Radeon Software Adrenalin 2019 Edition 19.11.3 Download & Discussion

Radeon Software Adrenalin 2019 Edition 19.11.3 Highlights
Support For

• Fortnite™
  • DirectX®12 API support

Known Issues

• Radeon RX 5700 series graphics products may intermittently experience loss of display or video signal during gameplay.
• Radeon RX 5700 series graphics products may experience stutter in some games at 1080p and low game settings.
• Performance Metrics Overlay may cause stutter or screen flashing on some applications.
• Toggling HDR may cause system instability during gaming when Radeon ReLive is enabled.
• AMD Radeon VII may experience elevated memory clocks at idle or on desktop.
• Performance Metrics Overlay may report incorrect VRAM utilization.
• Invoking Radeon Overlay may cause games to lose focus or become minimized when HDR is enabled within Windows®.

Important Notes

• AMD Ryzen™ Mobile Processors with Radeon™ Vega Graphics FAQ for Radeon Software Adrenalin 2019 Edition can be found here.

Package Contents
The Radeon Software Adrenalin 2019 Edition 19.11.3 installation package contains the following:

• Radeon Software Adrenalin 2019 Edition 19.11.3 Driver Version 19.30.31.18 (Windows Driver Store Version 26.20.13031.18002)

The Radeon Software Adrenalin 2019 Edition 19.11.3 installation package can be downloaded from the following links:
By clicking the Download button, you are confirming that you have read and agreed to be bound by the terms and conditions of the End User License Agreement (“EULA”).
If you do not agree to the terms and conditions of these licenses, you do not have a license to any of the AMD software provided by this download.

Download:
https://www.guru3d.com/files-details/amd-radeon-adrenalin-edition-19-11-3-driver-download.html

 

-> https://twitter.com/CatalystMaker/status/1196517363404939264?s=20

Faster than 11.1 & 11.2
around +100pts
-> https://www.3dmark.com/compare/fs/20988662/fs/20952523/fs/20939208#

The Outer Worlds = OK
Jedi FO = OK
Forza H4 = OK
Terminator Resistance = OK

 

Meh.
Good for those that play Fortnite.

 

19.11.3-KB. Links, Highlights

 

Just Installed 11.2. good thing im not playing fortnite.

 

Fortnope. Will install the driver in any case, thanks for the heads up!

 

Whats Fortnight?
Installing now...it usually takes a Fortnight to get these new drivers working.....

 

I'm confused, didn't the previous drivers support DX12 games already?

 

Support = Optimised

 

Is anyone manually OC their 5700 XT with these drivers?

 

Going back using (for the moment) Normal GPU's, they seem too update the drivers every 2 weeks. With the Workstation GPU's, every 2-3 months.
Drivers working just fine, not getting anymore random blue screens using the previous ones and this one. Thanks.

 

Same settings as I've been using for a while though once stability is confirmed in newer drivers I will see about hitting the actual limits of the GPU.

Default the settings and save as a XML for comparison test a few games and monitor either Wattman, GPU-z or similar what the card is doing, temps and all that.
Enable manual mode in Wattman and enable overclocking and the graph should calculate from the bios GPU settings.

In my case this resulted in a target of 2100 Mhz at 1260 something mv.
Bioses can also be checked here for example: https://www.techpowerup.com/vgabios...X+5700+XT&interface=&memType=&memSize=&since=
(One's the silent bios and one is the default and for the Pulse with 4 listed entries two should be the 5700 version and not the XT model.)

Compare with what the average reported clocks were for the game or games tested and then for how I did it used that as the maximum boost clock value.
Test again see that it's hitting this value now and how it behaves, generally the boost will be above and below with 50 Mhz or so sometimes a bit more with some spiking to higher happen on occasion. (These are important for voltages.)

Initial testing landed at 1900 Mhz and from that starting on voltages down from 1260 to 1050 reducing by 50mv at a time and testing which was stable at least until The Outer Worlds released and it hit a driver crash very quickly.
Upped voltage to 1100mv instead which was stable, game itself is quite crash prone though but it no longer takes out the driver or DWM on hitting a GPU crash.
(Also pushes the GPU temps to their limits even with a higher fan curve starting at 30% and going up to 70% where it starts getting quite audible.)

Not strictly required and it doesn't actually work too well in newer drivers but I also set a minimum speed of 1000 Mhz and 900mv and the mid ("game") clock gets to be 1450Mhz in between min and max boosting.

Power slider next after that dropping it down to -20% first and comparing and noticing a drop in how much the card is hitting landing around 1600 though even a smaller decrease shows a immediate drop here though it does certainly lower the power draw and the reported junction temperature by a bit so tuning can be quite useful here.

Up next and at full 50% and then 20% seeing no differences with the card hitting around 1950 Mhz and then at 10% it drops a bit so just set that to 20% to remove extra heat and reduce the power draw which doesn't really do anything for the card unless I could stabilize a higher setting at 2050 to 2100 but without a lower ambient temperature that's not happening.


Left memory alone because it's sensitive and the various titles all react a bit differently some crashing and some showing artifacts and it might even cause issues when just being idle on the desktop with the display cursor getting corrupted as a first indication before the GPU hits a loop of recurring DWM crashes and driver resets after that.
Up to 900 - 920 does show a gain though when it's stable and if ECC doesn't start trying to compensate but it seems above 950 can also work via soft power play registry tuning and there's just some ranges in memory that are unstable but higher could be fine so the 3 - 5% or so performance potential well I just kept the defaults and that will be good enough for stability as the important factor here.


Fan curve well manual 30% and up to 70% is how I ended up with but it's sensitive often overshooting and boosting at random and being a bit weird in RPM and % for actual speeds which going by reports from Igor's Lab this broke in 19.7.3 and hasn't been fully reliable since though it is serviceable and better than Vega was for the stepping effect between different points here when trying to build a fan *curve* out of that.

Also noticed the fans dropping below minimum if I go full General Winter and open the door and window letting the GPU temp drop to around 20 Celsius idle instead of 30 and chilling the entire room down to kinda cold levels which was interesting to observe almost going zero fan despite this card not actually supporting silent running from it's bios settings.
(Tri fan GPU's and overall better airflow and it might allow for a nicer reduction in junction too and also memory junction temperature.)


Overall I've cut down 160 some mv from 1260 to 1100 for GPU core temperature and GPU power is up from 195 default I think it is (180 something for the silent.) for this GPU and 20% so 230 ish also only for the GPU core so say 250 something total maybe a bit more but still less than what I was doing with the Vega though I could have lowered it there too I suspect. (Around say 350w total I think when going full +50% and some undervolting.)

Performance wise well I'm still hitting the clock speed so nothing really changed but for those able to hit 2000+ Mhz there's going to be a reduction of 3 - 6% or so from title to title depending on total performance and what the card is doing.
Can also go upwards of a 8 - 10% performance hit and really reduce the power draw and voltage allowing the GPU around 100w at -50% though it will reduce the boost clocks to well it's going to vary but 1400 - 1500 maybe a bit higher.
(Silicon and all that, varying performance and results after all.)



EDIT: It's simple repetition and re-testing I suppose, nothing super advanced but default settings and checking what the card is doing which can also include getting readings for temperature and such and then a max boost value from that possibly a bit higher or lower depending on intent and slowly working the voltage value down.

I stopped at 1000 and a bit extra for stability (I thought.) and so 1050 initially and after hitting instability up a bit more and that seems to be doing it.

950 to 1000mv is what I've seen others use around 1800 - 1950 Mhz so lower can be possible but might not be 100% stable everywhere so testing is helpful but eh overclocking in general requires testing for stability so nothing new here even if it's undervolting that is what I am going for.
(There's no overclock possible when the defaults max the settings either. )

Different cards and different results though, dual fan pulse here to somewhat different thermals from the blower default and the bigger tri fan designs.



EDIT: Right so temp wise well 90c junction down to 70c junction usually, Outer Worlds being again the one title that pushed towards the 100c cap and where fan speeds like to go up but that was useful for increasing the curve a bit for certain titles that well the card gets a bit hot for.

Oh and hardware wise this is a 60Hz display so 60 FPS being what I try target or thereabout so GPU demands and needs are going to differ for 75, 90 or 100 and higher although for 2560x1440 and 60 FPS this card is doing really well.

And if anyone decides that those ultra settings aren't worth the tradeoff I can see it hitting higher comfortably possibly even doing well under 4k even if sustained 60 FPS might not be 100% achievable in the latest more demanding games.
(Well it's going to vary too, optimization and all that and some settings just having a unusual high performance trade-off when pushed higher.)



EDIT: There should be a nice chart on the AMD Reddit too testing 1800 to 2100 and varying voltages and power slider settings and the noted stability, speed and power draw results I'll see if I can't find it.

EDIT: Oh and GPU results will be very variable so there's reports from users struggling with 2000 on the Sapphire Pulse until 1150Mv and others where it keeps stable.

Nothing new either but taking other reported values and using might not work well compared to some day or two of testing different titles, different graphics API's and noting down what works and finding the values that way.

GPU clocks spiking up a bit can also make it worthwhile to keep some extra headroom on the undervolt settings just in case.

And with the manual defaults basically maxing the settings out it's undervolt or nothing with a mild performance hit possibly depending on how the GPU is actually hitting these target values or what it lands on.
(RAM and CPU limited here so it's not going to be ideal but it's still doing really well as long as the drivers don't hit any issues, can't really complain on the performance even with those limits.)


EDIT: One of the charts from Reddit.
https://www.reddit.com/r/Amd/comments/dj0eya/oc_5700_clock_power_scaling/

And another.
https://www.reddit.com/r/Amd/comments/cf27xf/navi_powerperformance_study_5700xtstockuvoc/

(A big increase in power and when dialed down to the efficiency settings in that graph performance hits a near -8% for that test which is pretty low comparably to the -40% power draw reduction.)


EDIT: Well the power draw and voltage mostly serve to reduce noise by not having to dissipate as much heat from running the card at higher voltage and more wattage going into it and also allowing the GPU to run cooler which can even be a performance increase my maintaining closer to the boost clocks without throttling or with some tuning possibly hitting maintainable higher settings.

Or tuning it down to a even bigger reduction in power and resulting temperatures for a mostly pretty small performance decrease.

So for how I see it effectiveness and having the GPU being near silent and running cool for a ~5% performance decrease on balance seems pretty good but it can be tweaked with some tuning and testing.



Well that was quite some editing but I think that covers most of it now.

Not exactly a guide or anything like that but my own results and what I did which is fairly basic really and some URL's for other tests and charts showing that the GPU is pretty adjustable but results will vary from card to card though when scaling it down the performance drop is fairly low but it is also possible to hit a slight increase in performance while having some undervolting on the GPU core.
(But as long as the card isn't throttled by anything or hindered in how it performs it's not a but can make some differences to temps or boost values.)

 

recording on the right has lower detail.

 

What do you mean by that?

I see they look the same.

 

and thats why when comparing visual we don't rely on our eyes.

 

Looks like RTG forgot to optimize for 1440p

 

This is DX12 when it's working correctly - looks like Fortnite is more of a Gimmick attempt:

https://imgur.com/0jf8ezC

 

HW video decode still resulting in crash on multi monitor setups.

 

Oh boy, how long will it take them to finally fix all the bugs with Navi?