• Welcome to TechPowerUp Forums, Guest! Please check out our forum guidelines for info related to our community.

Repair R9 390 no voltage

Joined
Oct 13, 2021
Messages
34 (0.04/day)
Hi there!
Im trying to fix this sapphire r9 390 8gb,
It was working card and I leave PC overnight on, the next day the card has black screen(turn off screen).
Then I realized that memory controler drmos was toasted with fuse 12v pcie slot. Well now I have
12v from replaced fuse,
6.10v for vcc drmos's
5v(from linear regulator)
1.8v from apl5930
0.95v for display
12v from 8pin(and on fuses)
But I dont have PWM that comes from IR3567B controller. I try to measure controller but on pwm(output) I have 1.6v DC, not a signal, that goes to IR3553.On the microcontroller I have his vcc 3.2v and Enable I think 3.2v either.
So I dont have voltages for memory and gpu.
On the picture some drmos is missing(one on memory and two on cpu power supply) . Can somebody help me what to do next and how to get pwm signal from microcontroller.

Anyone to help please?
 

Attachments

  • IMG_20211125_205759.jpg
    IMG_20211125_205759.jpg
    1.2 MB · Views: 364
Joined
May 8, 2020
Messages
578 (0.41/day)
System Name Mini efficient rig.
Processor R9 3900, @4ghz -0.05v offset. 110W peak.
Motherboard Gigabyte B450M DS3H, bios f41 pcie 4.0 unlocked.
Cooling some server blower @1500rpm
Memory 2x16GB oem Samsung D-Die. 3200MHz
Video Card(s) RX 6600 Pulse w/conductonaut @65C hotspot
Storage 1x 128gb nvme Samsung 950 Pro - 4x 1tb sata Hitachi 2.5" hdds
Display(s) Samsung C24RG50FQI
Case Jonsbo C2 (almost itx sized)
Audio Device(s) integrated Realtek crap
Power Supply Seasonic SSR-750FX
Mouse Logitech G502
Keyboard Redragon K539 brown switches
Software Windows 7 Ultimate SP1 + Windows 10 21H2 LTSC (patched).
Benchmark Scores Cinebench: R15 3050 pts, R20 7000 pts, R23 17800 pts, r2024 1050 pts.
Does the system power on with the card plugged in? (in other words, does it generate any type of short which triggers SCP?). Replacing just the fuses is never a good cure, as the fuse gets burned for a good reason, that is to avoid further damage to the card components). Do you have oscilloscope at hand to check how VRM behaves? Most probably the vrm controller is toasted, swap it for a known good one. And measure the vram continuity (should be 500ohm+) and the gpu (4-10 ohm, dont know good value for Hawaii, but should be 4ohm or more). Also measure the continuity on the pcie slot, specifically the signalling lanes , they should all be of same value, about 400ohm (again i dont know about the known good value for a hawaii chip), and if u find a pin with infinite continuity or that significantly differs from the rest, then u got gpu toast (ie card not worth repairing). Here is pinout, u need to check the continuity between gpu data pins (pink and purple on the sheme) and the ground (grey). https://tweakers.net/i/5GdOKzVcUfA7...age/kZ7jY09LC3jiZBDkudT1SMza.png?f=user_large

As with any repair, start with basic stuff, and make ur way to the top, checking the continuity and voltages, replacing the minor parts if necessary. But once u hit gpu or vram related problems, just recycle the card, as it wont be a profitable repair (unless u want to perform some microsoldering practice lol).
 
Last edited:
Joined
Oct 13, 2021
Messages
34 (0.04/day)
Does the system power on with the card plugged in? (in other words, does it generate any type of short which triggers SCP?). Replacing just the fuses is never a good cure, as the fuse gets burned for a good reason, that is to avoid further damage to the card components). Do you have oscilloscope at hand to check how VRM behaves? Most probably the vrm controller is toasted, swap it for a known good one. And measure the vram continuity (should be 500ohm+) and the gpu (4-10 ohm, dont know good value for Hawaii, but should be 4ohm or more). Also measure the continuity on the pcie slot, specifically the signalling lanes , they should all be of same value, about 400ohm (again i dont know about the known good value for a hawaii chip), and if u find a pin with infinite continuity or that significantly differs from the rest, then u got gpu toast (ie card not worth repairing). Here is pinout, u need to check the continuity between gpu data pins (pink and purple on the sheme) and the ground (grey). https://tweakers.net/i/5GdOKzVcUfA7...age/kZ7jY09LC3jiZBDkudT1SMza.png?f=user_large

As with any repair, start with basic stuff, and make ur way to the top, checking the continuity and voltages, replacing the minor parts if necessary. But once u hit gpu or vram related problems, just recycle the card, as it wont be a profitable repair (unless u want to perform some microsoldering practice lol).
Thanks for your reply. I will check data lines and post again. I have oscilloscope (but yesterday for some reason I try to measure pwm on old motherboard and Im only saw two lines, not a signal,so Im not sure if it working properly.. I'll find it out soon) . ..
Well, I measured resistance first and then I was removed drmos and short dissapear. I put new drmos, and measure about 70Ω on memory controller(and other measures was ok all power supply lines), but after try on PC to see whats happening, the voltage is missing. Only have this voltages that I wrote up. So my drmos, controller and pwm controller has the necessary voltage for their vcc, but pwm signal is missing( at drmos I can see 1.6v DC voltage, with oscilloscope and multi meter but not a signal one).
P. S. The other resistance that I measured seems fine, this is the second card that Im working on so with the first one has the same resistsnce. But I sell the first card that I fixed and worked properly.
 
Joined
May 8, 2020
Messages
578 (0.41/day)
System Name Mini efficient rig.
Processor R9 3900, @4ghz -0.05v offset. 110W peak.
Motherboard Gigabyte B450M DS3H, bios f41 pcie 4.0 unlocked.
Cooling some server blower @1500rpm
Memory 2x16GB oem Samsung D-Die. 3200MHz
Video Card(s) RX 6600 Pulse w/conductonaut @65C hotspot
Storage 1x 128gb nvme Samsung 950 Pro - 4x 1tb sata Hitachi 2.5" hdds
Display(s) Samsung C24RG50FQI
Case Jonsbo C2 (almost itx sized)
Audio Device(s) integrated Realtek crap
Power Supply Seasonic SSR-750FX
Mouse Logitech G502
Keyboard Redragon K539 brown switches
Software Windows 7 Ultimate SP1 + Windows 10 21H2 LTSC (patched).
Benchmark Scores Cinebench: R15 3050 pts, R20 7000 pts, R23 17800 pts, r2024 1050 pts.
Thanks for your reply. I will check data lines and post again. I have oscilloscope (but yesterday for some reason I try to measure pwm on old motherboard and Im only saw two lines, not a signal,so Im not sure if it working properly.. I'll find it out soon) . ..
Well, I measured resistance first and then I was removed drmos and short dissapear. I put new drmos, and measure about 70Ω on memory controller(and other measures was ok all power supply lines), but after try on PC to see whats happening, the voltage is missing. Only have this voltages that I wrote up. So my drmos, controller and pwm controller has the necessary voltage for their vcc, but pwm signal is missing( at drmos I can see 1.6v DC voltage, with oscilloscope and multi meter but not a signal one).
P. S. The other resistance that I measured seems fine, this is the second card that Im working on so with the first one has the same resistsnce. But I sell the first card that I fixed and worked properly.
You probably need to replace the damaged mosfets first (there were 3 drmos units which were faulty and you removed them, if i understood correctly and based on your photo) , if the vrm fase count is incomplete then the controller may or may not work properly. Also replace the controller itself with a new one. You measured 70ohm on memory? (between ground and vmem)??? thats too low, (but too high for a short), which probably indicates bad memory chips.
 
Joined
Oct 13, 2021
Messages
34 (0.04/day)
You probably need to replace the damaged mosfets first (there were 3 drmos units which were faulty and you removed them, if i understood correctly and based on your photo) , if the vrm fase count is incomplete then the controller may or may not work properly. Also replace the controller itself with a new one. You measured 70ohm on memory? (between ground and vmem)??? thats too low, (but too high for a short), which probably indicates bad memory chips.
So I must repair all phases and then check. On another board I was removed toasted drmos and card was working properly without one phase. .
Yes, 70Ω between gnd and mem controller(sw pin from drmos).
I saw video and repair of same card and he has about 60Ω-70Ω so I guess this is not a problem. I must search for a donor card because here this chip is very expensive and must wait some weeks, from china I must wait over a month.. So...
Im trying to figure out where is the problem to know what to order exactly
 
Joined
May 8, 2020
Messages
578 (0.41/day)
System Name Mini efficient rig.
Processor R9 3900, @4ghz -0.05v offset. 110W peak.
Motherboard Gigabyte B450M DS3H, bios f41 pcie 4.0 unlocked.
Cooling some server blower @1500rpm
Memory 2x16GB oem Samsung D-Die. 3200MHz
Video Card(s) RX 6600 Pulse w/conductonaut @65C hotspot
Storage 1x 128gb nvme Samsung 950 Pro - 4x 1tb sata Hitachi 2.5" hdds
Display(s) Samsung C24RG50FQI
Case Jonsbo C2 (almost itx sized)
Audio Device(s) integrated Realtek crap
Power Supply Seasonic SSR-750FX
Mouse Logitech G502
Keyboard Redragon K539 brown switches
Software Windows 7 Ultimate SP1 + Windows 10 21H2 LTSC (patched).
Benchmark Scores Cinebench: R15 3050 pts, R20 7000 pts, R23 17800 pts, r2024 1050 pts.
really not sure if 70 ohm is a good reading for a card that has 16 chips of gddr5 memory. A 580 8g that i was working with a couple of weeks ago had ~600 ohms. as i said before, replace all of the minor parts, double check the readings (with the component on and off the board). I also saw a similar post on another forum, appears to be a similar card https://hardforum.com/threads/sapphire-r9-390-8gb-issue.2015390/. and he says that 207ohm is what it reads on vmem, which is much higher than 70
 
Joined
Oct 13, 2021
Messages
34 (0.04/day)
I will check it tomorrow, and I think my replaced drmos is blown again, I guess that 70Ω is caused that. I will check it again tomorrow and post some photos
 
Top