram 1111 mhz . I fear with that ram is not enogh for image and video processing . But that is the easiest way right now . I learn to repair this vga when i have accumulate enough knowledge ! . When you say about thermal component , I have no idea ! .But now i feel better ! . Thank you . I learn from university the Software , Computer Science . This is the last field in IT that i think i must complete my knowledge ! . Thank a lot for your guidance ! .

[doanhanam]

This is two thermistors has lowest value . Just 6 ohm . Picture circle . But the seller of electric component says that it is the zero thermistor . She does not have it ! . Any recomendation for me ! . Nice day ! .

[colesdav]

I would have expected them to be near the VRMs.
They should be positioned as close as possible to the VRMs to measure the local temperature of the VRMs on the other end of the PCB.
I thoml those are just resistord, but I cannot read the text on the PCB shot clearly.

[doanhanam]

Ah , I have just scanned the back side of the pcb . And jave them online for very detail info . From website  http       makecool       .   ddns   . net   / forcolesdav/                        front.bmp         .

http       makecool       .   ddns   . net   / forcolesdav/                        back .bmp 

Please erase the space . I do so for easy .

      I see the Mostfet and ductor and capacitor . They makes VRM . And near that i see the themister .. Certainly it is resister . In fact no "thermal sensor " as i think because it does not exist . The board uses themister to be the sensor ! . I have checked forest and circle as above picture  . But now the problem is i can not buy that thesistor because the seller think that is 0 ohm ! . In the surface , big O makes her to think that ! .