Computing 'paradigm shift' could see phones and laptops run twice as fast — without replacing a single component

[admin]

By letting different processing units — like GPUs, NPUs and hardware accelerators — work in parallel, rather than in sequence, systems can be up to twice and consume 50% less energy.

Computing 'paradigm shift' could see phones and laptops run twice as fast — without replacing a single component : Read more

 

[Steve.Nordquist]

Nice brief title. I'd have been like 'Leave your P, E, Arm22, Ryzen, CUDA and Annapurna schedulers (and process leak greying) and follow the TPU NPU way; who shall not receive manifold more.' But there's also room for a Team Rocket chant, I think.

Anyone done a line reading of TFA or the schedulers on StackExchange or such?

 

[JDSostrom]

Multitasking has been the hallmark of mainframe computing since the early 1970s and late 60s. The IBM mainframes were at the forefront of multiprocessing. It is nice to see PCs and Cell phones finally catching up.

 

[Jan Steinman]

It wasn't clear how SHMT was implemented.

Was this an OS-level change? Or did it require re-programming of individual applications in order to do multi-threading in some new way?

If the latter, I think it's a non-starter. Not many are going to be willing to pay for such performance increases. Most people are content with "good enough."

Now if Apple or Microsoft or the open-source Linux community can achieve these gains from existing software, that's interesting indeed!

 

[adamz]

The Commodore Amiga did this as a PC in the 1980s, with great success. Why not apply that logic to modern devices?