Graphics Cards

[theredspirit]

Ever since I got RX 580, I have been getting "wattman settings were reseted" on start-up or sometimes I get blank screen during POST. Sometimes this thing boots just fine. I will keep it short, does it exceed PCI-E power specification or not? RX 480 did that and RX 580 is just overclocked rebrand of it. I really hope that someone will give me some hope in this thing, because RX 580 is nothing but problems (Crazy amounts of heat, after 1 hour my whole room's temperature rises a few degrees and in area near desktop becomes Sahara, plus hot air is very dry. Power consumption. Yet thermals stay reasonable, around 70-75C. Performance is good, but it's a big turd otherwise)

Note: On my another computer RX 560 craps out in same fashion. Not gonna blame it on Radeon, since that thing has very weak mobo and it's possible that CPU may be causing those things. Even A4 6300 for that mobo was a bit too much. Once voltage was set in UEFI, problem went away.   

[arcticwind]

It would really help to know some spec's of your system...

https://community.amd.com/thread/196209

Like motherboard, power supply, windows version and GPU driver version...

In general the GPU will use (about) 75w of power from the motherboard's PCIe slot, the remaining power needed to run the cards comes directly from the power supply. Do you have the power cables connected GPU?  

[theredspirit]

Hardware specs:

Asrock 970 Pro3 R2.0

AMD FX 6300 stock with turbo enabled

Corsair LPX 1333MHz DDR3 4x kit

Powercolor Red Dragon V2 RX 580 8GB

Fortron FSP 700-50ARN PSU with silent fan mod

Cooler Master K280 modded (additional 80mm top exhaust fan)

Samsung 840 Evo 128GB SSD

Seagate Baracuda 1TB 7200 rpm

WD Caviar Blue 320GB 5400 rpm

Kingston A400 128GB SSD

Asus AC56 wifi PCIe card

Asus Xonar DG sound card

USB 3.0 PCIe controller

DVD drive

Memory card reader

Akasa Vegas Green intake and rear exhaust fans

Scythe slipstream side exhaust

Scythe Mugen 4PCGH with two Noiseblocker XLPs

Display:

BenQ BL2420PT

Software specs:

Windows 10 Pro 64 bit

Latest Adrenalin 19 version (yet issue existed with Crimson 18 versions too)

Update:

Computer gets frequent BSODs, since Asus AC56 wifi card driver reinstallation. BSODs if just after boot, I try to enter Youtube. 

[elstaci]

First what PSU are you using?  a weak or underpowered PSU can cause problems like you are having. Since it is happening before Windows loads any drivers.

Second: what are the exact BSOD Errors are you getting?

If you can upload a minidump file I can find out what driver is causing the problem. Also if you can run DXDIAG and upload that would be helpful also.

NOTE: Try stress testing the GPU and CPU and PSU using OCCT and Furmark and see what happens. You might be able to narrow it down to hardware or software issues.

[theredspirit]

I already wrote what PSU it is. It's Fortron FSP 700-50ARN. It doesn't have any other name as it is OEM PSU mostly. To me it came without box or anything, but at the time I was able to find reviews of it. It meets 80+ silver specifications and overall is decent. As I understand, it was aimed at European market mostly and to keep costs down, it never got 80+ official rating. At worst it will be 80+ Bronze. Again, at the time as I knew back then it was above Corsair CX series PSUs.

Note: I'm still not very familiar in this forum and that PC is semi-functional, so I may not upload that.

[elstaci]

Sorry, I missed that in your reply of your computer specs. Your PSU is plenty powerful for your computer and to run your GPU card unless it is failing.

Do you have any Monitoring software installed like Open Hardware or HWHardware that shows all temperatures and voltages of your installed hardware?  If not I would suggest installing it and see if you see any abnormal temperatures or voltages or other data.

[theredspirit]

I have wattmeter plugged into wall constantly. With OCCT and Furmark running and stock settings whole computer only eats ~350 watts. With CPU overclocked to 4.1GHz at 1.4volts, PC uses 425 watts max. That PSU is never utilized fully and it runs at best slightly warm, so nothing is frying inside it. From what I read PSU electronics can survive maybe 60-85C and at least from innacurate and subjective impressions, it doesn't reach that. I remember I had non 80+ PSU, which is cheap and most likely 70% effective and looked like low end power supply, it got much hotter at full loads and spun fan much more. 

During this time gap, I just tried to troubleshoot my wifi card and success, no more BSODs. What I did, was removed no longer needed drivers, with card unplugged. Then as Asus website says, let drivers to be downloaded by Windows. I just plugged in a cable. This worked beautifully. Even if power connector was plugged in correctly I replugged it again. At least for now, I have no Wattman settings reseted stuff. 

I did some testing before all these problems with OCCT and Furmark (it makes my GPUs pull the most watts, so I think it's the most stressful test, but it causes pretty much any modern GPU to throttle down clock speeds). I will just write my results here in list (note: fan curves, except GPU and case fan setup was changed a lot during this time, so results aren't scientific):

FX 6300 + GTX 650 Ti (tested separately)

Furmark. Works at full speed temperature is in 70-80C limits. Without overclocking, maxes out at 75C.

OCCT. At stock works cold. Maximum temperature is at 45-46C.

Personal comment. At highly demanding tasks, rear exhaust exhausts slightly warm air, nothing to worry about. With 4.1GHz overclock at 1.3875V PC worked correctly and pretty much just as hot.

FX 6300 + RX 560 4GB (no 6 pin version from Gigabyte) (tested separately)

Furmark: GPU tries to stay at around 70-75C at all costs and revs up fan to 100-115%. Throttles to 920-970Mhz, even if it could just heat more. If I set power limit in MSI Afterburner to max, it pretty much never throttled. I was puzzled. Decided to investigate this nonsense in Unigine Heaven. I ran card with stock settings and with maxed out power limit. Difference in performance was 7%, but difference in clock speed was much greater. Due to this poor scaling, I decided that it's just some power saving crap, which increases efficiency meaningfully, at low cost of some performance. In games card ran at almost full speed too. I wasn't happy, but I understood it. Had thoughts about RMAing it. Felt stupid for not getting full RX 560 with 6 pin connector, because I thought that revision 2 is just fixed revision 1.

OCCT. Nothing has changed here as RX 560 isn't very hot card and roughly dumps just as much heat as GTX 650 Ti.

Personal impression. Exhausted air is slightly warm, nothing to worry about. For fun CPU was overclocked to 5.150 GHz at 1.55V and still heated up only to 59C in OCCT test, but was unstable. Never overclocked it that high again, as my motherboard is low end, doesn't have VRM heatsinks and CPU cooler is tower, so it doesn't blow air on VRMs. After experimentation case was modded and got 80 mm exhaust fan over VRM area, but as closer to rear as possible (to keep standard airflow model).

Wattage: Pretty low overall. Around 200-230 watts during gaming. At idle, it's 65-75 watts. During web page scrolling it's ~100 watts. During webpage load it's ~110 watts. 

FX 6300 + RX 580 GB

Furmark. I expected to see throttling, but card acted differently. After 70-75C, it just drops the load, keeps low stock clock speed. Doesn't boost and also doesn't max out fans. Dunno how exactly, but it tried to keep itself less load. I suspect it changed clock speed itself, just didn't show it to me.

OCCT. No anomalies.

Furmark + OCCT. Once GPU gets hot, it dumps a lot of heat. CPU temperatures keep rising and stop at 69C. That's a lot and this Asrock board is annoying with stock settings as it keeps CPU at full load running at 3.2-3.5Ghz. That's just stupid design. Once I set clock speed to 3.5Ghz in UEFI, temperature reaches more than 72C and starts to throttle a lot. All way down to 2GHz without any shame. I tested out my 4.1Ghz overclock with 1.4V and yeah it heated up to 75C and started it's shameful act again. CPU cooler is seated properly, tightened well, paste was changed half year ago or something to Arctic MX-2. I decided to swap Scythe CPU fans to Noiseblockers and that proved to not be helpful. I also used Noctua resistors to slow down 2 case fans, even after their removal, CPU heated a lot. Even remounted case fan from intake to exhaust position, it didn't help.

Personal impressions (OCCT+Furmark). Exhausted air is hot, very hot. Case heats a lot. Warm spots are close to exhaust and on motherboard side of case, but after some time they even out a bit. After 1 hour of this testing, my whole room heated by several degrees. Air felt very dry. Being near computer caused my eyes to be very dry. I was sweating like it was 32C and summer. Soon I started feeling a bit dizzy, so I opened window to get some cold winter air. Gotta say that, I was extremely disappointed in RX 580 and thought about swapping it to GTX 1050 Ti. Radeon - The way it's meant to be cooked.

Wattage. Very high at all times. Around 250-300 watts during gaming. At idle, it's 75-100 watts. During web page scrolling, wattage is around 120 watts. Web page loading is at 130 watts. Furmark + OCCT makes it use 350 watts. Furmark + OCCT at 4.1GHz OC on CPU, makes PC use 420-425 watts. This is straight up horrible. My other PC with Athlon 845 and GTX 650 Ti, only uses 50-75 watts at idle, web page load - 70-80, during gaming - 100 and with stress tests running - ~120 watts. It's no surprise, that I don't like RX 580 and just used the other PC for daily usage. Power usage of my main is just too unreasonable and heat output is too bad.

Observation. After GTX 650 Ti was put in Fractal Design Define R4 in other PC, it ran 10C colder than it did in Cooler Master K280. It could be that Cooler Master case is just not great at cooling. That other PC doesn't even have CPU cooler like Mugen 4 PCGH. It only has Alpine 64 Plus. 4 fans are installed and they run at 7 volts. And fans are just something that came with case and my used Cooler master cheapo sleeve bearing fan and 800 rpm Scythe Slipstream. That system was pretty much just thrown in together without too much enthusiasm, hell it doesn't even have SSD or 8GB RAM, but runs much better than my main one.

Haven't tested stuff after fixing up my PC, but at least it works now. Voltages were always pretty normal, I haven't found any anomalies with them. Asrock motherboard doesn't have LLC (load line calibration), so it drops voltage during high CPU load. It can't correct that, but it doesn't need to correct it as everything works well. 

Anyway, can someone here answer how AMD handled RX 480 PCIe power specification limit violations and how can RX 580 work at higher clock speeds, yet with pretty much the same power connections needed? What at worst could happen if those limits are exceeded?

[elstaci]

Great troubleshooting and Diagnosis. 

I found your Wattage measurements really interesting while testing and using the various GPU cards you had. It shows you have a High Quality PSU installed since the voltage stayed within normal range during such heavy stressing. I, personally, never heard of the particular PSU brand you purchased. But it seems like a decent brand.

Usually if you had an Overheating or PSU issue, OCCT or Furmark stress tests would have caused the issue to occur, in some instances, almost immediately. I have known several Users mention while running OCCT PSU Test, as soon as the Test started, the computer would shut off by itself. That is because the PSU Test stress tests both the GPU and CPU at the same time putting maximum stress on the PSU. Normally they found out it was due to overheating of the CPU or some other type of failure to hardware unrelated to the PSU.

I really enjoyed reading your comment concerning the tests you made and the results. very informative.

I read from another User that his WiFi driver was also causing BSODs in which he thought it was his GPU drivers.

By the looks of your Diagnosis, Your GPU card and computer is working fine even under heavy stress. You seemed to know about how to Overclock. something I am not familiar about. But forcing the CPU voltage above 1.5 volts for long periods just to Overclock, in my opinion, is damaging to the CPU and will eventually cause an early death. At least in Ryzen CPU/APUs.  Having spikes above 1.5 is normal, but those only last for milliseconds or even nanoseconds which is by design and not harmful.

As for you last question, maybe someone else has the answer.

Glad to hear you find the reason for your computer crashing in BSODs.

[theredspirit]

Wifi stuff is often problematic and annoying. Honestly, I could never figure some stupid TP-Link adapter problems and it just kept not working correctly until replacement.

Oh, those FX CPUs can handle a lot. Sure, 1.55V is high, but in reality it only was 1.49V due to lack of LLC. I only ran it for short time. I have seen many people running FX CPUs on LN2, 6GHz and doing other crazy stuff. They just never die. It truly helps that FXs were a bit spartan and use older lithography + have long pipeline. They are designed similarly to Pentium 4s and those things could handle some abuse. At 1.55V I was more worried about motherboard as that thing doesn't have any cooling. Also that board was replacement for another board. On previous board I also tried overclocking and pushed it, until I noticed that it started browning in VRM area. I measured temperature and it was 149C in one spot. Of course I didn't overclock it anymore and after 1 year that board started to die. Ever since the purchase of this CPU in 2014, I collected information of its overclocking capabilities. Turns out it's normal or slightly better than average chip. With decent motherboard it could run at 5.5Ghz on air, if GPU isn't like RX 580.

BTW Ryzen is nothing like FX. I wouldn't really overclock Ryzen at all as they come maxed out from factory and 100 MHz OC over turbo to me is just not worth it. I just do overclocking for fun and Ryzen is not fun to overclock. Pretty much all of them reach the wall at around 3.9-4.1GHz. Even Ryzen X models they are nearing that wall. Their wattage is much higher than non X models, meaning that they already surpassed their maximum efficiency clock speeds and voltage, but it's just mild violation of that. FX CPUs have this thing at over 6GHz and even then it's not so bad as Ryzens'. FX CPUs could be pushed to 7GHz and even 8GHz, but this is LN2 category already. 5GHz is nothing, if you don't mind power usage and high motherboard buying requirements. FX CPUs also are great at undervolting and can be very efficient too. AMD after Vishera messed up Bulldozer architecture and late Athlon 800 and AMD A 7000 models couldn't overclock well at all. Meanwhile 6000 series APUs and 700 series Athlons could be overclocked just as well as FXs, if someone was willing to invest into good motherboard and big air cooler. Just like FXs, those things undervolted very well. When I had A4 6300 running, I managed to undervolt it 300 mV on stock clock speed of 3.7GHz. That was a massive improvement. If if was unlocked, I think it could have reached 5GHz only with Alpine 64 Plus. Sadly, it isn't. I tried overclocking with raising FSB, but that just didn't go well. Soon something crapped out. FSB overclocking nowadays is nothing like it was in 2004. Now motherboards are too complicated for that, have too many sensitive buses and lack locks.

Fun fact: Athlon X4 845 is the fastest Athlon made for FM2+ socket. It beats 870K. That's because it is Carrizo based chip, which is one of the last Bulldozer improvements. I heard that Athlon 880K exists, but never seen it anywhere.

[elstaci]

Hey, I can vouch for the FX CPU Series. I have a FX 8350 installed and it works like a charm. It does take a beating. When I replaced the old CPU Stock cooler for a much meatier CPU Stock cooler (years before I got the FX 8350), this is before all the monster air and liquid CPU Coolers started to come to market, the fan didn't start. Within seconds my computer would shut down with the FX CPU before even POST started. For a moment I thought I had a defective CPU. Than I noticed that CPU fan was jammed by an obstructing cable that I didn't initially noticed. After rearranging the cable, the Computer booted up without any issues.

Later on the same meatier CPU Stock cooler was not cooling well anymore (the CPU cooler was several years old at this time). When I ran a backup program that put a heavy load on the CPU (above 90%), it would overheat within an hour''s time and shut the computer down. Even after replacing the Thermal paste. Replaced with one of the monster inexpensive but very good Cooler master Hyper 212 EVO with two fans. No more overheating, no matter if CPU was at  100% load for more than an hour. It maxed out at 59c to 60c.

So even though the CPU overheated more than once, it never got damaged and still is working like a charm. And I have a computer loaded with hardware and USB devices.

[StephenD]

My guess is they didn’t, I must run two cables from the power supply, not daisy chained, otherwise it takes the pcie bus down at full throttle.

[arcticwind]

What motherboard BIOS version are you running? Update it to the latest version 2.80 if necessary.  Version 2.60 had PCIe improvements on it.

And as elstaci mentioned give us the error code from the BSOD. 

[theredspirit]

UEFI version is P2.80. It came with it.

BSOD doesn't have code, but it writes "page fault in nonpaged area".

[theredspirit]

Sadly, got blank screen on boot. Had to do hard restart and got that same "wattman settings restored" stuff. That sucks.

[theredspirit]

Still having issues. Decided to test out this PC without XMP profile on. Will report later.

Note: Why this thread is renamed to "System instability caused by wifi card?" ? Looks like some lazy moderation in these forums.

[theredspirit]

Had first blank screen. Yet up until this event, I think failures became slightly rarer. At least I still haven't got wattman reseted message without blank screen. So may maybe RAM could have contributed to that. I just double-checked HT Link and NB frequencies, they are set to auto and according to CPU-Z, they are ~2400MHz and ~2000Mhz. HT Link frequency is correct for 970 chipset, meanwhile I don't know about NB frequency. My old 760G chipset board also had default of 2000MHz, but chipset was different. So, I don't know for sure. Anyway, it looks like I'm not looking at relevant stuff in UEFI and just trying to randomly diagnose what causes this stupid problem. 

My motherboard has 16 PCIe lanes and I have GPU and wifi card in PCIe slots. I don't know exactly how do PCIe lanes works, but from what I know it's either by expanding lanes or by cutting them in half for each device. Wifi adapter only requires one lane and the slot it's in is electrically 4x slot. So, if lanes aren't slip in half, then GPU should have 12 lanes available and wifi card 4. I dunno if this can cause problems, but I have PCI wifi card not in use, so I can try uninstalling PCIe card and testing out PCI card. Even if it's a solution, I really wouldn't like to have PCI wifi card as it's bandwidth limited. Also that thing with changing XMP settings to JEDEC and affecting stability hints that problem may have nothing to do PCIe lanes, but rather RAM, Northbridge or software. Afterall, this PC worked perfectly fine with RX 560.

Note: XMP settings only change Command rate to 2T, which should be more stable than 1T. So technically disabling XMP and expecting magic is bogus. But this is just with limited abilities of CPU-Z. Maybe some less known subtimings are also affected. 

[Matt_AMD]

Anyway, can someone here answer how AMD handled RX 480 PCIe power specification limit violations and how can RX 580 work at higher clock speeds, yet with pretty much the same power connections needed? What at worst could happen if those limits are exceeded?

RX 480 reference design was using a 6PIN PSU connection.

RX580 uses a 8PIN PSU connection which can draw up to 150W + up to 75W come from the PCI-E lane. 

[theredspirit]

So, in the worst case scenario it should use maximum 225 watts. I just looked up Tom's hardware power consumption results and they had Sapphire's RX 580 8GB Nitro +. It has more power connectors and better cooler. Anyway, my RX 580 only has 8 pin connector and a bit worse cooler. Still, single 8 pin connector should be good enough. During gaming and testing GPU works fine, but it sometimes fail at boot up and doesn't recover. So, I suspect spike in power consumption, which I can't confirm or deny, because my wattmeter is too slow for that. Tom's Hardware test did notice small failure to fit into PCIe connector specifications, but they say it shouldn't be a problem. Apparently I have a problem, but I'm not sure with what exactly. It just looks like failed GPU overclock. In case of RAM failure, PC will behave differently. Most of the other components will fail differently too. Software isn't at fault as this happens during POST. It's either motherboard or graphics card. Temperatures are fine. Graphics card itself physically looks like it's in great shape. GPU failure would be visible in gaming and other heavy workloads, VRAM failure will be visible in obvious artifacting. Motherboard could only be blamed for lack of proper voltage delivery, except it actually runs in less than 16X mode, and if it doesn't run in that, I think that power can be cut to lower mode's. Another thing that could be blamed is PSU, but this unit is solid. I think I actually tested it with multimeter and it was perfectly okay, not to mention that it always made this PC work quite long time and didn't have much trouble with CPU running at 5.150 GHz. I think we can just throw it out of suspicious component list. This Asrock motherboard is sure low end unit and it like to drop voltage, when pushed far, but it doesn't drop voltage too much. But stuff happens during boot up, when GPU isn't at full load and only CPU is somewhat loaded. 

The conclusion is that it looks like my low quality motherboard doesn't like pushed to it's full potential and RX 580 is pushing PCIe slot over its limits. 

I guess, I need some advice here as I'm currently stuck. I could get GTX 1050 Ti now, but it's a major downgrade and it looks like nVidia will release their new lower end cards soon.

[Matt_AMD]

The conclusion is that it looks like my low quality motherboard doesn't like pushed to it's full potential and RX 580 is pushing PCIe slot over its limits. 

Not sure how you come to that conclusion, but i disagree. 

It sounds like you have some faulty hardware, but it's just a case of troubleshooting which hardware it is.

[theredspirit]

Any suggestions? Because previously PC worked fine and not a single component shows signs of failure. CPU works fine, RAM may be a bit suspicious, but there's no good way to test it (memtest86 isn't a stress test, OCCT doesn't load it well, prime95 doesn't load it well either), PSU works fine, motherboard is a bit suspicious, drives work just fine, cards seemingly work fine too, cooling isn't a problem. 

[Matt_AMD]

During gaming and testing GPU works fine, but it sometimes fail at boot up and doesn't recover.

What does this mean?

[theredspirit]

It means what it says. You should just read what I wrote before. Here's a short version for you, I get blank (grey) screen when PC boots up. No video output and PC doesn't recover itself from that. It needs to be hard reseted (via button) to try to start up again and usually it boots up just fine.

[Matt_AMD]

I would firstly try the GPU in a different PCI-E lane and then, if possible, try the GPU in a different system to see if you can reproduce the same behaviour. 

This sounds like a isolated local system issue. 

[theredspirit]

What do you mean by putting GPU in different PCIe lane? If you meant different PCIe slot, then I can say that AMD 970 chipset doesn't support many lanes and due to that it only has one full 16X slot available. However, my motherboard has another PCIe 16X slot, but electrically it's only 4X. It would be a major bottleneck, but maybe just for testing purposes it will work just fine. I don't know, but it looks like asking for more trouble than there already is.

I only have one another system that supposedly could make any use of RX 580, but like I said before, it also has similar problem (except black screen fixes itself during windows loading), but with RX 560 (no 6 pin version) and power supply there is only Thermaltake 450W Litepower. It may be just too weak. GTX 650 Ti system I have is for completely different purposes, it only has 4GB ram, no SSD and also has some history of instability. I resolved it by disabling XMP profile, but I'm not completely sure about that yet. It also has Windows 7 and pretty much all games there are at least from the last decade. I really don't have any other means of testing it out in other system.

It may sound a bit crazy, but cheap motherboards may be very overclocking not friendly. XMP is overclocking or at least running beyond JEDEC specifications. Turbo clock speed is also running beyond specifications of base speed. It may be unwritten rule that cheap mobos only ensure that stuff runs at bare minimum. Asrock specifically in my experience fares the worst and always has those negative surprises. I would have never considered 970 Pro 3 R2.0 motherboard in my life, but it was just one of the last available AM3+ mobos in store and in ATX for factor. The only alternative was Gigabyte 970A-DSRP. It's just as bad as this Asrock thing.

Currently I'm tying out RAM running at 800MHz to see if things change, because disabling XMP kinda had some results, so I want to see what happens in case of major changes. 

I just want this thing running without need of upgrading parts. Because if I will need to do that, this whole thing is pretty much just needs to be replaced. At the very least I would need new motherboard, RAM ,CPU, case (I overtighted screws in some places), more fans, some new cards (because PCI is pretty much gone for good now). Also it doesn't matter how much you back up, when reinstalling Windows you will have things like software settings refreshed, all game saves are gone, some other stuff is gone. It's just not a solution and in my case, I have OS on SSD, Windows directories on one HDD, Games on another HDDs. The files are just spread everywhere and at least without one drive other will not work correctly or just gonna have lots of trash in them. So, yeah it would be easier to just buy new computer entirely and reconfigure everything from scratch slowly and by having this thing by the side. I have no patience and sanity to do that now. I wanna get this thing working for at least 2-4 years more and replace it when may have more finances and will actually see huge improvements over this thing.   Rant over.

The weird thing is that this computer ran RX 560 and GTX 650 Ti without problems. It only has problems with RX 580. 

[theredspirit]

Could be that I found a problem. It may be unstable CPU clock (FSB). During some boots I noticed slight drops in it and something sensitive may be causing me problems. Setting manual CPU clock does nothing. I'm probably stuck with this nonsense. RX 580 just made bad design even worse, so at least RX 580 is fine. This or some voltage fluctuations during boot. (May be caused by loss of efficiency in VRMs due to more heat)

Also I think I figured out heat problem. Due to poor ventilation in case, heat rises. Since heat can only escape through exhaust fans and I only have two, it goes from GPU and hot air is sucked into CPU cooler. CPU coolers only work due to assumption that it can get cool air, but when air it gets is hotter than CPU, then CPU cooler is good CPU heater. In other words case is unable to provide enough ventilation and hot air isn't dissipated quick enough and my motherboard is whack. 

Today I tested out computer with Cooler Master Hyper 103 and I saw very similar CPU temperatures compared to Scythe Mugen 4 PCGH. Pretty much confirmed my theory. As in past when I used Hyper 103 it was noticeably worse than Scythe. Just for kicks I also tested out Athlon X4 870K stock cooler (AMD 90w cooling solution, basically the one with red Cooler Master fan on top and single heatpipe), it fared very badly. Only in less than 2 minutes temperature was way over 70C and kept raising. had to cancel testing. The cooler that came with FX 6300 is just a slab of aluminum and just in theory I think it should be worse than heatpiped AMD cooler. 

I think that if I got 120mm AIO and cut holes for it on top of the case and as close to front it may work, but I really don't wanna do such things. Or I can try getting NZXT graphics card water cooling mounting kit and 120 mm AIO, then GPU's heat could just be directly exhausted outside without traveling through CPU, but again I don't wanna do this. It's just too expensive and I was never a fan of water cooling in general.

[hardcoregames_]

Use GPU-Z and monitor the video card temperatures and see if the card is too hot.

Sometimes manufacturers use crap thermal pads that do not last long at all.I have had to refurbish a lot of cards over time with MX-4 to keep them from self destruction.

.

[theredspirit]

Furmark and OCCT serves me well. BTW there is a reason why manufacturers use thermal pads instead of thermal paste for VRAM and VRMs. 

[hardcoregames_]

pads are easier for pick and place machines to put on a AIB during manufacture

[theredspirit]

I would say that pads are better for bigger gaps. Thermal paste would just dry out if gap is too big. Anyway, it's brand new card and I'm not gonna do any of this nonsense for a while. The only good fix for heating is good cooler. Thermal compound is at best a band-aid solution. Still it doesn't matter as I would need to replace my case for anything truly meaningful, but then again I would need something to put into case. I will just use it till it dies, but it's a shame that this blank screen nonsense is probably gonna stay (except after my 4GHz overclock it seems that it went away).

[hardcoregames_]

There should be no gaps, the surface to be cooled and the heatsink must me as close as possible.

TIM is not gap filler

[theredspirit]

Seriously, I have never seen VRMs or VRAM chips with thermal paste. There's always a pad. Even crazy Arctic GPU coolers always give you pads instead of paste for those areas. I'm pretty sure there's a good reason for that, besides cost cutting.

[hardcoregames_]

My old EVGA GTX 260 SC had thermal pads on the regulators. 

My GTX 750 only has a pad on the ASIC itself, same for my GTX 1060

[theredspirit]

What is ASIC on GPU? Isn't ASIC a standalone device for mining only?

[hardcoregames_]

the ASIC is the GPU part as opposed to the RAM and regulators

[theredspirit]

So it's just a fancy name for GPU, right?

[hardcoregames_]

The term ASIC is more general and it applies to the feature chip on modems, fax cards etc etc etc

[theredspirit]

Something like "Assignment specific integrated circuit". In this case it's just fancy name for GPU and more likely than not it can simply cause unwanted confusion. If we call things as ASICs, then pretty much the only non-ASIC things are CPUs (central processing units, historically used for various tasks, but later parallelized and accelerated to do them faster). I feel no need to use such words, when just simple graphics processing unit is such a handy term, which doesn't get deep into the underlying stuff. 

[hardcoregames_]

Motherboards can have several devices in them over and above the CPU socket. Even low end boards have sound chips etc.

ASIC is an acronym for Application Specific Integrated Circuit. Modems in the old days were simple analog devices but eventually the sector changed to using software defined radios. Software radios use a digital signal process (DSP) to handle the signals.

Ethernet is now a DSP based device to handle the faster speeds. Wi-FI also uses DSP for wireless over a noisy spectrum. Today the ASIC is more sophisticated but it is still designed for a simple task.

The video card sector want to focus their marketing on GPU to make it seem to be more elite. At the end of the day the GPU is simply a bank of ALU components in parallel. Lots of cut and paste in the design process.

[theredspirit]

Thanks, for answer, but I already know what GPU is. I actually have read books about hardware engineering and have had interest in computers for half of my life, so I had to dig deep into that stuff. Still motherboards aren't my strongest field of knowledge. I still don't know how exactly cheapness affects motherboard's design, besides less features, less money on UEFI, worse VRMs, worse cooling, less style. I barely know anything about clock generators or voltage delivery for components that aren't CPU. It looks like in this case my problem lies here. If you have some idea or advice say it, I'm interested.

I think that I may just be able to crank up voltage a bit and this problem will be gone, but anyway it mysteriously fixed itself after 4GHz overclock.

[pokester]

they do the pads because then they don't have to have the gpu, memory and vrms at identical height. The pads will smash down to be whatever the height of the GPU is. The pad suck but the core of them transfers heat okay but you get uneven applications across the surface. I literally just pulled my RX 580 off last night to reapply my thermal paste as the stock paste was dried and probably cheap stuff to begin with. It dropped my in game temp by about 8 degrees. Back to the pads. I reused mine. I put a little of my compound on the chips and vrms then across the tops of the pads. I'm sure they are making better heat transfer than they did before now.

Now about your other issues. I don't really think the following is your issue but you could try as it would be easy to eliminate the possibility. Use a different 8 pin  plug on that power supply. Usually if you don't have another 8 pin your card probably came with a molex or pcie to 8 pin adapter. You could rule out a rail issue on that power supply this way. Again I doubt it's the issue.

Now your machine could come down to being an issue with a motherboard. However most of us with RX 580's have stability issues at stock driver settings. Many RX 480 user who didn't have issues also now do since the release of the Wattman drivers. This issue is that the power ceiling is too low, the gpus run hot do to throttling even though the max temp isn't that high. In game you get screens go to black, you still will hear audio. The screen may recover and repeat this or you may get knocked out of game if the driver crashes then returning radeon settings to defaults. 

If this sounds like part of your issue I can help with that. This is a Power Ceiling issue and Temerature Fan Speed issue. 

Now the down side of what I will tell you is your fan will be more noisy, but the card will run better.

Go to Radeon Settings/Gaming/Global Settings/ Wattman / Global Wattman:

Go down to the " red Power Limit slider"  slide it to it's maximum.  They you will want to set a fan and temp curve to kick those fans on high much sooner. I will include pics of my settings on my RX 580 and it runs rock solid with them. It doesn't run worth a darn on defaults. 

if you have older pre 2019 drivers the settings would be like this:

Good Luck in figuring out your issue. Hope it doesn't involve more money!