Incorrect. You can test this out for yourself, if you have the patience and the will.
Pick your favorite spot in the game you play most, make a save game (so you can reload it from the same spot while looking at the same direction) while using MSI Afterburner. Take a screenshot, note the FPS, the GPU utilization.
Now load the same scene again, this time with AMD's ReLive performance metric instead of MSI Afterburner. You'll see a slightly lower FPS and GPU utilization (capped at around 97%) because of how the ReLive performance OSD is implemented. The performance hit when using AMD's ReLive performance OSD is similar to recording the gameplay. This is something AMD can definitely improve upon.
However, if you are using some sort of a FPS cap and your card manages to hit that cap at all times, this isn't going to matter at all for you, and you won't be able to come to any conclusion from performing the above test.
[SPOILER]
There are a few people in this section of this forum who can not stand the sheer mention of the slightest bit of flaw/criticism of anything AMD. Such behavior doesn't help AMD's cause, and nor does it help the consumers as companies can get away with providing sub-standard support where there's obvious room for improvement. If AMD can improve their software support, then they have a better chance of being more competitive.
A practical example of this would be the drivers AMD managed to provide at launch for Forza Motorsport 7, which at the time made a VEGA 56 perform near equal/better than GTX 1080 Ti. This prompted nvidia to optimize their drivers further, something we wouldn't have seen had AMD been trailing behind competition.
There's no doubt room for improvement, AMDMatt also confirmed ReLive and Instant Replay had a known issue with stuttering in 18.1.1 so who's to say there can't be other problems with the current implementation of the overlay and it's performance metrics and the overall CPU and GPU load might be lessened over time as well though it will always be there from my understanding of what Unwinder said for Afterburner where it takes a little bit of performance to draw and display this information.
(Though it seems to be pretty minor but I don't have much more than framerate information displayed either whereas AB offers much more if you toggle everything which might incur a bigger hit, CPU monitoring especially I believe.)
There's no doubt room for improvement, AMDMatt also confirmed ReLive and Instant Replay had a known issue with stuttering in 18.1.1 so who's to say there can't be other problems with the current implementation of the overlay and it's performance metrics and the overall CPU and GPU load might be lessened over time as well though it will always be there from my understanding of what Unwinder said for Afterburner where it takes a little bit of performance to draw and display this information.
(Though it seems to be pretty minor but I don't have much more than framerate information displayed either whereas AB offers much more if you toggle everything which might incur a bigger hit, CPU monitoring especially I believe.)
The issue I am reporting was present in the first Adrenalin release too. I have MSI AB displaying a lot of metrics and it's not preventing the GPU from hitting 99% use whereas the ReLive performance OSD caps the usage to 97%. I think you and I had this very same conversation before. And I can bet that this isn't going to get looked at until some TechTuber points it out in a video, and gets posted on Reddit.
Hmm, BF1 dx12 runs awesome here with Vega multi GPU, very smooth and pegged @60 fps mostly at 5K even with TAA. But I got a problem with the menu settings in fullscreen to be visible and the worldmap is only partitially visible so I need to switch to windowed mode temporary to set/start things up (probably scaling issues with newer W10 fullscreen optimizations?) and switch to fullscreen after that.
The game runs so smooth I prefere it to run in it in dx12 at the cost of losing sweetfx. The same counts for ROTT, Siper Elite 4 and Hitman, they run great in dx12 in multi GPU.
I wish Deus Ex MD would run in multi GPU too, but I never get this in dx12. (did they drop it, they had a multi GPU test build a time ago?)
Also Fallout3 and NV now running insanely good with crossfire and the Fear profile with the more heavy enb's and tons of texture mods with the unsave memory settings hack on in enb local even at 5K! Now I wish I could find a way to let FO4 perform that way 😉
Yeah AMDMatt used to post here but I don't think he's been active for a while so I check other forums occasionally such as Reddit, AMD's own community forums (Those can be a mess.) and also Overclockers UK and the Overclock forums though that one is mainly for bios information and editing.
https://forums.overclockers.co.uk/forums/graphics-cards.3/ (That particular bit wasn't even in the driver topic but the Vega 64 GPU topic.)
http://www.overclock.net/forum/67-amd-ati/ (Handy information for bios tweaking, much of it also found in topics on this forum now and I don't know how things have gone for Polaris and Vega GPU's recently if they can be modified at all or as much as earlier models.)
https://www.reddit.com/r/Amd/ (Little involvement from AMD directly far as I can tell, a lot of CPU information as well in addition to GPU discussion.)
https://community.amd.com/community/support-forums/drivers-software (AMD's community forums, mostly community managers and such posting and members helping members so don't expect direct tech support or anything, there's a separate issue .)
EDIT:
Romulus_ut3:
...
I think you and I had this very same conversation before.
...
And this, hopefully that was a 17.12.x driver topic at least and not this very one as while short term memory can be a bit all over the place it shouldn't be quite that bad, yet. 😀
But yeah if nothing else works making a bit of noise usually sees a reaction so if the Relive section and the OSD continues to be problematic that might see AMD doing more to fix reported issues though it depends on if it's picked up by other sites or just a few forum discussions and similar though that might attract attention as well.
EDIT:
I wouldn't mind having a changelog for what they actually do tweak in Radeon Settings or the Catalyst Next software and it's related components, it's updated every driver release but the patch notes will mainly focus on the driver itself and for the December update it's mainly about the added features.
AMD is probably aware of other unlisted issues and same as NVIDIA not every change whether it's a fix or known issue makes it into the patch notes but that's probably nothing new. 🙂
One of the reasons I kept trying to find this profile file information and gleam any changes from that but due to how it's stored well at best I can find game names and that's about it.
(Knowing a game is supported isn't bad though and Crossfire profiles are clearly marked even if the flag properties are a complete unknown so until someone actually tests it then it's hard to say how well scaling works and issues such as micro-stutter or other glitches.)
Mostly just a hobby user though, actually understanding the driver is probably hard without any direct involvement with AMD and NVIDIA and then features of NVAPI and AMDADL and others are likely also under NDA or not immediately available to the public.
(The Linux drivers have a fairly large bit open source though so that might be a possibility, not like a common user like myself would have much of a hope of actually understanding much though.)
EDIT: So it's mostly keeping up to date with known issues and what other users discover and test, this particular GPU isn't going to get much more support but stability problems whether driver wise or in the software such as for Wattman can still be problematic so it's good to know about it if for nothing else than a preventive measure. 🙂
For example there was a couple of crashes in one of the last 16.12.x drivers, found a post on the AMD community forums about it recently when I was searching for any issues with Watch_Dogs 2 and then 17.1.x seems to have resolved the problem.
(That's from last year though so not a big issues now.)
Black screen issue with the R9 290 GPU or the Fury display corruption was also something I followed on various forums, took way too long for those to get resolved but it seems to have been resolved almost fully after a few months of time.
EDIT: Although I guess a regular user wouldn't really *need* to track multiple different websites if AMD kept the known issues list updated a bit until the next driver was released.
Just how it is I suppose, not like NVIDIA re-releases that PDF document either with further known issues though they do have a forums a well and also a part specifically for security concerns and fixes in some of their drivers.
Black screen issue with the R9 290 GPU or the Fury display corruption was also something I followed on various forums, took way too long for those to get resolved but it seems to have been resolved almost fully after a few months of time.
Whenever myself or anyone else reported the same issue in this part of the forum, we always came across some people who were like "I am not having this issue" (therefore implying that the issue isn't real or that it's negligible?) and when the issue went on to remain unresolved, reporting the same issue over and over got people suggesting that those of us who are having this issue should just stop making stuff up and should buy nvidia.
Yeah I had a 290X with the black screen problem which got RMA'd after waiting for a few weeks and testing some newer driver and that ultimately resolved the issue, then the Fury GPU which I'm still using (Vega disappeared after a few days and pricing was already above MSRP by a fair bit but nothing like the situation now.) well it started showing that display corruption glitch and that took as I recall at least 3 - 4 months before it was mostly resolved in a later driver update though AMD never clarified what either of these issues were caused by and some users still have traces of this such as a quick flicker or temporary screen glitches.
https://community.amd.com/thread/188642
Trying to search around and finding info was not very effective, threads mostly either involved people speculating or blaming the user and others just kept doing RMA's and hoped the next card would be unaffected.
I'm guessing the driver patched something with the GPU itself though even for a wide spread and widely reported issue like this one I don't believe AMD ever provided any info on what caused the glitch so there's just a lot of speculation.
Even smaller games can have a loyal following of fans that won't see any issues or anything wrong with it so I'm not that surprised when problems with Intel, AMD or NVidia pops and people still defend them even if there's something that's obviously isn't working as intended.
Vega is actually seeing a bit of this now with AMD officially backing away from the previous plans for primitive shader functionality requiring software specific implementation and then there's how the spun the Vega Frontier as a mining card rather specifically as they're now selling off what's available from that stock.
And outside of these examples there's still something off with Wattman and OverdriveN upgrading earlier GPU models to where even the Polaris architecture seems affected negatively even if it natively supports Wattman and Overdrive Next.
Something with how the reported value can be off with up to 5 or 6 percent or so I believe it was, not good for sensitive overclocking and voltage adjustments so people do bios edits instead or rely on third party utilities.
Vega also has something with it's I2C access or how it was called having a chance to crash the GPU requiring a hard reset for tools such as HWInfo which I believe later versions worked around but I don't know if the core issue itself has been fixed yet as of 18.1.1 or not and then there's memory tools reporting several page faults with it's memory which is hopefully a software issue and features such as HBCC have also had stability issues but I believe that has since been fixed.
Fast-boot support is also still iffy for Windows 10 to where Wattman will trigger as if the GPU has crashed and reset to default.
Let's see, here's a post about fast boot and Wattman resetting.
https://forums.overclockers.co.uk/threads/the-amd-driver-thread.18643923/page-786#post-31533630
And I2C disconnect.
https://forums.overclockers.co.uk/threads/the-rx-vega-64-owners-thread.18789713/page-219#post-31514733
Those memory errors.
https://forums.overclockers.co.uk/threads/the-rx-vega-64-owners-thread.18789713/page-228#post-31560995
To actually have some source references to a few of these reported problems, there should be something about fan speed occasionally ramping up because of some hiccup but it goes back down again pretty quickly though for a few seconds the card can be pretty loud until it corrects itself and then also that issue with reported values from Wattman being a bit off from what's actually being used by the card but I couldn't find the posts so it might have been from a different forum.
For that last issue in particular the Fury and newer GPU models also adjust power draw and clock speeds a bit across multiple power states and various ways for this to be throttled so getting exact values is a bit harder. 🙂
(I think I read somewhere that the AMD Fury also lacks a thermal sensor for the HBM so that could be one reason why it was locked from overclocking even if the stepping method worked pretty well for several owners at least for 500mhz -> 545mhz and had a small boost but it's only using stock now even if the user tries to mod in the values directly via bios.)
And for voltage it varies due to how it's generally automatically calculated from the chip and it's overall quality so to use the Fury as a example it usually has a core voltage ranging from I think it is just above 1.2v somewhere to a value of up to 1.268v though that's only for the Nitro model from Sapphire which is also the only model capable of having a higher than 1.3v core voltage without the GPU shutting down or by adding via offset to get around that limiter.
My own GPU reports as 1.218v by default for p7 at 56% ASIC quality (Fury non-X is pretty low overall for these, even Fury X can be a bit low with Nano usually having the highest reported ASIC quality.) with other users reporting higher or lower, usually there's headroom for both under and over volting but since overclocking only yields smaller gains it might be more worthwhile to under volt the card and reduce power draw while retaining the same performance level, I think Vega is even harder factory over clocked and can see some pretty big gains here if the GPU tolerates a larger drop down in voltage without becoming unstable just because AMD set the default values to a very high default value or something like that. 🙂
Well lots of text again but to summarize yeah there's some known issues and AMD has some work left with the Catalyst Next software and it's added features in Adrenaline and some fans get way too defensive to a particular brand or product ha ha or well there's all sorts of examples of this I guess.
I tend to be pretty fond of AMD myself since I've used them for so long now but I try to keep up with what Nvidia is doing too and keeping choices and such for future updates open but these price increases has really hindered upgrading the current system and who knows when that will get better again heh.
560 and up, almost entirely out of stock again with mid to late February or early March listed among retailers though there's only a few GPU's for the higher-end models being delivered so they'll be out of stock shortly again.
And enough has been written on the prices already but Vega 56 should never have been four digit US/EUR currency pricing ha ha.
(Sapphire banking heavily on purchases I guess with Nitro kinda disappearing never to be seen in stock again but here's a "lesser" Pulse version of Vega coming out in February but the MSRP is somehow 800 dollar now.)
EDIT: Well I write too much but it's fun to discuss and read about these things and it doesn't hurt to be aware of any potential issues although there's probably other undiscovered problems too.
Hopefully AMD will stop breaking older API's in newer driver releases at least and Catalyst Next can improve little by little going forward with hopefully the new features from Adrenaline being stabilized and fixed sooner rather than later.
(Some optimization might be possible too for GPU and CPU load though there's always going to be some hardware demand of course whether it's monitoring or recording.)
Hah well unfortunate as it is there's probably a bigger chance of the driver and software improving than the pricing going back down to something reasonable anytime soon.
(Both would be ideal of course and sooner rather than later but unless something really big happens this year I don't foresee pricing changing much though the introduction of new GPU's might shake things up a little bit at least if anything's planned later this year.)
(How many months before AMD added some of the "legacy" setting to CNext? And Legacy CLI is still there even if it doesn't really do much anymore.)
Sounds a bit like nitpicking I guess but it doesn't hurt to point and be aware of these flaws and the time it took AMD to fix and react to some of the requested changes. 🙂
(Or ignore them entirely and do some changes anyway which caused problems for all kinds of hardware for a while.)
^^ Post it on ATI/AMD WEB 😀
Im back to 17.9.3 Oct.2 -> One game is not Playable (SWBF2 have too much corruptions, it all appear after Dec. Patch lol , not fixed fully till today)
Im now not in a hurry to switch for Adrenalin.
===
Same was with 17.2.1 WHQL -> Next was 17.7.x
Yes 5-6 months then was great 17.9.3 beta & WHQL 🙄
Well, ranting aside it is a complex piece of software and it could be worse, both AMD and NVIDIA have their little quirks in the current drivers from reading up on the release notes and what users are reporting. 🙂
There are third party tools for working around some of the current issues as well or alternatives such as MSI Afterburner/Rivatuner Statistics Server for in-game on-screen-display support with various recording and capture tools too but it's nice to see it supported by the driver and it's settings software directly as well.
On to the 18.2.1 drivers next I guess and whatever they'll bring, I'm assuming readying for Far Cry 5's release is going to be a thing and Vega and it's Rapid Pack Math 32-bit to 2x 16-bit float operations although I believe the current drivers already also have a profile for the game. 🙂
And then we'll see what else gets fixed and updated I guess, or broken in the worst case though hopefully it's nothing major like DX9 going bye-bye for a few weeks this time. Doubt there's much more that AMD would get from older GPU's but GCN 1.2 and older will probably be supported for at least some time yet before GCN 4th and 5th gen (And newer.) take over entirely though depending on how this is all set up in the driver there might not be much gains from dropping GCN 1.0 or 1.1 compared to Terrascale and older.
^^ Post it on ATI/AMD WEB 😀
Im back to 17.9.3 Oct.2 -> One game is not Playable (SWBF2 have too much corruptions, it all appear after Dec. Patch lol , not fixed fully till today)
Im now not in a hurry to switch for Adrenalin.
===
Same was with 17.2.1 WHQL -> Next was 17.7.x
Yes 5-6 months then was great 17.9.3 beta & WHQL 🙄
Frostbite and their continued problems with DirectX 12 for both AMD and NVidia I guess? Well for NVidia it seems The Division is the current problematic title for DX12 though as I remember from Battlefield One DX12 didn't really yield anything but some minor games and occasional stuttering though Battlefront 2 is much more recent and probably has some changes in the underlying engine compared to previous games. 🙂
EDIT: Though my own understanding of both the software and hardware side of things is still not all that good, mostly trying to stay up to date and reading but it's probably a pretty complex affair if you were to dig into the driver code more in-depth and of course games or other software using lower-level API's such as Vulkan and DirectX 12 though it will be fun to see this picking up again as there's some good gains to be had but it's going to take time before we get there. 🙂
(A handful of games and then it kind quieted down again but it took time for DX11 to break through too.)
Im using Reshade+Sweeet for every game.
BF1 & SW BF2 i like them in DX11.1 (crispy lol)
FH3 & FM7 are in DX12 Mode.
After tweak Wolf also is working like a charm -> so im not in a hurry now.
They will eventualy Fix everything in time 😀 Patience is Virtue
Btw, after resurrecting an asus 7870 dcii v2, I could check 18.1.1s on CGN 1.0. Unfortunately the UVD bug is still present, and can be triggered immediately on that card, as it always did. Could be due to ancient vbios, as Romulus_ut3 reported it didn't happen anymore.
As I don't need a decent OC for its use (assuming this bake will even last...), I've just flashed the bios from the pre oc'ed model with same custom pcb so the UVD bug doesn't bother at all (these non boost cards have the highest UVD clock equal to the only 3D clock). Still, when it kicks, having it at 450/1250 is still better than the 60W idle power draw of the HD4890 that was being used.
Im using Reshade+Sweeet for every game.
BF1 & SW BF2 i like them in DX11.1 (crispy lol)
FH3 & FM7 are in DX12 Mode.
After tweak Wolf also is working like a charm -> so im not in a hurry now.
They will eventualy Fix everything in time 😀 Patience is Virtue
Sadly for me, BF1 is an instant crash with Resade. It sucks.
BF4, and BC2 work with Reshade for me. BF1 is amazing with reshade. Such a shame.
Id kill to have reshade on FH3. It would look good on Apex, but I dont play apex because its BS and has no 21:9 support.
Sadly for me, BF1 is an instant crash with Resade. It sucks.
BF4, and BC2 work with Reshade for me. BF1 is amazing with reshade. Such a shame.
Id kill to have reshade on FH3. It would look good on Apex, but I dont play apex because its BS and has no 21:9 support.
Wait, i will upload my Setup:
-> https://mega.nz/#F!0IFjjSTS!3f9W2AyhnMOImeBPszFL5w
check Reshade 2.3 + Sweet 2.0 (im on this with BF1/4)
or Origin Fix, and trust me it will work NP 😀 😛
For ON/OFF is pause/break key
Screene = prt sc
Btw, after resurrecting an asus 7870 dcii v2, I could check 18.1.1s on CGN 1.0. Unfortunately the UVD bug is still present, and can be triggered immediately on that card, as it always did. Could be due to ancient vbios, as Romulus_ut3 reported it didn't happen anymore.
As I don't need a decent OC for its use (assuming this bake will even last...), I've just flashed the bios from the pre oc'ed model with same custom pcb so the UVD bug doesn't bother at all (these non boost cards have the highest UVD clock equal to the only 3D clock). Still, when it kicks, having it at 450/1250 is still better than the 60W idle power draw of the HD4890 that was being used.
Since we use an outdated unsupported GPU, we don't deserve anything, not even a fully functional driver, Hurrr Durrr.
[SPOILER="Not for those who aren't affected by the UVD bug"]
Intelligence among the community is dropping rather sharply.
https://i.imgur.com/rrohsC9.jpg
Apparently asking AMD to implement proper power management for it's cards is asking them to throttle performance. 😀 What saddens me is that this person is a part of this community, and claims to be a hacker, programmer, a 3D Modeler, A Cad designer and a web developer too, but has no notion of power saving.
Then there are people who suggest that I should flash my BIOS to get rid of the UVD mode which is kind of like cutting your own head off because of a headache.[/SPOILER]
With that said, disabling ReLive usually gets rid of the UVD bug. I didn't have any issues with Adrenalin 17.11.1 and 17.11.2 as far as the UVD bug is concerned, though whenever I enabled ReLive, I had the issue.
Oh, I never install relive, so it's not related in this particular case, nor a friend's 280X which stills suffers from it and has the latest Asus bios.
Oh, I never install relive, so it's not related in this particular case, nor a friend's 280X which stills suffers from it and has the latest Asus bios.
From the 18.2.1 thread. 🙂
LocoDiceGR:
From Reddit about the 100% GPU bug that someone mentioned.
AMD_Kyle RTG Engineer - 5 hours ago:
Hey just to let you know we do know about this issue - and it turns out to be pretty complex to resolve.
As a workaround for now you can kill the AMDDVR process which should return the GPU utilization to the correct levels.
And why I quoted that even if you said DVR was not installed/skipped well that's because with Adrenaline they made a change so even if you use "skip" for AMD DVR and don't install the full package the software will still run dvrcmd.exe and launch two new services. (Also sets this as a scheduled task to run on boot up.)
Renaming the file kills this behavior, removing the scheduled task is only partially effective since running CNext will have it run dvrcmd.exe if it's not renamed and then the scheduled task comes back I think though in either case you get both the DVR services again.
This might interfere with AMD's OSD functionality though, I have not used it and keep it mostly disabled but there might be a reason for why 17.12.1 and on does this and you have two DVR services running even without DVR installed. 🙂
EDIT: DVRCMD.exe is found here:
C:\Program Files\AMD\CNext\CNext
Thanks for the tip but it's unrelated (no gpu load on idle). I can trigger the uvd bug on demand after stopping both remaining Relive processes, as with any driver after 16.7.2. As always, for my test I only need to scroll FB for a while under chrome until clocks switch to UVD, then at that clock, launch a clip on mpc-hc and let it play for a bit. Close chrome, close mpc-hc, and clocks will be stuck in UVD's lowest state at 0% gpu load. (same on 18.2.1)
I've never seen that particular 100% gpu bug on Adrenaline after exiting stuff, even after opencl assisted encodes which seem a common cause for it.
[SPOILER]
There are a few people in this section of this forum who can not stand the sheer mention of the slightest bit of flaw/criticism of anything AMD. Such behavior doesn't help AMD's cause, and nor does it help the consumers as companies can get away with providing sub-standard support where there's obvious room for improvement. If AMD can improve their software support, then they have a better chance of being more competitive. A practical example of this would be the drivers AMD managed to provide at launch for Forza Motorsport 7, which at the time made a VEGA 56 perform near equal/better than GTX 1080 Ti. This prompted nvidia to optimize their drivers further, something we wouldn't have seen had AMD been trailing behind competition.
https://i.imgur.com/xJuumcy.jpg Source: [H]ardOCP
[/SPOILER][SPOILER="Not for those who aren't affected by the UVD bug"] Intelligence among the community is dropping rather sharply. https://i.imgur.com/rrohsC9.jpg Apparently asking AMD to implement proper power management for it's cards is asking them to throttle performance. 😀 What saddens me is that this person is a part of this community, and claims to be a hacker, programmer, a 3D Modeler, A Cad designer and a web developer too, but has no notion of power saving. Then there are people who suggest that I should flash my BIOS to get rid of the UVD mode which is kind of like cutting your own head off because of a headache.[/SPOILER]
With that said, disabling ReLive usually gets rid of the UVD bug. I didn't have any issues with Adrenalin 17.11.1 and 17.11.2 as far as the UVD bug is concerned, though whenever I enabled ReLive, I had the issue.