Graphics Cards

[xrafael95x]

Hi guys let me make this story short and also included what I've done so you guys can help me, please

I got this RX 580, I was using it with a crappy 800w PSU, I was using a 6 pin connector because that is what the PSU had. I Oc the card to 1400mhz and Undervolted, all was great I can play games all normal, One day what I was expecting happened, the PSU died, I bought a used Corsair cx 750w 2017, and the problems started to happen, I get a black screen that forces me to restart my pc, I get this problem in rainbow six siege and in uniengine heaven, when the game is about to start, sometimes I can play sometimes I can't, today was the first time I was playing and the black screen happen during the game, if I OC to 1400mhz I get the black screen. In the case of uniengine i could OC to 1400mhz and I ran the benchmark, weird thing is that always it's in the same scene that I get the black screen, one thing that I want to make note of, is that I play tomb raider for 3 or 4 hours and all good, I don't get any problem, so I don't know what to think if its the PSU or the GPU that could have been compromised by the damaged PSU. Before this new PSU all was good.

Things I have done:

DDU and install an older driver (19.5.2)
Install a new windows in another HDD
Change the thermal paste

Reset Bios

Cleaned Pc slot and reinstalled the GPU
Use a different PCI line of the psu (it has 2 lines)
Change the power outlet

Systems specs:

i7 3770
rx 580 4gb
8gb ram 1600mhz
500gb hdd
windows 10
Corsair cx 750w

[mstfbsrn980]

You have already stated that the cause of the problem is PSU. But you need to know that in order to be able to make undervolt, PSU should be of good quality. So the potential to undervolt with a poor-quality PSU is weak. So reset the WattMan settings, do not undervolt and try again.

[xrafael95x]

I havent touch anything, all is default, i have not undervolt or anything

[mstfbsrn980]

There is probably a problem with the PSU. As a workaround, lower the GPU clock speed by 100 Mhz, set the power limit to -20% or -30% with WattMan and do not undervolt. These changes will increase stabilization for the GPU and the PSU and reduce crashes.

[xrafael95x]

Will try this, i have 50%+ like the answer below and i still get crashes

[pokester]

I would guess that you have an issue with your card. It may be possible to lower voltage and or clock speed to get it stable but then you are not getting the performance you paid for from that card. You may wish to just pursue and RMA on your GPU.

Did you run the OCCT software from OCBASE_COM I suggested? It can run tests on both GPU and PSU. You can see if voltage is not right with this software. 

[pokester]

Your card is Polaris not Vega. They are not known for being good for lowering the power. In fact you typically need to raise the Power Limit slider to +50 to make it work. Also providing incorrect power by using the wrong plug could potentially damage a GPU with that issue alone, not even counting what a faulty PSU could add to the equation. You can run a program called OCCT from OCBASE_COM that can run some tests on both and see what, if anything it shows being an issue. I highly suggest you raise that power limit to 50 and return to stock voltage and remove the overclock to test stability. If you get it running right again, then play with tweaking the OC. 

[mstfbsrn980]

I have no intention of arguing but it is not right to suggest that + 50% power limit for such problems. Most nVidia GPU power limits are zero. Some OC nVidia GPUs come with a maximum +25% power limit not +50%. This is because increasing a graphics card to maximum power limit may damage it. Also, the vast majority of AMD GPUs come with a high power limit according to potential, and the vast majority of users are having problems for this reason. AMD GPUs still do not determine GPU dynamic core voltage based on 12V value from PSU, the driver cannot already adjust it and crashes occur.

Also performing this OCCT tests do not solve the problem. Title owner should not undervolt, should underclock and the problem will likely be solved. If the problem continues with all trials, PSU is already problematic. With OCCT and +50% power limit recommendation may kill GPU...

[pokester]

First off I was not arguing with you in any way. Just offering an opinion of my experience with the 3 Polaris cards I actually own and the countless reports of the power limit raise helping stability. This is a very commonly known thing that often helps. Everyones mileage of course can vary depending on the different factors affecting the situation. It doesn't make anyone's advice wrong. I did not address you in any way negatively nor offer uncommon advice. I responded to the OP and the OP or anyone can choose to try or not try any advice they receive. Regardless if one avenue doesn't work it is prudent to try another. Especially when the bulk of my advice was to check the equipment with free software that can show if you have an issue and return the product to default settings until isolating the issue. If you don't think that is a good idea, that's fine you don't have to and neither does the OP. Now as far as a +50 killing the GPU, no it will not. It is not increasing the voltage in the same way as raising the voltage they are two different things. It just raises the power ceiling of available sustained power when needed. This is well documents by AMD themselves and is also what the two AMD mods here also frequently recommend. The power limit in Wattman at it maximum is within operating range for the card. Now a 3rd party app might allow it to go further depending on the available range in the bios of the card. Again follow whatever advice you want and I wish you luck. 

[mstfbsrn980]

You write not by technical knowledge but by trial and error. If the stabilization increases when you increase the power limit to the maximum, the GPU core voltage is because it remains constant. But this does not mean that graphics cards will always provide the same stabilization or not damage the GPU.

Graphics cards do not only contain GPUs. The mosfets that power the GPU are present on every graphics card, and the sum of these particles is positioned to the specific GPU wattage by manufacturers. Since Mosfets configuration increases the cost of graphics cards, it is generally not suitable for the maximum power specified by VBIOS. You will understand very well what I mean by looking at quality motherboards and cheap motherboards of mosfets.

If you are using a GPU with quality components with the maximum power limit and you have no problems, you are probably not playing all the power consuming games or working hard and you don't spend a lot of time on the computer.

I tried to convey my knowledge here, not my experiments,, with my weak English.

[pokester]

"You write not by technical knowledge but by trial and error."  Wow that is unnecessarily insulting.

Again it is not my intent to slander you or keep anyone from following your advice, in fact I still have not even read what you wrote as advice. I only answered the OP to the best of my ability and again that advise may or may not help for this situation.  I only responded to what you seemed to take exception to with what I said.

You are absolutely correct all I know is my own experience and what I have seen advised and confirmed as working by other users time and time again.

I make no claim to be and engineer on AMD products.

Aside from reading AMD published specs on hardware I know really know nothing more about them than what you get form reading published specs and white papers.

I am however an IT Manager with 40 years in systems management and service, with degrees in computer science, electrical engineering as well as graphic arts. I am not exactly a novice either. 

Again it is the same advice given by the AMD engineers in these forums time and time again. Again it is only raising the power ceiling to increase the power limit slider and it is within NORMAL RANGE. Maybe amdmatt or ray_m can clarify this for us?

I also agree with what you are saying as far as this can when the power ceiling is sustained increase the power across the board on all the parts. Of course that is how it works. It would have too. That being said the reason you need in many case to raise the power limit is because many if not most cards released are not released at reference specs they have factory overclocks in most cases. The AMD "vanilla drivers" are written for default reference parts. So in many cases the power limit slider tweak gives cards what they need to be stable. It is why AMD put it in their Wattman adjustable parameters to begin with. So users have the ability to adjust for those disparities between reference parts and what they actually purchase or try to achieve with over clocking on their own. 

I do not claim to know more than you or any other user. I however am also not condemning you for your advice.

These are USER TO USER forums as clearly explained in the forum header links. 

The expectation is that you are getting advice from other users in these forums which are not professional engineers working for AMD. It is however great when more knowledgeable users can offer advice that others users may not posses or when AMD engineers do get involved. 

[mstfbsrn980]

My intention is not to argue. I want to remind you again. You shared wrong information about the power limit. I was hoping you would give technical information about the limit. Anyways. I hope I didn't bother you.

[pokester]

Now worries. I disagree with the power limit being "WRONG" information. However I would urge the OP try things at defaults before making any changes and going from there. The power limit setting however IS AN EXTREMELY COMMON THING THE HELPS POLARIS CARDS. Again I did explain how the effect of the power ceiling works. You just disagree with that being good advice and that is okay. It is your right to do that. Again I have not called your advice wrong.
The power delivery on the card is dynamic. Raising the ceiling only increase the power ceiling available. It does not mean it will be all the time using more power. It is not the same as making a voltage change.  It just raise the available power when needed by a small amount that is still within normal operating range.  This again is made available as not all released cards follow reference specs.  This can often be the difference on a card hanging or not hanging. I hope that is more CLEAR for you.

[mstfbsrn980]

You shared your knowledge and now I just want to share my experience. I used the RX 480 with two different 750W PSUs. The first PSU was able to withstand 1100mV +20% power, while the second PSU was able to withstand 1075mV +50% power. So according to my experience GPU doesn't work with + 50% power with every PSU but lowering power always increases stabilization. Your suggestion only increases core voltage stabilization. It certainly does not increase PSU stabilization. However, driver stability may increase partially with your suggestion. 

[pokester]

That is comparing apples and oranges.  While some 480's do allow a +50 most are +25 max. The 480 was only designed for power from one plug. While the GPUs are mostly the same as the 580's the vrms, board power delivery are different. They have 2 pin connectors. I believe the reference design is two 6 pin while many aftermarket cards are a 6 and an 8. All will handle power differently. By the same token the 590 handles more power than the 590 yet the GPU is still Polaris. Things improve over time with fabrication and a die shrink in the case of the 590. In the scenario you describe the PSU is the issue. I addressed the OP using a bad PSU as being a problem and having also potentially damaging the GPU to begin with. So in your case and the OP's yes an improper PSU is a bad thing. 

[mstfbsrn980]

It is also wrong to arrive at the judgment that + 50% power is applied with a GPU that comes with two power cables. Because the capacity of one 12V power cable may not be enough for GPU watt. Also, I have no intention of judging you. My purpose is not argue. I just wanted to share and I hope I didn't bother you...

[pokester]

It's not an argument. It is a fact of the designs. Nothing I said is an OPINION. You can find the cards as I described them yourself, very easily. Your answer of one cable to, two pins being bad is correct. This is a bad thing and I never said or suggested otherwise. You should always use 2 separate dedicated connections to your GPU. Regardless once again this makes it a PSU issue and or user error in connecting the GPU to the PSU. Not that the description of the GPU's I gave is wrong. 

[mstfbsrn980]

It doesn't matter if a GPU comes with two different power cables. Mosfets structure is very important like that...

The result you are trying to arrive was incorrect, I think. All the information you provided was correct about Polaris infos. I'm glad I didn't bother you.

[pokester]

You keep bring up things like I challenged what your are saying. Yes all the components of the cards receiving the right power are important I not only did not say anything different I actually already pointed out components being different as the key reason there is a difference in the 3 generations of top end Polaris cards. You are supporting everything I have said now in 3 different ways and acting like I said something contrary.

I never said a word about the card coming with 2 power cables. It could come with with any number of cables and if you don't hook it up to the PS I said you should use 2 different dedicated lines into the power supply correctly it is an issue. So NO nothing I said once again is incorrect. You just read something into what was said that was never said.  

Anyway none of this banter helps the OP in anyway. There appears to be no end to whatever self created dispute you seem to want to keep going. So I am done with this conversation. 

[pokester]

I should also note the OP liked my comments which I appreciate as I do my best to help the OP. Unfortunately I'm sure my advice helps some and not others. I am glad the OP liked my advice and hope it helped. Most people understand that advice may or may not help and just appreciate other users trying to offer assitance. 

[xrafael95x]

I Will download the software you said, ut i dont have any kind of OC or UC, everything is stock.

[xrafael95x]

Forgot to say yesterday I was testing with stock clock and volt but +50% of power limit and got crashes while playing rainbow bow six, I will try lower the clock and -25% of power limit, i also made the test for 10 min on the OCCT, with the setting i just said and its good running it, the real test if the rainbow six

[xrafael95x]

I'm more worry about the GPU damaged that the fact i can't OC, i just want to make sure what is causing the problem, my common sense, says if the crappy PSU would have damaged the GPU, the GPU wouldn't work, but as electronics are so weird I don't know what to think

[xrafael95x]

I was using -20% of power limit and 1320 of core clock still got the black screen, now im lowering to 1280mhz and 0% power limit.

BTW: everything this happen when my pc restart and starts again, it says radeon wattman have been reset to default by an unexpected system failure

[mstfbsrn980]

Reduce all core clock speed levels by 100 Mhz (may be 50 Mhz) and set the power limit to -25% (may be -30%) with WattMan. OCCT 3D test creates a load on the GPU with only one frequency and hints that only the GPU is healthy. Apply WattMan changes for the GPU and restart the system. Then check that the WattMan settings you set are stored and try for the crashed game.

If these attempts are not successful, the new PSU seems to be necessary. Good luck...

[xrafael95x]

Ok after lowering to 1280 and 0% power limit I had no problem, but guys it easier for me ask for a refund and buy another PSU the problem is to find a good PSU like the CX 2017, so I don't what to do, because I'm worried that if the GPU is the one that is faulty I would change the corsair for a worst PSU and at the end of the day it wont solve my problem. So if there a way to test what is faulty or not?

[mstfbsrn980]

I tried to tell you about my experience and that you cannot apply + 50% power with a poor quality PSU. Also, with this type of PSU, boost frequencies will also be pushed, so I suggested you, should drop core clock speeds. Because poor-quality PSUs experience drops of 12V and the PSU cannot provide a stable voltage for the GPU. Corsair is not a quality brand, I think. Try a different brand. I used the same PSU and GPU 12V statistic has always dropped, and I could never do undervolt with this Corsair.

[xrafael95x]

according to the internet the cx 2017 is a really good made psu, maybe its faulty 

[mstfbsrn980]

The quality indicator for a PSU does not depend on a single variable value. Also PSU quality varies depending on usage (like how many years it was used). However, each hardware has its own protection systems and your system is unlikely to be damaged by this type of PSU. PSU quality is already the reason why some AMD users are experiencing stabilization problems. That's why AMD offers the WattMan software.

For these reasons, I suggested you WattMan software. I tried to explain how you can stabilize your system before trying a new PSU with WattMan. If you had damaged hardware(s), the symptoms of it would be very different.

[xrafael95x]

Well with ask for a refund a get an evga nova that is what i can get 

[xrafael95x]

and you know what I think its the PSU, I'm not sure but i think its the PSU causing this, why do I think this? every time the game goes black because is doing a transition from the lets say the menu to the game itself, I get the crash, so I think the PSU can't be fast enough to give power to GPU

[hardcoregames_]

set the power limit to +max and monitor the BIOS for a while to see if the voltages are all stable

[xrafael95x]

i did set at the power limit to max and still get crash but i didn't monitor the bios, when you say this you mean use GPU-Z to see the voltages?

[hardcoregames_]

xrafael95x wrote:

i did set at the power limit to maz and still cget crash but i didnt mintor the bios, when you say this you mena use gpu z to see the voltages?

I routinely reinstall windows due to problems so I suggest you try that.

I use 7-zip which has a console version, this can be fed to a BAT file to backup everything wanted 

[xrafael95x]

i already did that, and didnt help