Topic on User talk:Aemony

D0x360 (talkcontribs)

There is plenty of evidence you can check for yourself. I tested it on 3 systems, 2 Intel and 1 ryzen. All had 32 gigs of system ram. Over the course of 3 hours on every single run of the game it would go from using 12% up to 68% and im 1 case it was at 84%(of 32 gigs). All memory would then be released the second the game was either quit or forced to end via task manager.

The information about increasing the page file size during shader compilation is further evidence of this. You shouldn't even need a page file with 32 gigs of system memory if all you are doing is playing a game.

I am a programmer, I know a memory leak when I see one and that aside the forums are full of people describing memory leaks without even knowing it.

Aemony (talkcontribs)

As mentioned in the change, the reports are inconsistent enough to be unreliable and the inclusion of it in the section has no real informative value other than as unnecessary speculation.

Saying that it is potentially caused by a memory leak helps nobody -- especially not when reports are widely different with users of 4 GB of VRAM and 8 GB RAM reporting minor number of crashes and people with 64 GB of RAM reporting multiple crashes.

That whole "High memory usage" section in itself is technically also problematic as "unexpectedly high [...] usage" is a subjective statement to begin with when there is no official baseline to compare again. Further on, the references listed is filled with counteracting reports further questioning its inclusion on the wiki page. But so far I have looked past it while at least trimming the more speculative parts of the section while waiting for further development on how to proceed.

Further on, based on testing by myself as well as, again, reports from other users it is not even guaranteed that the crashes the game experiences is caused by the memory usage, and we have yet to actually see sufficiently reliable reports or findings that actually suggests the memory usage is an issue as well as what is might cause.

Like, take my last session just now where I played the game for more than two hours -- no crashes, no issues, only about 3 GB of RAM used towards the end of the session.

It's basically speculations everywhere combined with clueless users whom might not even know what they're looking at and so blames "unexpectedly high" (what does that even mean without a baseline?) RAM usage when crashes caused by unexpected D3D12 results causes the game to "crash". Like really, you can't even manipulate the number of back buffers used by the swapchain without the game going "Oh my god, this is unexpected! I believe I have crashed!" when in reality it would function just fine. It's more than likely that half the crashes experienced by users is due to the game's fucked up D3D12 usage that treats any minor unexpected result as a "crash".

Random tangent over -- What will most likely happen is that when further evidence or confirmed reports arise we'll remove that whole "high memory usage" section and throw it under a general-purpose "the game crashes" section instead. But we're not there yet, and so we need to minimize the speculation present in the PCGW article.

On another note, I'm still waiting for a reply from you on Rose's talk page in regards to in what way the game isn't "proper HDR".

On a third note, the talk page of the article is for these kinds of discussions -- not individual editors/administrators talk pages.

D0x360 (talkcontribs)

I'm new to this but I will have it down soon enough. Having some health issues that are slowing me down. Anyways... I confirmed the memory leak on 3 different systems (all with different hardware and tested with multiple driver revisions on a fresh install of win10) using tools designed specifically to look for them. Also the fact that every system showed over 70% memory use after 5 hours on a system with 64 gigs of memory and no other major tasks running is good reason to investigate hence the use of tools and why I'm so adamant about it. I was a programmer and behavior is consistent with unreleased memory. I agree it's inconsistent but likely not as inconsistent as you seem to believe. It's the leak that causes attempts to read or write to non existent address space causing a crash. It also causes performance issues because despite already loading data it needs into memory the leak makes it invisible to the game (at times) so it has to load the same data twice in a row.

The point of bringing it up is to draw attention so hopefully the developer fixes it. Otherwise crashes will continue. Also if users know about it then they can restart the game every hour or so to help avoid crashes and/or performance issues.

True HDR... It's determined by the measurements of contrast, nits and color. This game doesn't meet the criteria for HDR because it's color space is 8 bit and their are some major inaccuracies with said color (at times) which isn't the main reason. The main and defining reason is nits. It doesn't meet the criteria for the HDR spec.

Aemony (talkcontribs)

> The point of bringing it up is to draw attention so hopefully the developer fixes it.

Developers do not visit nor use PCGamingWiki as a resource. Guerrilla Games are basically only active on the subreddit and gets their information from the community there, as well as the multitude crash reports that users sends to them.

This isn't a place to throw darts on the board hoping against all wishes that *someone* might eventually see it. It's for confirmed and reproducible situations with valuable information for end users -- which speculations aren't.

And again, there's very little correlation between the crashes and memory usage, and your findings on 3 systems does not match the reports of multiple others. Right now the *vast* majority of crashes seems to be related to D3D12 calls as well, which this speculation further doesn't help with.

> It's the leak that causes attempts to read or write to non existent address space causing a crash.

Patch v1.02 fixed a crash related to memory incorrectly being overwritten.

And as for non-existent address space -- we already cover this under the "Seldom crashes due to game trying to write to nonexistent memory address" section.

> The main and defining reason is nits. It doesn't meet the criteria for the HDR spec.

Again though, by what specs are you measuring? The HDR visualization I provided in the other thread shows nits around or above 4000 (the purple areas).

Just saying "it doesn't meet the criteria" when the game makes use of HDR, makes use of *thousands* of nits, and you don't provide any actual example of said "major inaccuracies" doesn't make it HDR.

PCGamingWiki is not a site where individual users inject their personal belief into the articles. If 8-bit HDR is how Guerrilla Games have decided to provide HDR on PC, then that's how they've decided it as. *At most* what we can do on the PCGW article is set the HDR state as limited instead of true with a minor note about it, but we shouldn't outright state it as false when by most accounts it's still HDR -- just not the more preferred 10-bit HDR.

D0x360 (talkcontribs)

Im not just injecting my personal beliefs but I will concede on the HDR issue since it's only 8 bit and that's mostly nvidias fault.

As for the memory leak.. yes they fixed a memory bug during shader compilation however they still have not fixed the memory leak that occurs during gameplay.

Also the point about informing people in hopes that it helps get the issue fixed... I don't expect devs to read this site (although they definitely do, I've asked many) and it's not throwing darts to inform users of a significant bug. The benefit of people knowing something exist is they can talk about it and when people talk about certain bugs they tend to get focused on.

I don't see how informing people of a bug is a bad thing. It is confirmed and easily reproduced. You can check yourself with tools that literally check a program for a memory leak. You can also just see it happening as time goes on. There is no reason you would see 60+% memory use on a system with 32+ gigs of RAM. The latest version of the game used 37% of 128 gigs of system memory after 4 hours of running and also was flagged positive for a memory leak.

I'm not just some guy arguing because I think I'm right. I used to work for Sega of America and Epic. I have a degree in computer science, I am in charge of IT for a multi million dollar company called DEKA.. I am speaking as a professional not as a gamer.

D0x360 (talkcontribs)

Had to make an edit to my last post that completely changes the meaning. Auto correct removed the word "NOT" in the first sentence.

Although since nobody replied I guess the issue about the memory leak is closed. Although I tested the game yesterday and the memory leak is still present. That was a tool based check then I let the game run for 6 hours of a system with 128 gigs of system ram and when I checked in using monitoring tools there was 86% of system ram in use by absolutely nothing which is in line with a memory leak. Address space is reserved but not properly released. It doesn't get attributed to any application but it can't be used due to the error during release.

Reply to "HZD Memory leak"