Navi RDNA Owners Thread, Tests, Mods, BIOS & Tweaks !

The UVD clocks might of been some random fluke

So then it will be silicon lottery gpu dependant on how low you can go on the SoC but is worth looking into tweaking it to shave a few degrees off. I am gonna guess since I under a full block I am see a bit larger temp drop then just with air...but still 2 to 3 degrees is nothing laugh at. The closer Navi is to 45 degrees on the core the better it will perform.

 

This was happening to me with Vega and it's happening again with Navi. This was recorded in August using Relive and drivers 19.7.5.

It's not a game bug, as my friends with Nvidia cards don't get this problem when they reach this part of the game.
It also happens in other areas throughout the game.



I haven't retested using newer than 19.8.2 drivers. I was wondering if someone else could give it a shot and share their results.

Cheers.

PS. I think that visible visual glitches are due to immature drivers or some sort of problem with Navi video encoder.

 

Stock.

https://i.**********/Hxj6YSKd/firestrike.png
https://i.**********/nz50C4b7/timespy.pngp/img]

 

https://i.**********/R0h91TNR/bench.jpg
I'm on par with stock RX5700 XT, except for CPU score hehe.

 

https://www.3dmark.com/3dm/39559781?

On my Vega thats the max I can do. Need a better cpu and it was stock

 

Back to the TDP Limit SoC what I am seeing is about 1.25 degrees difference per amp up or down in temperatures. I told myself I was not gonna test anything this week but actually game.

 

How does that affect core clock boosting in your experience?

 

For me it does not...since I am on a full waterblock but on air depending on how many amps you can drop down it should boost a bit higher since temps play a part in the boost behavior. While 2 to 4 degrees may not seem much it is as far as boost goes. My GPU is already boosting to its absolute limits on the waterblock with my temps. In fact it tries a lot of time to boost way higher then it can actually do and I get crashes ...I am talking well above 2350 and many times over 2400. I was just posting what it seems to be doing without adversely affecting performance.

TBH I can't really say I understand amps, wattage etc formulas all that well, was never my thing.

This is pretty good explanation https://www.quora.com/What-causes-a...ws-through-a-wire-voltage-or-amperage-changes

And here is the math I just did with that article:

At stock it is 14 amps on both XT and non xt. You may or may not get down to what I did at 10 amps.

Stock
p = 1.35 v*14 a p=18.9 watts

(p = 1.35 v*13 a p=17.55 watts
p = 1.35 v*12 a p=16.2 watts
p = 1.35 v*11 a p=14.85 watts)

My GPU with the TDP Limit SoC lowered
p = 1.35*10 p=13.5 watts

This little bit of math explains the lower temps all by itself since most of us are familiar with heat generated from pc parts and why.

As shown I am having to deal with 5.4 less watts of heat being generated.

While 5.4 watts may not seem like much but in transistors it is a huge factor in the heat generated in such small parts. Also keep in mind as these parts heat up they also bleed heat into surround parts through the pcb.

We can best see this on an Air cooled navi running a monitor at 60 hz vs over 100 hz. The heat bleed from the ram running at full clock vs downclocking. It raises the idle temps of everything.

 

No one?

 

I do not have the game.

 

I am having issues with the Gigabyte RX 5700 XT Gaming OC. Randomly crashes (monitor disconnects, system freezes) and have to hard reset. According to bluescreenview it is atikmpag.sys/dxgkrnl.sys/ntoskrnl.exe. So far tried 19.9.1 and 2. It happens when idling/browsing not while gaming. Any suggestions besides DDU?

Edit: Happens while gaming as well.

 

If you have a FreeSync monitor try disabling free sync itself and testing, from earlier info from mtrai the next driver might have a fix for a current incorrect set "Chill" registry value where it's a invalid Dword instead of a value of 0 (on) or 1 (off) though on deletion and manually setting this that just sticks until the next boot where it has to be redone again.

Drivers need work is my view on it, few more months or so should make a good difference at this pace but that's not exactly fun to hear when you're dealing with system stability so hopefully it's a easy thing to mitigate until it's properly fixed.

EDIT: Well the issue was replicated, code and testing and getting that into say a 19.9.3 driver or the 19.10.1 driver (Third September driver or first October driver whichever is out next.) well it might still be a while before it's shipped in a release driver.

EDIT: Oh and the driver fixed some hardware acceleration issues mainly around media playback, Chromium and a few things like Discord can still cause system crashes from what I am reading up on for this situation so if possible disabling hardware acceleration might still be helpful.

 

I do have a freesync monitor. I will reinstall the driver and test it with FS disabled.

 

I do not have the hardware to test it myself but it popped up frequently in the AMD Reddit though from mtrai's prior comments in one of these topics it could also be related to how Chill is now kinda always on but toggled via hotkey and the incorrect registry string instead of if it's on or off which when resolved seems to at least fix some of the associated problems with Chill but also FreeSync though there might be others.

There's that post again where I found the discussion around Chill and it being replicated already by AMD so that points at a nice possibility for a inclusion in a soon to be released driver rather than more long term.
https://forums.guru3d.com/threads/n...-mods-bios-tweaks.427125/page-19#post-5709130

From this discussion here and the registry info if that's something to try too and if possible also keep the benefits of adaptive sync or AMD FreeSync from the display.
https://forums.guru3d.com/threads/n...-mods-bios-tweaks.427125/page-19#post-5709116

I find the hardware to be quite impressive but the software to be in need of work though I've only compared against a few of the reported problems and the main recurring issues already have quite a few other users both from Reddit and other forums and this forum too confirming many of the same issues so that already pretty much confirms that there's some work remaining though I like the progress so far but it's not fun for users affected by the more severe current problems that can even go into blue screens or other system stability issues and however long it can take for this to get fixed.

This time it's hopefully purely in the software itself though and no hardware problems that need to be worked around or adjusted via the software or micro code updates or even bios though even so it can take a while for these to get replicated and fixed and then for a driver containing said fix or fixes to come out although this applies to both AMD and NVIDIA although I have more experience from AMD having used their GPU's for quite some time now so I have less of a grasp on the situation for NVIDIA other than the most common recurring issues getting brought up.


DirectX 9 framerate or frame time instability, Wattman (No surprises there as there's outstanding issues for older GPU models too for a variety of things.) 1080p or maybe just low settings in general underperforming and then the BSOD issues, hardware browser acceleration and media playback issues, 240Hz issues, 75Hz issues and this Chill issue though it applies to all AMD GPU's at least since the functionality was changed to this hotkey system instead and from that to multi-monitor and FreeSync and ReLive I believe as well for the software side.

Lots more to read up on, just have ideas and theories so far and it's unlikely to be more than that but from 19.7.1 to 19.9.2 they're progressing nicely with the fixes bit by bit though there's been some regressions and some of the problems might not be 100% fixed under all situations either just yet.
(Think there's a problem with sensor readings in Wattman but overclock stability should be far improved over the early drivers particularly with memory though memory depending on manufacturer could have it's own little quirks so there's not just the software itself either.)


It's not like the driver or Navi10 support or functionality is entirely broken either it just needs a bit of work and focus on the bigger issues causing blue screens or system crashes or shutdowns and forcing hard resets and such and that sort of thing which hopefully won't be too long until this can be resolved in full and then the less severe problems can also be worked out over time though I couldn't say how long this might take.

I tend to jump on early because I'm curious and I know there can be problems but having to do some 6 - 8 month wait for the software situation to be in a better state is kinda dumb too for regular users who just want to plug in the GPU and have it working as advertised and expected within reason so it's not really excusing there being problems even if some are to be expected for new product launches and whatever else Navi10 and RDNA here brings specifically.


Well situation is what it is, hopefully either of these workarounds help for now and then the next driver or so can make it redundant and get it fixed in full and this problem can be done with and over.

 

New Test in VLK for The Surge 2 in Vega thread

 

5700XT Pulse PCB analysis via Gamer Nexus and Builzoid following the various GPU reviews and general teardowns that Gamer Nexus has done for a few of the 5700's such as the Nitro+ though from my understanding and current read-up the Red Devil and Red Dragon are also pretty solid as is the reference model itself though the cooler has it's limitations unless a custom one is utilized or going for a water block replacement.
(Strix seems solid as well but at a higher price point though now with Nitro+ I guess there's two fairly high priced 5700's)

Haven't seen too many reviews of the Nitro+ itself so far though I think there' a few available by now and then user impressions and results for overclocking, noise or just undervolting and regular testing in various games and overall experiences with the card though availability is probably a bit limited for these first shipments.


And yeah Surge 2 seems to be performing well though some users run into problems with it but when it runs it's quite fast likely thanks to Vulkan instead of DirectX 11 as the API it's running on. (And D3D12 is there too though it's not used by default and switching it over might not be entirely functional if forced.)

 

With freesync disabled, still happens. Bluescreenview says it's caused by dxgkrnl.sys Could it be a power issue? Although it doesn't happen without the driver installed. It's a new build with a new power supply; Corsair RM750x. But I am using extension sleeved cables . I will probably try removing them and directly connect the PSU PCI-E cables instead.

 

So 19.7.5 seems to be the working driver for me. No issues so far with it. BTW, how can I check the manufacturer of the memory chips? GPU-Z is not listing it.

 

You can't, only by removing graphic cooler.

 

ASRock Radeon RX 5700 XT Taichi OC+ Review

" The ASRock RX 5700 XT Taichi OC+ is the fastest factory overclocked Navi card we've tested so far. We measured it to run almost 2 GHz real clock on average in our real-life, mixed gaming load.
Temperatures are good, too, and the cooler includes idle-fan-stop."