ED:Odyssey Update 7 Performance Issue and proofs

[Morbad]

Is the reduction linear?

Where the GPU utilization is maxed out, yes. However, it's not maxed out as often as it was prior.

I'll have a look tonight but, IIRC, my CPU is waaaay underutilized. I'll also post some screen caps. ...way less work than making a video.

EDO has always had bizarre CPU utilization trends and it's going to look under utilizied on anything reporting CPU utilization unless you reducing the polling interval to ~100ms or so.

With Update 7 it seems to scale less well with cores than before. I think there is some sort of new scheduling or work distribution issue with the main game thread. It's much easier to become CPU limited, reducing the number of logical cores assigned to the came doesn't hurt performance until fewer than four cores are allocated (on my system) and during frame rate dips/stutters that were not present in Update 6, I see a secondary worker thread spike in CPU load while one of the render threads bottoms out.

Almost feels like they were trying to get the game to run better on min-spec systems, but just broke the thread syncronization in the process.

 

[ubermick]

Many of us have been saying its a CPU bottleneck.

I dunno. I've got a 5900X that barely breaks a sweat with the game. I think the highest CPU utilization I got was something like 36%. Whereas my RTX3080 is being absolutely whipped to death, hitting temps absolutely ZERO other game I own (granted not many, but F1 2020, RDR2, and even Cyberpunk) come close to. In fact every other game barely cause the fans to kick on.

(Haven't tried Update 7 yet, nor do I plan to for another day or two)

 

[Sighman]

I got a definite boost in stations, haven't tried settlements yet.

Proof below - same station, standing in the same spot and looking the same direction, same PC, same graphics settings.

Running around concourses at 60fps instead of 45 feels great.

 

[MCCART]

It's much easier to become CPU limited, reducing the number of logical cores assigned to the came doesn't hurt performance until fewer than four cores are allocated (on my system) and during frame rate dips/stutters that were not present in Update 6, I see a secondary worker thread spike in CPU load while one of the render threads bottoms out.

Almost feels like they were trying to get the game to run better on min-spec systems, but just broke the thread syncronization in the process.

I saw, generally, the same results. I was going to mention it in my earlier post. Then I remembered that I have SMT enabled and that might taint the results. Someone's tried all of this with SMT disabled, right?

 

[Morbad]

This is a 100ms polling rate:

First two-thirds is mostly a completely CPU limited scenario, later portion is predominantly GPU limited.

Task Manager, Process Explorer, or anything else that polls at 1000ms intervals never shows any of those spikes getting past 70-80%.

I dunno. I've got a 5900X that barely breaks a sweat with the game. I think the highest CPU utilization I got was something like 36%.

Overall CPU utilization is irrelevant and per-core utilization is averaged out to the point of being completely obfuscated at common polling rates. If your GPU utilization isn't nearly maxed out, chances are that you're CPU limited to at least some degree, and reducing CPU clock would demonstrate this by performance falling...probably not linearly, however.

I saw, generally, the same results. I was going to mention it in my earlier post. Then I remembered that I have SMT enabled and that might taint the results. Someone's tried all of this with SMT disabled, right?

Yes. The image above is with SMT enabled, which further obfuscates utilization, but turning SMT off doesn't prevent the Windows scheduler from bouncing loads from core to core, or keep the game engine from scheduling jerky loads in the first place.

 

[Crank Larson]

I dunno... My 3090 can get up to 35 FPS.

Flying over a planet even without boosting isn't fun as the rendering can't keep up with the ship:

Twitch

Twitch is the world's leading video platform and community for gamers.

clips.twitch.tv

 

[MCCART]

Yes. The image above is with SMT enabled, which further obfuscates utilization, but turning SMT off doesn't prevent the Windows scheduler from bouncing loads from core to core, or keep the game engine from scheduling jerky loads in the first place.

I was thinking SMT disabled AND affinity set to 4 cores. The results might not change but the measurement should be clearer. Good catch with the polling rate. I use HWMonitor when I'm watching performance. I have no idea what the polling rate is. It's not adjustable and the info on it doesn't seem to be readily available. I have HWInfo installed but it's too much work to fiddle with. Remember... lazy.

 

[FIREKNIGHT2029]

Nothing changed for me. I’m still getting the same crappy performance I was getting before the update lol. I’ve gone back to playing Horizons until they fix this game’s performance, if they ever fix it.

 

[Morbad]

I was thinking SMT disabled AND affinity set to 4 cores. The results might not change but the measurement should be clearer. Good catch with the polling rate. I use HWMonitor when I'm watching performance. I have no idea what the polling rate is. It's not adjustable and the info on it doesn't seem to be readily available. I have HWInfo installed but it's too much work to fiddle with. Remember... lazy.

HWiNFO's default polling rate is 2000ms.

This is SMT off, no manual affinity, same 100ms polling rate:

Cramming everything into four logical cores results in higher per-core CPU usage, for obvious reasons, but doesn't really change performance until I reduce it to three cores.

 

[FutronBob]

Despite this being the 2nd patch that mentioned optimizations to the glass window shaders, the problem still persists where looking through a window can drop your framerate by as much as 50% compared to looking at the exact same scene through a doorway 3 feet to the side.

Considering that virtually every concourse shop has a giant window next to its entrance, its very likely a big performance drain across the board.

Here's hoping 3rd time's the charm and they can finally nail down this issue in update 8.

 

[BadCoffee]

I had a good opinion of this update until I tried to actually play a mission. I have a performance boost in stations... and a performance catastrophic failure in settlements. Forget frames per second, in settlements I get seconds per frame. And changing the quality to low does nothing to fix it.
It's unplayable and I don't use that term lightly.

 

[cmdr_Giboulot]

I did a test for comparing day patch and day+1, already some improvement and almost as good as before

But there is no feeling of "Optimization" gain...

 

[Agony_Aunt]

I really don't understand it at all.

Its not like the game is doing something special with graphics. There are much more graphically demanding games out there that don't tank FPS like ED does.

All i can assume is FD don't have the in-house knowledge to resolve these issues.

 

[LKx]

[...]

All i can assume is FD don't have the in-house knowledge to resolve these issues.

I will wait Patch 8 before deciding if EDO is F.U.B.A.R.

 

[CMDR M Kozak]

I really don't understand it at all.

Its not like the game is doing something special with graphics. There are much more graphically demanding games out there that don't tank FPS like ED does.

All i can assume is FD don't have the in-house knowledge to resolve these issues.

Either that, or they had hoped to push this (massive) body of optimisation work in between the PC and Console launch (also explains why there was originally announced a large gap in time between launches).

This, however, also means that someone thought that PC players would not notice a 25fps average framerate. I still dont buy the story about this all being a surprise to them. Imagine the luck it takes to put together an entire QA department worth of PC's with (hopefully) differing configurations - and NOT experiencing the poor performance.

 

[Agony_Aunt]

I will wait Patch 8 before deciding if EDO is F.U.B.A.R.

Well, they did say some big performance gains coming in update 8, but time is stretching on now. I'm always willing to cut developers some slack, but its getting a bit tiresome now.

I'd probably have already moved on from ED a long time ago, but there is nothing else like it out there. So, in the meantime, playing GTA and other games.

 

[Dragsham]

No issues at all.
Runs flawlessly.
Oc 3060 vr rig.

 

[Rubbernuke]

Despite this being the 2nd patch that mentioned optimizations to the glass window shaders, the problem still persists where looking through a window can drop your framerate by as much as 50% compared to looking at the exact same scene through a doorway 3 feet to the side.

View attachment 264024View attachment 264025

Considering that virtually every concourse shop has a giant window next to its entrance, its very likely a big performance drain across the board.

Here's hoping 3rd time's the charm and they can finally nail down this issue in update 8.

Now call me crazy, but in most cases FD could remove the glass and most people would never know it, especially since you can't jump

 

[Winterwalker]

Imagine the luck it takes to put together an entire QA department worth of PC's with (hopefully) differing configurations - and NOT experiencing the poor performance.

My understanding is the game is tested with a software that emulates a range of hardware, not on a range of actual hardware. It's not impossible some weird conflict can't arise in (any, even test) software, blindsiding QA and potentially even devs, right through develoipment.

Like Dragsham, EDO runs fine on my machine too so, because it's not universal, for me it's definitely possible some change in office routines for Covid cause a QA test loophole. After release, with the pressure on fix delivery that adds, that can turn into a scramble .. and "Jurassic Park". Hopefully not and as said, I find it playable - and pretty - too.

 

[CMDR M Kozak]

My understanding is the game is tested with a software that emulates a range of hardware, not on a range of actual hardware. It's not impossible some weird conflict can't arise in (any, even test) software, blindsiding QA and potentially even devs, right through develoipment.

Like Dragsham, EDO runs fine on my machine too so, because it's not universal, for me it's definitely possible some change in office routines for Covid cause a QA test loophole. After release, with the pressure on fix delivery that adds, that can turn into a scramble .. and "Jurassic Park". Hopefully not and as said, I find it playable - and pretty - too.

Happy for you.

Update 7 was a small improvement for me, but stuttering has increased. Doubt I will use ODY for VR but load up horizons instead as long as we can.

its worth noting that, for ODY to run without me getting too motion sick, I have to run the game at a 1080p res and then upscale - even though I have a 1440p monitor. Native res stil kills the game, and causes crashes.

Avg. FPS at settlemens is still around the 45 mark, but now it can spike to 60ish from time to time. The spiking causes the stutters though so... not really a gain.