WebGL Powered Challenge

[gartnera]

I threw together a little demo where I compute a challenge on the GPU with webgl: https://github.com/gartnera/sha256-webgl. Demo at https://sha256-webgl.agartner.com/

This should make the challenge computation faster since it can be parallelized on the GPU. I have not fully validated the performance of this process.

Another interesting thing I discovered was that chrome cannot solve this challenge when running without a GPU:

[JasonLovesDoggo]

I've found a Difficulty greater than 4 always returns "undefined" and 4 just completes too quickly.

[gartnera]

Yeah nonce -1 indicates that the search timed out. I've tweaked the UI a bit so that undefined should no longer display.

I've tuned some parameters so that up to 6 should now work (at least on desktop).

Thanks for testing.

[ejrydhfs]

What if it was used to mine crypto to force scrapers to pay for their costs? And making webgl or webGPU a requirement to increase hardware costs since server CPUs don't have integrated graphics and usually scraping servers don't have dedicated GPUs either? However all humans will have a GPU that is integrated or dedicated to see webpages and will thus most likely support webGPU

[gartnera]

What if it was used to mine crypto to force scrapers to pay for their costs?

1. there would be a lot of overhead coordinating what work to do
2. mobile devices are not powerful enough to do a serious amount of work

server CPUs don't have integrated graphics and usually scraping servers don't have dedicated GPUs either

Yeah that was the most interesting thing that came out of this tbh.

webGPU

webGPU is not supported on all browsers yet. But it would be significantly easier to do this with webGPU.

[Xe]

SHA-256 is a hack and will not be around forever.

[gartnera]

Sure I think switching to another hash algorithm is reasonable.

I do think the most interesting part is that it prevents computers without GPUs (servers, scrapers, etc) from solving the challenge by default. Obviously a determined actor could do the calculations in a separate process but this would at least block dumb scrapers running the default chromedriver.