Specs? I'm on a i5-7500, 16 GB, 1060 Ti 6 GB and although performance is lackluster, my game doesn't do that.
Hows performance after patch 11 ?
- Thread starter [VR] Garfs
- Start date
heat soak can cause gradual decline. As can issues with local graphics library/drivers. your gpu and/or cpu will not hit quite as high boost clockrates when heatsoaked and if you're not hitting a framerate cap, that'll definitely be noticeable.
though, before you go chasing local gremlins... be sure that you're comparing re-entering the same scene, leaving the instance and coming back and just doing that over time with nothing else in between. It's been shown in the past that simply viewing certain menus can alter the framerate of a scene after the menu is closed. Or viewing certain scenes with a given shader or looking at a building and then not, may cause the framerate to change even when no longer viewing that scene/object/effect.
though, before you go chasing local gremlins... be sure that you're comparing re-entering the same scene, leaving the instance and coming back and just doing that over time with nothing else in between. It's been shown in the past that simply viewing certain menus can alter the framerate of a scene after the menu is closed. Or viewing certain scenes with a given shader or looking at a building and then not, may cause the framerate to change even when no longer viewing that scene/object/effect.
2070, 16gb, i7-9750H. It's a laptop, mind you.
It only happens after playing for several hours, but wouldn't heat be basically at its max after ~30 minutes?
The only time I have experienced something like this is while looking at the GalMap. I was in the GalMap, in the trading UI, and it just kept getting worse to the point that I was all the way down to 21fps. I hadn't been running the game that long but something had clearly gone wrong with either a hung subroutine or infinite loop or memory leak or something. It was only in the GalMap, though. Framerate would recover outside of it. Haven't seen it since.
Yeah, as mentioned, that's probably heat soak.
Would heat soak go away immediately, just by turning the game off and on again? Honestly I've got no idea what that phrase means.
VRM quality/temps have rarely been an issue on AM4...the CPUs cannot pull that much current, and even mid-range ITX boards are completely overbuilt. This ASRock board can deliver 600A to the CPU with the power stages at 125C. I've never seen the VRM temp over 71C (and that took ambient temps that threatened me with heatstroke) and can't get any AM4 CPU to pull more than ~180A before becoming thermally limited at the CPU itself, even with a radiator that is larger than my case. You need subambient cooling to push an AM4 CPU to levels were even a mid-range ITX board is going to become power delivery limited.
Even my $80 board, which was the absolute cheapest board that I could find that had a passable VRM, is well within safe limits while loaded 24/7 with the overclocked 3900X that's in it.
LGA-1700 is a different matter, which is why the price floor for usable boards is more around the 150-200 mark, or a bit higher for ITX stuff, but again, power delivery has come a long way and once you get out of the trash-tier stuff, you have very little to worry about for practical daily OCs where the CPU isn't going to be pulling more than 300w or so.
I'm not advocating skimping out on board at all, just buying the most sensible hardware for the task at hand. Stuff that will be unconditionally stable, reliable over the long term, and not a hindrance to performance or most overclocking that doesn't involve record setting attempts. The vast majority of of people that are spending more than 250 dollars on a motherboard are ripping themselves off wholesale; there are plenty of cases where putting the money elsewhere, even memory, would be a better decision.
The situations where memory helps generally aren't situations where GPU helps. I need fast memory because I have (almost) exactly the GPUs I wanted.
There is no GPU upgrade that will appreciably improve my CZ experience, because I am almost never GPU limited in CZs. I can jack up resolution further, but if I had a GPU (which don't exist yet) that could handle that at the frame rate I'm looking for, I'd be CPU/memory limited again.
Likewise, I am using the, or nearly the, fastest gaming CPU that currently exists for the platform I am on. The game doesn't need more than 8c/16t. Cores and cache on a separate Zen 3 CCX don't really help. Boost clocks with a 5900X or 5950X would not be significantly higher. The 5800X3D is about the only potential upgrade path (because it further masks memory latency with a triple size L3), and that's still a month away. My only realistic avenue was faster memory.
To make a better informed decision about my upcoming CPU upgrade. I'm looking at the 5800X3D, which has a huge cache, but is limited to fairly conservative clock speeds, or a 5950X, which is mostly the same deal, gaming wise, as what I have now, but would at least allow me to do things like transcode with better quality on the fly and give me better performance in non-gaming tasks.
Regardless, I've already indirectly measured hit rate by testing memory performance. If cache hit rate was good, main memory performance wouldn't matter this much.
I haven't had any issues with the game performance degrading over time on any of my systems, but I can't rule out some kind of memory management issue with game.
you'd think a laptop which usually has insufficient cooling would hit temperature equilibrium pretty quick when maxed out. But the cpu and the gpu both have dynamic clockrates. And the voltage regulators also have switching rates that are temperature managed. These things all come together and reach different temperature equilibriums at different times if at all. You may hit it fast in your gpu, but your cpu may take longer or vice versa. Or perhaps it's not the core of the cpu that hinders its ability to hit a certain clock rate but the vrm and other power components have to be dialed back because they're getting too hot, and this in turn limits the clockrate the cpu is allowed to be set to.
only way to be sure is to test it with a sufficiently intense benchmark (or alternative game) that taxes the system roughly as much as elite seems to be that you can run on a loop and see if a similar behavior develops or not.
with the bugs odyssey has (simply viewing menus or looking at a building etc can alter the framerate even when no longer doing those things), i wouldn't take it's behavior at face value without a second opinion test.
heatsoak is when the temperature of the system has allowed everything that matters (dynamic performance components) to get hot enough to start limiting boost performance. This happens at different rates for different components and cooling systems and loads where the cooling becomes sufficient to reduce temps to below max at times.
(edit, Some cooling sytems can keep a system out of a heatsoak state entirely...where temperature equilibrium is lower than the temp where any throttling occurs and some loads on the system are not consistent and could allow periods where the cooling system drops temps temporarily. However avoiding heatsoak is very unlikely with modern 7nm or smaller components. They're just too small to keep from eventually throttling when under constant load for long periods )
Last edited:
5000 series also can do fun things when paired with amd graphics cards. What do you have that's equiv to a 5950x though? Cuz anything short of another 5000 or 4000 series amd cpu would likely be a fairly noticeable upgrade - unless you are coming from intel land. Tests seem to show it would be a 25% improvement over my 3900x . That's nothing to scoff at.
I'm not really sold on the x3d tech yet. I'm worried we're going to hear more about the inability to cool the cache of that cpu and it's gonna come with some significant compromises for many real world use cases.
This is mostly a myth. Re-sizable BAR helps AMD graphics cards about the same irrespective of platform.
That said, in the system in question, I'm running a 2.6GHz RX 6800 XT with...
a well-tuned 5800X.
I'm aware of the improvement over the 3900X: https://forums.frontier.co.uk/search/580008/?q=3900X&c[users]=Morbad&o=date
The compromises are worked into the voltage caps, the lack of overclocking support (though the exact details of this are not known), and the lower peak clock speeds.
EDO does seem to be relatively memory performance dependent, and I have other games that are likely in the same boat, so I'm significantly interested in it as a drop-in upgrade for this box, which is gaming oriented (I have plenty of other computers).
yeah dude you mised a LOT of progress on optimization. it runs pretty well. in space at least it runs as one would expect going from horizons to odyssey. as in your going from a game that requires less horsepower to a dlc that requires more horsepower.
Example of what i mean:
bad= Hoz:200 fps to ody:20fps
expected= Hoz:200fps to ody:130-140fps
I noticed degradation of performance after hours of play too.
A game restart is fixing it (quit to desktop type of restart).
Most notably it happens during on-foot CZ after chaining CZ after CZ and after longer session of mining in rings.
If i dare to make a connection - i'd say it could be related to Night Vision.
I use NV almost non-stop if im in a CZ at night or when mining in a ring that's in the shadows.
But my gaming sessions are usually at least 2 hours. Sometimes way more than that
I'll pay more attention during longer sessions to see if I can reproduce anything.
Why upgrade when you can increase your FPS with this simple trick.
well true, they did add it to 3000 series and later cpus ...but i think it's still limited to 6000 series gpus. my 5700xt is not given smart access memory powers.
That's what we know about the short term immediately aware compromises. It's unknown what the longevity impact of the heat between the vertical interconnects are going to be like. The rolling back of OC support makes people very skeptical that a real concern has been identified that calls lifespan into question.
if you're chasing the individual percentage performance increases that much by upgrading a 5800x to a slightly higher end 5000 series cpu, why not just invest the money in a more hardcore cooling system that would improve gaming performance far more than just bumping core frequency up a couple hundred Mhz or adding some cache at the expense of those Mhz?
Put the money into a portable air conditioner that directly chills mineral oil that you've submerged your motherboard and cpu in and gpu in. One rated for the heat you'll be dealing with is about 200 bucks and infinitely more available in stock than cpus tend to be. Plus you can use the system in future hardware upgrades when the arch improvements warrant them.
that's what i'd suggest if you need the extra handful of frames from odyssey. that or waiting for 6000 series cpus that will offer ddr5. It's your cash and all but if you're intent on spending it, the oil submersion option is way more fun (and more effective) than a barely incremental increase in gaming perf from a 5800x to a 5950 or whatever. do eeeeeet.
What exactly is this?
Oh, there's a new bug since U11 (or the hotfix) where the vendors haven't spawned in when you first disembark (relog fixes it). Curiously though, my FPS goes up from 45 to 60 when they're absent. Rather large hit considering all the other NPCs are still there.
That's almost implicit in, and hopefully addressed by, the voltage caps and lack of OC support.
I'm not terribly concerned about early failure while the part is operating near spec. At ~1.5v peak maybe I would have been, but not ~1.35v.
The 5800X3D is apparently good for ~15% more performance in more memory dependent games, which EDO and some of my others likely are. Even 10% is a big difference when one is on the edge of smoothness.
I'm already into diminishing returns on CPU clock and another 200MHz I could get with better cooling would only be worth a few percent at this point. There are also no substantial cooling upgrades that would fit within my space and portability constraints. Even if this were a full sized and stationary system, I can think of much better ways than oil immersion cooling, but they still wouldn't allow boosting past 5.05GHz on a 5800X without being subambient.
I also have no intention of replacing this motherboard any time soon, so I'm limited to the constraints of the platform on hand. The last low hanging fruit without swapping out the CPU was the memory I just installed.
I haven't seen this yet. Vendors have always been present for me without a relog.
Last edited:
So far I've only observed it in surface ports and it happens when I get there with my own ship, Apex or Frontline. I've also found a single settlement which occasionally fails to spawn any NPCs either.
Anyone know of any currently active high-intensity CZs at large (++) settlements?
Military, mining, or labs would be ideal. I have enough examples of small and medium, but I've been having trouble finding this combination.
Still trying to get a more complete picture of the possible CPU (or memory) limited performance reduction at settlements with U11.
Military, mining, or labs would be ideal. I have enough examples of small and medium, but I've been having trouble finding this combination.
Still trying to get a more complete picture of the possible CPU (or memory) limited performance reduction at settlements with U11.
I've had this happen to me frequently when I've left Odyssey running for many hours (4-6+):
WARNING: VIDEO CONTAINS RAPIDLY CHANGING COLORS AND FLASHING / FLICKERING LIGHTS.
This has happened even in Update 11: the first incident was when I was docked at a station in my Courier while waiting for Mission Board refreshes over the course of several hours for Influence farming. The second happened a few days ago, when the right-hand panel on my mining Cutter bugged out after around 2 or so hours of mining, with each tab drawing over the previous one and resulting in an unreadable mess that only cleared on a full shutdown.
There's definitely something strange going on...