Is 2.5 now CPU multi-core?

From looking in MSI Afterburner and Windows Task Manager, it looks to me that 2.5 has added CPU multi-core support.

Am I reading this wrongly?

If not, then I'm surprised that no one at ED has mentioned it.

could you share a screenshot of the task-manager?

some users have already stated that they saw improved load distribution, but it was always a bit vague.

no official words from ed neither...

could someone from ED clarify the nature of multi-core in 2.5?

on another thread i've seen a screenshot showing load evenly distributed over 4 cores, that would not only mean good multi-threading, but might even hint towards multi-core-rendering, which would be HUGE!

(and even more baffling, why it was never mentioned)

DCS has been multi-threaded for a long time. Anything they have done a major rewrite on was, if possible, split off from the main thread. The problem is that a lot of stuff remains in the "main" thread.

DCS has been multi-threaded for a long time. Anything they have done a major rewrite on was, if possible, split off from the main thread. The problem is that a lot of stuff remains in the "main" thread.

You sure about that?

When I looked at the performance for my CPU cores in v.2.2, it was one core maxed out, and the rest vaguely used. As commented above, the loading now appears to be evenly distributed across cores.

Does make me wonder how someone with say an 8 core CPU is getting on.

Any thoughts or experiences?

The answer to the above has some serious implications about what's the "best" CPU choice for the game. Was just looking and couldn't help but notice that for a similar price to a decent i5, that you could get an 8 core AMD Ryzen. Those additional cores would be VERY useful for not only any multi-core application, but also for Vulkan, which is supposedly coming.

In 2.5 I checked in Task Manager and I have 4 cores (8 logical processors). When DCS was activated ALL the processors jumped into action. Good to know Im not just working on one processor like some other folks have reported.

Multiple effects going on. Suffice it to say, most of DCS is still running in those same 2 threads. These threads however, may jump from one core to another core so you'll see multiple cores being used. Some object loading and stuff may be done in parallel over all the cores though. Don't be fooled, DCS is still mostly the way it has been.

Multiple effects going on. Suffice it to say, most of DCS is still running in those same 2 threads. These threads however, may jump from one core to another core so you'll see multiple cores being used. Some object loading and stuff may be done in parallel over all the cores though. Don't be fooled, DCS is still mostly the way it has been.

I get the logic you're suggesting. All the same, we appear to be seeing genuine loading in the other CPU cores, i.e. that DCS is spreading load in an effective manner. Shouldn't we expect that to be beneficial.

Furthermore, doesn't that also mean that the newer multi-core CPUs (e.g. 8 core units) might be more effective than a higher speed quad core unit?

I get the logic you're suggesting. All the same, we appear to be seeing genuine loading in the other CPU cores, i.e. that DCS is spreading load in an effective manner. Shouldn't we expect that to be beneficial.

Furthermore, doesn't that also mean that the newer multi-core CPUs (e.g. 8 core units) might be more effective than a higher speed quad core unit?

It will be better for loading of assets, obviously. But all the unit AI / flight models / collision detection / line of sight calcs etc are done on that one thread and at the moment, a CPU with more cores isn't going to help those things.

Just had another look at my CPU usage whilst flying. None of my CPU cores ever hit 100%, and that's with an ancient i5 2500@4k, and with a 1080, that in theory, should result in the CPU being the bottleneck.

you still get a judder when scripts run and triggers fire..

It will be better for loading of assets, obviously. But all the unit AI / flight models / collision detection / line of sight calcs etc are done on that one thread and at the moment, a CPU with more cores isn't going to help those things.

DCS 2.5 I see CPU 0-7: 84,68,81,68,89,81,86,99% (turbo disabled)

Lemoen, there must be more to it. See the post by bbrz or other post scattered in related 2.5 topics.

If people see, 8 logical cores utilized over 50% each, there must be more going on apart from asset fetching.

Could it be, that the speedtrees are seperated from the main renderer and get prepared on other threads before being pushed to the GPU?

Also, if we assume, that the utilization seen f.e. by bbrz is due to valid calculation and NOT caused by buggy code (or ED mining bitcoins on our systems:smilewink: ), than this definitely IS a valid reason to at least consider a higher core count over single thread performance when deciding about a cpu upgrade...

Additionaly there was an "official" statement that ED was going for a server/client archtecture even for singleplayer, which would untie AI-calculations (server-thread) from the main renderer (client-thread).

It would be great if someone from ED could quickly chime in, just to give us some very brief headups on how dcs handles multi-thread right now, since i am quite sure, that at least on caucasus map, there has been changes that are not yet fully understood by the community...

I see 50% utilisation max, besides for some transients over 50%, which I assume is loading or windows or something, on my i5 (4 cores). This along with people complaining about stutters, low GPU usage etc tells me nothing has changed.

Okay just an observation DCS seems to still only/mostly use 2 actual cores one hyperthreaded with some miscellaneous activity on other cores although that is likely OS, Chrome or itiunes or something else like task manager or msi afterburner etc. It would be great to get a response to this question and if DCS will utilise more cores/threads.

I see 50% utilisation max, besides for some transients over 50%, which I assume is loading or windows or something, on my i5 (4 cores). This along with people complaining about stutters, low GPU usage etc tells me nothing has changed.

Agree,. :)

I see 50% utilisation max, besides for some transients over 50%, which I assume is loading or windows or something, on my i5 (4 cores). This along with people complaining about stutters, low GPU usage etc tells me nothing has changed.

Do you get one core above 90% and the other on 50% max, or do you have one core at max 50% and the others way lower?

If you can't get any core above 50%, you obviously won't see multi-threading to bigger extend, because why would you?

You may not realize that there are many CPUs that may have significantly lower single thread performance than your i5 has, so they would benefit from multicore way more than you do. I f.e. have an "old" i7-860, which very well may only have half the single thread performance of a modern i5 or i7. It might only be when one core gets enough load, that threads get distributed to other cores.

Users have been reporting that changing DCS core affinity on HT-processors in windows to physical cores yields way better performance. I can not see, how this could work with a applciation that has such limited multi-thread capabilities.

tldnr: Could the users who see no decent load-distribution over multiple cores, make sure that they are not running a cpu with high single-core performance, therefore giving DCS no reasons to distribute load in the first place?! i think every i5 / i7 no older than 5 years can be considered high single-core performance in relation to dcs recommended system specs...

I just upgraded from an i5-2500k at 3.8, to an i5-8600K at 4.6. The difference was night and day. I thought the OC 2500 on a single core would be great for a while yet. But I'm afraid to report, its still down to single core performance in DCS.

What you miss about core performance is the bursts from different types of activities that can cause little bits of micro stutter. If the speed isn't good enough the bursts of CPU requirement are noticed, but most of the time are not.

I think for the first time recently I saw GPU slow down in VR rather than CPU after I turned settings up to close to max in VR. Previously it was all the microstuttery stuff.

I have not seen any performance benefit for multicore in 2.5 and there is no reason to expect it. Any performance benefits are likely from just reinstalling and running clean tests.

Do you get one core above 90% and the other on 50% max, or do you have one core at max 50% and the others way lower?

 

If you can't get any core above 50%, you obviously won't see multi-threading to bigger extend, because why would you?

You may not realize that there are many CPUs that may have significantly lower single thread performance than your i5 has, so they would benefit from multicore way more than you do. I f.e. have an "old" i7-860, which very well may only have half the single thread performance of a modern i5 or i7. It might only be when one core gets enough load, that threads get distributed to other cores.

Users have been reporting that changing DCS core affinity on HT-processors in windows to physical cores yields way better performance. I can not see, how this could work with a applciation that has such limited multi-thread capabilities.

 

tldnr: Could the users who see no decent load-distribution over multiple cores, make sure that they are not running a cpu with high single-core performance, therefore giving DCS no reasons to distribute load in the first place?! i think every i5 / i7 no older than 5 years can be considered high single-core performance in relation to dcs recommended system specs...

I checked when I finished my update, for me at least it is using all the 4 cores in the exact same manner. It is so beautifully optimized for the four cores, I wanted to cry :D

I just upgraded from an i5-2500k at 3.8, to an i5-8600K at 4.6. The difference was night and day.

Did you run both the 2500k and 8600k in dcs 2.5, specifically on the new caucasus?

Right now, i'm still guessing that the speedtrees may be responsible (in a good way) for broader distribution, that some users see.

Check out user bbrzs statement "CPU 0-7: 84,68,81,68,89,81,86,99%"

Maybe bbrz could give more info on his system and in what ingame situation he measured these numbers...

I checked when I finished my update, for me at least it is using all the 4 cores in the exact same manner. It is so beautifully optimized for the four cores, I wanted to cry

Nice :thumbup: What are your system specs, if i may ask?

Do you get one core above 90% and the other on 50% max, or do you have one core at max 50% and the others way lower?

2 cores at 100%, CPU at 50% (4 cores, no HT).

In VR I once popped system manager to window to observe the CPU cores utilization and I could say that all got something, first few were about 50% and then rest lower.

GPU hits 98-99% all the time, even in the main menu and CPU never reach 100% anywhere.

6c/12t, almost full tilt in dense scenarios like a city ( Ka-50 Free Flight for example ), when I fly over not so dense areas 4 cores ease off and the 2 for DCS itself remain where they were, 1 full tilt and sound so lala. The other 4 cores seem to feed the GPU.

Yeah, It is nice to see that DCS as a whole can leverage at least 6 cores ( real cores, ProcessLasso used ).

I wonder what a 5G-not_yet_available 8c/16T would look like when coupled with a 1080Ti.

Vulkan will make you wanna cry for a 8-12c or even more cores CPU, that's my bet.

Vulkan will make you wanna cry for a 8-12c or even more cores CPU, that's my bet.

I'm ready :D

Are you guys sure your getting more than 2 real cores at ~50% and 1 virtual core at like 90 to 95% for DCS because as far as I can tell when I'm only running DCS and task manager that's all I get with maybe some slight usage of the other 9 cores. Exactly the same CPU utilization I got with 2.0, 2.1, 2.2. and now 2.5 around 12 to 15% utilization.

If I run MSI afterburner there is a heck of a lot more core activity an no improvement in DCS performance.

Same system and rendering RAWs or video and I get all 12 cores at 90 to 99%

As for graphics my old 980Ti used to crank along at nearly 100% GPU utilization now with my 1080Ti it's lucky to get to 60% GPU utilization with better performance in that I can run settings higher and maintain decent frame rates mostly 90FPS off the ground and never below 45FPS near ground in built up areas.

If you get more CPU core usage in DCS I like to know how do you achieve it?

thanks for all the reports.

i think, that it is very obvious now, that DCS is capable of doing "proper" multi-threading with major tasks (most likely even the renderer) being distributed over multiple threads.

i could not however see a pattern, that could explain why some users experience no real benefit compared to previous versions.

in this early days of 2.5 it could of course very easily just be a little bug, that prevents some systems from going "full cores".

if you see load distribution like in 1.5 and 2.2 (load mostly on one single core), make sure, that v-sync is off (in options AND in gpu driver suite) and choose a low render resoultion, to make sure that the GPu is not holding you back. also make sure, you are in exclusiv full screen mode and maybe set dcs task prioroty to "high" in task manager.

Please be aware, that a GPU utilization below 99% could still mean, that your graphic card is the bottleneck, f.e. because lack of Vram bottlenecks the gpu, or because other percularities of the gpu architekture create overhead with certain aspects of the dcs renderer.