Low FPS on a high end PC -> shadow issues

[DonDregon]

Sorry, I probably should have been more clear.

Yes the GPU is necessary, and it is recommended to have quite a good one. Just the difference between a 1060 and a 1070 is noticeable, though not significantly different. What I meant was that the 3080 won't see a notable improvement compared to the previous generations because the game is primarily CPU-bound still for SI and AI mechanics. For this particular title, it most likely does not perform significantly better than a 1080 or 2080.

I am almost constantly inundated with people asking "Why can't I get 144+FPS at 4k on this game?", when they're sitting there running it on a 6th or 7th-gen Intel. Even Gen4 AMD CPU's aren't seeing significant improvements over the Gen3's simply because the game can't really utilize all the extra cores.

Back in Alpha 18 I did a direct comparison on varying clock speeds here. It shows how much of a difference small changes in CPU and RAM can make with no change to the GPU. Somewhere in General Discussion here there is a similar thread where I also included benchmark scores on the different setups.

Hahaha I understand, btw I'm quite old to know that some games use some resources more than others and experienced a quite good gaming back those days when 144Hz monitors didn't even exist, I only seeked to avoid those drops to 20fps and having the game running over 45-50 always. BTW I've managed to reach that, read the long comment before if you are interested on knowing more.

For the CPU concerns I let the system to choose the freqs by default, not planning OCing it at this point, I did OCed a bit the mem clock (native speed is 3200) and now I've both RAM and CPU mem clock running at 3466 (it breaks down if I try to go further).

I may pick a 5600X on a near future just to upgrade my wife's CPU which is older using my current 3600X, I suppose that I can get a little improvement with the 5600x in comparison to the current 3600X, let's see.

 

[Naz]

because in order to make a strict core association to a given software to work well you'll need to edit the core usage for ALL running apps including system processes which is a titanic task and suitable for drawbacks, otherwise you can get better fps but some other app can malfunction such discord, chrome and whatever. I've 3 monitors and I use all of them with different things so It's a bad option to me, this is of course a TWEAK that can or cannot be achieved using 3rd party software but it will definitely affect the entire computer performance at least during game sessions on 7d2d.

Thank you for returning with something more constructive. First up is this statement. You do not have to edit any other process affinity because you altered another's. I've never had any problems setting affinity on multiple apps and using 4 monitors, not sure why you think this is the case. It most definitely does NOT affect the entire computers performance while 7dtd is running. 
 

Then I've found a thread about Unity games, that seems to work for both 7d2d and Escape from Tarkov

Now this tweak you linked actually holds some water. I've ran some benchmarks to test both the affinity tweak, boot config tweak and running both tweaks.

These test were run on my other machine with a 6700k (4 cores 8 threads) & 1080ti at 1440p. You can see regardless to what we do we're still firmly CPU Bound even with a older 1080ti. However both tweaks improve this

- Editing the boot.config seems to not combine with core affinity tweak so I'll say that editing the boot.config is the way to go as it has less drawbacks than core affinity has.

As we can see from these results this part obviously isn't true, running both tweaks here yielded greater performance than just either 1 and again there are no drawbacks from assign 7dtd 4 core affinity, not that I've ever experienced in thousands of hours anyway.
 

Other options:

I saw there's a -threads launch command that can be set in steam launch options which I guess can deal to the same result than using the affinity thing.

I mean if you set up the game to run on cores 0, 1, 2, 3 it will use (if muli-threading enabled) the threads:
0/1, 0/2, 1/1, 1/2, 2/1, 2/2, 3/1, 3/2

So if you set the launch command to use those threads, the result should be the same than saying "use this specific cores".

* Note that this is my guess, I said I'm a software engineer and it's true but I work on a different field so I may be wrong if any of those options have underlying effects that I know nothing about.

There is also another option like -high which makes it a high priority process but well, check the details here

I've tried:

-threads 4 -high

-threads 8 -high

-threads 4

-threads 8

Result: -threads 4 deals a performance loss, -threads 8 does not make any difference and -high does not make any difference as well.

This was an interesting read, however he's talking about csgo, you're comparing apples to a bacon sandwich, so doesn't really apply to a fully destructible voxel game on an entirely different engine. Also when i refer to affinity i mean launching the game with a cmd affinity start-up shortcut or task manager, i haven't tested the -threads launch option in steam so i can't say if this will still apply using that method.

However his closing statement has some relevance here.
My recommendation would be to not set the launch option, unless you can objectively measure different values are beneficial to your performance. I cannot stress this enough, don't do anything unless you can objective measure the effects. These kinds of tweaks are dependent on all kinds of things, such as your hardware, in game settings, stuff running on your machine and god knows what. Any 'guide' you find online that will tell you to set this to a specific value (be it 4/8, or the number of logical/physical processors) is likely full of snake oil.

I've published tons of data here to support my claims of the benefits of 4 core affinity, i'm not pulling recommendations out of thin air. Try things and benchmark them yourself to validate it.
I also agree with the statement that More people should test things and properly, and i don't mean taking a screenshot of the framerate. Run a proper comparative controlled benchmark. A "Spot Check" can give you a quick indication however you're comparing 2 frames to each other, you don't know what was running in that split second it took to take the screen shot, maybe a horde was spawning, maybe the slightly different camera angle changed the load on the system because it has more shadows or reflections, maybe the state of the particle effect on the weapon affected the result. Even taking the screenshot can affect the result if for some reason it was saving to a hdd that had to spin up freezing the game for a split second. A proper benchmark run can account for these things and give a more realistic result.

I'll do some more benchmarks on my 5950x and 3090 system to see how the boot config tweak works on a different system, but first results are really good even better running them both.

 

[DonDregon]

Thank you for returning with something more constructive. First up is this statement. You do not have to edit any other process affinity because you altered another's. I've never had any problems setting affinity on multiple apps and using 4 monitors, not sure why you think this is the case. It most definitely does NOT affect the entire computers performance while 7dtd is running. 
 

Now this tweak you linked actually holds some water. I've ran some benchmarks to test both the affinity tweak, boot config tweak and running both tweaks.



These test were run on my other machine with a 6700k (4 cores 8 threads) & 1080ti at 1440p. You can see regardless to what we do we're still firmly CPU Bound even with a older 1080ti. However both tweaks improve this

As we can see from these results this part obviously isn't true, running both tweaks here yielded greater performance than just either 1 and again there are no drawbacks from assign 7dtd 4 core affinity, not that I've ever experienced in thousands of hours anyway.
 

This was an interesting read, however he's talking about csgo, you're comparing apples to a bacon sandwich, so doesn't really apply to a fully destructible voxel game on an entirely different engine. Also when i refer to affinity i mean launching the game with a cmd affinity start-up shortcut or task manager, i haven't tested the -threads launch option in steam so i can't say if this will still apply using that method.

However his closing statement has some relevance here.
My recommendation would be to not set the launch option, unless you can objectively measure different values are beneficial to your performance. I cannot stress this enough, don't do anything unless you can objective measure the effects. These kinds of tweaks are dependent on all kinds of things, such as your hardware, in game settings, stuff running on your machine and god knows what. Any 'guide' you find online that will tell you to set this to a specific value (be it 4/8, or the number of logical/physical processors) is likely full of snake oil.

I've published tons of data here to support my claims of the benefits of 4 core affinity, i'm not pulling recommendations out of thin air. Try things and benchmark them yourself to validate it.
I also agree with the statement that More people should test things and properly, and i don't mean taking a screenshot of the framerate. Run a proper comparative controlled benchmark. A "Spot Check" can give you a quick indication however you're comparing 2 frames to each other, you don't know what was running in that split second it took to take the screen shot, maybe a horde was spawning, maybe the slightly different camera angle changed the load on the system because it has more shadows or reflections, maybe the state of the particle effect on the weapon affected the result. Even taking the screenshot can affect the result if for some reason it was saving to a hdd that had to spin up freezing the game for a split second. A proper benchmark run can account for these things and give a more realistic result.

I'll do some more benchmarks on my 5950x and 3090 system to see how the boot config tweak works on a different system, but first results are really good even better running them both.

That's great so it means many users could get a boost performance combining those tweaks!

I've no tool to bench this and 7d2d has no built-in bench, I'm just trying to help with what I could found and what I've tried so far walking around the same like 200x200 with this tweaks enabled/disabled/combined looking at the fps count and I'm telling you the result. Btw how does it come that you get better result limiting the game to 4 cores on a 4 core CPU? Doesn't limit the cores to 4 means "use all cores" anyway?

Also I would like to know which tool do you use to bench this images out so I can use it as well to post the results 

 

[DonDregon]

This was an interesting read, however he's talking about csgo, you're comparing apples to a bacon sandwich, so doesn't really apply to a fully destructible voxel game on an entirely different engine. Also when i refer to affinity i mean launching the game with a cmd affinity start-up shortcut or task manager, i haven't tested the -threads launch option in steam so i can't say if this will still apply using that method.
 

It's talking about steam launch commands. These launch options work for almost all the games. They may not sometimes work with third party applications and nor they are always compatible with everyone present in the Steam store, it does not matter the matter of discussion being CS:GO, I wanted to try it with 7D2D as well and I told you the results, you may test it isolated to see if it makes any difference to you as you benefit of this affinity tweak, I can't say much because setting affinity for 4 real cores (8 virtual cores) does not add any improvement on performance nor stability to me as far as I've tried.

 

[Naz]

That's great so it means many users could get a boost performance combining those tweaks!

I've no tool to bench this and 7d2d has no built-in bench, I'm just trying to help with what I could found and what I've tried so far walking around the same like 200x200 with this tweaks enabled/disabled/combined looking at the fps count and I'm telling you the result. Btw how does it come that you get better result limiting the game to 4 cores on a 4 core CPU? Doesn't limit the cores to 4 means "use all cores" anyway?

Also I would like to know which tool do you use to bench this images out so I can use it as well to post the results 

Yeah no worries I use msi afterburner and rivatuner server statistics. The reason it helps for a 4 core cpu is because it has hyperthreading, so on top of it's 4 physical cores it has 4 virtual cores. So windows and applications see it as an 8 thread cpu. I haven't tried the tweak on a 4 core 4 thread cpu, but I think it's safe to say it wouldn't do anything. 

 

[Naz]

can't say much because setting affinity for 4 real cores (8 virtual cores) does not add any improvement on performance nor stability to me as far as I've tried.

This is because you assigned the game 8 threads, just assign the 4 physical cores not the virtual threads. Physical cores start at core 0, then core 1 is core 0's virtual thread, then core 2 is a physical core, core 3 is core 2's virtual threads etc so forth. So you just want core 0,2,4,6

 

[Naz]

Benchmarks from my other system

Specs
cpu: AMD 5950x 16 Core 32 Threads
Ram: 64gb 3600mhz
GPU: RTX 3090

All setting's set to ultra, except motion blur which is disabled.

This time the affinity tweak came out on top, but combining the tweaks still wins the day. Like i've said many times 7DTD doesn't like having tons of threads available to it, the more threads it has the worse the performance is. This is why the affinity tweak won this round because the game had access to 32 threads it couldn't use efficiently.

If you want to run a similar benchmark, choose a new game on nav's map. Since i use the "settime day" command to reset the time of day after every run, i wouldn't recommend doing this on a world you intend to play on, since settime is notorious for breaking things. i also use GPUZ and it's sensors tab to log the avg gpu load and clock speed, just reset the log just before you start a new run and remember to screen cap the app after every run to somewhat accurately record the averages and add them to the benchmark txt data.

I'd also recommend you follow my benchmark run around Diresvile if you want it to be comparable, i'll leave the exact path i follow below (note the image is from a14 and the working stiffs before the bridge is now on the other side of the street)
 

Spoiler

I used to use food and coffee to continually sprint, but it's no longer necessary. You can how enable debug mode in the console (dm) and hit Q the H. Q is god mode and H is fly mode, first go into god mode for unlimited stamina and then disable fly mode so you can run like normal.

 

[DonDregon]

This is because you assigned the game 8 threads, just assign the 4 physical cores not the virtual threads. Physical cores start at core 0, then core 1 is core 0's virtual thread, then core 2 is a physical core, core 3 is core 2's virtual threads etc so forth. So you just want core 0,2,4,6

understood, I'm taking a try just now and let's see

 

[DonDregon]

@Naz I've set the affinity config as you said (0, 2, 4, 6), instead getting 81-125 now i'm getting 98-119 while most of time seeing numbers between 101 and 111 so I can say both combined are pretty good, even lowered the highest peak I'm getting more average fps and better stability.

Now I'm gonna test in-game video options to check if some make more difference than before and conclude this entire stuff.

@SylenThunder as a moderator, do you thing this entire stuff qualifies to be a pinned "something" as instructions for everyone to reach the most performance possible on the game or something similar? I can write it down pointing Naz's post about 4-core limit plus this boot.config tweak as well as video options tweaks and Naz can provide benchmarks and maybe add further details for example (if you want so as you've already screenshot some).

 

[SylenThunder]

@SylenThunder as a moderator, do you thing this entire stuff qualifies to be a pinned "something" as instructions for everyone to reach the most performance possible on the game or something similar?

Have done this before, but there isn't a flat set of settings and configuration that works on all systems. Add to this that it's a moving target. Each update changes values. Sometimes quite severely. 

I do keep track of a lot of the tips, and when the game is closer to the end of Beta, I fully intend to put something like this together. It's a lot of work though, and as it is right now, I'm about a year behind on updating some of the support FAQ. I need to do a full re-design there, and am thinking of splitting it into two separate threads. Maybe three.

Seems like every time I start to get caught up, life throws a wrench in the mix.   Soon as I get done with restoration from the tornado damage, and take care of a few other tasks, I should have a fair amount of free time again. At the current rate of affairs, that will probably be around the time a20 drops. So then I'll be busy with stuff again.   (Actually might work out in favor of things here though. I'm already gathering some things together. Like chess, I'm typically thinking several moves ahead.)

 

[Naz]

Just wanted to follow up with some more info on the bootconfig tweak. Seems it isn't compatible with EAC, so unfortunately players can't use it if they play on a EAC enabled server. It would have to be part of vanilla which is a shame since it's a really simple change and on initial tests provides a great deal more performance.