Re: BF2042 CPU Running At 100% Constantly For No Reason

The game's player count has doubled which unfortunately means the CPU is going to be the bottleneck in most systems now. I'm a game dev. Hear me out.

BF3 and BF4 featured 64-player support. However, BF3 on X360 and PS3 couldn't handle the doubled player count of 64. Now, 2042 supports 128-player counts. Xbox One, Xbox One X, PS4 and PS4 Pro can only run 64-player matches. Only PC, Series X and S, and PS5 can run with 128.

This is simply because of how game code is utilised. It is purely a mathematical calculation that is run on the CPU side. Doubling the player count will quite simply double the CPU cost per frame for synchronising positions. I'm going to guess that position synchronisation logic is the bottleneck out of all CPU tasks with this game given the player count limitations of the past and present platforms. There's isn't a simple fix for this, unfortunately.

FYI, I'm running a 6800 XT which in most cases can net me 144fps. My 3600 however, struggles to achieve 90 in a 128 match. This is not a bug, it's simply how the game is designed.

You want to tell me that mine I9 9900K is too slow to reach 80 fps in that game on low details just because there's 128 players on map ? Please stop this 🙂

Ok - But why DICE recommend then a 4790K quad-core CPU? Why also hexa/octa-core owner describing the same problem? If that is really by design, then its a fatal decision. Then they should drop that 128 players feature.

---

@LeMairyHuff wrote:

The game's player count has doubled which unfortunately means the CPU is going to be the bottleneck in most systems now. I'm a game dev. Hear me out.

BF3 and BF4 featured 64-player support. However, BF3 on X360 and PS3 couldn't handle the doubled player count of 64. Now, 2042 supports 128-player counts. Xbox One, Xbox One X, PS4 and PS4 Pro can only run 64-player matches. Only PC, Series X and S, and PS5 can run with 128.

This is simply because of how game code is utilised. It is purely a mathematical calculation that is run on the CPU side. Doubling the player count will quite simply double the CPU cost per frame for synchronising positions. I'm going to guess that position synchronisation logic is the bottleneck out of all CPU tasks with this game given the player count limitations of the past and present platforms. There's isn't a simple fix for this, unfortunately.

FYI, I'm running a 6800 XT which in most cases can net me 144fps. My 3600 however, struggles to achieve 90 in a 128 match. This is not a bug, it's simply how the game is designed.

---

@LeMairyHuffwarzone has 150 players on a bigger map than any bf2042 map, and the destruction in bf2042 is so minimal (dice should be ashamed of bf2042's destruction and still calling it battlefield, after all previous games so far) that it doesnt count anyway, and i have no issues with warzone at all, neither at the beginning of a match when all 150 players are here (ie. more than in bf2042 at any time, even if there are bots filling missing real players slots), nor later when many leave.

moreover, bfv with 64 players and 60 tick runs at more than double the cpu-rendered fps than bf2042 with double the player count and 45 tick does for me (while the gpu renders it like a 5-years-old perfectly optimized game would run, ie. great, since it doesnt look any better anyway), and the engine is basically the same, so if there were no optimization issues, the performance should scale accordingly, which it just doesnt.

and as already mentioned by several people, there are reports of people with top-end intel 9/10/11-series i9 hexa-core cpus, and their amd counterparts, experiencing bad (cpu-rendered) performance, while the *recommended* specs for bf2042 state a quad-core i7 from 2014 to be enough.

so, there definitely is a huge problem with the game code, see eg. jayztwocents comment on bad game optimization; i doubt he is the type of guy who would just blame it on the code despite knowing better....
worst of all, most likely the meanwhile dated frostbite game engine simply isnt capable of handling bf2042 without overstressing most cpus, so there unfortunately isnt much optimization headroom anyway. this even more shows just how bad the whole bf2042 development is and misleading the system requirements are. i have lost all hope with this game already since the beta, technical playtest still had the "alpha excuse".

The explanation as to why Warzone doesn't have issues with the player count is because rarely on warzone are there more than 10-20 players in the user's vicinity. It's not just player positions you have to consider though.

Technically, everything which is not in view of the Camera viewport (what the player sees) are culled (not considered for rendering). This is suitable in warzone and very effective because you rarely see upwards of 10 people at once. All of those other 130 are therefore not considered for rendering and have literally 0 negative impact on performance. In Battlefield, it's very often you see 30 or more in front of you. Although you may not notice 30 people, there are 99% of the time, more people on screen than you realise. However, it's not as simple as that either.

For example, there are more vehicles present and rendered in BF 2042 at any given time than Warzone. How many quads, Hovercraft, helis, Cars, Tanks, Jets? Ok there are a few more vehicles maybe 10 on-screen at any time. But... then the position and rotation of all of those items have to be scaled as well. When a tank shell fly's across the map? Everyone has to render it and all of the subsequent systems this triggers. For example, bushes may be pushed animated to the side by a grenade explosion or an ammo box animation has to be rendered for everyone near the ammo box. You can see where more players really add a lot more to the strain than just the players position themselves. 2042 is on another level compared to Warzone.

The tick count is an odd one. Actually, it can't really be factored in because even if the frame you're rendering is a frame not rendered using data directly from the server, the game still renders using predictions when data isn't available. So for example, if every other frame is guesswork (120 fps, 60 tick), your client is guessing where everything will be. This will be arguably lower or higher depending on what goes into the guesswork.

You can't add more players and have performance "scale". You have to lose some performance at least. It's not "free" and there's no technology that will make it free. However, it can be helped, for example with culling. Other common methods are "silencing" physics systems which don't move (turning their expensive calculations off), rendering objects at lower quality when further away (LOD). I'm very surprised we haven't seen Variable Rate shading implemented as an engine feature yet.

Moreover, everything associated with Physics is done on the CPU also. I noticed their vehicle physics systems have improved dramatically over previous games which will be more costly on performance.

I agree about the recommendations part. They are completely inaccurate. I think more optimization needs to be done and more tricks for rendering and passing only the needed data to fulfil a frame need to be added. But then when you look at the amount of missing "features" or "implementations" compared to other BF games, content was certainly cut for release. No questions asked. The gun customisation menu is a mess and would not expect that to be in an RC Build. The most confusing system I've ever used.

Nowadays the norm is releasing unfinished games and this is another great example.

@LeMairyHuff  warzone has game modes like rumble, king slayer, clash with 100 players all playing at once with unlimited respawns, condensed in an area far smaller than any bf2042 map (parachuting into the match actually lets you see basically everyone in the match at once, and the performance doesnt suffer), with vehicles. and there are bullet physics (velocity, drop, drag), explosions, terrain physics, weather effects, shadows, lighting, sound as well, and it still runs great. really dont want to defend warzone, never been a cod player, im a pure bf fanboy, just switched to warzone to bridge the gap between bfv's development end and bf2042, and now that bf2042 is out i played more warzone in the past 10 days than i did bf2042...

if the tckrate was doubled vs. bfv, i could understand that the client-side performance would suffer (this is eg. why cs/css/csgo never introduced 128 tick official public servers, but have them in most official leagues where there is a limited number of clients, all with equal high-end sytems), but the tickrate is 25% *less* than bfv in bf2042! and packet extrapolation worked in cs in 2000 already without overstressing single-core, single-threaded cpus 21 years ago, no reason why today's multi-core multi-threaded cpus with considerably higher ipc shouldnt handle it with some 2x-4x more players in a match. its just the engine, they forced 128 players mostly for marketing reasons without making sure the engine can handle it resp. developing a new frostbite engine that does, and even portal with 64 players doesnt run well, if it only was 128 players, portal should run as well as bfv does, but it doesnt at all, it is just slightly less unplayable, but still unplayable, compared to how it should run to be playable.

I wasn't aware of that as I don't keep up to date with that game. However, I can guarantee the visual quality of 2042 is a step above Warzone. There's simply less going on to make everything look nice because the details are not the same.

I'm more aware the engine for Warzone will be more performant than Frostbite. Probably because it's simpler and most of its features aren't a WIP at any given time. I don't doubt that EA/DICE will be writing just as much engine code as game code to keep up with the upgrades they try to present with these games. Moreover, no one asked for 128 and from a player perspective, you won't be able to notice much of a difference. I doubt anyone wants to run at half the frames just to see a couple more explosions and helis flying in the distance. This isn't good game design.

The data per tick rate has doubled, which of course will degrade performance and all networking stuff is of course, CPU driven. Hmm CSGO can do that because literally everything on a given map is static. You don't even have to synchronise bullets because they don't have physics. They are implemented using Hitscan which costs nothing in comparison to 2042s simulation. COD used hit-scan until Modern Warfare (2019).

It all comes down to redundant and overcomplicated code. Frostbite is probably massive and overly complex compared to the likes of COD. You have to think about what games it's shared accross. Fifa, BF, Madden, NFS...

Not enough thought goes into larger games these days because they know they can get away with it. I would not work for a AAA like this which doesn't get to independently create a game they want. It's all money-driven.

If you enable the console using "`" key, type drawfps and set the value to 1, you can see the expected frame rate from the CPU and GPU seperately. My GPU is constant 130+, my cpu sits at 60 which is laughable.

An example from earlier. I run a 6800XT Nitro+ and 3600 @ 4.3GHz, 3600MHz DDR4 dual-channel linked with IF clocks perfectly:

6k1cbdW.png (166×112) (imgur.com)