16.101.X [Pre-GCN driver pack]

Fortnite and frostbite don't work because their fixes cannot properly detect the newer driver.
It's not about the drivers missing fixes.

If you had read just two posts before your first, you'd have noticed my very tentative idea.
If you want more information about just any possible complicacy of >14.4, go here.

 

Can't install the Windows 7 version of the recommended pack on Windows 7 Ultimate 64-bit. It says that the driver does not support my current operating system.

 

I have investigated the fortnite sky glitch, it is a very strange bug, seems to be a problem with the shader generation, placing dlls from the 15.301.2600.1004 driver (which shares the same dll versions 16.101 but is not signed) in the directory with the game exe solves the issue, would appear it is a bug with the dll's paired with the 16.101 driver branch(fortunately fortnite logs the pregcn detection event and shows that it applies the workaround regardless of the driver version, since it uses the device id), you can drop 15.11.1 dx10-11 dlls from the convenient amd dll thread https://forums.guru3d.com/threads/amd-driver-dlls-repository-and-performance-survey.399547/ into the same folder as the game's exe as a work around if you have this driver installed and do not wish to revert.

 

Oh, this is very bleak thing.
I wonder if we'd loose anything by rebasing on that driver instead (did you mean 15.301.2601.1002 maybe? I cannot find that other version anywhere)

It makes total sense then for the dll version to be meaningless tho. While investigating an OpenGL bug, for as much as counterintuitive I found out the *kernel branch* version of the package the dll I was swapping came in.. to be more indicative than the actual GL_VERSION release_number of the dll itself (which is, coincidentally, the last set of digits you see in "Version" property), as for tracking down fixes.
I don't think that has anything to do with "pairings" (since the wrong dll with the fixed driver caused the problem, and the right dll with the bad driver fixed it). It probably just is they bump the version when they touch some "big component" of the spec or something, but all of the tens of MB of the icd dlls aren't just that.

 

the only thing that bugs me is that there appears to be no easy way to disable the pre-gcn fixes on ue4 , it crosses my mind that perhaps the workround is actually causing the problem if its fixed on this version, no way to know for sure. some other ue4 games have issues but they are minor ( can turn off a setting and it goes away kind of thing)

ive wanted to use the 15.301.2601.1004( dell driver)but its not properly signed(old amd cert , not devportal signed), which prevents it from being permanently installed.

while i have my old card installed ill do some more testing, btw the latest adl libs still work(18.8.1) could probably get the new radeon settings /overlay working. not that there's really any point. afterburner is much better imo

 

Well, there's nothing wrong with 16.101.2901.1003 kernel driver.
If really any, that one could be used as a source for user component(s)

As for disabling workarounds.. I mean, there isn't a thing that I can think of, that shouldn't be possible to at least spoof somehow.

 

"Refreshed" Driver pack,
now has the opencl libs from 15.201.2401.1010, adl /blb from 18.8.1 and Dx10-11 libs from 15.301.2601.0

 

Cool I guess (I assume you couldn't get to trick Fortnite or Bf1?)
Waiting for somebody with more time of us to provide a comparison/benchmarks.

p.s. I don't think you are "spoofing" 16.2.1.. with its actual branch revision. It *is* it

 

performance is very close, honestly the best performing version is the 15.12 crimson version, but thats a bit dated.

I previously had the crimson version as 17.1.1, this does have an effect, for example fortnite complains if the version is 15.11 and previously some other games complained about 16.2.1 , I figure at this point its more likely that leaving it at the higher version will only cause problems, thus reverting it to 16.2.1 which is the version set by default on 16.101.[/QUOTE]

 

I thought there were a lot of reports of better performance (lower cpu utilization and/or multi-thread efficiency) for 16.x?

Anyway, do you mean that there are games that specifically look for the "15.301.1901.0000" (or less than or equal) string?

 

Is this driver compatible with the update 1803 of windows 10?

Since I updated I had no chance of use both my gpus, only the integrated one (intel 2nd gen) is working... If I use the discrete one (HD 6770M) or even if I try to install its official drivers from windows update (!) everything freezes and crashes and DDU in safe mode becomes the only way to reuse the pc.

EDIT: I tried the latest release. Everything is ok, except that the driver itself is not installing, even with driver enforcing off and in test mode. Any tips?

 

Installed on my old laptop, everything going right without any problems. BTW nice release, thanks user1

 

15.300.1025.1001 (crimson 15.12) is basically an earlier version of 16.101 the reason was that carrizo and stoney had a bug that made the newer branches unstable (15.301 and newer i believe), they didnt fix it until september of 2016, which is why 16.101 existed at all publicly, 16.101 has some advantages over 15.300 , but generally all of the 15.300+ branches perform similarly, within margin of error for the most part.
16.101 does have some improvements, but they are not huge. at least on my machine 15.300 seemed to have slightly better performance , but were talking edge cases like being cpu bound which is unlikely at this point.

in other observations, most curiously my rx 560 has more opengl cpu overhead for whatever reason, on the same system, very odd. perhaps a quirk of the gcn uarch+fglrx driver, might be a regression , ill have to see about testing on an older driver to see whats up with that.

there are games that look for the crimson version string (fortnite), the device id(also fortnite/Ue4), and some that look at the release version (bf1 beta) really down to the developer's choice
I figure that leaving the version numbers alone will probably be the best option for now, to prevent issues cropping up in the future , since it seems that if a developer decides to help a brotha out on pre gcn, it will only likely serve to cause problems with driver detection.

If its not installing, there is likely some other issue at hand, really can't help, I would recomend disabling the intel graphics if you can, I'd imagine on a laptop that old the battery is likely toast anyway not much point to use hybrid mode.

 

Oh I see.
In this case I'm totally on board of the last release.

p.s. higher opengl overhead could be as well just them "fixing stuff"

 

Damn, I installed those on a nearly freshly formatted machine... So I don't really know what's wrong with my pc.

Unfortunately enough, looks like it's impossible to disable the intel gpu, because all the "gpu traffic" is routed to it and then to the amd one only if the system thinks it's worth it.

Btw the intel alone is working good, and I think it's still making use of the amd one, even if it's listed as "basic microsoft video card", because I can play Skyrim without issues, and the intel one isn't supposed to be able to be so smooth... I'll investigate and let you know (if it's of some interest).

 

@user1 I know you're providing these drivers as is, but I am kind of lost. Here's my problem.

I have this HD 5770 taken out of a Mac Pro 2012 machine, and on Windows 10 x64, none of the Drivers recommended by AMD on their website works. It gives me the message "We are unable to find a driver for your system. No supported AMD hardware was detected" upon initializing the installer. Even this repacked driver package of yours behave the same.

However, if I go into device manager and point to the drivers manually, it just installs the Display Drivers and nothing else. No CCC or Adrenalin Control Panel.

What's shocking is that Windows Update manages to find and install drivers for the said card, with a working CCC. But the OpenGL titles don't perform up to expectation at all with these drivers. What can I do to install your modded driver that will boost my performance a bit?

In case if it helps, here are the device IDs for the card:

Code:

PCI\VEN_1002&DEV_68B8&SUBSYS_00CF106B&REV_00
PCI\VEN_1002&DEV_68B8&SUBSYS_00CF106B
PCI\VEN_1002&DEV_68B8&CC_030000
PCI\VEN_1002&DEV_68B8&CC_0300

HD Audio:

Code:

HDAUDIO\FUNC_01&VEN_1002&DEV_AA01&SUBSYS_00AA0100&REV_1002
HDAUDIO\FUNC_01&VEN_1002&DEV_AA01&SUBSYS_00AA0100

Here's more info on the set of drivers Windows Update installs for my card:

Spoiler

​

Any help or input will be much obliged!

 

OpenGL being slow on closed amd drivers is nothing new under the sun (if any, on osx all opengl sucks hard so I guess developers debuff everything they can in the first place)
Similarly, it's no surprise device manager just install the *driver* and none of its accessories.

Said this it's odd it's not installing.. DEV_68B8 is already in the .inf, and even the wupdate driver doesn't seem to differ in that.

 

ok so first off i can already see that the problem is the device id, , the installer doesn't use the inf to decide if the hw is compatible. it uses msi files for various things like ccc and the driver installion which contain a list of supported device ids, using the Orca msi tool, you should be able to add your card to the supported device list on each of the msi files. (ccc , audio driver , video driver ect all have their own msi files)

as for opengl performance, you're not gonna see any major differences, opengl performance has always been a weak point, there will be little to no difference between the windows update driver and the driver repack interms of performance most likely.

 

Because of the 1809 update, I'm doing a fresh install again. Which Intel driver were you using when you wrote this message? Thanks!

 

My laptop is a full AMD system, so no intel driver needed. Sorry