@JoeJ can relate

@calin : you wrote “the flattening of the curve is exponential”. That's not possible because exponential means an increase in the rise of a curve. But an asymptotic flattening may well be.

Yesterday I compared processors from 2010 with ones of today (e.g. i5). Their performance indizes rose by >10, random comparison site. That's well in line with graphics cards. So there is that.

out

Green_Baron said:
Yesterday I compared processors from 2010 with ones of today (e.g. i5). Their performance indizes rose by >10. So there is that.

Completely invalid statement. An old Xeon beast will beat low end modern CPUs. Completely invalid statement.

Invalid without stating the model, the year, the parameters etc. Needs quotations from somebody graduated from MIT or at least a user of stackoverflow forums with reputation of 10K+.

https://www.tomshardware.com/reviews/atom-d510-pentium-4-nettop,2649-4.html

2 cores 1.6Ghz vs 1 core 3.2GHz…

old vs new…

Green_Baron said:
Yesterday I compared processors from 2010 with ones of today (e.g. i5). Their performance indizes rose by >10, random comparison site. That's well in line with graphics cards. So there is that. out

What's the test cases?

My statement from yesterday was about i7-930 vs. Ryzen-2700. I compare them with my own applications, where i do not use AVX which old i7 does not have. It's a fair comparison with multi threaded workloads that saturate all threads to 100%.

The speed up roughly matches expections from core count and frequency increase. So a bit more than 2. If Moores Law would be true (meaning doubling performance after 2 years, not transistor density, if that's ok), the speedup should be 30 after 10 years?

Game performance confirms those results, also the few other apps i'm using. (I'm not disappointed - always wanted 8 cores and now i can afford this thanks to AMDs offerings.)

Synthetic benchmarks might tell another story, but i ended up ignoring them. To me they never relate to real world performance - neither for CPUs nor for GPU compute.

What can show much bigger increases than 2 in my case is AVX utilization and the larger caches, but that's special cases that should not affect a practical average too much.

Now where we have competition from AMD, i expect better progress on CPUs again. But at some point we can no longer make chips smaller due to physical limits and we're close to that soon.

@Calin For what it's worth, my dad is 100% convinced that space aliens are real.