VRAM leak over time?

Again if you read that thread, you’ll see that myself and others are able to monitor the VRAM increases and can time the crashes almost to the second.

Meaning, I can see that every time I teleport from one location to another, VRAM usage jumps by 2-5gb. Every time you idle with your inventory open, VRAM increases.

This happens every time, until eventually you reach your cap, whatever that is, and you crash.

I have a 4090, so not having enough VRAM isn’t the problem.

Someone also discovered that if you change your texture settings, whether that be an increase or decrease (doesn’t matter), the VRAM resets back to it’s proper value, then begins its climb again.

This means you can extend the duration you can play without crashing by switching between different texture settings.

It’s pretty clear, you can monitor it, time it, and it’s 100% consistent.

3 Likes

Guys it’s just another badly optimized console port… haven’t you gotten used to it yet? :stuck_out_tongue:

Yeah that’s not really a discovery. You’re just forcing it to flush the whole texture pool and then it loads in what’s around you; which would be relatively little. As soon as you move around to new places and see new things, it’s going to add more to the pool.

See:

If disabling those features doesn’t fix your problem, it means you’ve got something else weird going on with your OS environment. Might be software or driver related.

And I know plenty of other people playing with 4080s and 90s that are not experiencing any sort of problems. If this problem was solely the game’s fault, it would affect everyone with the same hardware, but it’s not. This points toward something specific going on with your machine and others with similar software/driver configurations.

Could i ask, what shader and texture setting you use?

Unfortunately, I think I suffer the same issue here. I have a 4090 gpu, and with high resolution settings I can play for hours without any problem. As soon as I set the textures to ultra, the game crashes after 10-15min. I’ve tried everything, ran VRAM tests on my GPU (with OCCT, everything looks fine). Its a pity cos there is a huge graphical quality gap between the high and ultra setting. At least as I said, in “high-textures” I have 0 issues.

I don’t use DLSS or Reflex on D4.

Unlikely to be OS or Driver related, as I’ve clean installed, but obviously it could be.

I never argued it was “solely” caused by anything.

It is likely affecting certain intel CPUs, based on the thousands of posts I’ve read on this exact issue since it began after the 1.0.3a hotfix.

There was a similar issue with Apex Legends for years, to do with WHEA Event 19 warnings.

To quote someone in another thread:
Apex: Legends is one example of a game that had to have its code tweaked after this issue was discovered, because the game would crash over and over on those CPUs until it was fixed. Minecraft is also known to trigger it. It’s something related to how those CPUs handle garbage collection if I recall correctly but I’m not an expert. Some people got around it by increasing voltage to combat v-droop on their CPU under load but maybe you got around it by reducing your frequency.

They fixed the issue on their end. That does not mean the problem was “solely” caused by them.

I’m not sure why you’re so desperate to point the finger away from the game and developers and onto us as consumers.

I’m a software engineer and security expert, and I know what I’m doing on my machines.

I also know, when I play dozens of games for many hundreds of hours, including Diablo 4, and then they add a hotfix to patch a crashing bug people were having, and it causes mine and thousands of others games to crash, that the two are likely connected.

And once again I will say, if you did your research and actually read through the posts and threads like I have, you would know that this issue has started happening on Xbox and PS5 as well, since the latest hotfix.

So, your argument somewhat falls flat, based on that alone, unless you want to start arguing that the hardware or software setups of people’s consoles are to blame.

5 Likes

no wonder i keep getting random crashes on my ancient 970gtx.

I kept a close eye on my VRAM usage on my last session by using AMD’s logging function. My finding is that it’s not reaching the limit that causes issues, rather it’s being at the limit and going to an intense area, such as a town. My layman hypothesis then is that while VRAM is recycled to some degree, it’s not done quickly enough for every situation. In a town the game tries to recycle too much at once but fails, and chokes.

Approximate VRAM usage in GB. Capacity is 12. Texture quality: High.
Gameplay is perfectly smooth unless noted otherwise

Before launching the game: 1.3
Character selection: 6.9
After loading into the game (town): 8.8
Five minutes after launch, after teleporting to a dungeon: 11.5
Running a couple of dungeons back to back: fluctuating between 11.7 - 12
Thirty minutes after launch, after teleporting to town: 12. Unplayable intermittent freezes

Windows 10 22H2
AMD Ryzen 7 5800X
32GB RAM
Radeon RX 6700 XT 12GB, Driver: 23.5.2

2 Likes

From only a couple hundred people, out of millions of players.

It’s actually highly likely, seeing how it’s only affecting a small percentage of players.

CPUs don’t handle garbage collection, they blindly run code given to them. It’s on the software to handle the management of the memory and then feed those instructions to the CPU for it to run.

Yeah that’s what everyone says. If that’s the case, then you’d know that the games don’t actually directly handle memory, Windows does. When a game sets things like pointers and ranges to specific addresses, those don’t actually line up with the true physical ranges. Windows virtualizes those ranges and plays middleman for everything that needs doing(even moreso if hyper-v is on). If there are other things running, they can interfere with the memory management. Especially if they are kernal level, like drivers or security softwares.

Hard crashing, due to things like errors from specific deadlocks or infinite loops, are a different story. And even still, it’s not happening for very many people, out of the millions playing on them, because they are likely caused by perfect storm conditions. If there were something really broken with the game client, consoles would all feel it especially so; since they are mostly all running the same hardware between respective platforms. Like if it’s an Xbox Series X, they are all going to pretty much be identical, PS5 to PS5, same thing.

1 Like

Mine is set to high settings, and nothing customized under it.

I have ran Ultra, and I have tried 4k. But it plays those settings ok too.

1 Like

Replacing my previous reply with a far more succinct one to see if I can understand why you are here and why you have a history of arguing these points in this way.

  1. Do you think this issue myself and others are having was caused by the 1.0.3A hotfix?

  2. Do you think if I was able to roll back to the previous update, my errors would go away?

I will expand a bit on this myself, even though this isn’t aimed at me.

I wouldn’t say it was caused, considering this thread has been here for a while now. We have a VRAM use issue. And we have no idea what the hotfix was actually trying to fix. They haven’t specifically said.

Since that isn’t even possible, I have no idea. What I do know, is that a few players reported improvement after patches, and a lot more players reported same or worse results.

So, its still a mixed bag. I think they are guessing with the hotfixes, or are targeting specific things that they neglected to tell us.

*shrug

1 Like

There are two groups when it comes to this issue.

The first group had this issue for a long time and Blizzard addressed it with hotfix 1.0.3A.

Then you have people like myself who had no issues with Diablo 4 through the beta and slam, and with the main game launch. However, after hotfix 1.0.3A the game crashes every 10-30 minutes

Perhaps you should be using this thread then:

-[Main Thread] Game Crashing specifically after patch 1.0.3 (June 27th)

1.0.3a build #42936 (PC) - June 30, 2023
- Addressed multiple instances where the Diablo IV client would crash in game

And this was actioned off the back of a mega thread, where people were complaining about the crashing and VRAM usage

My issue is caused by a VRAM leak.

As I’ve said, I can track and time the crashes.

seems impossible to play now… crashes every 30mins or so.

Whats strange is how once the freeze and then follow up tank in performance happens, switching the game to window mode instantly fixes it and never has the problem.

So whatever changes with the game running in fullscreen has something to do with it. Hopefully someone can figure out what it is.

My spec are 5800x3d - 32gb ram - rx6800 16gb vram - nvme ssd

edit: o btw my game doesn’t actually crash the performance just drops to unplayable.

1 Like

rtx3090+latest drivers, vram usage stay around 20-20.5GB no crash

This game has crashed/restarted my PC at least 6 dozen times since release 1 month ago. The amount of times I have been forcefully knocked out of a dungeon, dealt with rubber banding, heard my fan whir like a jet engine over time, etc is simply insane. They never fixed the VRAM leak and it still persists to this day. Hotfix 1.0.3A did nothing to improve this problem. This doesn’t happen with any other game. This doesn’t happen with any other program. It is only D4. It is not the OS, it is not faulty hardware. It is D4 that is the problem and I’m frankly getting really sick of this. I’ve got around 225 hours of gameplay currently despite these crashes and I can sometimes have the game active for 10 hours before a crash, sometimes only 3 hours, sometimes 1 hour, sometimes it can even last for over 24 hours. High intensity grind sessions of jumping in and out of town/dungeons seem to be what kills the game the quickest. When I was exploring open world my game times lasted longer. When I reached end game and started heavily grinding NM dungeons it became much more frequent.