5 ms·
When GPUs became vastly more powerful over the last ~10 years, it made big neural nets practical. I wonder what an equivalent jump in CPU power will unlock.
by piinbinary 7y ago
When GPUs became vastly more powerful over the last ~10 years, it made big neural nets practical. I wonder what an equivalent jump in CPU power will unlock.
- Scarbutt 7y agoMore electron apps.
- ralusek 7y agoOnly if people make more usage of worker threads.
- dkersten 7y agoNot faster electron apps, just more. One per core.
- adzm 7y agoThat's more of a memory bottleneck though.
- mantap 7y agoYeah indeed chromium is a more efficient 2D rendering engine than what's available natively and performance of JS is comparable to Java. Electron apps being slow is mostly a developer laziness problem, not a technology problem. Making a fast electron app is actually easier than making a fast C++ app IMO, when you take into account the libraries and tooling available.
- mrec 7y agoI think the difficulty is in making a lightweight Electron app, not a fast one. Games run fast too, but they can get away with hogging resources because they're typically the only app you're using at the time.
- aoeusnth1 7y agoThis is a good article on the technical reasons Finda uses electron to make a portable, yet fast & lightweight desktop search app: https://keminglabs.com/blog/building-a-fast-electron-app-with-rust/ https://keminglabs.com/blog/building-a-fast-electron-app-wit...
- megous 7y agoIt's not more efficient than directly using skia or cairo. Not on ARMv7 anyway.
- Filligree 7y agoSadly, I'm afraid you're right. I love my 1920X for making a modern desktop snappy. I hate Discord, chrome et. al for needing one.
- udhbeeui 7y agoMax spec KDE with ALL batteries included on a 12 year old laptop computer with a new cheap Crucial SSD even with the old UDMA-6 is ultra low latency and very snappy. KDE is an excellent Mac replacement for power users.
- pizza234 7y agoAt this stage, generic computing is unlikely to get any more significant boost (under the assumption that generic computing is something not necessarily [highly] parallelizable); instead, we'll likely see specialized processors. Therefore, ultimately, the market/industry will decide what to accelerate - and the technology to support the acceleration.
- lonelappde 7y agoThe reason GPUs exist is because CPUs don't scale.