Wattman is still so extremely bugged. Infuriatingly bugged. Broken. Multiple issues!
Not once can i get any overclock to hold since buying my Vega 64 Strix in February. By "holding" i mean when i restart my PC. Any one of the many settings in Wattman that have been changed like undervolting/overclocking goes back to default while other settings remain the same. Random settings at random PC restarts have gone back to default. This some times does dangerous things like having 50% power offset still work and the undervolt has gone back to default. I do not want my GPU to draw 420w+ and create that kind of crazy temps to match while gaming TYVM. That is after getting lucky when applying settings initially and them actually applying. Using profiles just makes the issues even worse. After knowing my card inside and out for over 6 months and still no stable overclock while using monitoring software for countless hours to learn this horrible behaviour. The 20% to 30% FPS increase i can get in some games is awesome. Especially when temps, noise and power usage/consumption are all improved over default.
The best that can be done to get a overclock that is stable is:
-Use a lot of monitoring software to watch things like watts and volts being used under load to indicate as best as possible that stable settings have even been applied
-Do a fresh full manual overclock every time in a specific order to get the best chance of having all changes apply. This will last as long as the software is running. Either from a crash or a PC restart
-Profiles themselves do not even have a reset option that works. Once it is a manual profile it is always a manual profile unless manually entering all default settings
To play a game while using an overclock i will always go through all the following steps. Which i have just made as a fast setup and stable for all games. Go in Wattman. Get all settings back to default by selecting either "balanced" or "turbo". The reset button does not always default all things. Those settings luckily do still work when NOT using a profile. Give states above 1050mv that exact amount starting from the lower states to the highest. I prefer 1030mv in 90%+ of games. My memory goes to 1150MHZ fully stable and no artifacting always IF the undervolt has applied properly. Anything competitive i no longer even do that specific light/stable overclock before each play session.
I see the same issues reported for over 2 years on all AMD GPU's released. I am ready for a refund for a graphics card i really do like and wish would stay the same settings i apply each time i use it. It is far to unreliable to use any custom profile in any competitive gaming since no overclock to this day will hold. This behaviour makes me like my old 290x even more. I see the same behaviour in things as far back as the 580 or earlier, all the way up to the latest Navi release being the 5700XT at time of this message. Basically every issue shown in the live Overclocking Vega 56 video by Steve from Gamers Nexus. That amount of work is needed for every time a game is wanted to be played and the faith in it not crashing is just as high always.
Please fix possibly the most annoying software ever and please fix the smaller but very annoying things like not having zero RPM option when trying to increase the fan curve. I always use default fan curve for this very reason. Only when the software/driver is fixed will i get another AMD graphics card. And that is the brand i want most if it will work. If not i will be sad buying Nvidea.
Sorry for your issues. Can't say as I don't share your opinion of Wattman. The drivers IMHO went from stellar to far from it since Wattman came on the scene. Unfortunately you are only preaching to the choir. Nobody from driver development or in management at AMD has ever been seen in these forums. They are USER TO USER ONLY. It does explain this in the obscure links in the driver headings. If you want to let AMD to know what you said I suggest you copy and paste it to the following support ticket link: Online Service Request | AMD
Turn off Fast Startup in the power options. It fixed this issue for me on the 5700 xt.
It basically is just a checkmark for using hibernate or not, which you can also disable hibernate via command line.
I never had a problem with it losing any of my settings on a restart regardless, only a shutdown. If you have that problem maybe its Vega 64.
Again, not a single issue on missing my settings since disabling it. I wish AMD Drivers wouldn't need to have us do that though.
Thanks for the reply. I did try that out and recently reverted it back since it made no change for me. I basically can do either a restart, shutdown or simply restart the Radeon software after closing it in task manager. Some settings i can find on occasions in monitoring software like GPU-Z indicate that not all things are applying depending on what it feels like that time.
It was many months of me even trying to figure out what was causing my random crashes after spending the entire day to be so confident i will never crash. Then the next day or even weeks later, crashing could occur. Realising that i was getting crashes on overclocks that i know are good and also from the very minor ones also made me notice something wrong. I went through all the things. Nothing seemed to affect it like different settings or temps. Without the monitoring software i would have just had occasional crashing that frustratingly seemed like i was using settings too high for such a long time. It wasn't too high and was the reverting issue. The GPU temp sensor shown by Wattman is about the only useless sensor i find. That temp always stays around the same with the default fan curve and i do notice a change in the fan speed once at operating temps. The LED temp feature of the Strix is another minor indicator to know that my overclock is fully applied and thus being stable. All other temps and things like volts and watts are all able to display the problem of things not applying. It also took me a while to do many PC restarts and figure out just the Radeon software needed to be restarted. I now do a PC restart if i get a graphics card crash of any type since on some occasions the software has not recovered so well always. Over repeated PC restarts and constant monitoring of software reveals the issue to me. I like to be around 260w using my undervlot/overclock. When i start the PC and then monitoring software at any time after something like a restart or shutdown. I can put the GPU under load in the same stress test or game and It might be drawing anything around 220w, 240w, 260w, 280w, 400w+ depending if it wants to apply all the overclock settings on that occasion. The VDDC is also a great indicator for undervolting i find. The behaviour of how high it jumps i can mostly tell if that has applied. Without monitoring software it is just a crash that will occur again and not much more is noticed. 1 time of the fans being reset after only 1 day of using manual fan curves was heard and seen. Temps of many things apart from edge temp displayed by Wattman with Vega 64 are way higher also but not indicated to me, until the card is at operating temp.
I should add that i can either have things not apply by it, displaying just that part of the overclock has gone back to default. Or it actually shows my undervolt/overclock the way i want them but is actually performing that under a default setting. That leads to a crash and from that point on it becomes a fresh lot of settings. It has very random behaviour and has happened with every setting the program has.
I can get a stable overclock that i am 99%+ confident with at any time. I just have to follow a lot of steps to be as sure as i can get, that all things are behaving like they should with the guaranteed overclock i am using. Not being 100% ever means that i only use default for competitive things. Definitely online sim racing with seasons lasting many months and a single crash during a race could easily ruin the entire season. And if i am wanting to use an overclock i want it mostly for the competitive stuff. I just have to put in by either manually entering all things and i must be using something like GPU-Z to test to see if everything is operating at numbers relating to the stable overclock. Once i have made it to 10 mins under load then that is enough time to know that if it is not crashing and never get a crash for the entire gaming session. I can go a bit further and save the profile but i MUST load that profile or i will not always get all things apply if left untouched.
I have wrote a lot about it already. Though i have just said enough to just show how to see the problem. Even me 6 months later of non stop trying to fix this still had many days that i said i am fully confident i have got rid of the issue. Then a few weeks later and crash again. The time spent learning this card to realise that some of the crashes did not make sense when higher ones are stable for weeks on end. Always using a 50%+ power offset, i had 1050mv and 1050MHZ memory. That works and is stable in all my games. Even when 1 of those settings is reverted i can not really even notice as i will not crash with one of those being turned to default. My memory is stable above 1150MHZ but having memory that high would result in a crash eventually. It was also fine to use any time i would apply it for that day. Heaven is able to crash when i use my lower memory overclock very weirdly initially. If my undervolt is not set properly then i will see my Watts not sitting at the 260W region. 240w or 280w being displayed by GPU-Z indicates to me whether the graphics card will do a crash if i push the memory between about 1050MHZ and 1150MHZ. If it is sitting at around 260w then my memory will go as high as artifacting. Which on my memory occurs at about 1155MHZ to 1175MHZ depending on the application. I then get very pleased if i am seeing memory artifacting instead of crashing way earlier which in itself was annoying to go though and figure out.
I like my card a lot and ideally i just want the overclock settings to stay the same always. If there is another AMD card that doesn't experience any of these problems ever. I will definitely take this card back for a refund and get that. I do not care if it is a lower or higher spec card. Paying extra for the Strix version is more annoying. I wish i payed less for the reference model. Many of it's features go to waste because of these issue. I can use zero RPM mode at least but i must use a default fan curve. I do that to and i do not see a point in having a better cooler. I even use the LED feature on it set to temperature to further indicate to me if i am running a stable overclock and can often save the crash if i am fast enough to change things. The default fan curve also allows me to hear and see the fan % increase above where it should sit for that particular overclock for added indication. Until i can have stable overclock with just a general setting then i can finally move on to possibly using gaming profiles and do any kind of fine tuning. I have many profiles saved by game name. I can use saved profiles loaded up freshly but i can not use any profile that is left unchanged beyond a program restart of some kind. I doubt i will make it to 10 PC restarts/shutdowns with out 1 of the settings reverting back to default. and may or may not display that reverting behaviour in Wattman.
I really hope they fix this quickly. I can totally see how this issue may not have been noticed by AMD before releasing the software. Even if they tested for 6 months they may not have noticed anything wrong and any reasonable person would give there full approval. I definitely did many times, some what recently. Changes made while the software is running do stay the same until the program restart, PC restart or shutdown. The amount of hair pulling to these issues just to start out with, was nothing short of many months until i realised a problem and also discovering what it was.
Some links to other people reporting the same issues i get:
Resolve the Wattman reset problem (1st paragraph)
WattMan profile resetting / loading (multiple people with the same issue)
Global Wattman custom settings reset after pc reboot
Same problem. It happens even when i just restart the program after closing it in task manager, restart the PC, Start the PC after a shutdown and crashes of any type. I had no luck trying all the suggestions people had both on this forum and elswhere on the net where many more people report the same bugs and problems
-I loaded my main profile that involves minimal changes and is totally stable inside the games i spend so much time in
50%+ power offset, 1050mv on P5, P6 and P7 states, 1150MHZ memory
(no other changes. P4 shows 1050mv as default so 4 of the P states all show 1050mv. This is not stable in every application. The things it does not work in like Heaven requires just a 5 to 10mv further adjustment to smooth out the load line to make it fully stable in those applications. Between those 2 profiles is where i can get full stability in every application i have ever tried)
-I was in a game for about 5 hours and not an issue at all. (As expected)
-I shutdown the PC for the night
-I turned my PC on
-I logged in to the same game
-About 1 minute in and the game crashed as if it was an unstable overclock
-I logged in to the game again and about another minute in and it crashed again
-I restarted the PC. Opened Wattman for the first time since setting the undervolt/overclock. It shows all things set to default as it should after a crash. I loaded the same saved profile again. Went in to that game and not a problem at all again.
-I open Furmark and GPU-Z to see my watts sit around 240 quite often after my PC is turned on. (bad and expected by me now)
-I am getting my 260w+ which is my indication that this is the correct undervolt/overclock i know that is stable and is also what i initially made.
-When i am seeing the power usage at around 240w i will also see the VDDC not even go above 1.000v
-When i restart my PC it is common for me to go back to the 240w region and volts will be quite low compared to what it should be. Thus resulting in a crash and that also removing the bad overclock from being seen.
-When my VDDC is sitting around 1.02v to 1.06v then i am able to tell from that also that i am likely to not crash
A lot of things do not make sense but getting this far to learn all these things allows me to at least know what i am always in for. I intend to play a game and again i am stuck fixing this. I have not renewed some gaming memberships or played those games since getting this graphics card after my 290x died. I have to somehow give up on it and just use no overclock at all. This never ends and has been like this since day 1, 6+ months a go. Or maybe at most, i use the settings i used to use. The same as my main profile but with 100mhz lower on the memory (1050MHZ). That keeps me out of the weird crashing region regardless of what MV i use. Not by coincidence, It does not have any crashing during months of testing. And does that by not caring if the undervolt has gone weird. At 1050MHZ there is no crash regardless if i change the volts to anything at all not unreasonable 950mv to 1250mv i know always works. I haven't needed to test below 950mv as the VDDC stays above that under load even when it is showing low.
Because of all that and not being able to overclock like it should. Especially when it cost extra to have a Strix version and i can not get any benefit at all apart form having zero rpm mode on the desktop. The cooler i find pointless because i need to choose between zero rpm mode or using a custom curve. I can not use both since AMD has that greyed out if not using the default fan curve. I am glad the default fan curve is good enough but my case fans totally silence the graphics card always. I still have almost 6 months left on my Warranty. I am "fairly" confident the PC shop i got it from (PCCG in Australia) would agree this product is not meeting the standards that it is required to be. This is not at all the product that i expected. I do not know entirely if i will return it. Being in this position still 6 months later is probably enough to say i should. The timeline of it being reported a good 2 years back and no fix for the time i have had this card is probably a good indication i can most likely never expect a fix for this. I do not know what graphics card to replace it with. I would pay extra to get Navi if that definitely did not do the same thing or the PC shop agree to a refund if i were to encounter the same issues ever during the remaining warranty period i already have. The problems i have with my card are well beyond reasonable. Currently i have no choice to use Nvidea. That is horrible since i use eyefinity triple screen setup and would lose a lot of that functionality.
Even if i got nothing fixed at least others may be able to use some of this info to save a lot of the time wasting.
If anyone has a graphics card and knows for sure that none of this behaviour ever happens over say a few weeks of heavy use. I am very interested to get it.