The microblog: 2022.06.15 23:28:18

2022.06.15 23:28:18 (1537185254707896320) from Daniel J. Bernstein, replying to "Ruben Kelevra (@RubenKelevra)" (1537072571442774020):

Are you waiting for your computer during these unspecified workloads? If so, shouldn't you be asking the software providers for multithreading and vectorization to make the code an order of magnitude faster, even if this makes Turbo Boost drop to 1%, as in your x265 example?

2022.06.15 23:33:54 (1537186663356506112) from Daniel J. Bernstein:

Given how many major applications that users care about are already multithreaded and vectorized, it's wrong to cherry-pick unoptimized single-threaded applications as pictures of total system performance. This error will increase as more and more applications add optimizations.


2022.06.15 15:43:42 (1537068333090078722) from "Ruben Kelevra (@RubenKelevra)":

Okay, let's take a highly optimized multi-threading software: x265 The run with turbo is still faster – as expected the difference isn't that much, as long term workloads are not expected to be better with turbo. :)

2022.06.15 15:47:46 (1537069354285678592) from "Ruben Kelevra (@RubenKelevra)", replying to "Ruben Kelevra (@RubenKelevra)" (1537068333090078722):

So back to the beginning: Short workloads will feel 50% slower, as most application loads/webpages etc are just CPU bound and not IO bound these days. Long term, very highly optimized multithreading applications may run just slightly slower, like 1% slower.

2022.06.15 15:48:35 (1537069561157238784) from "Ruben Kelevra (@RubenKelevra)", replying to "Ruben Kelevra (@RubenKelevra)" (1537069354285678592):

But what the user will feel is a 50% drop in performance, as long term workloads do not really matter for most people. If it takes 2 hours or 2 hours 30 for a video encoding isn't really bothering many people.

2022.06.15 16:00:33 (1537072571442774020) from "Ruben Kelevra (@RubenKelevra)", replying to "Ruben Kelevra (@RubenKelevra)" (1537069561157238784):

Here as an example how my typical workload looks like, my graphs from yesterday: It clearly shows that the CPU can maintain a lot higher frequency in the workloads than 2.4 GHz base and typical desktop workloads doesn't use all cores.