I will likely buy the “1700” in the next few months. How do theses CPUs look from a dev’s point of view in how it will affect the game’s performance? 16 threads seems really nice in terms of splitting up the workload (goes up from 4 of my i5 3570k XD ).
new cpu architecture … leaked specs …
There’s nothing in there that would show, anywhow, what actuall perfomance these chips will have. I too hope they will be great and one may be able to guess the performance, but no one is able to compare.
What is needed is for the actuall processor to be used in a realworld scenario and compared with others in the same situation … a benchmark.
There have been quite a few synthetic benchmarks, but the CPUs haven’t launched yet so obviously nothing can be shown for real-world game performance. But one thing i’m looking out for is benchmarks on games that haven’t been optimised for more than 2 or 4 cores. So far it seems it is comparable to the i7-6850K which costs twice as much and competes in certain situations with 1000$ ones. The X-series CPUs from intel still have higher single-threaded performance but they cost way more too.
i have i7 6700k and next option for gaming would be i7 7700k, even if it is 7-8% performance increase.
why? cos there are still games that are single core heavy like arma3 or some flight sims (i don’t know about IB) and intel is insurance to run every game with good performance while AMD unfortunaly has poor single core performance.
Won’t know for sure until the benchmarks come out but overall I’m pretty excited about Ryzen. I won’t be buying one anytime soon because I have a fairly new machine w/ an i7 6700k but it’s something I’ll be keeping a close eye on over the next 2 years.
As for its impact on games I can only speak about I:B. Our game is not particularly well threaded at the moment however over the last 1-2 months we’ve been investing heavily in threading the game so the 8 cores in the Ryzen should give you a significant boost over raw single core performance with I:B in the near future.
As far as I understand the cores won’t make a difference for IBS, since all threads form a single process run on a single core, that’s certainly the case now and you guys do have some multi-threading present…
This goes against my personal experience developing (hobby) physics simulations and games. Not saying you’re wrong, just that I haven’t experienced such a limitation myself.
Would you care to elaborate on why that is the case? In my experience, threads spawned by a process are free to run on any core and normally seem to be split up as evenly as possible across cores.
EDIT: I’ll add that most of my knowledge in the area should be taken in the context of Java and (only recently) Rust, but I would have thought that the language in use would not affect how the threads are scheduled.
I might be wrong, have been monitoring IBS CPU usage though and it hardly ever goes above 15% (on a 8 thread CPU) even with some 40 threads running.
Ah I see, my bad, I thought you were speaking generally about multi-threading and not specifically about I:BS.
I’ve not done any monitoring on I:BS (was only able to play during the Alpha weekend) so I can’t really comment on how it performs on that level.
Yes i was leaning more towards IB threading performance. I don’t expect anything special from current state of development and was proding for info on what gains could be extracted by having 16 threads.
As I’ve already explained, we are using multiple threads, but only the main thread is usually busy. Other threads wake up and become active but only for a short amount of time; for example, the sound thread might run for less than a millisecond while the main thread goes for 16 ms ( 60 fps ). Same for physics or networking. So yes, we do use multithreading, but only for a fraction of each frame, because there’s usually not enough work for those additional threads to run 100% of the time. Our next step is to split the main thread into multiple sub tasks, that’s where you’ll see real performance gains on multi core machines.
pcper got a new video about it, most of it is early analysis of the known-so-far info: