This just seems like explaining the difference between GPU-bound vs CPU-bound scenarios, and pointing out how generational improvements have made a big impact there.
Also, probably something about diminishing returns on the computation needed/accepted for games. Or the other side of the coin, the rising floor (with newer hardware) of what negligible CPU usage can achieve (also something like multiseat or server usage).
EDIT: I should say it both reaffirms how good of a deal I got on a 2700 in mid-2019, and is a good thing to remember on how much better new (Zen) cores are. (probably not upgrading any time soon though and still using a 1050Ti as well, though I don’t really need an upgrade anyway)
To your point of how far new zen cores have come, I have a fun story from work.
In short, in my specific use case, my 7840HS (8 zen4 laptop cores) was at parity or outperforming a 1950X (16 zen1 desktop cores), in a fully multithreaded task.
The workload was essentially a bunch of RISC-V simulators running independently in parallel through a makefile, so the individual tasks benefit greatly from increased IPC.
I’m not sure the entire gain in performance comes from IPC, but it’s probably the majority and that is still very impressive.
I got a great deal on a 1950X setup due to a NewEgg sale, and it’s been a powerhouse for various “serious” tasks, but each individual core is pretty anemic and that does hurt it in a few games.
If laptop cores are matching or exceeding that, I feel confident in sticking with AMD for the next little while.
Those laptop cores are kerbstomping the zen1 cores according to these benchmarks.
Double the performance for single-core geekbench, which matches what I observed as well.
You are not logged in. However you can subscribe from another Fediverse account, for example Lemmy or Mastodon. To do this, paste the following into the search field of your instance: [email protected]
No game suggestions, friend requests, surveys, or begging.
No Let’s Plays, streams, highlight reels/montages, random videos or shorts.
No off-topic posts/comments, within reason.
Use the original source, no clickbait titles, no duplicates.
(Submissions should be from the original source if possible, unless from paywalled or non-english sources.
If the title is clickbait or lacks context you may lightly edit the title.)
8 cores 16 threads is currently a sweetspot if you’re a gamer but also do some compute here and there.
This just seems like explaining the difference between GPU-bound vs CPU-bound scenarios, and pointing out how generational improvements have made a big impact there.
Also, probably something about diminishing returns on the computation needed/accepted for games. Or the other side of the coin, the rising floor (with newer hardware) of what negligible CPU usage can achieve (also something like multiseat or server usage).
EDIT: I should say it both reaffirms how good of a deal I got on a 2700 in mid-2019, and is a good thing to remember on how much better new (Zen) cores are. (probably not upgrading any time soon though and still using a 1050Ti as well, though I don’t really need an upgrade anyway)
To your point of how far new zen cores have come, I have a fun story from work. In short, in my specific use case, my 7840HS (8 zen4 laptop cores) was at parity or outperforming a 1950X (16 zen1 desktop cores), in a fully multithreaded task. The workload was essentially a bunch of RISC-V simulators running independently in parallel through a makefile, so the individual tasks benefit greatly from increased IPC. I’m not sure the entire gain in performance comes from IPC, but it’s probably the majority and that is still very impressive.
I got a great deal on a 1950X setup due to a NewEgg sale, and it’s been a powerhouse for various “serious” tasks, but each individual core is pretty anemic and that does hurt it in a few games. If laptop cores are matching or exceeding that, I feel confident in sticking with AMD for the next little while.
Those laptop cores are kerbstomping the zen1 cores according to these benchmarks. Double the performance for single-core geekbench, which matches what I observed as well.