The GPLv3 GPU Is Now Available

[Luke]

Circuit desing? A homemade GPU wpould be one HELL of an achievement

Have you actually looked at what you'd be getting for your money? Basically, it's an exercise in FPGA programming... educational, but with little real-world value. As far as I can tell from the article and linked blog, there's no hardware (simulator only for now), no drivers, and the whole thing is equivalent to first-generation GPU designs from the late 90's.

It's great if you're into FPGAs and circuit design... but it's not an open-source competitor to modern hardware, and is never likely to be.

If someone actually managed to fab a silicon chip on ANY fab size to a GPU that would simply light up the monitor and display normal desktop stuff, it would be one hell of an achievement. Don't think anyone would care whether or not it accelerated anything, anymore than they did when some hobbyist made an 8 bit CPU and system around it that actually functioned from discrete transistors.

Even if fabbing the chip went out to an engineering shop that does such things, once the chip was on a homemade board, plugged into a board and running it would be a huge deal for them. Following the whole process of GPU development back from the beginning might be exactly the approach to follow

[RealNC]

Does it run Crysis?

[DanL]

Does it run Crysis?

Only the slideshow version.

[computerquip]

I think an open-source hardware is interesting if you're wanting to get into driver development, especially if it's cheap. I'd buy one if the money is right.

[Delgarde]

If someone actually managed to fab a silicon chip on ANY fab size to a GPU that would simply light up the monitor and display normal desktop stuff, it would be one hell of an achievement. Don't think anyone would care whether or not it accelerated anything, anymore than they did when some hobbyist made an 8 bit CPU and system around it that actually functioned from discrete transistors.

Oh, sure... it's a huge deal for anyone interested in circuit design. But it's not - and probably never will be - a workable substitute for even the cheapest modern GPU, so anyone enthusiastically wanting one should be realistic about what they'd be getting.

[Kemosabe]

If only 3D printers could do it ....

[brent]

I think an open-source hardware is interesting if you're wanting to get into driver development, especially if it's cheap. I'd buy one if the money is right.

Nah. Hardware (if anyone makes some) isn't going to be cheap due to low volume. If you want to start with driver development, look at the popular choices like AMD and Intel.

[ssvb]

We can only welcome this news. This is all about open source hardware, which is only starting to take off the ground. Something like the OpenRISC CPU core combined with this GPU could be one of the next stepping stones. So far all this stuff can be perhaps only debugged/prototyped on FPGA development boards such as Arrow SoCKIT. The power consumption, performance and board cost are naturally not going to be competitive with the current mainstream mass produced consumer ARM and x86 devices and their integrated GPUs.

Later on, a proper ASIC might be also a possibility. After all, bitcoin miners got their ASIC hardware already. This is just a proof that not only multi-billion mega-corporations can do this

[chrisb]

Oh, sure... it's a huge deal for anyone interested in circuit design. But it's not - and probably never will be - a workable substitute for even the cheapest modern GPU, so anyone enthusiastically wanting one should be realistic about what they'd be getting.

People said we would never have functional, commercially used open source CPUs, but then projects like Leon appeared and proved them wrong. It is not impossible, just hard to be motivated when we do not have the tools or ability to fabricate and test the designs at home. IMHO designing a CPU is no more complicated than the Linux kernel - fabricating a CPU is another matter.

Leon: http://en.wikipedia.org/wiki/LEON

[smitty3268]

it's DOA.

FTFY.