RAM Management - Serious Question

[I give up 152]

Mate, with 8GB you cant play (maybe with luck you can load Stratis (with settings in High) but in Altis is not possible) having pagefile disabled.

I've played for quite some time with i7 930 having 6GB of RAM and never was able to play with pagefile disabled, I am perfectly aware about how the pagefile operates and how it is being used in this game and also how the load of it heavily dependent (above everything) of graphics quality (textures).

 

Also, in these situations when we do not have physical RAM enough (with this game or any other application) and the system needs more memory and it starts to use pagefile, it helps alot having a fast Hard Disk (SSD or RAMDISK) and it helps more than having a fast RAM, whatever maybe.

 

Despite of what some are saying por may say, Pagefile is used (and managed) by the system as virtual memory for when your physical memory is not enough for the demands of the application and in most cases (which is the case of ARMA 3), the real load and use of the pagefile starts when your physical memory is about to reach the limit (which is 1 GB bellow the max RAM installed (average)). Here the user does not have interference its all managed by the system.

 

Why (when we do not have physical RAM enough and we need pagefile (virtual memory)) it helps having the fastest Hard Disk that we can?

As probably some may know pagefile (and virtual memory) is nothing more than some free space of your Hard Disk that the system is allocating to use as memory.

 

If after some time your system (pagefile) starts to use the DIsk as if actually was physical RAM, I guess it is easy to understand why is important the Hard Disk performance (I guess thats why the guy said that he is having pagefile on  RAMDISK).

 

Basically there are 2 ways in how we can, let's say, control Pagefile in matters of size.

 

We can just let it managed by the system and the system will use it according with the needs. In this case we will lose free space in Hard Disk according to pagefile usage, Of at any point pagefile is having a usage of 6 GB (lets say), it just takes 6 GB of free space in our Hard Disk that we cannot use for anything else because is not free anymore. Soon you close the application that was using pagefile we have the space used for pagefile in Hard Disk, free again.

 

Or we can set pagefile with defined/limited size. in this case when we set it with lose automatically (as free space) the same amount (GB) that are setting for pagefile. Pagefile limited/defined for 10 GB means less 10 GB of free space in our Hard Disk, whatever the circumstance.

 

Now, when we have physical RAM enough there is no valid reason for Pagefile use.

No matter where Pagefile can or may be allocated (SSD,RAMDISK or neighbour closet), physical RAM is always faster and provide better performance, even the slowest one.

 

This take us back the reason of the thread. RAM USAGE.

 

And where is being used? 

I believe that I've said this a couple of times, but I'll repeat.

 

We know that the RAM is being used because we do not have available anymore, we know it was loaded with game usage (that's why the video(s)), we know where and how was loaded in the game(that's why the video(s)), but which process, application, resource or thread is using it? That's the question and the reason why I have started this thread.

 

There are no traces of where is being used, NADA, NIENTE, ZERO, NOTHING. 

The RAM is in use and we know what application has loaded and is using it, we know that we do not have it available anymore (as free RAM in our system), our System tells that the RAM is in use, but according to Windows is not being used by any resource, thread, process or whatever. 

The only process related with ARMA 3 (arma3.exe) is using about 2GB, which is far from the 12 GBs loaded and in use. All the other process or resources are using the average RAM, as usual/normal and all together, Steam, MSI Afterburner, Graphics Driver, Sound Driver and System processes are using around 1 GB. Where is being used the remaining RAM? Anyone knows? I dont.

 

Hope this time the people around can understand, (I am using the best translator that I could find), but I can learn English, if needed.

 

Btw, this was my guess at this time.

https://forums.bistudio.com/topic/180882-arma-3-memory-leak/

[St. Jimmy 272]

Btw, this was my guess at this time.

https://forums.bistudio.com/topic/180882-arma-3-memory-leak/

So this doesn't happen with singe GPU but only with multiple GPU? Now I might understand bit more what you've been talking about earlier in some threads about GPU memory and stuff. Also that could explain why people don't usually recommend using more than a one GPU in Arma. That memory leak can put many system on their knees because people don't usually have more than 8GB of RAM. Would be interesting to hear if this is AMD only issue or does this happen with Nvidia.

How many GPUs you're using in those tests and does the amount of memory used differ between 2 and 3 GPUs for example? So maybe 3 GPUs somehow triples the memory usage or something along those lines.

[I give up 152]

Yes, the behaviour is not the same under Nvidia or AMD, but is not related with multi gpu configurations (as I thought at that time).

Nvidia loads much less RAM and I suspect it is because of the exclusive feature from Nvidia named "Shader Cache"., Still not sure if is this one the reason.

[I give up 152]

Also, and this is merely my opinion.

32 bit breaking barrier was indeed an awesome feature, at that time.

Mainly because, at that time, most of computers were running with 4 GB of RAM (or lower) and in fact was a valuable way to surpass the physical RAM constraints and also because indeed it helped and allowed those machines to use much more RAM (for textures cache) than the one reserved in physical or virtual memory, for a 32 bit application. It really helped and I am grateful.

But in these days, 4 GB of RAM is something obsolete, maybe is time to rethink the 32 bit breaking barrier (of virtual and physical memory) and make it more appropriate for the current hardware configurations and rethink how the memory (cache) is being managed in this game.

[I give up 152]

sry double post.

[x3kj 1247]

Yes it loads alot of ram... so what? Why should it not?  Show us the actual performance impact.

[I give up 152]

Yes it loads alot of ram... so what? Why should it not?  Show us the actual performance impact.

One more expert.

First,

You show me any other game where this insane amount of memory is being cached and to make your task simple you can use as example even the x64 games.

[Freghar 73]

And didn't you say you're not using page file? Well then you should get load in your RAM. It has been pointed many times in these years that if you've low or disabled page file and you don't have enough RAM, the game crashes because of "out of memory". The interesting part in here is that you seem to actually maybe see the RAM load, when usually nobody hasn't seen any big RAM loads. But usually those issues has been posted by people that have 8GB or less and you can get the RAM load to 6GB pretty easily so I guess there isn't "enough free room".

The crazy part is that this shouldn't be needed. Memory-mapped files are a simple concept that has nothing to do with pagefiles - actually, a pagefile is a memory-mapped file, by the kernel. :) The concept itself is really simple - you just let the page fault handling routine do all the work. That's why the relation to pagefile is baffling as, traditionally, memory-mapped files cannot be "paged out" to pagefile as they're at the same "level".

I mean - it would make sense with anonymous (not file-backed) memory - if the system was low on RAM, the memory could be paged out to make room for other memory-mapped content. This is in fact what might be happening with multiple GPUs and not enough RAM.

As I try to understand memory management on Windows more, I find it pretty complex and somewhat stupid, compared to Linux memory management. Windows doesn't even have a decent page (disk) cache, you have to use external software like this to get the basic functionality. It also apparently divides memory (even with win7+) into multiple non-overlapping areas for fs cache and "programs" at least, instead of sharing the space.

On top of that, it seems to have poor page eviction policies (that are impossible to tune?), paging out running programs to make room for more cache, to the point that a separate DynCache service is apparently needed to stop this madness.

I would like to properly answer some of what Arma does, but I really can't as Windows memory management is just freaking weird. And not documented!

My theory so far (based on some sources) that memory-mapped files on Windows bypass the "Cache Manager" and - by extension - the page cache. That's maybe why I can see disk activity when loading textures upon returning to a specific place multiple times (which shouldn't happen as the cache should kick in). Or maybe the 50MB of cache the kernel is reporting is really it, which obviously isn't enough. Increasing the minimum working set (sysinternals CacheSet / DynCache linked above) doesn't really help.

@Freghar

Why are you using RAM disk for page file? I've read there's no real point to do that. You could just disable it and have the same effect or is there some other wizardy going on?

Read the previous posts - I explained that unfortunately I see difference with/without pagefile (on a ramdisk or not) and I indeed did the last set of benchmarks without pagefile (highlighted in bold).

[Freghar 73]

As a wild possibility, the system could keep multiple files/ranges loaded even if Arma is not using them due to how the memory-mapped files API seems to work (according to the documentation). There seems to be a two-layer approach of opening the file itself and mapping parts ("views") of it into the virtual memory.

Now, if Arma doesn't close the files during runtime of a mission, but only unmaps the views from VM, the system could theoretically hold the files open / mapped. When the mission ends, Arma closes the files, which makes Windows release the resources.

If Arma did close the file(s) on unmap, it would use much less RAM, but at the expense of disk traffic, assuming the OS doesn't cache memory-mapped files.

Very wild guess, although it makes at least some sense (come on). :)

To reiterate - the "used" RAM isn't really "used", it could be easily purged as all the data are stored on the disk in files - this is normally what Linux would do with a low swappiness value and when under memory pressure. Windows might prefer paging out active memory. :(

[I give up 152]

As a wild possibility, the system could keep multiple files/ranges loaded even if Arma is not using them due to how the memory-mapped files API seems to work (according to the documentation). There seems to be a two-layer approach of opening the file itself and mapping parts ("views") of it into the virtual memory.

Now, if Arma doesn't close the files during runtime of a mission, but only unmaps the views from VM, the system could theoretically hold the files open / mapped. When the mission ends, Arma closes the files, which makes Windows release the resources.

If Arma did close the file(s) on unmap, it would use much less RAM, but at the expense of disk traffic, assuming the OS doesn't cache memory-mapped files.

Very wild guess, although it makes at least some sense (come on). :)

To reiterate - the "used" RAM isn't really "used", it could be easily purged as all the data are stored on the disk in files - this is normally what Linux would do with a low swappiness value and when under memory pressure. Windows might prefer paging out active memory. :(

Mate, you've convinced me.

I am going to uninstall and wait for ARMA 3 on Linux.

Peace. best of luck.

[x3kj 1247]

One more expert.

First,

You show me any other game where this insane amount of memory is being cached and to make your task simple you can use as example even the x64 games.

 

 First, other games dont matter. Show us the performance impact of high ram loaded vs low ram loaded in arma3. Then you can claim that this is bad.

[I give up 152]

Obviously other games doesn't matter, but you have none. The question is you know why? If you dont, most likely, here is not the right place for you to learn.

Anyway, everybody knows (I think) that high memory consumption increases computation time and vice versa.

Unless ARMA 3 is different. It is? You tell me.

Still, the reason of this thread is not only about RAM load, is mainly about its management (as the tittle of the thread indicates), Do you how it is being managed? If you do just tell me, i'll say thanks.

[I give up 152]

I am going post for the last time in this thread.

I have opened it in the hope that we could have some clarification or even some improvements in game performance, but all I've got is useless spam.

 

I could have named this thread Pagefile (virtual memory) because the subject applies in to both (physical and virtual), I have chose physical in the hope that could make it simpler.

 

With that said, also want to say that I am perfectly aware (and documented) about of what is happening.

32 bit Breaking Barrier was indeed a revolutionary change and in fact it helped at that time. It helped with physical RAM (4 GB average at that time) and it helped to surpass the 3GB limitation of x86 systems since it is quite clear that due to ARMA architecture more than 3GB are needed in matters of cache or cached data.

 

With ARMA 2 (OA) textures (shaders, lighting, shadows, etc) have a relative small size (because it s DirectX 9) and the impact of these on files cache size practically goes unnoticed. With such small size it can be managed without major issues. So the impact in performance was benefic.

The issue is.

With ARMA 3 and because it is DirectX 11 the same data is considerably larger consequently the cached files/cache used (virtual or physical) are/is considerably larger and with this size there is no way to manage such data properly under a x86 archicture, consequently the impact in performance is negative, sometimes really nasty.

Just as example, Chernarus and Takistan (A2) together use less cache than Stratis alone.

 

I am sure that game developers are aware of this, I guess thats why at some point they gave some support to Fred and his malloc, which for obvious reasons was condemned to fail. There is no way for this game (or any other) to deal with large blocks, at start it may be possible that we can see a performance increase, but after a short time of usage what we see is a drastic degradation in performance, for obvious reasons.

 

Its up to developers, maybe they want to find an appropriate solution in matters of cached files for the current version, or maybe they just prefer to keep letting people complaining about CPU and GPU bottleneck

 

Anyway, peace.

[Freghar 73]

I am going post for the last time in this thread.

I would hereby like to thank you for the provided humor, it was indeed entertaining and enlightening.

For example,

Anyway, everybody knows (I think) that high memory consumption increases computation time and vice versa.

Everybody knows (I think) it's the other way around - https://en.wikipedia.org/wiki/Space–time_tradeoff.

(Or just from the other referenced thread.)

 

Exactly, the main reason for bottleneck in Arma 3 is the Hard Disk. Arma 3 due to its architecture "needs" Hard Disk.

Measurements seem to disagree - during runtime, reads peak around ~300 IOPS with rotational HDD doing ~110 random IOPS and a regular SSD ~80000. Most of the time, however, the IO pattern is not random and doesn't go above 30 IOPS.

Arma 3, after loaded, requires a continuous Hard Disk usage and this happens practically every time we move the mouse or press a key.

Again, not true, measurements easily confirm it.

About RAM speed, with current technology higher frequency means higher latency, the CAS for 2133/2400 are way high when compared with 1600 or even 1866. This means that the gains that we may have in frequency are lost in latency.

Although popular belief, not exactly true - http://www.crucial.com/usa/en/memory-performance-speed-latency. Yes, higher throughput = better, lower latency = better, but comparing the two against each other on a theoretical level is nonsense.

Especially given how DRAM actually works - prefetching, pipelines, etc. And that DDR4 catches up on higher frequencies in latency.

Not to mention the relative unimportance of RAM speeds with discrete graphics these days (lots of L2/L3 cache) for most applications - few still benefit, would be nice to see Arma benchmarks.

Now if you really want to improve the performance for Arma 3 do not waste money in CPU, GPU and RAM.

For Arma 3 any mid range CPU, any mid range GPU (ATI/Nvidia) and any mid range RAM (1600/1866) are perfectly fine.

Like I have said several times, if you really want to get better performance you have 3 options.

1. Get a PCI Express SSD.

2 Get a RAMDISK.

3. Get a couple of SSD in RAID0.

(ordered by performance)

This is completely false. Don't take my word for it, measure it. SSD helps for terrain that changes quickly, but single-thread (CPU) is still the main bottleneck to the point where OC'ing my i5 turbo from 3.7GHz to 4.0GHz gives me 6FPS extra in a busy firefight where HDD/SSD activity is negligible.

Please bratwurste don´t highjack this very useful thread with your wrong statements. Thanks in advance.

You ignore Chavez´ hint the significant fps increase from high ram clocks is pretty consistently affirmed/validated in this thread.

Ah, another "troll", right?

If you know RAID 0 and Arma 3, you should know that the game streams data directly into Hard Disk (not in to RAM as some believe), streams in a continuous and peristent way things like terrain or new objects.

Nope, measurements showed no correlation, it's the reads that are visible.

Is 500MB/s of read/write speed enough to achieve this? NO. We need more. How we get it? By the methods described above.

Confusing throughput with latency. 500MB/s is much more than enough.

These are the facts and what really happens, all the rest are fairy tales.

Mhm.

But first try to run Arma 3 and load Altis with pagefile disabled.

If you can do it without errors then you dont even need a SSD, because the data will be loaded in to HDD only one time when you open/load the game/island for the first time, after that is all about RAM. Doing this successfully will place you in the same level of the guy that have discovered the powder.

Not true again, Arma doesn't load the entire island into RAM and data are indeed loaded from the HDD - even the same data multiple times. In ideal world, it would not be, but that's not what the stats show.

By the way, a bit inconsistent with your previous "facts".

If you cant, then is because Arma 3 in fact needs and use system pagefile for file swap. In this case the myth, that the game do not stream (textures/lod/terrain/objects) from Hard Disk swapping files between system pagefile and physical memory, is dead.

Again with the firm declarations - any references to sources?

While with DDR3 is like you say, DDR4 is a completely new architecture where the bandwidth (data rate) gains a new relevance over latency.

If you in fact read the JEDEC standard and the previous ones, you see that DDR4 is just another small evolution over DDR3, adjusting throughput, power consumption, adding 3D stacked die, adjusting the signalling, ... few bits and bobs, not a whole new architecture.

That's why DDR4 wasn't here in 2013 - nobody wants it because there are no practical benefits that would justify the cost (currently, this will change with higher freqs).

 

And how much you wanted to have per card? Are you saying that with SLI/Crossfire the second card should add more VRAM to first card or vice/versa? Should be like adding one more RAM module?

Also the stack definition (the one we see over the web) is completely wrong when applied at this subject. That's what I am saying. Look at nonsense threads over the web where people just do not have a clue about what their are saying.

SLI/Crossfire working principles are not supposed to stack (according to your definition of stack) VRAM or anything else and honestly dont even know why people talks about it or why should "stack".

When we combine multiple GPUs (obviously all the GPUs must have the same characteristics) they will operate in parallel with the workload being distributed equally by all of them, this with applications that have support (and profile) for it.

Means,that we have all combined VRAM available for use, being the usage distributed between the cards.

The same applies to clocks however the clocks subject is a bit different, if with combined VRAM we can use it all, with clocks is not possible to take full advantage of it. That's why when we compare performance between single and multiple GPUs, the second gives less than 50% of performance increase and the third less than 30%.

And btw, DX12 is not related with this subject simply because this is a matter of hardware architecture, consequently with DX12 nothing will change in these matters.

Now please, some one can be kind enough to explain what is "VRAM stack" or the purpose of it with SLI/Crossfire tech?

I'm just going to leave that without additional comment as the last paragraph is the punchline.

You should perhaps reconsider your life philosophy, the current one seems to be around making things up out of thin air and declaring them as facts, which makes you look knowledgeable, but when it backfires, you have to use the lalala-I-dont-hear-you-troll technique, which kind of discredits you.

You have unfortunately chosen the wrong area - in science/technology, things can be objectively right (facts) and wrong and when you state something as fact, it can be proven true or false, objectively, with no subjective bias of opinion. References are a big part of that - when I'm wrong, I can usually point to where I got the original info from and blame/fix that source. When I'm right, it only backs up my claims.

 

You look a nice guy and because of that I will give you a few tips and you can do the benchmark (for free).

So .. perhaps marketing would suit you better?

If not, please don't mix basic knowledge together, the mix doesn't produce deep knowledge, even if it might confuse and trick the less knowledgeable. When you make assumptions, mark them as such ("I think this means ..."), not as facts.

 

/nuff said.

Indeed.

PS: Since you seem to be using "virtual memory" incorrectly in a consistent fashion,

I could have named this thread Pagefile (virtual memory) because the subject applies in to both (physical and virtual), I have chose physical in the hope that could make it simpler.

you might want to read about virtual memory.

[St. Jimmy 272]

Not to mention the relative unimportance of RAM speeds with discrete graphics these days (lots of L2/L3 cache) for most applications - few still benefit, would be nice to see Arma benchmarks.

Sorry I go offtopic now.

Watch this benchmark. Notice how well Broadwell performs considering it runs much lower GHz and it's DDR3L. It beats the Skylake and Arma 3 isn't the only game Broadwell performs very well. The thing that Broadwell has is the 128MB L4 cache.

That article also shows the importance of RAM speed and latency in Arma 3. Go few pages back in it to see.

[Freghar 73]

Sorry I go offtopic now.

Watch this benchmark. Notice how well Broadwell performs considering it runs much lower GHz and it's DDR3L. It beats the Skylake and Arma 3 isn't the only game Broadwell performs very well. The thing that Broadwell has is the 128MB L4 cache.

That article also shows the importance of RAM speed and latency in Arma 3. Go few pages back in it to see.

Good to know, thanks. If Arma is indeed RAM-intensive, a large eDRAM cache can certainly do that. The question is whether it makes sense to buy a "better" mainstream for premium price or go for the "enthusiast" segment with quad-channel memory and bigger L2/L3 caches.

(Currently waiting for Broadwell-E vs AMD Zen.) :)

edit: Are you sure they used dedicated (discrete) graphics for the test? .. Integrated (APUs) hugely benefit from higher RAM speeds as they essentially use RAM as their VRAM (and DDR3 is pathetic compared to GDDR5). The other benchmarks of integrated graphics make me wonder.

[heavygunner 179]

Good to know, thanks. If Arma is indeed RAM-intensive, a large eDRAM cache can certainly do that. The question is whether it makes sense to buy a "better" mainstream for premium price or go for the "enthusiast" segment with quad-channel memory and bigger L2/L3 caches.

(Currently waiting for Broadwell-E vs AMD Zen.) :)

edit: Are you sure they used dedicated (discrete) graphics for the test? .. Integrated (APUs) hugely benefit from higher RAM speeds as they essentially use RAM as their VRAM (and DDR3 is pathetic compared to GDDR5). The other benchmarks of integrated graphics make me wonder.

Quite sure they sed a dedicateed graphics.

Also RAM defintely helps in most games, when you are limited by the CPU, which ofc isn't the case in most games.

Arma however is mostly CPU limited as you know.

 

As an example they made some tests with the I5 2500K and different RAM, which gave more FPS in CPU intensive games/moments.

 

[I give up 152]

Sorry I go offtopic now.

Watch this benchmark. Notice how well Broadwell performs considering it runs much lower GHz and it's DDR3L. It beats the Skylake and Arma 3 isn't the only game Broadwell performs very well. The thing that Broadwell has is the 128MB L4 cache.

That article also shows the importance of RAM speed and latency in Arma 3. Go few pages back in it to see.

I am forced to post again because some useless wikipedia spam troll keeps posting and quoting out of context (not you).

 

Yes mate,  DDR4 can in fact improve performance with ARMA 3, under certain circumstances.

 

DDR4 with Haswell will not bring significant improvements for ARMA.

 

DDR4 with Skylake will bring noticeable improvements, mainly due to CPU memory controller which is the first one that can really take advantage from the higher frequencies of DDR4.

 

Why? 

 

In first place because data rate in DDR4 has an improvement of 50%, when compared with DDR3. 

Data rate on DDR3 has a max of 2133 Mb/s and in DDR4 has a max of 3200 Mb/s.

 

In second place because an i7 4470K it has a max memory bandwidth of 25 GB/s and an i7 6700K has a max memory bandwidth of 34.1 GB/s, this is basically a 50% increase when compared with Sandy Bridge.

 

And in a game like ARMA 3 heavily dependent of memory performance can bring some improvements.

 

And why is the can?

 

For performance improvements we need to have at least 16 GB of RAM, here the game cache will mainly rely and it will be provided by the physical RAM and under this circumstances RAM performance is a crucial factor.

 

If we have only 8 GB of RAM, at some point the game will start to cache using he Pagefile (virtual memory), at this point RAM ceases to be the main factor and the Hard Disk (where Pagefile is managed) starts to predominate. Under this situation (having 8 GB of RAM) is better to have a fast Hard DIsk than a fast RAM.

Here the improvements from DDR4 are not noticiable, still there are some.

 

All the rest are fairy tales from clueless wikipedia users.

[I give up 152]

Oh well, trying to figure whats happening, I just went into my basement, cleaned the dust of my 10 year old 5870 1GB Vram and connected in to mine top notch machine.. and the result... just superb.

Like, 500 Mb of Vram usage, stable 60 FPS and average RAM Load.

Conclusion, want to run the game without major flaws? Get a 10 years old graphics card with 1Gb Vram.

 

https://www.youtube.com/watch?v=-dEgB_xepg4

[St. Jimmy 272]

Oh well, trying to figure whats happening, I just went into my basement, cleaned the dust of my 10 year old 5870 1GB Vram and connected in to mine top notch machine.. and the result... just superb.

Like, 500 Mb of Vram usage, stable 60 FPS and average RAM Load.

Conclusion, want to run the game without major flaws? Get a 10 years old graphics card with 1Gb Vram.

Give the mission for repro.

System specs in both situations would still be nice to know... We now know that you've 32GB of RAM and in that video you've 5870 but what other AMD card you've in the situation where things aren't good? And you likely have page file disabled or is it enabled? You could edit that stuff also in the first post.

Did you change any graphic settings or launch parameters?

This could be AMD only problem.

 

You also could post the other video where the performance was bad in that situation. You had a topic about it or something. In this topic you've shown RAM load after flying all over the map but you didn't show that now with your video. You could help a lot but you're too cryptic.

 

Some basic stuff needs to be answered first before the issue can be looked. Dwarden would likely happily forward these stuff but he very likely needs to get the same questions answered if not more.

[Valken 621]

Very interesting thread... Here I was thinking I have to UPGRADE MY PC (get faster RAM, faster and bigger GPU) when I should pull out my CFX 6950s and drop in an older 1 GB GPU card! I might just enable the onboard video on this i7-4970k and see how that does instead.

[Dwarden 1124]

if you want to help nail the 'earlier' discussed "memory leak(s)" please

 

go here https://forums.bistudio.com/topic/160288-arma-3-stable-server-156-performance-binary-feedback/page-67?p=2999681#entry2999681

get the performance build v4 running and provide some 'crashes' (due to out of memory etc.) and share logs/dump files so we can track/fix those ...

 

while this will not answer most of your questions ;) it will sort the 'bug in spotlight'

 

and yes I know the usual excuses follow:

WIP large pages compatible allocator, allocator improvements (SSE2, tweaks) and other (e.g. ability to disable heapManager for custom allocators)

WIP 64bit

can't provide ETA, first line more likely sooner than second :)

[I give up 152]

Well, I dont really have errors or crashes, mine only issue is the huge amount of ram loaded because after some point the game starts struggling.

So far I am sure of one thing, more gpu vram means more ram loaded and having 6 and/or 9 GB of Vram, the amount of ram loaded is just insane.

I have played for long time with a GTX 580 1.5 GB vram and never had issues with ram, also with a single R 280 3 GB vram, even the ram load being higher, there are no issues related with.

Still not sure if this issue happens due to crossfire (multi gpu) or if also happens with a single gpu. I am waiting for a R 390 with 8 GB vram, then I will be sure if is related with vram or if is related with crossfire (multi gpu).

[jumpinghubert 48]

Oh well, trying to figure whats happening, I just went into my basement, cleaned the dust of my 10 year old 5870 1GB Vram and connected in to mine top notch machine.. and the result... just superb.

Like, 500 Mb of Vram usage, stable 60 FPS and average RAM Load.

Conclusion, want to run the game without major flaws? Get a 10 years old graphics card with 1Gb Vram.

 

https://www.youtube.com/watch?v=-dEgB_xepg4

There is something missing for a comparison. You forgot the second apple. The video is a little bit like a joke: your running around at the place with the lowest object density and make the "conclusion": hey it runs like a charm! lol

 

P.S.

why ignoring Jimmy´s questions?

[CaptainAzimuth 714]

Wrong thing. Ignore this.