Yeah, but at least you get what I’m trying to do.
I so wish I could get my hands on some of these systems that are not working well to see if I could figure out a why.
And here I am , 2 laptops, 2 desktops, can’t reproduce it yet. Driving me mad. lol And you are worried about what temp scale I use. LOL
Well most software reports Celsius by default and most websites use that as well for temperature measurements/limits. Anywho, when I’m not zonked out on painkillers tomorrow I’ll look through this thread and see if there’s anything I can come up with myself.
FWIW, D4 is absolutely terrible with garbage collection and how it handles textures. All you have to do to trigger the texture dump/swap situation is literally go anywhere in the world, open the map with TAB and close the window and watch it stutter in all its glory. The game’s cohesive as a unit but unoptimized to a pretty staggering degree (i.e. it was rushed out like every other Blizzard game in the last five years).
Hopefully between the two of us we can spitball some plausible scenarios for moving forward.
My main computer does not do this. I might get a hint of that on my laptop, but the AMD I built tonight has some stagger. However its wimpier too. And I was forcing it to run above its capabilities.
Speaking of that system the GPU driver version that Windows 11 default loaded was 456.71.
Game requires 466. something. But I forced it to use that driver anyway. And it played. So… Interesting.
A couple people in this thread have done so already, but if everyone could post their system information it would help immensely.
We need the following:
We’re basically looking for a pattern, or at least a specific set of conditions that might trigger some of the issues found in this thread.
Is your 1080 OC an EVGA per chance?
Same thing as some others, the game worked mostly fine for me and the memory issues only started sometime later. Game seems to be running better with textures at medium for me so I’ve been leaving them there.
At first I thought I caused the memory issues because I moved my desktop recently due to house sitting for six weeks and had a whole lot of issues upon doing so. The game seems to run fine on my laptop which has lower specs than my desktop.
OS: Windows 10
CPU: AMD Ryzen 7 1700
GPU: NVIDIA GeForce RTX 2080
RAM: 32GB (G.Skill)
The only thing that changed before and after is I had to reseat a bunch of my hardware for other problems and re-set up my boot drive. I don’t remember if there was a diablo update in between but it’s likely. I didn’t update the video card (after the day 1 update) until after the issues started happening. I was also behind on a windows update but after updating nothing changed.
CPU: Ryzen 7 5800X
RAM: 32gb DDR4 3600Mhz
GPU: Gigabyte Radeon RX 7900 XT GAMING OC
Driver Version: Adrenalin 23.5.2
Windows Version: Win10 21H2 19044.3086
Got the memory leak issue since the first pre order Beta. Sometimes i can play for 3 hours straight without a single stutter but the stutter start when the full Vram capacity of my gpu is used (20GB). I’m always monitoring my vram usage when playing diablo 4, when the game starts in medium textures the game is using 5 to 6 gb of vram. After about 1 hour i’m at 9 to 10 GB used (i’ve done this test this morning) the memory used goes up faster if im porting to different parts of the world to do my nightmare dungeons and even faster for every different dungeons i do after about 3 hours played my vram was at 17GB used by the game. When i stopped playing i switch my textures quality to high and the Vram used dropped from around 18GB to 7Gb used. When you switch the quality the game freeze for the time it need to clear the vram of the gpu it goes down for 10 to 15 seconds and when its back to a normal Vram usage the game stop freezing. Its clearly a pure memory leak the game just stack data on vram but never drop it i really think you can do whatever you want you’ll not be able to do anything its all about the game. i have also checked with 10 different people from my clan with all type of hardware for low end to high and EVERYONE got the memory leak, and for everyone the vram usage drop when you change textures quality. Nothing can be done its all in blizzard hands.
CPU: intel i7 13700k
RAM: 32gb ddr5-6000
GPU: gamerock premium 4070ti, 12gb
Driver version: 535.98 (the latest one, 536.23, made the issue even worse so had to rollback)
Windows version: win 11 home, v22h2. Os build:22621.1848
In my case it randomly crashes the whole system, not just the game. (Display goes black, pc unresponsive)
Cpu temp, gpu temp around 50c
Ram usually caps at 15gb so no issue there
Vram always nearly maxed out, im guessing thats somehow the cause.
Doesnt happen outside of diablo usage…
CPU: Intel I9 7980XE
RAM: 128gb DDR4
GPU: RX 6800
Storage: Samsung 980 pro
OS: Windows 11 Pro (22621)
Settings: Everything lowest
Starting the game immediately allocates 5gb of VRAM, at the character selection screen it climbs to 8gb and finally it’s about 11gb when entering the game.
At this point it appears like the game is using the lowest mip-level of all textures and never finishes the load, regardless of how long you wait. However when travelling for a bit the game will allocate ~100mb of VRAM and then suddenly you can see the better mip-levels. This happens for each new biom / dungeon type.
After some time I can visit an area that was already loaded, and it’ll display the lowest mip-level again, allocate ~100mb after a bit of travel, etc…
This goes on forever until the 16gb of my GPU are reached. This is when the next “load” will freeze the entire computer for about a minute, frees some video memory, and the entire thingy can repeat itself.
Yesterday I created support ticket, but haven’t received a reply yet.
Yeah F blizzard man, they didn’t optimize this game for $hit. Last Blizz game I buy.
I totally understand. I would be frustrated and angry too.
Mhhh, I think rebar might be the issue for me. Disable it and now I can now enable ultra textures without immediate freezes, some slow downs when 16gb are filled, but then it actually reclaims the memory without a complete freeze. Gotta do some long term runs in order to be sure though. So perhaps some kind of bottleneck with PCIe?
If this is the case, it would appear the game is loading textures improperly as well as not triggering garbage collection. It’s known that open the world map and closing it again will introduce pretty heavy stuttering for a period of time before it settles back down again. This indicates something not happening correctly on the GPU side.
On my setup Hardware Monitor only shows a max load of 14% on the GPU bus interface (PCIe slot). That’s ridiculously low for how much D4 is having to shuttle back and forth and would definitely cause problems. That’s 14% on a PCIe 3.0 x16 slot, or less than 2 GB/sec. If there is some sort of bottlenecking going on at the PCIe bus level that indicates the game is not utilizing transfer paths correctly and this needs to be addressed.
Resizable Bar (ReBAR) works by using available system RAM as VRAM. Problem is, the game has to properly utilize this feature. My setup is too old for ReBAR to function, which indicates why I can’t reproduce the “leak” scenario. With ReBAR active, if the game is not properly utilizing that feature it can cause an overflow situation which would BSOD the machine. And as to the rest, see above in this post regarding PCIe utilization. If it’s only utilizing 14% of a full PCIe x16 3.0 connection, that will severely impact performance on much more powerful equipment that expects as much bandwidth as possible.
Erm, no. Always keep in mind that you may observe similar symptoms for different problem. I was talking about the behavior I described above where the entire system basically just dropped down to the frames per minute. And in my case it made perfect sense to check for rebar.
You see, in windows all graphic APIs must “extend” video memory with normal memory and that goes way back because CPU & GPU basically need to be synchronized. Rebar just allows the graphics API (like Direct3D, Vulkan or w/e) to optimize the upload process by using a big chunk and skip a massive of copying, fences, etc… Just check out the nvidia’s article on GPU upload heaps to get all the glorious details 
Anyways, the problem now occurs because your GPU is basically locked while the upload is happening. So let’s assume that upload takes really long because you are dealing with 2.5 ghz CPU … well I guess you can see where the problem is coming from, right?
And again, this was the solution for my problem (hopefully), other folks are more likely to check for different things.
Btw, never received a response on my ticket 
Certainly, they are facing significant issues. One prominent problem is the occurrence of a memory leak, which becomes immediately evident when opening the map, resulting in a sudden drop in frames per second (30-60 fps). It seems they have implemented an “event tick” mechanism that continuously loads unnecessary data into VRAM when interacting with activities on the map. There is absolutely no reason to perpetually load these activities without limit, as it is a fundamental mistake in coding basic user interfaces. Similar performance issues arise when opening the inventory on vendors, where streaming textures and unloading them unexpectedly cause severe stuttering, even on high-end machines capable of running Cyberpunk 2077 smoothly at over 100 fps in 4K. Furthermore, there are problems specifically with DLSS 3.0 itself, as it fails to deliver a consistently smooth gaming experience, even when downscaling. It is possible that these issues are exacerbated by compatibility concerns between the game and AMD CPUs.
I find that curious. My system does not act that way when opening the map, in fact none of them do. Instead they actually drop in power consumption, heat, and utilization completely.
I would be more inclined to think it would cause an FPS drop to switch from the map, rather than to it. But I get some of what you are saying. Its just not reacting that way on my systems that I have been testing things on the past two weeks.
I will have to examine this more, and see if perhaps I’m missing the drop. Right now, I only see FPS changes when entering/leaving cellars/dungeons, teleporting, but they are brief and last only a sec or 2.
I should record a session of game play and share it, but the problem is the recording makes my game play look a bit jittery, as I play at 144FPS and it records at like 30 or 60 or something. lol But mine is butter smooth all the time.
Now on the laptop it has a bit less smoothness, but its not bad.
My dad’s HP Omen with the 2060 is a bit worse, but still not bad.
The AMD I built last night with a 2070 that I ran at full Ultra for an hour… Now that showed it struggling to keep up with the Ultra settings. But it was consistent and did not crash, or go crazy. Just struggled. Even once the VRAM maxed out, it just kept playing. And it was even running an outdated driver version that Windows 11 auto installed. (456.71, which is below requirements for D4) But it plays very nicely at High with a cap of 60FPS.
What happens if you turn the DLSS off completely? Just wondering.
My GPU usage dropped 60% by turning off DLAA.
Why are you running your FPS so damn high?
No wonder you are seeing the drops. Cap that down and control the amount of hit to the GPU and it would probably stop doing that.
Cap it to about 5FPS above your refresh rate if you can. This is Diablo, not Cod.
Mine has a refresh of 144, so I cap it to 150, for both foreground and background.
Lowers the load on the GPU, runs it cooler and I don’t get those hard spikes in the frame rate.
It appears to be essentially identical, but I believe there are multiple issues at play here. Primarily, the user interface itself seems to be causing unintended actions, and the game’s texture handling is also problematic. Typically, these issues are resolved gradually as we progress.