Question Soon-to-be opensourced FuryGPU running Quake will advance humanity!

[igor_kavinski]

New open source GPU is free to all — FuryGPU runs Quake at 60fps, supports modern Windows software

FuryGPU is currently capable of running Quake at 720p and 60fps.

www.tomshardware.com

The future is all set for a 10-year old kid years from now who learns the workings of a GPU from this gentleman's hard work and develops his own, more powerful design that will run Doom 3 (hyper-optimism alert!).

 

[zir_blazer]

While the work is amazing, I have a tendency to believe than trying to do CPU or GPU on FPGA is PURE MASOCHISM because they're the most complicated and less useful things you can do. You simply can't get much performance out of a FPGA that most people can actually afford. Yet, most hobbyst seems to want to do these...

There were multiple previous tries for open source Video Cards, the one I got on google first was Project VGA from more than a decade ago, but there should be at least another one that Phoronix covered, not sure if two more. And we're talking about some very basic framebuffer with video output, with no acceleration whatsover, albeit this Fury seems to be already on a more advanced state than that. On the side of CPUs, there are even custom ISAs - Agner Fog, which did some major code optimization manuals for a multitude of x86 generations and got at least once into Hardware sites news has its own CPU ISA as Softcore for FPGA and supporting Software (Compiler, Emulator. Not sure if there is a Linux port), ForwardCom.

So what I believe than should be done? Accumulate know-how and IP cores for peripherals instead. USB 3 is 15 years old or so, still widely used. Same with SATA and PCIe 3.0. There is a french open source developer, EnjoyDigital, which has that kind of cores open source. Whatever gets done with those could be integrated into other SoC projects, so they're not wasted effort like CPU and GPU tends to be.
So far, I have been thinking for years that a FPGA based PCIe 1x card with USB and SATA (And maybe even Ethernet) could be useful for virtualization/PCI Passthrough if it exposes Ports as individual PCI Devices and has a flawless FLR (Function Level Reset) implementation. Sadly, FPGAs that have transceivers capable of PCIe 3.0/USB 3.0/SATA 3 are not cheap enough, but it is theorically possible and you have most of the code and know-how already available to do it in a short timeframe if funding grew on trees...

 

[igor_kavinski]

Someone went even further! https://www.tomshardware.com/pc-com...hows-it-off-with-homemade-gpu-and-thumbsticks

 

[Hitman928]

Someone went even further! https://www.tomshardware.com/pc-com...hows-it-off-with-homemade-gpu-and-thumbsticks

Umm, maybe on the software side? He used off the shelf hardware and created a 3d engine that could run (basic wire frame at least) on it. Cool project, but the FuryGPU was way more impressive, IMO. This one did something lots of people have done (software 3d engine) but instead of running it on a full blown CPU, got it working on a microcontroller. The FuryGPU was an actual GPU created via an FPGA, with drivers and everything, that could run a commercially available game (an old one, but still could run it smoothly).

 

[DavidC1]

While the work is amazing, I have a tendency to believe than trying to do CPU or GPU on FPGA is PURE MASOCHISM because they're the most complicated and less useful things you can do. You simply can't get much performance out of a FPGA that most people can actually afford. Yet, most hobbyst seems to want to do these...

I agree very much here.

CPUs and GPUs can be a goldmine but only for the 1-2 years that it's competitive then it quickly becomes worth less than the actual value of the silicon used(which is almost nothing).

Old computers are a dictionary definition of "sum of it's parts worth less than individual parts". The capacitors, heatsinks, fans, and metal that are extracted from an old computer are worth more than the functional computer itself.

These projects are just for the sake of advancing knowledge and an advanced hobbyist project.

Although, I see a market for a complete open source CPU with the design fabbed on older silicon processes for the maximum privacy and security even if it's much slower and be able to run Linux. x86 would be better but you aren't going to have Intel ever agree to it.