[Success] AMD RX6000 Series working in macOS

[TheBloke]

Where do you guys place the odds of any 6xxx support?

I honestly am getting more and more pessimistic about it.

I think they might skip right to all Apple Silicon, even for GPU

For me 11.4 beta 1 is the tipping point. If it doesn't come then, I'm going to start to lose hope.

There's definitely some reasons to be hopeful for 11.4 - albeit with the caveat that we've been here before many times. The 6700s are out; the Linux and Windows drivers have had time to mature; they've had plenty of time to work on 11.4 given 11.3 has now had 6 betas and therefore presumably won't launch for at least another two weeks (including one or two RCs).

Then there's the 'leaks'. Octane developer says he was told it was coming (or was already there) in 11.3 beta 3. It wasn't. Did he misunderstand what he was told? I don't think so; he seemed certain. Did the Apple contact misunderstand the question or just get it completely wrong? Possible, but seems unlikely - though of course we have no idea what the position of that contact was, like whether they're a tech/dev contact or not. Or did the Apple contact really think it was coming, and then maybe it just slipped? In that scenario, which is at least plausible, that would make 11.4 the obvious candidate.

I don't know what to make of the Redshift comments. At first glance it seems great, but phrases like "as is common knowledge" makes me fear that they're just referring to the publicity generated by the Octane dev's comments, and don't have any first-hand, specific knowledge or proof that it's coming. It's possible they're simply reacting to the same hearsay we've all heard, and haven't realised that drivers have not in fact been released in any 11.3 beta to-date. I'd love to hear about any follow-up comments that might clarify that.

I do think it's looking pretty unlikely that there will ever be an Apple 6000-series GPU, and therefore the only reason from Apple's perspective to release them would be for eGPU support, and for third-party Mac Pro 2019 upgrades. They do sell an eGPU enclosure, so it's a technology they promote at least to some extent. Then again, the fact that it encourages third-party upgrades could also be seen as a reason for them not to add support, and the fact that Silicon Macs don't currently support eGPUs could be evidence that it's not a technology they plan to support for long.

So yes, like you, I am tending more pessimistic now. I've not yet given up hope, but I do think it could go either way.

 

[dehjomz]

Well all I can say is the Radeon 6000 kexts are changing. Different sha256 sums on the binary file inside the Radeon6000Framebuffer kext, for example, from beta 5 to 6, so some work is being done.

But you’re right in that if support doesn’t come in 11.4 then we’re probably never gonna see support for big Navi on Big Sur. 11.4 is probably the last point release on Big Sur before wwdc and macOS 12.0 beta is released. There still could be an 11.5 prior to then, but I doubt it. Big Navi support could come in 12.0.

 

[auzer1]

Speaking as someone who is on the redshift forums a bunch, I trust this intel. They are very aware of the situation, they are not misinformed about it already being out. They said they have not tried it or have no experience using it so cant say about the performance. However Redshift Devs had to work closely with apple to get a Metal version out just this last year. My sense is that apple threw a bunch of money at the problem after hearing so many Pros in this field complain about the Nvidia situation. Redshift devs are very responsive, trustworthy and I would guess they have good source intel. Of course this is all just my speculation.

 

[TurbineSeaplane]

I really wish they would go to a longer release cycle for macOS

It feels like they finally fix and add stuff...and then start all over with a new version that simply isn’t needed or wanted.

 

[TheBloke]

Well all I can say is the Radeon 6000 kexts are changing. Different sha256 sums on the binary file inside the Radeon6000Framebuffer kext, for example, from beta 5 to 6, so some work is being done.

Unfortunately I don't believe this means anything. Every release bumps the version number of every kext, and I think that is the checksum difference you are seeing.

Also I really wouldn't expect to see any incremental changes until the actual release. Apple may be working on the drivers internally but they're not going to release that work until it's all done. They won't push out half-finished versions in each new beta - that'd be really bad practice. They'll wait until they're ready and then push out all the changes at once. Then subsequent betas would contain any bugfixes/changes to that work.

The one change that has happened and that we've not discussed recently is the removal of the 6000-series framebuffers in 11.3 beta 1. That's another thing that could be taken as an ill omen. Why would they remove those now - removing the ability to get a GUI picture on a 6000-series - if they plan to add full support later? Why not do both operations at the same time?

One possible optimistic explanation for that removal is that the framebuffer defs in X6000 were just temporary, and that the drivers are coming a new kext, eg X7000. So removing those framebuffer defs was a cleanup necessary on the way to full drivers, and they happened to choose to do the cleanup first, before the drivers came. Or perhaps that's even evidence that drivers were meant to come in 11.3 beta (like the Octane Dev thought), but perhaps got delayed. So they cleaned up the framebuffers in X6000 expecting that to coincide with a new X7000 for Big Navi 2, but then X7000 was pushed to 11.4, and they didn't care (or forgot) about also pushing the cleanup to X6000.

I always expected that drivers, if they came, would be in a new kext; it'd be quite odd to add Big Navi 2 support to X6000, which is for Big Navi 1, and doing so wouldn't match their past practice. So that's all plausible. But that again would mean that any patches made to X6000 in the meantime - even if they were more than just version number bumps - are not related to whether we do or don't get 6000-series drivers.

TLDR: I don't think those checksum changes to the AMD kexts tell us anything

Speaking as someone who is on the redshift forums a bunch, I trust this intel. They are very aware of the situation, they are not misinformed about it already being out. They said they have not tried it or have no experience using it so cant say about the performance. However Redshift Devs had to work closely with apple to get a Metal version out just this last year. My sense is that apple threw a bunch of money at the problem after hearing so many Pros in this field complain about the Nvidia situation. Redshift devs are very responsive, trustworthy and I would guess they have good source intel. Of course this is all just my speculation.

OK well of course I hope you're right. I do still have some concerns about it. The wording "we've not tested it" implies either what I said before - that they think it's out already when it's not - or else, if you're right, that they know for sure they could test it, meaning they must have a private beta build (perhaps of 11.4?). However would they then be able to talk so openly about it? Wouldn't that breach an NDA? I guess it's possible Apple gave them permission to do so?

Oh well, as you say it's all pointless speculation. Something to do before going back to the waiting!

 

[pastrychef]

There's always the possibility that the testing of 6000-series drivers is on Apple Silicon Macs...

Also, I find it a little odd that these devs are not bound by any NDAs... That's very un-Apple-like.

 

[Andy1983]

There's always the possibility that the testing of 6000-series drivers is on Apple Silicon Macs...

Also, I find it a little odd that these devs are not bound by any NDAs... That's very un-Apple-like.

For context this is the question that was posed on the Redshift forum....

"Can the devs confirm (or deny) if RX 6000 support is coming & and if not is that because:
1) you simply don’t know
2) aren’t allow to share
At this point investing in Redshift on Mac is becoming a risky capital investment especially if hardware requirements become Mac on-system GPU only. We are talking thousands of dollars of possible in extra cost.
It would suit the Redshift team to at the very least try & mitigate this risk for it’s customers by sharing as much as they can. Even a “no” would be better than “we don’t know”.
It is hard to run a business on non-information."

So the poster did give the Redshift dev plenty of room to say he was bound by an NDA, or he has no information. Redshift and Octane have gone out of their way to support AMD on OS X, this is in the interest of Apple and its user base, so I guess they are privy to inside information.

I guess we just need to wacth this space and wait for the 12th April when we will find out more.

 

[TheBloke]

There's always the possibility that the testing of 6000-series drivers is on Apple Silicon Macs...

Ha, that'd be a real kick in the teeth for us Hackintoshers - not to mention many Intel Mac owners: if the drivers did come.. on Silicon only But hopefully they plan to do dual-binaries for any new code for a while longer.

Also, I find it a little odd that these devs are not bound by any NDAs... That's very un-Apple-like.

Yeah, that was my concern with the Redshift remarks: to be useful their comments need to refer to specific knowledge of as-yet un-released code, but any such code (or Apple communication regarding it) would almost certainly be under NDA. As they'd be quite unlikely to break NDA, and it also seemed unlikely that they'd get permission from Apple, it made me doubt if they did in fact have extra information.

I see the remarks from Beppe (the Octane dev) as being a little different, though. In that case I think it's plausible that he did not believe his remarks were breaking any NDA. He said he was told by Apple that drivers were available in 11.3 beta 3, and he only said that after that specific version was already released on the public seed. As he believed that the drivers were now publicly available, he may well have felt he was fine to talk about it. He had no 6000-series GPU to test it himself, so it was only later that he realised that drivers weren't released when Apple told him they would be.

So the poster did give the Redshift dev plenty of room to say he was bound by an NDA, or he has no information. Redshift and Octane have gone out of their way to support AMD on OS X, this is in the interest of Apple and its user base, so I guess they are privy to inside information.

Thanks very much for the further context. That is definitely interesting. I can certainly see - and would expect - them to be privy to inside information of relevance to their products. I do still find it a little surprising that they're able to talk about it, but I guess it's possible that Apple gave the OK in this instance.

Fingers' crossed for 11.4 beta 1!

 

[TheBloke]

I ran the screenshot provided by Harris0n through Google Translate on my phone:

Apple are going to sell a 6900XT? Well that would certainly be very interesting if true..

But do Apple actually sell discrete GPUs in normal sales channels?

I can't help feeling this is more likely an error than anything interesting, but I'd love to know more context.

@Harris0n do you have an more information you could send, like the original link for this picture?

 

[Andy1983]

I ran the screenshot provided by Harris0n through Google Translate on my phone:

View attachment 514259

Apple are going to sell a 6900XT? Well that would certainly be very interesting if true..

But do Apple actually sell discrete GPUs in normal sales channels?

I can't help feeling this is more likely an error than anything interesting, but I'd love to know more context.

@Harris0n do you have an more information you could send, like the original link for this picture?

Hmmm? does that say Minidp? Mini DisplayPort?

That is old technology no?