I don't see any other posts about this, so I don't know if this is just an issue I'm dealing with for some reason, but I have a decent gaming computer with 8 gigs of ram that can handle just about any game I throw at it... except Direwolf20 1.10.2. It consumes almost the entirety of the available ram just showing me the main menu. Is this a problem that anyone else is having, or does the new modpack just hate me?
Unfortunately, I'm getting this issue in spite of limmiting the jvm to 4 gigs. Xmx only limits what memory the virtual application can use, but has no effect on permgen, stack size, or native libraries' memory usage. I've seen posts mentioning that they've changed how models are loaded. If that's true I'm thinking that they're doing something that's causing the LWJGL native to leak memory like a sieve. I'm curious if anyone else is having this problem.
1) Recently I too have had the same problem with the only difference being mine is 6 Gb (excluding the graphics cards memory). Were you ever able to find a work around?
The rendering changes and the shift to models basically trades CPU for RAM to increase performance, this means that any pack with a reasonable number of mods is going to need at least 4 GB to even launch. The 2.5-3GB that you were seeing for just 20-30 mods is a lot more than you'd be using in 1.7.10 and earlier, you'd have been able to have 4 or 5 times that number of mods in the same amount of RAM. FoamFix does help and I've seen report of 0.5GB being knocked off by using it, which is why it's now in our packs.
2) For FTB devs: Can you look into if FoamFix is working correctly with Direwolf20 especially related to enderIO? I too have been getting 6 Gb to 8 Gb of committed memory instead of 4 Gb as well. Committed memory or Private Memory is the actual memory that should set the min RAM requirements of a game (whichever is higher).
Also based on what you said PermSize and XMX don't seem to really affect allocation anymore, do you have any tips on how to set up Minecraft or java to better allocate and utilize memory for these newer versions?
P.S. There is more I wanted to ask you Quetzi and the other mods about, but I will have to save for my next post. For example, some of the things you and the other mode mentioned 1.10 sound similar to Big Data Volume problems I have seen as a data analysis. If you could reply to my initial questions labeled in (2), in the meantime I would appreciate it very much.
I've experienced similar issues in DW20 all the way back to the 1.6 era. There seemed to be a memory leak caused by loading or unloading chunks. I heard rumors that it was caused by Mystcraft, but I don't think the source was ever confirmed, since it was never fixed. This kind of memory leak was tolerable to me, it made the game unplayable after a few hours of use, but then I just had to restart. The issue I have now makes the game unplayable as soon as I turn it on.
I can confirm that at one time Mystcraft chunk issue was real though there was more than one flavor. First, Mystworlds at one point had intentional instability built into them where chunks and eventually the world would crash after a certain time or conditions were met. Originally it was "in-game computer code" you used to create your mystic craft world, but then it later became an actual pseudorandom aspect where the instability was based on the seed number and the number of worlds that had be created so far. This second pseudorandom version did go into the 1.7.4 FTB version, but it should be disabled by default. Second and more likely what you experienced, Mystcraft does not like quarry blocks or chunck-loading blocks and when mixed with them creates persistent chunk issues. I would assume that FTB devs got it fixed, by now; however, it was a thing.
P.P.S. Yes, your private or committed RAM memory can exceed (the combined ram of your CPU and GPU). JVM and many languages typically write some memory to disk. It is sometimes bad thing when it does exceed your combined RAM, but its normally safe in short bursts.