Tiny Corp At "70%" Confidence For AMD To Open-Source Some Relevant GPU Firmware

[Brane215]

I'm sure AMD will move heaven and earth to sell those 6 XTXs... I know he might sell 1000 tinyboxes with 6 each. His words just sound so ridiculous.

That's a stupid argument. AMD might not care about their quantities but surely could care about their use case and profile of future potential customers that might see the advantage of open firmware.

They also are another drop in a bucket that might push AMD to do something one of these days. As AI comes in, people are making ever more uses for GPUs and NPUs and AMD might want to open the door as wide as possible to general public.
They need marketshare and AI is the whole world to GPU manufacturers, so why not ?

[Doomer]

That would be hilarious, if it wasn't so awful. And it shows just how far that gap in compute closed with ZLUDA. That being said, I'm not sure how much that clause in the EULA can be enforced, legally speaking.

I guess it's a way to attack devs who make these translations layers, I doubt they can do anything to consumers who use them.
they will just sue devs who can't fight back becaus it would cost a lot of money, it does not matter if it does not hold in court, most devs won't risk it.

That's why most tech makers artificially lock things down via software.

the current state of smartphones is depressing, we can't even use root without losing access to lots of services, this is truly depressing to see how low we've fallen.
these golden jails are any big corpo wet dream and we don't have a single reason to play along as these things are anti-consumer by nature, there is no such thing as a good closed ecosystem.

There might be an argument to be made that you're using the software outside of it's intended use

I buy it, it's mine, whatever I run on it isn't any of their business, they should not be able to still have a say on how I use the stuff I BOUGHT and OWN, otherwise it's renting.
reminder that renting does not imply a recurring fee, I can rent something for an initial fee and they can gracefully decide to not ask for any more dollar when I ditch the product, that's how the smartphone world operates for phones you can't unlock the booloader and therefore never really own.
It should be illegal to sell a service disguised as a product, I don't see anywhere on any website that if I buy a rtx gpu, I agree to never run specific pieces of software on it, same for my smartphone.

[Mystro256]

We also advised that the build process for amdgpu-dkms should be more open.

Strange, anyone with a basic understanding of dkms and package post scripts should be able to produce the package pretty easy. Granted, they might be talking about making it a 100% reproducible package instead, i.e. like a source package (dsc or srpm).

[Jabberwocky]

That would be hilarious, if it wasn't so awful. And it shows just how far that gap in compute closed with ZLUDA. That being said, I'm not sure how much that clause in the EULA can be enforced, legally speaking. ZLUDA doesn't modify the CUDA binary in the same way that WINE doesn't modify windows binaries. NVidia can't legally require you to not run other pieces of software on your device. That's why most tech makers artificially lock things down via software.

There might be an argument to be made that you're using the software outside of it's intended use, but again, as long as you're not hacking into it I don't think there would be a strong case.

I have tested ZLUDA and recent ZLUDA forks extensively. IMO the gap between rocm and CUDA is still wide open. ZLUDA just shows us another way the gap can be closed. It doesn't close the gap by any means yet. Nobody has stepped up to say they are going to close that gap. I am doubtful about zluda in the long run.

I agree with you about the legal stuff. Nvidia isn't happy about it. They will probably implement something that makes it more difficult for zluda to work.

[coder]

Well, I guess it's a really big chance for AMD, to actually take a bit of this market from nvidia

They're not taking anything from Nvidia. This clown is using gaming GPUs. His business model probably requires him to use such cheap cards. If he used Nvidia GPUs, he'd have to buy expensive workstation-grade cards, because Nvidia's CUDA license forbids their gaming cards to be used in such a manner.

and show Radeons are not just toys.

But they are toys, and therefore what he's building will be a toy.

People act like he's doing AMD such a favor, but all he's doing is moving a few low-margin gaming cards. If he really wanted to build a solid product for his customers and be a more valuable partner for AMD, the least he could do is use Radeon Pro W7900, which also have double the RAM.

[coder]

why do you think AMD, which makes both the best GPUs and (the best) CPUs, is worth 1/7 of nvidia which makes only (the best) GPUs?

It's funny to see someone so ignorant imply that someone else is ignorant.

[Railander]

It's funny to see someone so ignorant imply that someone else is ignorant.

it's a typo and i can't edit that comment.

[geerge]

I have tested ZLUDA and recent ZLUDA forks extensively. IMO the gap between rocm and CUDA is still wide open. ZLUDA just shows us another way the gap can be closed. It doesn't close the gap by any means yet. Nobody has stepped up to say they are going to close that gap. I am doubtful about zluda in the long run.

I agree with you about the legal stuff. Nvidia isn't happy about it. They will probably implement something that makes it more difficult for zluda to work.

If zluda gains any sort of traction they could just keep changing the API's etc to be a giant pain in the butt, so zluda would lag behind in software support and perpetually be a sub-par experience. But I don't think they'll bother, zluda is very nice for hobbyists to make use of some software their hardware couldn't previously run, but there's a reason we're talking tiny corp doing some small bespoke thing instead of a big corporation doing big corporation things.

AMD hardware is good at traditional scientific computing aka HPC workloads. ML (and with the xilinx acquisition FPGA/hybrid things) will continue to expand nicely. But the hardware still doesn't come close to competing with nvidia in lots of the standard space that cuda occupies. Even if they did I think GPU is still low on their list of priorities when it comes to fab allocation, if the thing was better they'd make more sure but it'd also raise the price as the card punches higher than before so the benefit is less apparent to the consumer.

[sobrus]

because Nvidia's CUDA license forbids their gaming cards to be used in such a manner.

Can you please provide source for this? I know that consumer cards are built to less strict quality and can get worse support, but never heard about CUDA license limitations. GeForce cards are quite commonly used for non gaming tasks, unless it requires FP64.

Although your point is still valid, Pro series card would be certainly better suited for professional machine.

[mb_q]

Nvidia made a ton of money on it, but nevertheless the idea of GPGPU is stupid, eventually some ASIC will always come, offering a leap in efficiency and speed. IMHO this is what AMD is doing, investing in Xilinx, putting those "tensor cores" into Ryzens... It should be noted that the chip development process takes several years, the current tech was probably more influenced by the crypto bubble and shader-based ray-tracing than the AI bubble.