Jump to content
Sign in to follow this  
Georgie

General OOM Issues

Recommended Posts

If you have a 32 bit process then you will run out of virtual memory at 2, 3 or 4 GB based on a number of conditions.

As long as you have that much RAM+swapspace left over (after OS and background junk) the limit you hit is the virtual memory limit and no amount of more RAM or swapspace will improve the situation.

I do not understand why it cannot be changed the Large Address Aware flag...that would work out may OOM issues

Share this post


Link to post
Share on other sites
Guest

Knaust, don't start that please. I can't discuss whether or not we use that, whether we've tried it, whether it does or does not fix anything. And frankly without the code in front of you, you have no way of knowing whether it would "work out many OOM issues" or not. Further discussion of it is, for the purposes of this thread, completely pointless.

Please, report your general OOM issues here, guys. Don't start theorizing about what might or might not magically fix it all. That didn't help in the last thread - the reports about the actual WEGO OOM issues were infinitely more helpful than the "discussion" that sprang up around them - and it won't help here.

Share this post


Link to post
Share on other sites
Knaust, don't start that please. I can't discuss whether or not we use that, whether we've tried it, whether it does or does not fix anything. And frankly without the code in front of you, you have no way of knowing whether it would "work out many OOM issues" or not. Further discussion of it is, for the purposes of this thread, completely pointless.

Please, report your general OOM issues here, guys. Don't start theorizing about what might or might not magically fix it all. That didn't help in the last thread - the reports about the actual WEGO OOM issues were infinitely more helpful than the "discussion" that sprang up around them - and it won't help here.

OK....please erase my post

Share this post


Link to post
Share on other sites
Knaust, don't start that please. I can't discuss whether or not we use that, whether we've tried it, whether it does or does not fix anything. And frankly without the code in front of you, you have no way of knowing whether it would "work out many OOM issues" or not. Further discussion of it is, for the purposes of this thread, completely pointless.

Please, report your general OOM issues here, guys. Don't start theorizing about what might or might not magically fix it all. That didn't help in the last thread - the reports about the actual WEGO OOM issues were infinitely more helpful than the "discussion" that sprang up around them - and it won't help here.

Phil, although I agree it's not the customers' job to tell you how to fix it, please remember some people are just trying to be helpful, not critical.

Share this post


Link to post
Share on other sites
Guest

I understand that. That's precisely why I'm explicitly telling you guys what will be helpful.

It's not even that I mind discussing these things, honestly - it's that I a) can't discuss specifics regarding the code or anything related, a fact which I had to repeatedly explain in the last thread, and B) any evaluation of the problem by people outside of it tends to spiral out of control, as it did in the last thread when several posters insisted that a general memory leak was happening when in fact it wasn't, and later on when the Large Address Aware flag was (repeatedly) brought to my attention.

This leads to conclusions being drawn, as evidenced by Knaust, who clearly had good intentions, assuming that something I had not discussed was somehow a fix for our problems. This type of response isn't confined only to these threads, either. I get dozens of emails and help desk tickets from customers saying "well if this is a fix why haven't you done it yet?", all of which take time away from actually dealing with problems. Helpful or critical, when you claim that X or Y will fix something, or is the problem, my time is impacted.

You can understand then, perhaps, why I want to shut down this line of discussion immediately and focus on getting actual bug reports from this thread.

Share this post


Link to post
Share on other sites
I do not understand why it cannot be changed the Large Address Aware flag...that would work out may OOM issues

If you have the necessary motivation it is easy to test yourself. Log the amount of currently mapped virtual address space (not swapspace which microsoft erroneously calls "virtual memory") until the game crashes. If you got more than 2 GB then the flag is set (you can go to 3 or nearly 4 GB depending on OS).

Share this post


Link to post
Share on other sites
If you have the necessary motivation it is easy to test yourself. Log the amount of currently mapped virtual address space (not swapspace which microsoft erroneously calls "virtual memory") until the game crashes. If you got more than 2 GB then the flag is set (you can go to 3 or nearly 4 GB depending on OS).

so what do these figures then tell?

CPU usage 60%

Memory usage 762.292 k

Virtual memory 899.668 k

at 36% "out of memory" (Read Data) crash time.

With all MaxQ settings & Sound On, data was almost the same:

CPU usage 60-90%

Memory usage ~770.000 k

Virtual memory ~900.000 k

at 36% "out of memory" (Read Data) crash time.

To me it looks the crash happens when the game actually tries to allocate future memory, but hasn´t yet used up the (physical) one available. Still had 700-900MB phys mem left at that time. Hard to grasp (..if one is not a programmer...).

Share this post


Link to post
Share on other sites

You don't have the right number in there. You need the amount of virtual address space in the CMBN process. In non-suck operating systems that is called "virtual memory" in the per-process accounting.

You don't run out of physical memory there (neither RAM nor swapspace are full), you run out of virtual address space.

Share this post


Link to post
Share on other sites

Activate this column in process monitor => "View" -> "Select Columns" menu.

Surprisingly enough Mickeysoft uses the right name for virtual memory in the per-process accounting.

virtual-memory-in-process-monitor.png

Share this post


Link to post
Share on other sites
Using this might interfere with the DRM of the game.

No, it seems to only use outside observable things like system calls and the memory map. I don't see anything that would require tracing like in a debugger.

Share this post


Link to post
Share on other sites
Guest

Process Explorer does cause issues with the DRM. To avoid unintended side effects stick with Task Manager.

Share this post


Link to post
Share on other sites
Activate this column in process monitor => "View" -> "Select Columns" menu.

Surprisingly enough Mickeysoft uses the right name for virtual memory in the per-process accounting.

virtual-memory-in-process-monitor.png

Maybe I wasn´t clear (see post #19 further above), but that´s exactly what I did. Data is from columns in the processes tab, which I configured the way you suggested before.

Share this post


Link to post
Share on other sites
Maybe I wasn´t clear (see post #19 further above), but that´s exactly what I did. Data is from columns in the processes tab, which I configured the way you suggested before.

OK. Then what did it run out of? Everything is < 1 GB.

This isn't the situation where it runs out of video memory, right?

Share this post


Link to post
Share on other sites
OK. Then what did it run out of? Everything is < 1 GB.

This isn't the situation where it runs out of video memory, right?

The answer to that question is that MS screwed up again. If you look at the virtual process size in the normal Control-Alt-Del process list it is reported incorrectly. It gives you the number of private mappings but of course you need the total size of all mappings combined because that's what you run out of.

The ProcessExplorer you can download from Microsoft displays the correct number for virtual process size. And people wonder why I don't use this crap...

Share this post


Link to post
Share on other sites

Phil, do I understand the status that you think you understand the issue here but you need to do some brainstorming and testing to see what if anything you can do about it? If that's the case, can you give us an ETA as to when you might have more info?

Share this post


Link to post
Share on other sites
Guest
Phil, do I understand the status that you think you understand the issue here but you need to do some brainstorming and testing to see what if anything you can do about it? If that's the case, can you give us an ETA as to when you might have more info?

It's more the case that we need to work out a timeline for when and how these things will be implemented. I'll hopefully have some info next week.

Share this post


Link to post
Share on other sites
Guest

RockinHarry, I've been talking about solutions to the problem with Charles. I'll let you know when we have a timeline for changes. Please keep the thread confined to discussing the actual issues - if you have a bug report, feel free to share. I'm fairly certain I mentioned this a couple of pages ago.

Redwolf, I have an NDA. If Charles really wanted to he could come out and discuss this with you. If he wanted to - which I doubt - I don't think he'd have a problem sharing it. It's not earth-shattering. I, however, have made an agreement I intend to honor. Let it go.

Share this post


Link to post
Share on other sites
RockinHarry, I've been talking about solutions to the problem with Charles. I'll let you know when we have a timeline for changes. Please keep the thread confined to discussing the actual issues - if you have a bug report, feel free to share. I'm fairly certain I mentioned this a couple of pages ago.

No bug reporting actually, since my problems probably more relate to the games actual hardware requirements, which I try to figure out for myself when it comes to large maps loading (3D view) and the OOM error as result. (...just the maps, with no OOBs and in the editor)

AFAIK, most people who have no problems loading mentioned maps to 3D, have a machine with 4 Gigs of RAM and a graphics card with 1 Gig.

Upgrading to 4 Gigs of Ram wouldn´t be a problem for me financially, but thinking about purchasing a better graphics card would be more so (80-90 Euros). Currently I have NO problems with other hardware hungry games, like ARMA2, MOW ect. so my planned investment goes straightly and purely for the ability to get the most out of CMBN. This would be 4x4km maps and at least 1 Btl. forces each side.

I won´t purchase any further module or expansion for CMX2, until I got that matter cleared (true hardware, OS and drivers requirements). ;)

Share this post


Link to post
Share on other sites
No bug reporting actually, since my problems probably more relate to the games actual hardware requirements, which I try to figure out for myself when it comes to large maps loading (3D view) and the OOM error as result. (...just the maps, with no OOBs and in the editor)

AFAIK, most people who have no problems loading mentioned maps to 3D, have a machine with 4 Gigs of RAM and a graphics card with 1 Gig.

Upgrading to 4 Gigs of Ram wouldn´t be a problem for me financially, but thinking about purchasing a better graphics card would be more so (80-90 Euros). Currently I have NO problems with other hardware hungry games, like ARMA2, MOW ect. so my planned investment goes straightly and purely for the ability to get the most out of CMBN. This would be 4x4km maps and at least 1 Btl. forces each side.

I won´t purchase any further module or expansion for CMX2, until I got that matter cleared (true hardware, OS and drivers requirements). ;)

It is not a hardware issue. Do not buy hardware to fix it.

Share this post


Link to post
Share on other sites
Guest

True. This isn't a flat-out "everybody gets this problem in the exact same way" issue, so clearly hardware has an effect (as well as software and usage patterns), but the obvious solutions - buying more RAM or a faster / bigger video card - certainly don't apply here. In fact I've seen far more reports of this from people with hefty amounts of VRAM.

We're working on it. As we solve pieces of it everyone will benefit regardless of hardware, so holding off on CM-specific updates would be wise. Until we've released some relevant updates, upgrades will likely not have an effect and may - depending on circumstance - actually make things worse.

Share this post


Link to post
Share on other sites
Guest
This topic is now closed to further replies.
Sign in to follow this  

×
×
  • Create New...