General Discussions

[kidwaved]

My current pc build - 

GPU - RX 5700 XT

CPU - Ryzen 7 2700

RAM - Corsair Vengeance LPX 8gbx2 3600 mhz

Cooler - Corsair H100i Liquid cooler & stock nzxt fan

Motherboard - MSI MPG X570 Gaming pro carbon wifi

PSU - Corsair RM750

I don't play many games as I'm quite boring so I stick to mostly Call of duty games. Here and then I will play some single player games however I've had enough with the crashes. Ever since I switched over to the RX 5700 XT from the gtx 1060 I've been having constant issues all the time. Don't get me wrong, performance wise this computer and this GPU is amazing, but it can't keep a game running for whatever reason. On cold war recently I've been getting constant crashes which is anoying as I keep losing progress as my game shuts down midgame, I'm sure you could understand my frustration. In order to fix the non-stop game crashes overtime I have kept upgrading my parts thinking it was that the GPU is too much for the rest of my PC to handle. But now I'm at a point where I have spent a lot on this computer and it's amazing, but I can't put my finger on why am I still getting crashes.. I hope someone can give me a permanent resolve to this.

[mackbolan777]

I'm in the same club. I have a Gigabyte RX 5600 XT with a similar issue but I can play for about an hour. I did notice the 20.8.3 driver is most stable so far. Stupid question, but did you run DDU graphics uninstaller before installing this card and install the latest driver or the one you prefer?

You can find older AMD drivers on here by searching "AMD previous drivers". One other thing that can cause this is a faulty PSU or connecting the PCIE to a peripheral output off that model. I have the RM1000, older but specs are fine. The "RM" line had a high failure rate. HWINFO(free download) is a good tool to read voltages while stressing the GPU or anything else. Within +/- 5% of 5/3.3/12v under load is ok. 

 If this is like mine, mid-game, random, I'm still stuck. My last post was about using my DisplayPort for video and an HDMI for audio only as a possible cause. A few months back crashes were bad and it was due to improper DRAM timings, which I fixed and it helped. You could try moving the card down a slot to see if that changes anything. I think that board has 2 PCIE slots. 

Heat is another issue. I noticed if my VRAM hits 70c, I crash or green screen. So I use the "manual" OC mode and disable "zero RPM" for the fan and set a custom curve. I also undervolted the GPU to 960mv but the 5700 XT needs a bit more I think. Undervolting some can reduce temps quite a bit. There are guides on how to do it or select it in Adrenaline "manual" performance then input that value, it should work fine. It will only calculate and do one OC thing, like "undervolt GPU" , "overclock GPU", or "overclock VRAM". To get more than one setting, including fans, you need to run the card in "manual" performance mode and input increased or decreased values. Even stock clocks I would set zero RPM off and undervolt the GPU. The GPU gets way more juice than it needs causing extra heat with zero gain.

Some thoughts, hope one helps.

[artur_aragao]

Folks!

I discovered something interesting here and it doesn't really seem to be my VGA problem.

I uninstalled the drivers and the VGA works stably. Of course, I don't have the features of the standard driver, but this already seems to be a good indication that the cause of problems is the Adrenalin driver.

I ask you to do the tests with the generic Microsoft driver. So far without black screens.

[artur_aragao]

I am amazed at what I discovered. I know that the friend had already recommended this to me, but now I'm on my way.

My setup is without black screens with version 20.8.3. Sinister!!!
Enter the system, rest in idle and return. Super stable.

Some nonsense AMD did with the drivers. It's fact!

[artur_aragao]

Arrrrgggghhhhhhh !!!!

Same **bleep**!
I'm tired of it!
I will return this video card.

[mackbolan777]

Get a better brand. MSI makes "pretty" things but their quality has been lack luster to say the least. Go with a Gigabyte, never had an issue with one except fan noise on one at 80% or more. XFX THIC III Ultra, great from what I hear. Asus makes a decent card too. Sapphire is over rated, Power Color, meh. But MSI lately is just awful. But you're plagued with an MSI board too.... running 32GB at 3600Mhz, try turning that down to 3200, especially if 4 sticks and see if it stops crashing. To run 4 sticks at more than 3200, you need a T topology board not that daisy-chain you have. Even 2 x16GB is tough on that type of board with a Ryzen. Think X570 chipset boards or expensive B550's for a T topology board. Or search "AM4 motherboards with T topology 2020", something should pop up. Anything made by ADATA is suspect. I've experienced intermittent performance issues with their SSD's, where I RMA'd 2 of them. I have older PC's with those sticks just intermittently mess up but pass testing. Sorry but I see some suspect things there, besides the card that would be intermittent. RAM brand, size, number of sticks, speed, board topology, brand model of board...

[delevero]

A guess try this

1). go into bios
2). select advanced mode ( F2 key )
3). find CPU Vcore Auto and change AUTO to 1,4volt
4). save and exit
5). try to play your games again.

The above made my own computer able to run games that i could not play yearsterday and i had 0 crashes the last 5-6 houres i play hitman 3 and another game.

[mackbolan777]

Setting Vcore to a solid 1.4v is insane and the next crash you'll have is a black screen with a no boot because you fried the silicon. You have other issues. Unless you're trying to heat your room, proceed.

Vcore depending on CPU should fluctuate based on the CPU frequency. The only time that gets set to a fixed number is for an all core OC, which is not typically cool to do on a Ryzen CPU.

Better off setting your PBO enable, setting your voltages to "motherboard max" or "motherboard" instead of different Amp settings or "Auto". Setting the SOC to 1.10 on 3000 series Ryzen's seems to stabilize then running PBO, even PBO +200. Setting LLC to in the middle or 3 if you have that setting or "allow OC voltages" is ok as well.

All of that is assuming you have a decent cooling solution. AIO or closed loop preferred. RAM speed, IF and RAM sub-timings can improve results greatly as well. 

[artur_aragao]

I've tried everything.
Nothing so far.

My source is yet to be delivered.
But I want to try something else, because I cannot accept that it is a problem from my source, just that it has worked with an RX480 for so long without any problems.

For a Ryzen 3000 what would be the best VCORE configurations. The others you mentioned only identified the SOC Voltage at 1,100v. The rest do not know where to find on the MSI mainboard that I have.

[delevero]

I dont think that a Cpu Vcore of 1,4v is insane for testing if it will make your pc stable and then you can reduce your power from there. Its a setting people can try and its withing safe margins.
Example if you set cpu-vcore in bios to auto then the automatic contant changes voltage froom x and all the way up to 1,55v
So 1,4 volts is within safe limits.

Advertised "maximum safe" values on the internet are 1.375V for Ryzen 1000 and 2000, while Ryzen 3000 tends to be 1.325V under
I dont know the value for the ryzen 5000 series but i asume its higher. ( the info i found say max 1,55v )

[artur_aragao]

I'm hitting the mat VERY disappointed. Really much.
1- I validated that only with Microsoft drivers does VGA work normally. Put the most current drivers 21.2.2 and the same zika occurs;
2- The source has arrived. Guess what? Even with it, the same thing happens. So, I bought a new font without needing it. I now have two. ;
3- I removed 2 clips from the 4 of the mainboard. Nothing! The same thing;
4- I removed the RX 5700 XT and put the RX550 and everything returned to normal.
After these bad experiences, I don't want to know more about AMD for VGA. I think you already know? I'm a fan, but things have limits. I don't like nVidia and I will be obliged to buy a VGA in the future when I can, as the current prices are not possible to have one.
I appreciate all the patience and attention.
I will finish filling out the form and forward it tomorrow. I want my money back and hope what I can get from nVidia.

AMD for GPU never again.

[cristi_trohin]

@artur_aragao There are a couple of things you can try...

If you want to stick to Adrenalin, try setting the boost clock (P3 in Tunings) to the RX 5700 XT spec clock, 1905MHz. By default, Adrenalin may set it higher.

Another thing to try, if you're willing to ditch Adrenalin, is to DDU the current driver, and restart Windows. Then wait for Windows to detect the hardware and install a driver from the Microsoft repository. It'll be an older one for sure, but it could be a stable one.

This is what I did yesterday when my son suddenly started having crashes on his RX 580 with the latest drivers. The driver that Windows installed had no issues, he played all evening without a crash.

[artur_aragao]

Friend,
I am grateful, but it was not for all these gambiarras that we were testing here that I bought this VGA. I gave up, I'm already forwarding it back and I just bought the 3070 from ASUS.
That's enough for me from AMD for graphics for quite a while.
I say this with a torn heart.
AMD needs to respect its customers more and what it has done is a total lie and lack of commitment. They need new engineers in the field or better hardware design. Something has not been said.

Good luck to all friends.

[cristi_trohin]

@artur_aragao Good for you mate! I would return my card as well if I could.

[artur_aragao]

Thanks!

[mackbolan777]

@delevero Your problem is promoting unsafe testing methods based solely on internet frankly, BS, and not facts. Most of these tests are done using LN2 or on "disposable" hardware. 1.4v, as you had mentioned it's use in several post that I flagged, is misleadingly dangerous for a "regular" user to do. They risk serious degradation of the silicon, VOID their WARRANTY, can burn out other hardware as well and your original post/repost was to just set it and forget it, not "test" anything or to navigate to a much lower, safer voltage. 1.375v constant Vcore is NOT a safe 24/7 idea, it will degrade the silicon, any all core OC will. Ryzen CPU's are not designed to do all core OC, that's why id you really did testing you would see IPC's dropping over time, the heck with benchmark scores, IPC's. "Internet advertised" safe limits by whom? Buildzoid? Even he will tell you he ran his 3600X at 1.375v for 4 months until it BSOD'ed nad wouldn't even post at stock clocks.

Just because you see in HWiNFO that your cores hit 1.4-1.48v or maybe a little more, it's per core up to 3, not all 6-16 cores art one time and it's for a second. More voltage has proven to be a mistake with Ryzen CPU's, especially the 3000 series. The lower the better. In fact, if you WERE to do an all core OC the way to find your max "safe" OC Vcore is to run Prime95, with all back round stuff off. The highest number on the left is the max voltage and it's usually ~1.28v, some rare chips might show a steady 1.325v not 1.375v. Then watch temps if after running that for ~30 minutes and you aren't going over ~80c, you might be "ok" to use that clock/Vcore. Even so, not allowing the CPU to cycle as intended will eventually degrade the CPU. How long? It depends on the chip. Minutes, hours, days, months, but less than a year. 

Never compare a Ryzen to Ryzen 2, they are completely different technology. Both handle memory differently, have different Vcore, handle all core OC differently and more. No the 5000 series does not want more voltage, in fact the whole performance ramp is designed to undervolt for higher gains.

So do us all a favor and until you have over 20 years in the business of building, testing various components, please stop posting irresponsible web based garbage. When or if you do gain that many years you won't think of posting such non-sense.

Go ahead and put 1.55v to the Vcore and see how long your CPU lasts, how long before the IPC's drop into a non-recoverable loss or you melt the socket. You go past the limit on an AMD CPU and there is ZERO forgiveness. Once it's toast, it's toast. I do not think anyone on here cares to replace their CPU because the warranty is voided by your instructions on how to abuse the chip.

Anyone got ~$300-800 USD, depending on CPU to waste testing this OP's theory's?

Raising Vcore has never stabilized any build I had an issue with, especially when not running an all core OC. Even with PBO enabled and +200, technically warranty voiding, raising Vcore doesn't come into play manually. AMD's built in chip microcode handles that variation perfectly. The CPU is capable or it's not, that simple. It's also not guaranteed to be able to use those settings, that's why it's on "auto" by default.  

[cristi_trohin]

@mackbolan777 I agree that when testing system stability, whether it's related to the CPU or the GPU, one should not increase the vcore. That's the thing you actually need to rollback if  you find instabilities. Upping the vcore is a last resort measure if you want to achieve a high overclock, but it should be done with the acknowledgement that it can actually break the hardware. Too much vcore is not an answer to system stability issues, especially at stock clocks. If the system is not stable at default vcore values, it could be due to faulty hardware, which should be RMA'd.

[artur_aragao]

And why not 4 modules? It's in the QVL of the mainboard.
It works steadily for days with an RX550 and an RX480.
Rams clock at 3600MHz for days. I put this **bleep** RX 5700XT and the plagues of crashes and WHEA errors infest my PC by turning it on and off.

I really appreciate the help.
I see that MSI's support has declined in quality, but the products never let me down.

I will try to play RAMs for less or even disable XMP.
For me, it is the source that cannot deliver power to the VGA. I'm surprised, but that's what it looks like. I'll only be sure tomorrow or Friday, when the new source will arrive.

[Anonymous]

Hallo 

ich hatte das Problem auch und habe es so gelöst.

Im BIOS den PCI Express von Auto auf Gen 3 stellen, in der Radeon Software Anti Lag und Free Sync deaktivieren bei allen Games!

Seitdem nur sehr selten gamebezogener Absturz ohne Black Screen...

[cristi_trohin]

@Anonymous what version of drivers are you using? The black screen issues were resolved by AMD last august, if I remember well. Try using the latest recommended driver. Optional drivers do have a tendency to mess things up.

[mackbolan777]

My German is a bit rusty but full PC specs help. Also check your PSU, if you have 2 PCIe power connectors on your GPU, use 2 cables not the split type.

[artur_aragao]

Esteemed,

I feel for you. The WHEA problem is related to n variants. Some cases are the CPU, others the VGA, memory, source, cables or drivers. It sucks to do the identification. For that, just patience.

After 2 months with headaches caused by problems with the VGA RX 5700XT, I decided on two well-thought-out days to go to nVIDIA with the ASUS ROG STRIX RTX 3070. I have no regrets. It was Plug n 'Play! Plugged in, installed drivers, didn't even need to configure the drivers, played. That simple.

I will not return to AMD on VGAs until it improves design and drivers. I do not indicate for ANY VGAs of architectures derived from NAVI (v.1, v.2 and v.3).

I removed the RX 5700XT, put the RTX 3070 on and all my problems are over.

I want God to direct friends where the problem really is and that they be resolved as soon as possible.

[mackbolan777]

So far my RX6800 is solid. With prices of cards and lack of availability, many want to try anything other than spend ~$1500 on a new card. Even the RX 5700 XT is going for ~$1k USD. Used around ~$800. So this is why most will fight as hard as possible to find work arounds, etc. As far as not returning to AMD, that's your personal choice and best of luck. I'm sure that Nvidia will do just fine. It's faster than the RX 5700 XT but stuck at the 8GB VRAM. The future is already moving towards the 16GB VRAM for VR and in some cases only a RTX3090 will pass for 8K. AMD will need to step up the game again.