My gpu got repaired

[NAM3ofVIOLATION]

I just got my gpu repaired today. It is a Sapphire Nitro R7 370 and The repairer told me that this gpu had a shorted ic chip which he repaired. When I bought this card it ran well but after a week it stopped giving display. At that time my pc had a bad lenovo stock psu and my home electricity also had power surges (high voltages) at that time. But now I bought new pc which has Ryzen 3 apu with EVGA 500w new psu. Should I use this gpu as that repairer told me it will not be repaired again if it get shorted this time. Or should I just sell it? Because right now it is running pretty well without any problems and temps are 72°C at max load. GPUs here are very expensive and i barely get to get my hands on this one. Please guide me in this process. I will also share furmark results if you need to guide me.

 

[silentbogo]

It depends on how "blown" that phase was and what they did.
If it was a simple case of blown mosfets - no problem whatsoever. I had an MSI 270X in my workshop with their infamous DirectFETs, and had to fix those 3 times before GPU died.
But if you had PCB damage on that phase and they had to physically disable it and drill the PCB to remove short - then it's a death sentence. There's only so many VRM phases you can remove, and without that phase all others have to work a bit harder (e.g. run hotter and have smaller lifespan as well). Basically your card will work, but you should absolutely avoid any overclocking and I'd also recommend lowering power limit to ~80%.

 

[The King]

AMD classifies Furmark has a power virus.

AMD has replied that FurMark’s slowdown is really intentional because softwares such as FurMark damage graphics cards and do not represent real world applications.

FurMark Slowdown by Catalyst Graphics Drivers is INtentional! | Geeks3D

FurMark Slowdown by Catalyst Graphics Drivers is INtentional!.

www.geeks3d.com

There are plenty of benchmarks available to stress your GPU.
Unigine Valley or Heaven.
Super Position
3D MARK

These are better benchmarks.
This fascination with Furmark to test a GPU is pointless if your driver is cutting power and slowing down the card.

 

[NAM3ofVIOLATION]

I don't know about the full thing but I can tell you that before that card burned I started to experience crashing in GTA V whenever I load the game it crashes 3 seconds after into the game. As I saw whenever the load was on memory of gpu it crashes with display driver. And when i gave it to repairer he said that the main chip is fine i don't know about the rest of the board honestly. And I never overclocked it. I always used it on stock. Do u want me to lower that settings as well?

 

[silentbogo]

Well, if you have power fluctuations - your GPU will act up, so that's expected. But yep, if dude said GPU is fine, then just in case lower Power Limit to 80% to avoid the same thing happening in the future.
Sapphire R7 370 has an abysmal VRM, so just in case I'd sacrifice a little performance for the sake of not burning it up again.

And DO NOT RUN FURMARK or any stress tests on it. 300-series is basically a re-re-re-fresh of 7000-series pushed to it's absolute limit hence it doesn't like stress tests and excessive heat. And with VRM repair situation it'll be even more dangerous, since it can't handle as much power anymore.

 

[NAM3ofVIOLATION]

Well, if you have power fluctuations - your GPU will act up, so that's expected. But yep, if dude said GPU is fine, then just in case lower Power Limit to 80% to avoid the same thing happening in the future.
R7 370 has an abysmal VRM, so just in case I'd sacrifice a little performance for the sake of not burning it up again.

And DO NOT RUN FURMARK or any stress tests on it. 300-series is basically a re-re-re-fresh of 7000-series pushed to it's absolute limit hence it doesn't like stress tests and excessive heat. And with VRM repair situation it'll be even more dangerous, since it can't handle as much power anymore.

So sir in short words, I can still use this card if power limit is 80%? I just want to play games and high settings doesn't matter to me . I'll not run stress tests on it but i noticed 1 thing. while playing Warframe when i switch to windows multitasking the colors of game suddenly flashes to purple when i switch but doesn't do anything more than that. is this a problem with hardware related too or it might just be in the drivers or in game?

and this card have issues with PCIE express slot too. because my gpu and motherboard both support x16 but gpu always run on x8. I tried digging in bios settings but tried everything even Resizable bar support but nothing helps.
my mobo: TUF Gaming Asus A520M Plus WIFI.
I don't understand that is this issue with the card or motherboard?

 

[HD64G]

GPU is on its dying phase again me thinks. Play with it being restricted in power draw as possible to prolong it and get prepared to get another.

 

[eidairaman1]

Sounds like the die is dying. If it blew in a week, there is no telling what other parts are failing.

A 380 is a 285 for sure but the 370 im not sure

 

[HD64G]

Sounds like the die is dying. If it blew in a week, there is no telling what other parts are failing.

A 380 is a 285 for sure but the 370 im not sure

Α 370 is a renamed HD7870.

 

[NAM3ofVIOLATION]

I have decided to sell this gpu asap. ty for helping in this situation

I'll stick to my vega 8 for some time now

 

[pavle]

A 380 is a 285 for sure but the 370 im not sure

A R7 370 is a HD7850. 1024 shaders, 64TMU's, 32ROPs, Pitcairn core. I had the HD7850, cute card, very power-efficient compared to Tashitty, I mean Tahiti (HD7970).

 

[Nike_486DX]

AMD classifies Furmark has a power virus.

FurMark Slowdown by Catalyst Graphics Drivers is INtentional! | Geeks3D

FurMark Slowdown by Catalyst Graphics Drivers is INtentional!.

www.geeks3d.com

There are plenty of benchmarks available to stress your GPU.
Unigine Valley or Heaven.
Super Position
3D MARK

These are better benchmarks.
This fascination with Furmark to test a GPU is pointless if your driver is cutting power and slowing down the card.

Furmark is still the correct tool for the job when checking the cooling performance. Even today with all the throttling limits and stuff it still requires the highest amount of W a graphics card would ever consume. For checking the overclock stability of gpu/mem there are other tools (games, 3d benchmarks etc)

So sir in short words, I can still use this card if power limit is 80%? I just want to play games and high settings doesn't matter to me . I'll not run stress tests on it but i noticed 1 thing. while playing Warframe when i switch to windows multitasking the colors of game suddenly flashes to purple when i switch but doesn't do anything more than that. is this a problem with hardware related too or it might just be in the drivers or in game?

View attachment 224997

and this card have issues with PCIE express slot too. because my gpu and motherboard both support x16 but gpu always run on x8. I tried digging in bios settings but tried everything even Resizable bar support but nothing helps.
my mobo: TUF Gaming Asus A520M Plus WIFI.
I don't understand that is this issue with the card or motherboard?

1) Some cards are actually pcie x8 or x4, even if they physically appear to have x16. So there is no red flag yet.
2) I fixed a r7 260x a year ago, there was a short on the pcb near the 6pin connector. i managed to drill it out (just scraped some of the top layer, no holes lol) and got it up and running (office rig). I had to get a bit ghetto on that one, ran some wires and stuff. But in the end no vrm phases were cut off, and the card works like clockwork till this day (some casual 1080p gaming at low and lots of spreadsheet and office shit u know). Sooo it all depends on ur specific case.
3) If the card works fine and stuff, then (if u really want to sell it) i would suggest selling it to a friend or something so that if any problem arises u will be able to sort it out without major issues.

 

[HD64G]

Furmark is still the correct tool for the job when checking the cooling performance. Even today with all the throttling limits and stuff it still requires the highest amount of W a graphics card would ever consume. For checking the overclock stability of gpu/mem there are other tools (games, 3d benchmarks etc)

Even a well optimised game that fully utilises the GPU can do so, no need to use such a program. For example TheDivision2, TheWitcher3 are games that reach 99% GPU utilisation when standing still. You leave it for 10 minutes and you get the max temps and noise and power draw of the GPU.

 

[eidairaman1]

A R7 370 is a HD7850. 1024 shaders, 64TMU's, 32ROPs, Pitcairn core. I had the HD7850, cute card, very power-efficient compared to Tashitty, I mean Tahiti (HD7970).

Tahiti was an All out card like a 780ti was, or a 980ti, 1080ti

Furmark is still the correct tool for the job when checking the cooling performance. Even today with all the throttling limits and stuff it still requires the highest amount of W a graphics card would ever consume. For checking the overclock stability of gpu/mem there are other tools (games, 3d benchmarks etc)


1) Some cards are actually pcie x8 or x4, even if they physically appear to have x16. So there is no red flag yet.
2) I fixed a r7 260x a year ago, there was a short on the pcb near the 6pin connector. i managed to drill it out (just scraped some of the top layer, no holes lol) and got it up and running (office rig). I had to get a bit ghetto on that one, ran some wires and stuff. But in the end no vrm phases were cut off, and the card works like clockwork till this day (some casual 1080p gaming at low and lots of spreadsheet and office shit u know). Sooo it all depends on ur specific case.
3) If the card works fine and stuff, then (if u really want to sell it) i would suggest selling it to a friend or something so that if any problem arises u will be able to sort it out without major issues.

Better off using occt.

Even a well optimised game that fully utilises the GPU can do so, no need to use such a program. For example TheDivision2, TheWitcher3 are games that reach 99% GPU utilisation when standing still. You leave it for 10 minutes and you get the max temps and noise and power draw of the GPU.

I used unigen tests for Total system stability (only way i knew 5.0 with correct volts would be stable) Ryzen Blender never crashed at 5.0 for me either, that was a stress test.