Radeon Rays 4.0 Released - Adds Vulkan While Dropping OpenCL, No Longer Open-Source

[angrypie]

No, but it probably has something to do with AMD's ray tracing hardware, in their upcoming RDNA2 GPUs. Maybe they felt that exposing their ray tracing code was revealing too much of their IP.

Well, they did patent it, didn't they? In 2017. Anyone interested can look at the patent filing and figure it out. Also I don't think there are many different ways of doing raytracing with current tech, which suggests they might have a cross-licensing agreement with NVIDIA (and possibly Intel). So no, that one doesn't fly.

[Giovanni Fabbro]

I have no idea. We can speculate, but I think the best option is to wait and see. It would be pointless to get worked up over nothing.

What I mean is that I can imagine this change was mandated by Sony (and probably reciprocated by Microsoft), as it was with the RSX in the PS3. This is why there was no GPU acceleration for Linux on the PS3 - it was always a licensing issue. Mind you, that involved NVIDIA, but still...

And lets not forget that Sony uses BSD, not Linux - they don't have to release any source code for it. That says something about how they treat licensing for their "semi"-custom hardware IP.

[Guest]

Except it's not. They just announced v3.0.

OpenCL 1.2 is used quite widely, even if you're not aware of it.

Honestly, even though I love open-source, CUDA dominates,... Academic projects, that I know of, use CUDA and not OpenCL. Maybe, as you say, people are not aware of it, but that unawareness also makes more industries to rather invest into well-adopted CUDA.

[Anarchy]

Buying hardware in the first day is never a good idea. I remember blue screens when some NVIDIA cards were recently released. In AMD for Linux you only have to wait, in the worst of cases, six months. But yes, that's the main advantage of NVIDIA, you can use its cards from the first day, but the wait for the AMD support is not dramatic.

And about DLSS 2.0, I think the most of people don't care about DLSS (see the RTX 20 sales), but with NVIDIA I have a horrible support to do screencasting, I don't have Wayland support and I don't have any desktop environment well supported further than GNOME (although I'm a GNOME user).

Obviously, If you are tied to NVIDIA's proprietary technologies, NVIDIA is better for you, but the most of people only want to see their desktops working correctly and their games running smoothly. Due the Linux gaming is very tied to Vulkan, old AMD GPU's has even more advantages than NVIDIA to play with Linux because the first NVIDIA architecture truly designed to support Vulkan is Pascal.

Just recently, I managed to test 4 5700xt gpus on two machines that previously worked fine with several other gpus from amd and nvidia, two of the cards had very similar problems the worst of which was black-screening at random times and flickering. There's tons of complaints by people on reddit and amd's forums about this. AMD still hasn't fixed their shit almost a year after the initial release of the product.

When I spend 500 euros on what is essentially a "high-end" graphics card, the least I expect is for your product to kill my computer three times a day. I sold the second 5700xt (after the rma) and got myself a 2060s, which has been working flawlessly for several weeks now. If AMD can't even achieve that kind of level of stability there's nothing to talk about any more. People buy nvidia cards not because they like expensive and closed-source, but because nvidia's shit works according to specs (at least most of the time).

[Nille_kungen]

Not dramatic, for people with laptops is dramatic, I will not buy a laptop today and wait six mounths untill it works.

Amd gpu drivers are always bad supported even last ati brand, Amd should support the things better

I haven't had problems like that with AMD laptops but i did need some overrides at boot in the beginning for my ryzen laptop.
It did however work with overrides until bios got fixed.

Just recently, I managed to test 4 5700xt gpus on two machines that previously worked fine with several other gpus from amd and nvidia, two of the cards had very similar problems the worst of which was black-screening at random times and flickering. There's tons of complaints by people on reddit and amd's forums about this. AMD still hasn't fixed their shit almost a year after the initial release of the product.

When I spend 500 euros on what is essentially a "high-end" graphics card, the least I expect is for your product to kill my computer three times a day. I sold the second 5700xt (after the rma) and got myself a 2060s, which has been working flawlessly for several weeks now. If AMD can't even achieve that kind of level of stability there's nothing to talk about any more. People buy nvidia cards not because they like expensive and closed-source, but because nvidia's shit works according to specs (at least most of the time).

What kind of setup do you have?
How many screens?

[TemplarGR]

Except it's not. They just announced v3.0.

OpenCL 1.2 is used quite widely, even if you're not aware of it.

OpenCL 1.2 rebrand edition was just announced, 9 years after the original OpenCL 1.2 release.

[Giovanni Fabbro]

Just recently, I managed to test 4 5700xt gpus on two machines that previously worked fine with several other gpus from amd and nvidia, two of the cards had very similar problems the worst of which was black-screening at random times and flickering. There's tons of complaints by people on reddit and amd's forums about this. AMD still hasn't fixed their shit almost a year after the initial release of the product.

When I spend 500 euros on what is essentially a "high-end" graphics card, the least I expect is for your product to kill my computer three times a day. I sold the second 5700xt (after the rma) and got myself a 2060s, which has been working flawlessly for several weeks now. If AMD can't even achieve that kind of level of stability there's nothing to talk about any more. People buy nvidia cards not because they like expensive and closed-source, but because nvidia's shit works according to specs (at least most of the time).

I would be interested to know who built each card. I've worked in computers since the IBM PC first shipped and I've seen lots of video cards die, not necessarily because of the GPU chip, but because of faulty circuitry on the board (circuit pathways, resistors, etc.), defective video RAM, overheating (sometimes due to shoddy cooling designs, sometimes because of dust...NVIDIA's heat issues and the crap solder caused the RROD and YLOD faults on both the PS3 and XBOX360 as well as a whole string of laptop problems around the time of late Windows XP up till early Windows 7 which is why they stopped making motherboard chipsets), faulty RAMDAC's (on old cards).

[andyprough]

Do you even know what Radeon Rays is? It's used by like 0.01% of AMD GPU customers, if that, and it's a value-add that's not required to use any of the GPU's underlying hardware. It's a far cry from them locking down their driver stack!

Wow, it's like some of you guys are going around and looking for the slightest excuse to express outrage.

I was wondering what would be the first post that actually talked about Radeon Rays itself, not just a bunch of hair-tearing by people who don't even know what it is. Post #17 - not bad for a Phoronix thread.

[tildearrow]

GPUClosed

[Giovanni Fabbro]

I was wondering what would be the first post that actually talked about Radeon Rays itself, not just a bunch of hair-tearing by people who don't even know what it is. Post #17 - not bad for a Phoronix thread.

Honestly, I don't really care. I'm just here for the argument.