Hacker News Reader

Comment by mathattack | Hacker News Reader

An open question... How long will selling computers based on increasing processing power make sense for anything but a very-niche audience?

My thinking is that at some point, it should be transparent where the computations take place. So rather than sell us 12 cores, it would be more efficient for Apple to sell us "5 years of access to 12 cores", or whatever the expected lifetime is, and we wouldn't know where the processing power comes from.

Right now accessing remote computing power takes a little bit of thought (AWS) but if anyone can make it easy enough for Grandma, it's Apple.

They may always be cases of having this power local (a place with bad internet access?) but that seems like a niche.

View on HN

For most tasks that you really want to throw 12 cores at (e.g. video rendering), the latency of sending everything over the Internet will be extremely noticeable. If you don't care about that sort of thing, you probably don't care about having access to 12 cores.

I was thinking that 12 cores would be to cut a compile time from hours to minutes, or shorten a scientific calculation, no? Or pricing a complicated bond.

Astute observation. We're already there.

Most people I know, including myself, are quite happy on a 6-7 year old unit.

Bar developers, graphic artists (2d,3d), video, mega-data there's sod all use for a 12 core Xeon. Don't start spewing gaming either as the Xbox one doesn't have a 12 core Xeon in it!

Don't forget about engineers and medical research. Moving data off-site for processing isn't always an option. Even when it is, there is money to be saved by running sub-sampled data locally and prototyping models before scaling up.

I'm getting a 12 core asap, but eager for more compute-focused options.

Why wouldn't it be theoretically possible to move the data off-site? Too big? Again I'm assuming there is some seamless way for the operating system to treat off-machine cores and in-machine cores.

Mostly just policy around certain types of data. That people who complain about such policies are also willing to e-mail around zip files is beside the point.

Edit: it seems like I read about certain algorithms that could perform calculations without direct knowledge of the underlying data, but I can't find a link. It only worked for a specific class of problems in any case.

http://en.wikipedia.org/wiki/Homomorphic_encryption

Thanks!

ok - so this is just for the corporate environment that there would be these issues? But couldn't they just have a similar getup sending all their processing to their own data center in the background?

My intuition has us going back to the dumb terminal mode, albeit with a very fancy screen.

Reply on news.ycombinator.com