Hacker News Reader

Comment by glitchc | Hacker News Reader

An ARM64 is "a very powerful computer"? The whole promise with ARM is better thermals and long battery life, not screaming performance. With the thermals/cpu not working, we don't even get that.

View on HN

M4 crushing benchmarks and AWS running Graviton 2 would disagree on the idea of "whole premise".

I always knew hardware discussions on HN were below the normal standards. But sometimes I am still surprised at how low it can go.

The finer grained performance per watt vs silicon cost set in the context of use case is just lost in HN hardware conversations like this that ignore total cost of ownership, vendor politics and such.

Everyone is buying the tool that does the job, or building that tool if they want to make that large investment...

There's a difference in cost per flop at home and in a data center.

I'm updating my wiring and air conditioning for a 7x5090 workstation because having that power for experiments under the desk is worth the cost (and fire hazard).

If I had to build 10,000 of those I'd be banned by NVidia from ever buying their hardware again.

What are you doing with the workstation that it’s worth the cost?

High dimensional computations for time series data in finance.

Private algotrading? Nice. May the odds be always at your favor!

If I'd have those GPUs, I'd use them to run some AI models locally. But since I am not rich and I don't make money from that, I won't buy those GPUs.

And it usually ends with some weird implied conclusion, like “AMD-Intel duopoly bad, Apple monopoly good”.

I wonder what I would discover if I do some archeology and look at HN discussion when Apple used PowerPC CPUs.

Considering HN started in 2007 I don't think your expedition will yield a ton from the heyday of Mac PowerPC. Apple shipped the last PowerPC hardware in 2006.

To me it seems some people developed some highly political or even religious views on tech. Truth doesn't matter as much as it used to. It's more of "I like the idea behind X, I think X is cool so X should be the best thing ever".

If you consider Geekbench a good benchmark...

It is a start. A wrong one, but still a start. /s

Reply on news.ycombinator.com