CPU bottle neck

Grim Dawn General Discussion
1

I know this has been beaten in to the ground. But I am hoping one of the devs can answer me?
My friend and I have Ryzen 9 3950x and Ryzen 7 7800x3D, respectively. He has an Nvidia 3090 and I have AMD 7900 XT.
How is it possible that we get lag or stutter? In single player mode, I get absolutely no lag. In multiplayer, I share in his lag. He gets the lag regardless of play mode. We have also tried absolutely every topic provided by google, these forums, steam forums, reddit, etc.

This is of course a problem because I love this game and I want to buy the expansion, but how can I convince my friends to come play it and buy it if the game is so poorly optimized for CPU utiliziation, that it becomes unplayable in multiplayer mode?

2

When do these stutters occur? I also had some, which seemed to happen when moving between (sub)areas or something. I’ve got a 7900 (cpu) and 7900XTX (gpu), so really shouldn’t be an issue… but :man_shrugging: I had no issues whatsoever when I had a 3900X and 5700XT. So there’s something weird going on.

Somewhere someone suggested to use GitHub - doitsujin/dxvk: Vulkan-based implementation of D3D8, 9, 10 and 11 for Linux / Wine
You can just download the files and dump them next to the exe.
I’m not 100% sure if it’s because of this, but my stutters are gone now. I’d give it a go!

3

It REALLY happens in the blood grove. But usually with bosses that have lots of effects. However, its totally not GPU related (obvi) because im watching my resource monitor.

4

My CPU is AMD FX 8320. GPU Nvidia 2070. The game plays mostly on 60 fps. It slows down a bit when there are hundreds of enemies on the screen. I also get a 0.5s freeze in one particular area in the blood grove, near the aetherfire. I also get a very minor freeze, barely noticeable when loading a new area in the overworld.

5

one
two
three
I can find more. :slight_smile:

I guess, he is a host in MP:

I’m not saying that multiplayer in GD doesn’t have any problems TBH, P2P MP is the one big problem itself, but 2/3 of all “lags” depend on a host’s machine (since MP is P2P) and on a network.

6

I didn’t say it was. :neutral_face:
Graphics also uses the cpu do tell the gpu what to do you know.

Also not saying it will definitely fix your problem, but it’s easy to try and who knows, maybe magic happens.

7

to me that sounds like the issue needs to be resolved for your mate first in SP, since it’s both not normal to have constant lag there, but then might also explain the MP issue?

There has been a couple ryzen specific “issues”/fixes mentioned over time, some guy posted a bug report where the solution turned out to be increasing powerplan or something (unsure if that applies to you guy’s newer? ryzen cpus), another either got it collaterally fixed updating their bios or enabling(disabling? sry legit forgot) a bios feature.
I say specific to ryzen since these were ryzen users and the normal/old troubleshoot stuff indeed hadn’t worked for them and they each had system specific solutions (not same ryzen CPUs i think)
The person with the bios thingy was literally night and day, he posted before and after videos and it was a complete change/fix.

alternative thing, since blood grove is mentioned specifically, maybe something as basic as shaders (or shadows?) settings is playing in. Mentioning that because it’s been a thing some people have mentioned in the past that XY setting was more prone to stutter and having them on either low or not lowest/not maxed helped

8

Measurements, videos, graphs. Provide the numbers when reporting performance problems. I have a 10 year old CPU and it works just fine, so help me understand what we are talking about.

9

Well first off, thank you all for the replies. We did not have these issues before we built new PC’s. I dont know how a Ryzen 9 and 3090 would have ANY lag in this game, period. Which is what he is using.

10

TLDR, i firmly think it’s solvable/there is hope for your friend/to find on their end, even if it might be a question of how much work it could require.

assuming it’s not a mere power setting and or deferred render setting or driver thing for the GPU, going by past Ryzen user mentions, i don’t think it’s like “poor hardware” related, but simply just system setting/interactions, which even if bothersome as heck/“nail in a haystack” then seems like it can be ironed out, because the HW is indeed capable(and so is the game)

i don’t have the full quote, so it’s still very basic/like i mentioned before in those examples
one user on an, what i presume is an earlier ryzen model since i was told later it’s not applicable to all now, simply had to increase the CPU powerplan (not talking overclock, i’m assuming windows/app setting)
other got their magically fixed when they, at the same time,
update bios+chipset drivers and enabled TPM
(since they did it at the same time they were unsure which was their direct fix, but that’s the note i jotted down atleast)
and while this/either of these 2 might not even be applicable to your friend, might be a different needle, but just like how your ryzen is working fine in SP?, it would make me believe it’s then fixable for your friend too/the case specific fix just has to be discovered for them too

this is not an excuse that GD doesn’t have some issues, old code be finicky or have its limitations, but when there are people with X issues, that manages to get it fixed, should mean it can be fixed, even if not every post affirms they found theirs.

To make another “weird bug” comparison, you have some people’s settings not saving “randomly” or after a game update. The fix is kinda known, but even then it varies for people which of it is, but also if they got it fixed/“followed all the steps proper” to account for the known, because there will strangely be someone here and there saying they did it, but still didnt’ work.

Since CPU interaction is so much more complicated i wouldnt’ be surprised if it can be harder to troubleshoot on a system’s varied basis, but probably likely still solvable even if it’s not just 4 fixed steps to go through. - and because it’s potentially so complex, i dont’ really fault people when they give up/post “the provided list of troubleshooting didn’t work for them”, becomes more about how willing/much effort one is ok with applying for “just playing a game” in the end.

11

I had the same experience. I also had some minor success with limiting the cores the game uses, but not sure if it was just placebo or not.

12

Ive asked him about his settings, but I think I need to remote in to his computer and take a look.

I appreciate the feedback!

13

One thing that I do want to mention is that you’re using two CPUs with a pair of CCDs. If you have the game going to a new core across the CCD, you’re going to have a stutter as the L3 cache needs to repopulate.

The best gaming CPUs right now are the 9800X3D and 7800X3D, not the R9s, because of the extra cache. Generally speaking, games don’t use up all of your cores like Cinebench does. An older title like Grim Dawn is probably not even maxing out a 6 core CCD, much less your friend’s 8 core CCDs.

What I’d suggest is you look at Intel’s Presentmon. It tracks data on performance. More importantly, it has a GPUBusy metric that can show data on when the GPU’s busy and when it’s not (read: CPU or RAM are a bottleneck). If you get data, I’d love to see it.

https://game.intel.com/us/intel-presentmon/

14

Theoretically, yes. But in practice I wonder how often that actually happens. I’ve had a 3900, which has the same issue (and probably worse even because each ccd is also split in two ccx) and never had issues. So I’m a bit skeptical.

Looking at actual perf data is a good suggestion though. I’m going to give it a go myself. Curious what I’ll find.

15

It happens more than you think.

I like Gamers Nexus’s site, so I’ll reference the data in it.

You can see Stellaris, FF14, and Rainbow Six Siege has similar generation X600/X700/X800 parts outperforming X900/X950 parts. Even F1 2024 has the 9700X with better 1% lows than the 9950X, which is the most relevant piece of data when talking about lag spikes.

It’s also worth mentioning that multiplayer is probably more relevant for lag, but that’s not going to be very consistent. And it won’t have cool graphs.

1 Like
16

Hmm… Thanks for the link. I had to look up the specs on the specific cpus. A 7600X is half a 7900X for example. So comparing those two on the gaming results it’s a bit of a mixed bag. Some games the 7600X is clearly better, others it’s very similar (usually slightly worse). And I guess on the 0.1% it’s more noticeable, which indeed indicates it has a little hiccup sometimes.

It’s a very different kind of lag usually. It won’t stall your graphics, it’ll just teleport the other dude (or monsters).

I tried the perf thingy. I get cpu usage spikes like these when I get my hiccups. (I think the raise at the end is because I started to take a screenshot with the windows native tool)

afbeelding
afbeelding763×411 69.6 KB

But not all usage spikes result in hiccups.

For me it’s always when moving between ‘areas’ or something. Not the big ones, but smaller parts of the map. I assume it’s spawning in stuff or something, but I don’t know for sure.

17

I’m not a Crate dev, but I am a dev. Mostly business stuff but I have dabbled in games.
I guess the primary market is seen as individual players and multiplayer is a secondary consideration. Multiplayer was not there day 1, so they didn’t design for it.
Not tried multiplayer myself but I took a quick look at the blurb.

Multiplayer is hosted one one machine and other players invited.
So the game is running on that one machine and it is communicating with the other player machines for inputs and sending them outputs - what’s happening.
Which is not free.

I would suggest you try hosting on different machines.
A productivity - high cycles and threads - cpu might perform better.