NEOX V Announced By Think Silicon As First RISC-V 3D GPU

[microcode]

Refused to communicate. Check comments to article about switching this project to POWER.

I don't know, but every time one of these libre GPU efforts comes around, the people involved expect to be congratulated before accomplishing anything. They don't need any permission to make a PoC on a FPGA, so when they don't even bother with that, it doesn't lend much credibility. Nobody cares that some dood threwa tantrum decided to "switch" his vapourware from one ISA to another because he wasn't being patted on the back for having an idea that literally everyone already had.

[coder]

I think it's good that GLOVE is open source (LGPL v3), and backed by a commercial entity using it for something real. At least most of us should be able to get behind that.

I don't believe their GPU is a serious threat to GPUs with purpose-built ISAs and architectures. I maintain that you can't make a truly competitive GPU with a general-purpose ISA. However, for those who care about having a standard ISA on your GPU so you can control the software stack on it, this could still be a win.

It seems to me that this is a positive development, even if nobody here is completely happy with it.

[coder]

^^^ OMG, why in the world is that post caught in the spam filter?? ^^^

(yes, obviously you'll have to wait until it gets approved in order to weigh in.)

[coder]

I guess this thing, unlike the Libre RISC-V, actually has a geometry processor... right?

Based on what? ...or are you just speculating?

Even in their aborted Larrabee GPU, Intel realized that you needed hardware texture samplers, for even a hope at competitive performance (which it still didn't achieve). We don't know that Think's GPU lacks any fixed-function units, but their press release certainly didn't state or suggest otherwise.

BTW, I think it's cool that Think Silicon is based in Greece. I imagine they must have a lot of under-employed talent available to work on enterprises like this. I wish them well.

[starshipeleven]

implementation also can implement forked standard, like microsoft with eee

Yes but this is hardware. It is a bit hard to fork a standard if you must give full register documentation for developers. If you don't give full register documentation for the new stuff then you are making your own little black box, not a new standard. See how OpenGL and Vulkan extensions work.

[curaga]

I don't know, but every time one of these libre GPU efforts comes around, the people involved expect to be congratulated before accomplishing anything. They don't need any permission to make a PoC on a FPGA, so when they don't even bother with that, it doesn't lend much credibility. Nobody cares that some dood threwa tantrum decided to "switch" his vapourware from one ISA to another because he wasn't being patted on the back for having an idea that literally everyone already had.

There's nearly half a million of EU money going behind the project soon. Plus funds from Purism and others. That, and having skilled people with expertise in various areas, gives credibility to the project.

[Jabberwocky]

this or we wave a CLA-WAR,PATENT-WAR,Anti-FLOSS-war and Microsoft talmud style divide your enemy into CSD vs SSD(clind side decoration/server side decoration) and Gnome vs KDE(CLA WAR give away your code to the project so the evil project owner can make your code into a closed source zombie.)

in my point of view the RISC-V is a attack on REAL FLOSS CPU designs

and the "NEOX V Announced By Think Silicon As First RISC-V 3D GPU" is maybe an attack on any real FLOSS GPU design.

in reality we need real FLOSS cpu and gpu design with not only open-source software but with also open-source hardware and even open-source manufacturing tools. what we get is a FOSS(without the letter L for libre) gpu with dirty patents and dirty CLA WAR and so one and so one.

Sad but true.

Due to lack of transparency this is my opinion at the moment: Si-Five claim to be open and free (even their model names use same terms). Everyone (Phoronix, Level 1 Techs, Linus Tech Tips, etc.) just took their word for it and nobody actually checked what type of code they released or what type of development models they decided to use. Many of them got the devices from Si-Five and was too focused on what the device can do or will be able to do... rather than to focus on what the device stands for. Si-Five (and RISC-V to an extent) are using community driven projects and community driven news for their own gain screwing FLOSS along the way. They can do what they want, but little transparency goes a long way especially if you claim to be open.

It's very simple. If you have conflict of interests announce it. If you mix open and proprietary announce it. MAKE A CLEAR DISTINCTION between what is open and what isn't.

Kernels and compilers make use of some optimizations between x86 architectures but for the most part it has been binary compatible. ARM is binary incompatible in so many ways and has shown uneducated people what the educated engineers already knew. The industry is going to need more symbiotic relationship between RISC-V and FLOSS to really challenge x86 in the long run. If they don't sort out these problems quickly they will have a massive train wreck! (Like Qaridarium described above)

[ahcox]

> ... OpenGL ES / Vulkan support is done using the GLOVE middleware. So this isn't a complete graphics processor but uses a similar route to what was proposed with Libre RISC-V but instead using Think Silicon's own GLOVE software code.

Michael, They get GLES from GLOVE but there is still a conventional Vulkan driver for their GPU under that:

GLOVE: GLES Over Vulkan

https://web.archive.org/web/20191210190406/https://think-silicon.com/tsi-products/software/glove/