AMD Radeon Adrenalin Edition 20.5.1 driver download & Discussion

For a margin difference. I had Vega VII, now have 5700 XT, and in the future i will have big Navi (I have the money ready )

 

16.6k gs in firestrike and 4.8k gs in timespy with a rx580. Didnt notice any change in scores with these and windows 2004 update.

 

A boost would probably occur only once AMD gets the HW scheduler to working.... it's currently disabled. Pity cause enabling it on NV and Intel did lead to some improvements....

 

I believe amd will implement this new feature better than the completion that's the reason for pending release. Do not forget new gpus this year. Amd will deliver new architecture without issues for the first time almost after a decade.

 

You guys should read my thread and take note of it. These drivers, along with the other 20's driver versions, WDDM 2.7 there's no difference in performance, and using drivers from 2019 below 20's, are about 15%30% faster compared to the newer ones lol. I'm using some enterprise 19.Q4.1 for my 570 8GB card. I get double the performance in some games. What might work for my setup, might not work on y'all, but who knows.

But if their older drivers out perform all their new ones lol, there's an problem hahaa

 

that's not strange... they stop optimizing older cards and might even drop performance on them

doubt you'll see same results on 5700xt

 

Oh yeah the famous AMD finewine... *disc scratch sound* oh wait
I keep seeing the "it's anything,e.g, PSU, windows 10 not being clean...etc except a driver issue" replies when the user having the issue/s states that going back to a previous driver fixes the issue/s... I'm starting to think it's a running joke or some kind of parody

 

Well i have no plans to buy a 5700 series card, so i wouldn't know. But i would say for anyone using a RX 570 to use an older set of drivers, if i see a big performance increase compared to the newer drivers, they should see it also.

 

I have no need, nor the money to buy a new gfxcard, still on a sapphire R390 Nitro+ /w backplate, don't see much reason to upgrade, as I'm old, and play old games.
With that said, this drivers, or the win 10 2004, I feel it's smoother.

 

Which are the ones with netter performance? Those drivers have wattman? Or Image Sharpening? Which W10 build?

 

Thanks for the reply, though, i think i finally found the issue. To my surprise it has nothing to do with AMD, it just started happening at the time i got the RX 5700.
I found some other people that had similar issues, and some of them tracked it down to Superfetch aka SysMain. I was pretty sure i disabled that function a while ago, but i appears that perhaps a windows update activated it again and after reinstalling windows it of course stayed on. Seems like iRacing and Superfetch don't like each other, since iRacing uses RAM a lot and generally saves data during a race. I've done three races so far and i didn't have an issues, will do more races to confirm.

In case it really was SysMain, i have to admit i blamed the wrong company :>

 

If it happens over time chances are the issue surrounding cached memory could still be a thing even on the newest version of Windows 10 where it's slow to empty this out so even on high-memory systems and newer systems with faster memory and other components this buildup can gradually introduce stuttering over time until a reboot.

Or until you use a utility such as EmptyStandbyList.exe to empty out standby / cached memory though super-fetch will immediately begin to cache running applications and the prefetch list though the buildup will be less than whatever you have in the task manager.
(Or how to describe it, handy way of measuring and checking what's built up.)

Two days of uptime after the latest reboot and I'm sitting at 12.5 GB cached with only some lighter OS activity and browsing so about half of the available 32 GB already but memory is meant to be used though there's plenty on the subject of this issue as it builds up to the limit of available memory and then doesn't empty out immediately or fast enough thus stuttering until forced either from clearing it out or a reboot.


EDIT: There's tools to automate it too which I suppose you could use, I prefer manual usage as it's there to be a effective way to speed up applications but something that could hit 10 - 14 GB usage over time on a 16 GB system would possibly hit this limit and then there's leakage, page file leakage and other problems or just very RAM heavy games or software in general.

Improper usage of D3D and trying to cache or fill or reserve ridiculous amounts of memory and more. (OS manages it in most cases, low-level API's like D3D12 or Vulkan probably expects that the developers follow new and recommended practices though.)

 

You can read this thread i posted
https://forums.guru3d.com/threads/h...-amd-adrenalin-drivers-check-this-out.432500/
Radeon Enterprise 19.Q4.1 drivers.

Yes it has Wattman and Radeon Image Sharping. I'm using Windows Enterprise 2004 19041.264

 

normal

 

The fine wine thing seems real. Plenty of documented evidence of the 580 overtaking the 1060, sometimes by a large margin, over time. Those two cards were quite even for a while. Setting a custom fan profile really does wonders for the temps on my rad 7. the amd driver prioritises noise over heat, which given the 7's deserved rep for being noisy is fair, but junction temps can exceed 110C, which is nuts. custom curve is a bit noisier but temps are much more reasonable.

 

Same goes for Vega cards they are both faster than their pascal counterparts. My 580 have similiar scores in benchmarks to 1660 non super. Question is will Navi have the same finewine but so far amd did a nice job improving the drivers and performance.

 

Is it just me or is there a vsync/enhanced sync bug in this driver? Tried a bunch of games like COD WW2, Tomb Raider and Deus Ex Mankind Divided, they all run at unlocked FPS. I have enhanced sync disabled.

 

I assume you are sarcastic.

There are various parts of the code that are really old and are broken for 9+ years already. After such long time I believe that only writing it from 0 has some chances to fix it, main problem is that AMD doesn't want to allocated the needed resources for such a task. And what was "rewrote" is plain awfull (don't even want to think how Linus Torvalds would react if he would see the code, good part is that it will never happen).
There are things added on top of a code that was problematic from the start. Things are not gonna improve until you fix what is broken, more and more weird things are gonna happen. It's also important how that code is wrote (if it's wrote in a similar way with how I code (no comments, code not structured, weird jumps) and if the person that wrote that particular code is still working for AMD; the idea is that the code is not wrote in a friendly way and the persons that wrote it no longer work at AMD (even if you hire back that person if years have passed since that person worked at that code he/she just no longer remembers what he/she did so it's useless) you better rewrite it than try to fix it).


People say about possible hardware bugs. Sure it's a possibility but that also say that they didn't really did proper testing on the prototypes (won't be a first, faildozer is another case when they just pushed on the market a product that well couldn't compete if you needed high single thread performance something that at that point was needed for dx9 & dx11, sure the multithread performance has it's own uses; from my point of view AMD just refuses to learn from past mistakes, basicaly similar with the 4 years old kid of my ex-girlfriend that was doing the same crap over and over despite the fact that it was painfull for him...).
Some of the hardware bugs can be fixed with microcode updates. In same way you can update a cpu microcode after windows has loaded you can do it with a gpu (basicaly the driver will load the newest microcode for your particular gpu). So if you trully want to fix something you can do it.

Main problem with AMD is that in the attempt to be competitive is rushing products that are just not ready for release (something similar with gaming industry, when some games are basicaly still in beta or even alpha state for 6-12 months after official release). When you do such thing quality basicaly sucks and you basicaly show no respect for your customers (in the example with the games they at least fix most of the problems, in amd case they don't really do it, every single gpu from red side I owned had a major not fixed problem, it went to the point I just decided to assume that this is the way AMD is doing and completly moved away from them).

There is also the lack of communication. NVIDIA does it way better. Going silent and not fixing the problems is making you a really bad image because the only logical reason is that you just don't care. You can improve your image by just creating the impression that you are working at a problem (when in fact you are not). It's all about what you can do with the client mind. if you make the client believe that he/she is important for you he/she will buy your stuff, if you make the client believe that he/she is irrelevant for you he/she will move away. This how you can have lower performance than competition and still sell better and have higher income compared to competition.
I'm willing to pay more for less performance as long as I don't have problems. With AMD sorry to say but the time I had to spend to try (try cause some are unfixable without full access to the driver code and full access to specifications; and ask yourself what is cheaper for me fix AMD code with AMD not helping at all or just buy a relative cheap NVIDIA gpu enough for my low needs) to fix the problems I faced is way more than the most expensive product AMD has/had.

Don't get me wrong. I'm willing to help you. You ignore me completly for years. At some point my patience will run out. From that point you can even send me your products for free and I will just refuse to receive them. When my patience runs out it's all over and there is nothing that can be done, I just no longer want to give you a chance (because you wasted every single change I gave you, so what could make me believe that you wouldn't waste the next one as you constantly did it?!?).

I know some people believe I'm an AMD hater. I'm not. I just used my logic that was all. And logic is cruel.

CGN 1.0 is abbadoned for years if you want to something use newer than Windows 7. This driver has the same problems I face since 15.7 , not that there is a single driver decent. I won't even bother to list them anymore because won't be fixed until support will be officialy removed from driver for CGN. They will say the lie that the architecture was optimised to max possible, when well they never really bothered to fix major problems.
Don't worry Linux situation is bad also. I have to selfcompile Mesa with a particular change not to face screen corruption. Apparetly Mesa devs consider anomalies on screen as acceptable, I don't. Really hard to just make that zerovram default behaviour, because there is screen corruption that can't be avoided else simple as that. Specs say nothing about this behaviour so you can do whatever you want, when you clearly have screen corruption without then there is a single solution. If they don't want, they don't want, they happy closed my bug without a single reason and without fixing the problem, you can bet I will not report anymore bugs in their case. I'll fix the crap on my own (at least I have the source code) if I can and that's it, they won't upstream my fixes anyway or come with another solution so no real point help.

When the HD 7750 will break I will just replace it with an gpu from competition (assuming intel igpu won't prove to be enough). Way too much stress for me to try to fix AMD crap. I could had used that time for way better things.

 

Anybody gets crush on CoD WWII with texture error just after cutscene while loading any mission? Also got on CoD MW Campaign Remastered couple times. (Windows 10 2004)

 

Maybe though we'll see.

(Via Wikipedia.)

That landed in WDDM 2.6 with the 1903 / May 2019 update.

If the info I read was accurate then AMD still doesn't support this in their drivers.


This feature here might be more important though, course you don't want thread conflicts and stalling and race conditions and the resulting stalling, micro stuttering or hangups or other hiccups so letting the OS (hopefully.) do it's thing and doing it well yeah it could have a significant impact after all.

Some software support and decent implementation probably still required though, titles like what Monster Hunter World (used to do?) did with just creating 32 threads instead of something more proper or aligned with the physical CPU capabilities or the coding relying on not breaking thread priority importance.

Or the user not breaking thread priority by assigning a manual mode through the task manager affecting things instead of separate threads in the process, can be important as well although more so for later titles I presume using far more threading and relying on up to hexa or even octa core CPU's to meet said demand on CPU (And when supported, GPU.) multi threading and it all working. (OS, drivers, software the whole thing.)


EDIT: Though if AMD's drivers are still primarily single core bound or maybe dual core at most then yeah there's less to do here for stuff like this.

Although DX12 and Vulkan would be a different affair.
Or could be, software side and developers being more in control might decouple some of this from the drivers and OS and let the software (Hopefully.) get it right.
(Someone thought it was a good idea not just on paper, in time though it'll probably prove to be.)