Is 2.5 now CPU multi-core?

i have hyperthreading turned off for DCS...

wondering if having hyperthreading on in the current 2.5 state would help its performance...

i have hyperthreading turned off for DCS...

wondering if having hyperthreading on in the current 2.5 state would help its performance...

As long as the OS is aware of hyperthreading and that it understands the CPU architecture, if it's doing this correctly putting the busy threads on separate physical cores, it should be okay, but I haven't tested yet anything with HT yet so no comparison.

And ofcourse the results I would post would count for the , it may not necessairly be same for completely different CPUs as OS Scheduler could do things differently there, or as someone else said, firmware/chipset possibly.

So far all these tests were without HT.

EDIT: I'm kinda exhausted after this ... might get back to it with a few tests with HT turned on, but that would just be more details that won't change much of the primary question of this thread.

You can start telling the software to run on a specific core but then you interfere with the OS scheduler and it will quickly become a whack-a-mole situation.

Well most of my later tests I used for this thread had affinity set to 2-3 for DCS and 0-1 for every single process visible to Process Lasso. I think I didn't post pictures with those yet, but that's what the youtube video had. (without HT)

But actually, I was playing multiple missions with Su-25T leaving these kind of settings.

Actually, this means my whole system and everything was running only on 2 cores for a few days now, oh, even I forgot and now realized this :doh:

But, I see absolutely no adverse effects or glitches in practice yet, ofcourse the perf and or responsivness is a bit down overall for the desktop stuff, technically.

If your application only has 1 thread, the OS can't make it run any faster on multiple cores than it would have run on a single core.

Yes, and that goes for any number of threads the application has,

That is the whole point what I was trying to make, albeit in a wall of text.:smartass:

I must point out that the task manager indeed shows correct numbers for what's it showing, it's simply showing what all other tools when you look at CPU Utilization would show, now ofcourse some people who understood it would not agree with my "fake" label.

But it's one more example of how generalization and consumer-friendliness many times sacrifices vital context and it shows in a number of ways in Windows GUI, this is true across the board, not just on windows, everything loses context when you simplify and shorten the description.

For example DXDIAG has a line where it says "Page File: ??? MB Used ??? MB Available"

If you take it literally, it would be all fake, however, if we correct for the improper labeling, then it would be all fine, because the number it self alone is completely fine, but it's showing something else, not pagefile usage.

Afaik pagefile usage can only be seen in Performance Monitor when manually selecting the counter, and it reports only the % used (that's how it is in Win7, I'm fresh with Win10)

What these numbers are showing is most probably Commit Charge (Win7) or Commited (Win10)

Infact is the Commit Charge that's the most important thing you should be looking at, whatever type of user you are. Task Manager in all post YR2K (dunno for 90') Windows versions shows "Physical Memory In Use" which is quite lower of a number than what the system is actually using in total.

This makes no sense because from a practical point of view, the programs and windows start acting unstable as soon as commit chagre limit is reached, you will get weird glitches, crashes, or a BSOD after at the limit. But looking at the big graph in Task Manager it would still show a bit lower at ~90% for example leading you to think you still have "~10% of memory left" which isn't true, so in that sense it would be "all fake" ... as in giving you a false impression, that 90% would just be the level of In Use, but it's just a number from a piece of the pie, it's only half the story.

Unfortunately this failtrain keeps going on the latest versions of Windows.

i have hyperthreading turned off for DCS...

wondering if having hyperthreading on in the current 2.5 state would help its performance...

i have a six core 4930k and i have had HT off for over 4 years.

this morning i turned it back on because OBS is struggling to keep up with recording..

short answer, both dcs and obs recordings got smoother.

so in my experience 6 cores is not enough for dcs and recording in VR anymore.

I'm going to continue testing but it looks good.

HT is staying on for now.

Guys if you run video capturing at the same time you are on ANOTHER PLANET.

I've tried my best to go through these posts to determine whether Project Lasso is a worthy venture for DCS in VR.......disturbingly, I fly for a living and for fun.........I am none the wiser TBH.

I can see it's a can of worms to those in the know let alone the rest of us.

It seems that there is a lot of experts with differing views on this, would anyone care to provide some generic layperson's advice?

Appreciate any help, cheers.

@WN

The ultra short story is your mileage may vary. Tinker with your system and if you get improvement good, if not, don't sweat it.

I've tried my best to go through these posts to determine whether Project Lasso is a worthy venture for DCS in VR.......disturbingly, I fly for a living and for fun.........I am none the wiser TBH.

 

I can see it's a can of worms to those in the know let alone the rest of us.

 

It seems that there is a lot of experts with differing views on this, would anyone care to provide some generic layperson's advice?

 

Appreciate any help, cheers.

in addition to the above from zhukov, most positive feedback came from X99 chipset owners along with 5th gen Intel CPU's and their cumbersome Turbo Boost behavior.

I have no real difference with or w/o PL, not on 6700k, 7700k or 8700k, but that's only me.

I opened this thread, but I believe that we do know the current position, i.e.

- DCS is still primarily single threaded, so the faster your CPU runs, the better, regardless of whether or not it has 4, 8 or 12 cores

- It can be misleading to look in windows task manager, as it may appear that the CPU load is spread across cores. From using MSI Afterburner, I'm personally convinced that core jumping makes it appear that way, but in reality, it's just not happening

- Some players have had success with hyperthreading, some not. I get the impression that it doesn't really add value

So my current belief is that we still need a highly clocked CPU, regardless of how many cores may not be available. Hopefully Vulkan will help spread the load generated from calculating the in-game visuals, and thus freeing up CPU availability on the primary core.

Thank you so much people, that’s an excellent summation. Back to the plane!

I can only encourage the talented developers at ED to make true multi-core a higher priority. It's hard to buy a competent CPU with fewer than 6 cores these days, with 8/16 threads becoming the mainstream norm in about 15 months time, as the new Sony and MS gaming consoles come out.

I don't exactly have any issues with running DCS it's perfectly smooth in most situations, but you know, it's always nice to get the most out of the hardware you buy, and particularly in multiplayer or scenarios with huge amounts of planes and things going on, there would likely be some benefits. But hey, I'm just glad we such as amazing simulator :-)

Imho, the solution is to fix it in silicon and not in sw.

There are/were prosperous looking projects, some have been taken over by Intel, AMD and others, that try to approach this from the other end, so the dev has nothing to do with this.

It's a month or two ago since I read about those approaches, forgot the project names and who bought them finally, but it is not that it can only be done in the software.

Time will tell

Yep, apparently Intel bought a startup company that was making headways on the software independent multi-threading front. If somethign comes off of those efforts, it will probably yeare before we can reap the benefits.

I noticed the same thing.

You'd think a change this major would be worth mentioning in the Changelog.

Did you show the 4 logical processors or the four cores?

Did you show the 4 logical processors or the four cores?

On my CPU (Core i5-7500 @ 3.4 GHz) those are the same. And Windows Performance Manager only offers to show "Overall Utilization" or "Logical Processors", at least on my system.

Yeah on my system I am trying to figure out why my VR is so bad and I was exploring my cores when I ran across your post. i wanted to know if all my cores were in action. It seems to me that a logical processor is not the same as a core. It says that a logical processor is the result of multiplying a core by the thread (hyper threading) Something like that. anyways I cant show real core load, just the load of the logical processors. Either way its not going to help my issue.

Yeah on my system I am trying to figure out why my VR is so bad and I was exploring my cores when I ran across your post. i wanted to know if all my cores were in action. It seems to me that a logical processor is not the same as a core. It says that a logical processor is the result of multiplying a core by the thread (hyper threading) Something like that. anyways I cant show real core load, just the load of the logical processors. Either way its not going to help my issue.

SMT or HT is that thing that add a whole other level of complexity to performance diagnostics in depth, when dealing with a lot of threads, cores, and factors in a single session of performance dataset, in basics you could just run two identical benchmarks with it disabled and enabled, but the general take is it's about paralellism, it won't raise maximum of what a single core can do, so it won't single-thread performance, and since DCS as many other games is limited by single-thread performance then it's not hard to figure out it won't do much.