Jonathan Blow

Are you really speaking to everyone here? Or is there a "you" I'm missing? People buy things they know they won't use all the time. Almost no one is taking their monster SUVs with 4 friends and a small kennel full of dogs into mountains. They buy in case. (Or to get more cores.)

You talk about not understanding how fast chips are... I think we mostly very much understand how slow our software is. That's the issue, right? We like fast vehicles. We know we usually stay around speed limits. How do we, consumers, encourage faster software?

There's a fundamental problem w/ the current software "stacks" that are built on opinionated concepts like garbage collection and "objects" that are laid out in memory patterns that are orthogonal to cache locality. Consumers have no control over this; it's the zeitgeist of SDKs

I guess my thing is maybe an effect of tweet length on discourse; the use of "you". I think when most give it thought, they at least recognize "a lot" if not insanely more. Maybe everyone thinks it's the responsibility of the person above them on the ladder. Ignorance vs apathy.

I think you're spot on with your ladder analogy. The problem Jon refers too is hard to grasp; it's the difference between nanoseconds and microseconds. I know this b/c I've observed it. We are probably <1% of all software developers. But 3 orders of magnitude is huge, right?

Maybe. I still think it's apathy over ignorance. (I can't believe I'm that smart.) I think it would help to push the metaphor the other way. "Imagine your computer is capable of one operation a second. Your software asks for X operations every X days. Your processor is bored."

I think you're correct from a consumer standpoint, absolutely. But there's a programmer standpoint too; developers in Java, Scala, Ruby, Python, Node.js, C# are executing under many layers of stuff happening. This is fine for buying something on a credit card. But if you draw...

Slow/bloated software is good news because it applies more pressure on Intel/AMD to make faster/bigger chips which we'll take advantage of when their speed tops out - code efficiency will improve as speed/core count increases dry up

Also, obesity epidemic is good because it encourages fast food manufacturers to create higher caloric density meals which we will eventually use to end world hunger.

I'm surprised you use Windows based on this. I know you've said Linux has issues, but I've found that X server + i3wm to be a significantly lighter and faster environment than Windows.

Yeah because no programs run on Linux

Not got a new CPU in almost seven years because all the efforts seem to be going towards GPU accel on everything. Even raytracing (which obviously would benefit from mulithread) is being pushed now from the GPU side.

1. Are chip/motherboard architectures themselves held back by legacy compatibility? 2. Presumably BIOS, OS, and app each introduces further speed bumps between user and max speed? 3. Any current combo of hard/software which can help us understand how much performance is wasted?

is it actually how fast they operate on 'everyday tasks' that people generally care about? it's certainly not where the optimization efforts go and it's not where the speedups of new hardware are seen you are not going to get for free faster DFTs, fluid sims, FEA, ML, etc.

Aren't most of those likely to be solved on the GPU rather than the CPU these days?

yea, to some extent I expect the CPU still needs to be fast to not bottleneck the GPU in some way though, in any case it's not going to be 'running 1000 email clients at once' or anything like that driving performance requirements, it's going to be doing a lot of something simple

The real use case for a new CPU: Spambots! :)