Folding home as part of a games?

[TheElpis]

HI was there every an attempt to be include folding home as part of some computer game? Given the huge amount of time people tend to play computer games and how strong their GPU tend to be and how often they tend update their system, it could provide huge amount of computing power. Lets say that Dota 2 and LOL players which games aren't that GPU intensive could contribute a bit. Seems to me that we could have a ton of new computing power. Obviously there are several issues to be solved, but seems like worth a try.

[bruce]

Folding at home was ported to the Play Station, but it wasn't part of a game -- you could run it when you were NOT playing a game. It contributed a certain amount of science, but it was a static platform, and other hardware outpaced it. Then, too, it produced a lot of heat and it's electricity consumption was high.

Many years before, folding was ported to the google browser. (It ran while you were looking at the web).

Most popular games are designed to take advantage of serious amounts of resources, in completion with FAH. There still are a few games around that don't use much resources, but when was the last time you played something like Battleship or Tetris for more than a few minutes at a time?

[foldy]

It would help if folding and gaming could work in parallel.
So if a game does not use all GPU resources then FAH could use the remaining GPU resources.
But it is not supported, so FAH or the game or the driver will crash.

[bruce]

It will not necessarily crash. It depends on what you're trying to do and what resources are available.

It's possible to run simple games concurrently with FAH -- and for them, integrating them into a single package makes no sense.

In general, however, GPUs are not as well designed for multitasking as CPUs are. CPUs have virtual memory so that if you can run one or more programs which allocate more RAM than you actually have. When that happens, the CPU moves the less active pages to disk and everything proceeds normally, if somewhat slower than when everything can fit it real RAM. If you intentionally limit the CPU's virtual memory to zero, you can still run one or more small programs which happen to fit within real RAM, but if the total exceeds your real RAM, something will crash.

GPUs do not have virtual memory, and if the total of all active GPU programs exceeds (real) VRAM, something will crash. Each program that uses the GPU is designed to it will run on your GPU, but the programmer has probably assumed the his program has full access to all of the (real) resources of your GPU.

It's very likely that there are resources other that VRAM which may be over-allocated when more than one program attempts to use the GPU concurrently. In any case, minor programs, such as moving a mouse and probably sneak in successfully between slices of FAH processing. Major programs probably cannot.

[JimboPalmer]

Obviously there are several issues to be solved, but seems like worth a try.

Starting in the 1960's preemptive multitasking became an issue in Computing. Allowing multiple processes to each see the computer as their own involved adding hardware and writing Operating Systems to allow applications to share resources as needed.
This wikipedia article lists some requirements for preemptive multitasking: https://en.wikipedia.org/wiki/Computer_multitasking all of which Windows, OSX, and Linux do just fine for the CPU.

No-one has written a multitasking Operating System for a graphics card yet. It would almost have to be the card vendor, or at the least, have the support of the card vendor. If one existed, gamers would resist using it, as they 'need' graphics speed,not OS overhead.

With the obvious exception of those using graphics cards for computation, like you and me, no market for a Nvidia Linux vs AMD Linux debate has yet materialized i am afraid. Until it can be implemented without increasing latency in gaming, I suspect it will not gain favor.

[bruce]

Very well said, JimboPalmer.

Before preemptive multitasking was developed for CPUs, operating systems depended on cooperative multitasking (which required that each long-running program politely share resources with anybody else that might want them) and before that, everything had to fit in RAM and individual programs that were too big for RAM had to manage their own overlay structure. These OS developments allowed the CPU to develop into a general purpose resource that could do many, many things.

Historically, graphics was originally processed by the CPU, but demand increased to the point that an optional specialized (called a GPU) was developed. It needed the ability to do enhanced parallel processing but required specialized programming.

I agree that it's unlikely that preemptive multitasking will never be developed for the GPU. The CPU does that very nicely. At one time, a GPU manufacturer suggested that they could restrict games and/or FAH to, say 90% of the resources, leaving the other 10% so the OS can update the screen. I doubt that will sell more GPUs.