HDMI Forum Rejects Open-Source HDMI 2.1 Driver Support Sought By AMD

[duby229]

That's true in theory, but often false in practice. Patent lawyers get paid a lot of money to try and make their patents as broadly applicable as possible, and patent offices are incredibly overworked and often approve things they shouldn't.

Of course, if something isn't exact enough that's a good defense if you want to get involved in a patent lawsuit. You can claim that the patent isn't valid and should be thrown out. But I suspect AMD, Red Hat, and others have no desire to spend millions fighting a patent lawsuit over this. Often the threat of a patent lawsuit is the goal more than actual lawsuits.

Except that I don't believe that applies at all in this scenario. This is clearly a case of somebody made something and realized elements of it could be patented. Which is how most patents are derived. Clean room reverse engineering would reveal specifications that an original work could be based on. At which point it's a license issue, not a patent issue.

Once an original work is created there is no patent issue.

[Sonadow]

People can start realizing that they are computers and not a religion, forget about trying to stuff their ideological beliefs down everyone's throats and just write software that allows people to use the hardware theu paid good money for to the fullest extent.

That would actually require people to think rationally.

[Sonadow]

That is the most delusional nonsense I have read in a long while. The by far most frequent complaint I have heard from Windows users over decades is:

"I need my devices X and Y to work, but latest driver for X is compatible only up to Windows version N while a driver for Y is only available for Windows version N+1."

Which is the same situation you find yourself in with any operating system if you have to rely on closed-source drivers.

That is the most ridiculous nonsense ever.

The stable ABI in Windows is stable for a reason. A driver from Windows 8 will still work in Windows 11 and my computers are living proof of that.

Not that Linux users will ever know what a stable ABI is.

[Panix]

The active adapter does nothing of that sort. It just translates between DP and HDMI signaling and protocol.
DP++ based passive adapters are more simple but otherwise there are only disadvantages of this approach. Complexity on host side increases, not supported via USB-C and/or MST hubs, etc.

This is not accurate. Only the people who contribute to the implementation must lack access to proprietary documentation, not every employee at AMD.

This is why an Intel employee (Alan Cox) was able to write large parts of the gma500 driver, and presumably one of the reasons why Intel didn't want Poulsbo code in i915, despite considerable amounts of code duplication between the drivers.


Blurbusters investigated this a while back. Most DP to HDMI 2.x adapters support FreeSync in hardware, just the firmware does not translate the EDID properly.

In some cases you could use firmware intended for another adapter to get it working properly. Alternatively, pass a custom EDID (generated with CRU or similar) and FreeSync may work.

Some reading:
https://www.reddit.com/r/linux_gamin...cable_matters/ (user report about working VRR with the CableMatters adapter)
https://twitter.com/BlurBusters/stat...26715474665473 (about Club3D CAC-1080/1085)

A reddit user claims 'VRR is working.' That's your proof? I wouldn't trust that - not with the prices of those adapters. Perhaps, one of the reviewers can test it? It seems to be a small niche market - Linux users who want to use amd gpu cards with HDMI TVs?
Some ppl on a github site claim (iirc) that if they update the firmware - it then works - so either a) you need to update/install firmware - unless, the company (Cable Matters?) is updating the version of their adapters to include the new firmware? That is, if this *solves* the issue in the first place - so the new firmware is able to 'translate the EDID properly?'

[Anux]

So do the windows get moved if you switch off some of your monitors?

No, only if I cut the power to the monitor or unplug the DP cable which is also the case with HDMI.

[TheLexMachine]

It requires testing certification and license payments, plus the hardware circuitry and software required to utilize it and the patent for PC usage/implementation of CEC is held by Fujitsu, so they have to be paid, until the patent expires in the the next few years. In ye olden days, HDMI output was achieved by a conversion chip that had all the HDMI stuff built in - this is how it's done in consumer electronics like TV and disc players and game consoles - and now, everything is built into the GPU itself, so it's not possible anymore and requires GPU space to be allocated to implement. The short answer is, it costs something and nobody is willing to pay for it.

A bit of an update on my earlier post, in that HDMI-CEC is apparently now being implemented in the newest Intel Arc GPU products that utilize HDMI 2.1, such as the upcoming NUC-type units from ASUS and others, as well as the upcoming Battlemage (Arc-2) GPUs.. There is no information on whether or not the next-gen AMD and Nvidia GPUs will have CEC in the GPU designs, but I would expect AMD to implement it, given how popular their APU and SFF builds are with people who connect their PCs to TVs.

[oiaohm]

The wikipedia is not always right. There is a key part of patent law Wikipedia nicely skips over.

So clean room does not typically work against patents.

https://www.brettoniplaw.com/requirements%20for%20a%20patent.html

35 U.S.C. §103 - Non-obviousness

A POSITA would be the typical scientist or engineer working in the relevant field.

Yes doing a clean room implementation is doing a POSITA test with Patents to Invalidate the patent. This is hard to setup at times. to make sure the POSITA people in the clean room at not tainted.

We can also compare with a hardware patent, e.g the Asetek one where they have a patent to have the pump inside the cooling plate, every single competitor to them that tried to go the "independent invention" route was sued out of oblivion which is why the existing competition all have their pumps at different locations. Proving obviousness is not as simply as you make it out to be.

Asetek gets away with it because of this bit in 35 U.S.C. §103 - Non-obviousness

Where there is an accepted understanding in the field that the combination of elements would have been undesirable

Anyone who has an Asetek design AIO fail knows this one way too well. Even so Asetek had to buy out 2 companies because their patents would not work because they had correct clean room implementations that could have nuked their patent. One had the pump on the CPU block 90% off to the Asetek. Asetek worked these 2 companies they were going to lose in court so completely acquired the company. Asetek has never solved the fact that over time the liquid inside a sealed AIO disappears. There are in fact older non sealed AIO designs that have the pump on the CPU as well. This is the problem the non sealed is the obvious design for quality AIO that lasts so that the user can top it up as the liquid disappears .

Yes patent law for you. You patent something that is broken garbage by design patent law will allow you to keep on enforcing your patent to punish those who use the garbage design.

Patent law its about obviousness. Not just simple obviousness either. It has to be obviousness that it a good design that a person would have came up with.

Clean room with copyright is not going to cause the other party to consider buying your company outright to protect their copyright. Clean room with Patent if they are a larger company they see you doing clean room on their patents that could prove a key patent they are making money from as obvious buying your company out to stop you comes really possible.

Clean room usage with patent is pure offensive. This is not to make a competing implementation this is purely to destroy the patent by proving it obvious good design. Yes once you destroy the patent you can just clone their product. This is how some china clone product makers get away with it legally because they destroy the patent that could be used to block them shipping around the world.

[billyswong]

No, only if I cut the power to the monitor or unplug the DP cable which is also the case with HDMI.

HDMI/DVI/VGA can keep windows position/monitor detection even if a monitor is powered off/cut as long as it is plugged. So the difference exists indeed.

[Anux]

HDMI/DVI/VGA can keep windows position/monitor detection even if a monitor is powered off/cut as long as it is plugged. So the difference exists indeed.

You're right, I did a short test with DVI and windows stay there even if I cut power. Not sure if I like that behavior, because unless you are to reconnect that monitor your windows are somewhat inaccessible.

The obvious solution would be to not cut the power to your monitors when using standby. Edit: or to power them on before wake up

[hartree]

A reddit user claims 'VRR is working.' That's your proof? I wouldn't trust that - not with the prices of those adapters. Perhaps, one of the reviewers can test it? It seems to be a small niche market - Linux users who want to use amd gpu cards with HDMI TVs?
Some ppl on a github site claim (iirc) that if they update the firmware - it then works - so either a) you need to update/install firmware - unless, the company (Cable Matters?) is updating the version of their adapters to include the new firmware? That is, if this *solves* the issue in the first place - so the new firmware is able to 'translate the EDID properly?'

Sounds like an opportunity for Michael to do a Digital Foundry style investigation of the matter

The 124 firmware for Chinese adapters doesn't enable VRR on Windows for some reason, so I doubt Cable Matters will update their description. VRR is also still quite fickle even without the adapter, the newest 6.7.6-201 kernel on Nobara seems to have a regression that causes VRR to malfunction in the same way it does when HDR is enabled even in SDR (refresh rate jumping).