HDMI Forum to AMD: No, you can’t make an open source HDMI 2.1 driver

D

DanNeely

Supreme [H]ardness
Joined
Aug 26, 2005
Messages
4,481
https://arstechnica.com/gadgets/202...you-cant-make-an-open-source-hdmi-2-1-driver/

Linux blog Phoronix noted in January 2021 that the HDMI Forum did not offer public access to the HDMI 2.1 specification. Alex Deucher, an AMD engineer who has long contributed to the company's open source offerings, has kept a related bug thread alive for at least two years, only to deliver the negative outcome yesterday.


In February 2023, Deucher reported that he was "working with our [AMD] legal team to sort out what we can deliver while still complying with our obligations to HDMI Forum." Two months later, he said that AMD got "the basic functionality up and running, now we have to go through each of the features with legal and determine if/how we can expose them while still meeting our obligations." Summer and fall of 2023 went by, with legal review still underway, and in October, the decision was "in the hands of the HDMI Forum."


On Wednesday afternoon, Deucher offered the current resolution:


The HDMI Forum has rejected our proposal unfortunately. At this time an open source HDMI 2.1 implementation is not possible without running afoul of the HDMI Forum requirements.
Click to expand...
Click to expand...

Display port just keeps looking better every day.
 
I'm not an expert on the details, but I hear that even among CPUs, it is true that only Intel and AMD are *allowed* to produce x86 CPUs.
Something about that seems fishy, and needs to be said, and this context is as good as any. However, that implies that Intel is the one preventing competition.

As for this issue; I wonder what's so secret about HDMI's code that it can't be released?
 
1337Goat said:
I'm not an expert on the details, but I hear that even among CPUs, it is true that only Intel and AMD are *allowed* to produce x86 CPUs.
Something about that seems fishy, and needs to be said, and this context is as good as any. However, that implies that Intel is the one preventing competition.

As for this issue; I wonder what's so secret about HDMI's code that it can't be released?
Click to expand...

There's also VIA and a few other manufacturers that make x486/pentium class chips for embedded purposes. There the issue is that Intel owns a ton of IP on the base x86 instruction set and vector extensions (MMX, SSE, AVX), and AMD has the same on the 64 bit extensions to it. The old stuff was never locked up the same way, but to do anything vaguely modern you'd need licenses for the relevant parts from both companies.
 
DanNeely said:
There's also VIA and a few other manufacturers that make x486/pentium class chips for embedded purposes. There the issue is that Intel owns a ton of IP on the base x86 instruction set and vector extensions (MMX, SSE, AVX), and AMD has the same on the 64 bit extensions to it. The old stuff was never locked up the same way, but to do anything vaguely modern you'd need licenses for the relevant parts from both companies.
Click to expand...

Or figure out something that works better without that intellectual property. Not impossible but unlikely, probably the closest is ARM, but even then they still have a ways to go to really challenge Intel or AMD.
 
Gideon said:
Or figure out something that works better without that intellectual property. Not impossible but unlikely, probably the closest is ARM, but even then they still have a ways to go to really challenge Intel or AMD.
Click to expand...

Either we as a civilization choose to have law that protects this sort of "Best way known way to process data" IP, or law that does not allow intellectual property of this nature to be kept exclusive or secret. It really depends on how important it is for any random company to be able to compete against Intel in terms of producing a well performing physical product, versus being able to become the next Intel by inventing something better, but different.

Both paths seem to have their uses, but I would advise against any path that's built upon a foundation of greed or selfishness of mere ideas.
Whatever will grant humanity the best future is what should be chosen. Same with AMD and this HDMI issue. Is this sort of situation something that we as civilization find tolerable? The inability to release an open-source version of something that interfaces with hardware? Is this what we actually want?

I'm failing to see how anyone other than HDMI Forum benefits from this, furthermore because there are already alternatives that are better, I wonder just what sort of schemes HDMI Forum has, when it comes to how it will compete against DisplayPort or other more free or more powerful alternatives. I wonder if they would try to patent the very nature of transmitting data through a wire if they had the power to make that happen. How far will these companies go to prevent competition from existing? I can see why EVGA left the graphics market, in that context.
 
Last edited:
Pretty sure that the HDMI specs are closed because of HDCP, and the agreements with the movie industry. It's the copy protection that they would be worried would get hacked/cracked if it was open source, even partially.

Just an educated guess.
 
No, I think the spec is closed because they want to collect the membership free that you have to pay to see the spec.

An open source software implementation would enable others to implement the spec without being a member.

To make sure this is understood, Displayport is unaffected by this nonsense.
 
DisplayPort has always been a better alternative imo anyway with its only limiting factor for me being hdmi is used much more universally by the tv industry.
 
This is a repugnant situation. Consider that many monitors, including the recent often-used-as-monitors LG CX/C1/C2/C3 OLED displays only have HDMI inputs, and that HDMI 2.1 is necessary in order to have the full feature set like 4K 120hz HDR 10-bit 4:4:4 VRR/Gsync/FreeSync, this is infuriating fron a usabiity standpoint with the ubiquity. I agree with Goat above that this is another issue with IP law run amok and yet another example of why it tends to serve powerful incumbents; note that the HDMI organization changed their demands relatively recently in a way that was not the case in the past. Furthermore, this is a penalty to - like sadly many things in this society - those trying to do things "right".

AMD should be able to advertise their open source fully capable drivers on Linux as a feature, just like their use of many other open source standards. However at the practical level, having watched for the last two decades or so, there are only a handful of users who are willing to GIVE UP functionality because of an ideological viewpoint. Lots of new Linux users love the ability to game on their Steam Decks or manual Linux install thanks to Proton, and more devs are porting their games to Linux native because this growing interest. However, when making the choice between their next GPU, if it comes down to "get Nvidia, use their proprietary drivers but everything works at full resolution vs "Get AMD which I like their open source drivers, but it doesn't work on my HDMI 4K monitor without limitations", lots of people will just go with the former - something that is no fault of AMD's, but rather the HDMI Forum. This is just one partial example but still, the issue shouldbe now how do we get around it?

Solutions at this point come down to either the technical to the legal. From the technical point of view, one stop-gap is perhaps a DP > HDMI converter, but so far many of those i've seen are either A) expensive B) maybe don't support ALL the features of a display like mine I listed above so its still undesirable. Is anyone aware of a converter capable of the full 48gbps bandwidth, 4K 120hz VRR HDR, wide color gamut 10bit ? Some claim to be able to do so but then user reports suggest otherwise. Maybe there are some good ones out there, but an active converter is probably only a stopgap. The other solution may have to be legal. Some hypothesize that the more permissive legal environment of Europe will rule for interoperability or at very least can be another "the users must manually download the video and audio codecs if their local jurisdiction allows it *wink* " situation . I saw some discussing that simply calling it HDMI Compatible (a la IBM Comaptible) may get around the legal issues and if that's the case, more power to them - no diferent than having a monitor that's not officially Gsync supported, but is "gsync compatible" because its using VESA ActiveSync / VRR as part of te FreeSync standard. There are other solutions too, like AMD just paying for the inclusion that allows them to release it FOSS if possible, or create a tiny proprietary add-on for HDMI support functionally; that is the strength of their GPU drivers on Linux where they can have a FOSS base and proprietary add-ons as it was, if necessary. Hell, maybe a mesh of the two where the add-on is in open source form in another jurisdiction as a HDMI-Compatible bit, yet a binary blob when included officially?

Either way, this cannot be allowed to stand. Those who will not find themselves deeply interested in the politics of free software and Linux will instead look at it as "something AMD can't do, but Nvidia can" from a pragmatic standpoint and that's unfortunate. Hopefully a variety of soluitions officially or otherwise will come soon.
 
RanceJustice said:
This is a repugnant situation. Consider that many monitors, including the recent often-used-as-monitors LG CX/C1/C2/C3 OLED displays only have HDMI inputs, and that HDMI 2.1 is necessary in order to have the full feature set like 4K 120hz HDR 10-bit 4:4:4 VRR/Gsync/FreeSync, this is infuriating fron a usabiity standpoint with the ubiquity. I agree with Goat above that this is another issue with IP law run amok and yet another example of why it tends to serve powerful incumbents; note that the HDMI organization changed their demands relatively recently in a way that was not the case in the past. Furthermore, this is a penalty to - like sadly many things in this society - those trying to do things "right"....
Click to expand...
It's a pointless rant. Displayport is a better interface anyway. When buying a PC monitor it would be hard to find one without displayport.

Use that.

It's not up to you, me, or AMD to take someone else's technology and decide that "this must be open source! For the Good of all mankind!!!!!!"
 
Yeah, seems to me, in this instance, just use DP. HDMI Forum can shove it, maybe I'll just look at devices with HDMI differently (less) now.

My DP/vga monitor certainly doesn't mind.
 
As an Amazon Associate, HardForum may earn from qualifying purchases.
1337Goat said:
I'm not an expert on the details, but I hear that even among CPUs, it is true that only Intel and AMD are *allowed* to produce x86 CPUs.
Something about that seems fishy, and needs to be said, and this context is as good as any. However, that implies that Intel is the one preventing competition.

As for this issue; I wonder what's so secret about HDMI's code that it can't be released?
Click to expand...
I'd venture to guess that had AMD not invented AMD64 at the moment in time that it did, Intel was probably getting ready to revoke it's x86 license from AMD.
 
pavel said:
Not with big screens - i.e. 50"+

It's not just the HDMI Forum - it's the TV companies/manufacturers, Hollywood etc. Why won't the TV companies include/manufacture even one display port on their TVs?
Click to expand...

Because it's just like what I implied! HDMI Forum is doing more than merely "protecting their IP." They're creating scheming deals and corrupt contracts to stifle competition and prevent us from buying things that we actually want. It would be easy to say, "Then just buy or MAKE a big-screen TV with DP!" But we know that won't happen. We know that we cannot do that, so to even suggest it would be raw malice and mockery of our helplessness as consumers in this dystopian era. It all comes back to that same concept that keeps haunting me: None of us in the world are allowed to make an X86 processor. Not in our basement, not in a lab, not in a factory. Only the people licensed to do so can make them, which means that if we were to begin to deeply ponder what choices we truly have in life, AMD vs Intel, AMD vs Nvidia, Samsung TV vs Visio TV, and so on. What's really going on here?

The darkest truths are the truths that are true, and let it be known that it is surely true that HDMI Forum, and other big-name companies and organizations, are in cahoots to deliberately keep us from creating or using alternatives. If we as society can deem so many things illegal, then why not this sort of thing too? Who is making rules and laws if not "We the people?" And if not "We the people," then what dark implications does that present?

westrock2000 said:
I'd venture to guess that had AMD not invented AMD64 at the moment in time that it did, Intel was probably getting ready to revoke it's x86 license from AMD.
Click to expand...
Bingo. But all that means is that evolution of technology is forcefully halted because we cannot improve upon what currently exists. We are supposed to start from scratch? Or to take an inferior product such as RISC, and claim that we should just make a deliberately reduced instruction set somehow on the same level as what Intel can do? Yet the 14900K from Intel is only roughly equal to their 13900K from last generation. In some benchmarks; worse. What a world we live in. And what a future we shall have.
 
RanceJustice said:
Either way, this cannot be allowed to stand. Those who will not find themselves deeply interested in the politics of free software and Linux will instead look at it as "something AMD can't do, but Nvidia can" from a pragmatic standpoint and that's unfortunate. Hopefully a variety of soluitions officially or otherwise will come soon.
Click to expand...
Nvidia doesn't exactly release open source drivers like AMD does. If it's in the source code then yea AMD has no foot to stand on. Better to look at Intel to see if their open source drivers support HDMI 2.1.
GoodBoy said:
It's a pointless rant. Displayport is a better interface anyway. When buying a PC monitor it would be hard to find one without displayport.

Use that.

It's not up to you, me, or AMD to take someone else's technology and decide that "this must be open source! For the Good of all mankind!!!!!!"
Click to expand...
As great as that sounds, I still see HDMI sticking around for some time. For most people it's a glorified wire. VGA stuck around for a long time because it just worked. I'd like to know why AMD can't do it but Intel can?
 
Mazzspeed said:
What's odd is the fact that I'm pretty sure HDMI 2.1 is supported by Intel and their open source driver.
Click to expand...
HDMI 2.1 is supported but not native.

Current Intel Arc Graphics (DG2/Alchemist) only natively support HDMI 2.0 but HDMI 2.1 can be an option with some of the cards where the board vendor chooses to use a PCON for converting a DisplayPort 2.0 output from the GPU to an HDMI 2.1 interface.
 
Lakados said:
HDMI 2.1 is supported but not native.

Current Intel Arc Graphics (DG2/Alchemist) only natively support HDMI 2.0 but HDMI 2.1 can be an option with some of the cards where the board vendor chooses to use a PCON for converting a DisplayPort 2.0 output from the GPU to an HDMI 2.1 interface.
Click to expand...
So there's essentially a built in active converter DP 2.0 > HDMI 2.1 built into the card? Does it support full bandwidth and feature set that a native HDMI 2.1 ( for instance, 4K 120hz HDR 10 bit color at 4:4:4 FreeSync/VRR etc) would or are there some limits because of the conversion, either at the hardware/converter level or the driver level?
 
RanceJustice said:
So there's essentially a built in active converter DP 2.0 > HDMI 2.1 built into the card? Does it support full bandwidth and feature set that a native HDMI 2.1 ( for instance, 4K 120hz HDR 10 bit color at 4:4:4 FreeSync/VRR etc) would or are there some limits because of the conversion, either at the hardware/converter level or the driver level?
Click to expand...
From what I know it has the full DP 2.0 feature set being converted over. So DP2.0 does everything that HDMI 2.1 does for 4k and below, so it should. The problem for VRR and HDR in HDMI 2.1 those are proprietary and there is no documentation that is available for 3’rd parties.
I honestly do not know if those functions work for Intel opensource drivers or not, and I can’t find much in the way of anybody having tested it.

But no HDMI 2.1 Native for AMD or Intel in the near future on Linux from what I can tell.

Update:
A cursory glance over at Phoronix wouldn’t indicate that
1. HDR does not work correctly on Linux at all currently, this is an Xorg issue and they are on it.
2. VRR over HDMI 2.1 is only working with Nvidia’s closed drivers.
3. Intel and AMD are currently stuck in HDMI licensing hell because of how the 2.1 license was structured and it seems to punish open source.
 
Last edited:
Lakados said:
HDMI 2.1 is supported but not native.

Current Intel Arc Graphics (DG2/Alchemist) only natively support HDMI 2.0 but HDMI 2.1 can be an option with some of the cards where the board vendor chooses to use a PCON for converting a DisplayPort 2.0 output from the GPU to an HDMI 2.1 interface
Click to expand...
Hardware aside, I'm fairly certain HDMI 2.1 is supported by Intel's FOSS drivers.
 
The article is a bit strange, why would anything AMD do here would need to open Source ?

Could not just have the relevant HDMI part in their drive be closest source ? Seem there is 2 people in the dance responsible for not making this happen.

Nvidia-Intel card push 120hz-4k on hdmi on Linux no ?
 
LukeTbk said:
The article is a bit strange, why would anything AMD do here would need to open Source ?

Could not just have the relevant HDMI part in their drive be closest source ? Seem there is 2 people in the dance responsible for not making this happen.

Nvidia-Intel card push 120hz-4k on hdmi on Linux no ?
Click to expand...
AMD's whole stack is open on Linux. The only closed bits they have are their pro drivers, which they don't recommend unless you're doing "pro" things. Much of the closed driver's source is shared with the open driver anyway, I should add.

Therefore, it makes sense for as much to be open as possible, especially in the kernel driver, so you don't have to install two different things to get it working (some distros don't include drivers if they include binary blobs, or depend on binary blobs in some cases--you might even have to enable an external repository).
 
Last edited:
Nobu said:
Therefore, it makes sense for as much to be open as possible,
Click to expand...
yes sure, but if it is not possible, simply have it close.... this would be on the distro side of things fault and choice like them refusing AMD pro drivers, not HDMI or AMD, not sure how popular that kind of distro would be in the want to plug my linux machine on hdmi 2.1 tv
 
LukeTbk said:
yes sure, but if it is not possible, simply have it close.... this would be on the distro side of things fault and choice like them refusing AMD pro drivers, not HDMI or AMD, not sure how popular that kind of distro would be in the want to plug my linux machine on hdmi 2.1 tv
Click to expand...

Closed source drivers suck on a technical level, starting with constantly breaking when the kernel changes.
 
uOpt said:
Closed source drivers suck on a technical level, starting with constantly breaking when the kernel changes.
Click to expand...
It's not so bad with dkms, but can be confusing when you upgrade the kernel and don't realize you need to reboot to use the newer kernel with the driver.
 
GoodBoy said:
It's a pointless rant. Displayport is a better interface anyway. When buying a PC monitor it would be hard to find one without displayport.

Use that.

It's not up to you, me, or AMD to take someone else's technology and decide that "this must be open source! For the Good of all mankind!!!!!!"
Click to expand...
Your point is mostly valid. HDMI is not AMDs alone.
However HDMI isn't a company. Its an agreed on industry standard with development on the standard and its specs performed by the core members of the forum of which AMD is one. The standard was initially developed by the likes of Sony, Panasonic, Thompson/technicolor and a few others. However following updates have had much more input from the 80 some other members of the HDMI Forum. AMD pushed new specs themselves such as 1.3.
 
LukeTbk said:
yes sure, but if it is not possible, simply have it close.... this would be on the distro side of things fault and choice like them refusing AMD pro drivers, not HDMI or AMD, not sure how popular that kind of distro would be in the want to plug my linux machine on hdmi 2.1 tv
Click to expand...
AMD has refused their pro drivers. They create them yes... but they are not cutting edge drivers. They are aimed specifically at old life time service setups. The pro drivers are updated a 1/4 as often if that. AMDs own work is on their open source driver. A few times a year they push a super stable version to the pro driver. (think of it like a windows signed driver. only in this case its expected to go on servers running airport and banking stuff... and is treated as such from a stability standpoint. If a performance feature isn't 100% stable its not going in the pro driver)

It is what it is really. AMD has went full in on open source good an bad. They haven't left themselves the ability to bolt on closed source firmware crap like Nvidia has. It is a big part of why the Linux community has helped AMD as much as they have. There is a much higher level of trust. Nvidia likes to talk up open source lately... while building themselves multiple ways to hook in closed source. AMD laid their hand on the table if you will. That is how you get companies like valve to invest 100s of millions making your product work on Linux. AMD has been acting in good faith in regards to open source.
 
Nobu said:
AMD has no firmware blob for the HDMI bits to reside in, for better or worse.
Click to expand...
There is a proprietary firmware blob regarding AMDGPU.

https://wiki.gentoo.org/wiki/Linux_firmware

Linux firmware is a package distributed alongside the Linux kernel that contains firmware binary blobs necessary for partial or full functionality of certain hardware devices. These binary blobs are usually proprietary because some hardware manufacturers do not release source code necessary to build the firmware itself.

Modern graphics cards from AMD and NVIDIA almost certainly require binary blobs to be loaded for the hardware to operate correctly.
Click to expand...

uOpt said:
Closed source drivers suck on a technical level, starting with constantly breaking when the kernel changes.
Click to expand...

Only when running the very latest bleeding edge kernels, stay a release back and you're fine.
 
Last edited:
Mazzspeed said:
There is a proprietary firmware blob regarding AMDGPU.

https://wiki.gentoo.org/wiki/Linux_firmware
Click to expand...
Ah, guess they added it with GCN. Been a while, should start reading OSS news again.

Still, it's another feature they'd have to add to the firmware -- if there's not enough room, nothin' doin'.

bridgman said:
AFAIK it's not a matter of just releasing the required microcode, it has to be written first and I believe there were some technical obstacles there - I think we hit a size limit in a microcode store. I don't know how bad the required hacks would be to run on amdgpu with the microcode images we developed for radeon but it might be worth revisiting that.
Click to expand...
https://www.phoronix.com/forums/for...re-holding-back-community-s-development/page2

This is an old quote regarding gcn1.1 APUs, but I wouldn't be surprised if it ever happens again -- space on Si is precious after all.
 
GoodBoy said:
It's a pointless rant. Displayport is a better interface anyway. When buying a PC monitor it would be hard to find one without displayport.

Use that.

It's not up to you, me, or AMD to take someone else's technology and decide that "this must be open source! For the Good of all mankind!!!!!!"
Click to expand...

Look, I'm all in favor of proprietary solutions and driving innovation, but if something is going to be an industry standard, that is a problem.

If something is going to be used as an industry standard - like HDMI is - then it should be free for all to use, without restriction. Otherwise you re giving some industry consortium a form of monopoly power to decide who can and cannot participate in the marketplace, and that is just wrong any way you look at it.
 
GoodBoy said:
It's a pointless rant. Displayport is a better interface anyway. When buying a PC monitor it would be hard to find one without displayport.
Click to expand...
I set up a brand new MSI Cubi for a client the other week along with two brand new 24" monitors, one monitor was connected via HDMI, one monitor was connected via DP. On a cold boot both monitors worked fine, do restart or a warm boot and the DP monitor would fail to sync even on POST - To the OS (Windows 11) the DP monitor was invisible. Swap monitors, same deal. Swap DP leads, same deal. Force shutdown PC and power back up, same deal.

In the end I used a passive DP > HDMI adapter and everything worked perfect.

On my own PC running a single 4k monitor via DP, if I do a reboot, the monitor cycles through all available inputs before syncing and locking onto the DP input - It's a PITA when you want to see the UEFI splash screen to access the UEFI/BIOS. This has been the case running a 980Ti as well as a 2070S.

Essentially, compared to HDMI, in my experience DP isn't without issues.
 
Mazzspeed said:
I set up a brand new MSI Cubi for a client the other week along with two brand new 24" monitors, one monitor was connected via HDMI, one monitor was connected via DP. On a cold boot both monitors worked fine, do restart or a warm boot and the DP monitor would fail to sync even on POST - To the OS (Windows 11) the DP monitor was invisible. Swap monitors, same deal. Swap DP leads, same deal. Force shutdown PC and power back up, same deal.

In the end I used a passive DP > HDMI adapter and everything worked perfect.

On my own PC running a single 4k monitor via DP, if I do a reboot, the monitor cycles through all available inputs before syncing and locking onto the DP input - It's a PITA when you want to see the UEFI splash screen to access the UEFI/BIOS. This has been the case running a 980Ti as well as a 2070S.

Essentially, compared to HDMI, in my experience DP isn't without issues.
Click to expand...
Sounds like that monitor isn't listening for the wake-up signal on the displayport input. Bummer. I'd send the manufacturer a message, maybe they'd update the firmware if it's not a hardware fault. Apparently it's common, though.
 
Mazzspeed said:
On my own PC running a single 4k monitor via DP, if I do a reboot, the monitor cycles through all available inputs before syncing and locking onto the DP input - It's a PITA when you want to see the UEFI splash screen to access the UEFI/BIOS. This has been the case running a 980Ti as well as a 2070S.
Click to expand...
is there not an option in the monitor to stop the cycling? that would be annoying... might try asking the oem like nobu just suggested.
 
You must log in or register to reply here.
Back
Top