SoylentNews is people
SoylentNews
from the MIAOW-sounds-like-one-cat-crying-out dept.
While open-source hardware is already available for CPUs, researchers from the Vertical Research Group at the University of Wisconsin-Madison have announced at the Hot Chips Event in Cupertino, Calif., that they have created the first open source general-purpose graphics processor (GPGPU).
Called MIAOW, which stands for Many-core Integrated Accelerator Of the Waterdeep, the processor is a resistor-transistor logic implementation of AMD's open source Southern Islands instruction set architecture. The researchers published a white paper on the device.
The creation of MIAOW is the latest in a series of steps meant to keep processor development in step with Moore's Law, explains computer scientist Karu Sankaralingham, who leads the Wisconsin research group.
(Score: 0) by Anonymous Coward on Friday September 04 2015, @07:31PM
Because games are all about graphics, and you need a more powerful GPU than CPU if you want to brag about your elite status as an elite gamer.
(Score: -1, Troll) by Anonymous Coward on Friday September 04 2015, @07:39PM
I'm going to glory hole in about an hour or so. Would you like to join me and let me fuck your ass?
(Score: -1, Troll) by Anonymous Coward on Friday September 04 2015, @07:46PM
No thanks. Your cock is too short to reach through the wall.
(Score: -1, Troll) by Anonymous Coward on Friday September 04 2015, @07:55PM
Your dad wasn't complaining last night.
(Score: -1, Flamebait) by Anonymous Coward on Friday September 04 2015, @10:08PM
He is a very considerate man.
(Score: 5, Interesting) by VortexCortex on Friday September 04 2015, @08:39PM
Because games are all about graphics, and you need a more powerful GPU than CPU if you want to brag about your elite status as an elite gamer.
Well, first off, most Linux gamers are happy when they have graphics that WORK. I'll spend 10x$ on a GPU if it has open source drivers that work. Also, as a gamedev who cares less about graphics than gameplay I'm glad for vastly more parallel compute power.
Protip: Tired of dumb enemy AI? That's because AAA gamedevs aren't running more complex and embarrassingly parallel AI on GPUs (AI devs get budgeted 1%-2% max processing/storage/RAM). Tired of the same old gameplay? Hardware advances will bring net types of games.
As more GPU becomes available the graphical fidelity hits a point where 10% less it's not much visibly different, and so more parallel processing can be dedicated to physics (gameplay) and AI (gameplay).
IMO, I like AMD's APU design. Shared memory architecture is where we need to go, because my physics code on the GPU can't talk to the NIC or input devices without going through the CPU/main memory. Currently I have to keep two copies of geometry in RAM - One CPU side doing collision detection and some physics, and one GPU side to render. The CPU side is typically lower res, but the GPU also contains the lower res data for reducing model complexity with increased view distance (LoD). Also, the GPU driver might have to "swap out" geometry or memory to the CPU side to render some complex frame. Devs try to avoid this at all costs, but "streaming assets" really is just "swapping out" to the CPU side. With shared memory architecture not only does parallel computing / heterogeneous computing become easier but you get a LOT more RAM to play with even without increasing the available memory. And, since the "CPU" side code of networking and gameplay relevant physics (which must be available for network sync, as opposed to visual-only physics, like particle effects) can then communicate directly with "GPU" memory then MORE visible things can affect your gameplay. The reason particles go through walls and don't do gameplay changing things (like any visible ember being able to catch your clothes on fire) is because it's so painful and costly to read back data from the GPU. APU and shared memory architectures -- Advances in "GPUs" increases possibilities for GAMEPLAY, even leading to new types of gameplay not possible on current platforms.
Back in the days before hardware acceleration we had more freedom to interface physics, networking and rendering via software rasterization. New and more advanced GPUs are bringing us back to that place where graphics and gameplay can co-mingle again. It's not just about bragging about my "elite status" as an "elite gamer", I want to implement some game mechanics that are not possible (in realtime) on today's hardware. In other words: Blame crummy console compatibility constraints for AAA's not doing more with GPUs, not gamers. Gamers just throw the best hardware they can at what game developers create. Don't dismiss new hardware as "muh graphics" simply because risk averse AAA studios don't use them to do more. The faster everyone gets high memory bandwidth parallel computing, the more feasible it is to recoup the cost of making new types of games.
I'm not sure what to think of MIAOW. I want open hardware, but I also want established vendors to have open sourced hardware. If AMD can benefit from MIAOW's developments then maybe this will be good for everyone. They're a long way off, but if something like MIAOW becomes the defacto GPU for open source platforms it puts less pressure on proprietary vendors to open their sources. Or, to put it another way: Proprietary hardware vendors have a window of time to capitalize on open source platforms. As open hardware becomes more competitive this window slowly closes and the proprietary shops will find themselves competing with "open". The forward thinking response from proprietary vendors would be to open more hardware designs, even manufacture and sell the MIAOW design. Focus more on the manufacturing process and less on protecting their bullshit driver code (of course factoring in the MPAA and Microsoft tends to complicate such decisions).