Hi all,

Running a 1,5 million cells (hex-tetra) simulation on a 4gb ram pc (actually it's win32 bit, so you know about the bottleneck), here are some details for the memory allocation error:

Data Type Kwords Words/Node Words/Elem Kbytes Bytes/Node

Real 275324.8 233.91 223.46 1075487.6 935.65

Integer 88721.6 75.38 72.01 346568.8 301.51

Character 3009.3 2.56 2.44 2938.8 2.56

Logical 65.0 0.06 0.05 253.9 0.22

Double 1208.0 1.03 0.98 9437.5 8.21

You see, process would take ~1,5gb ram so i don't think this one is a so expensive job.

I've been trying defeating this #001100279 *** Run-time memory allocation error *** from 1 week, I'm getting really frustrated because I need this for my thesis. I tried all topic suggestions on this forum, manual overrides, allocation factors, I even tried lowering solver details such as raising physical time scale.

I also run this on a second pc but nothing. The odd thing is that on my prof.'s laptop it works and we don't know why, with the same allocated request and same system.

I will be grateful to every kind soul who try to help me,

ReeKo

Hi, search for your error "Run-time memory allocation error " in this forum, there have been lots of dicussions on this previously. Try changing your memory allocation factor, try parallel processing if you have a license; 32 bit is the problem in your case I have such big cases in serial on a 64bit m/c but it takes an awful long time.

Sans

Sans, I'm grateful to you but.

///////

I've been trying defeating this #001100279 *** Run-time memory allocation error *** from 1 week, I'm getting really frustrated because I need this for my thesis. I tried all topic suggestions on this forum, manual overrides, allocation factors, I even tried lowering solver details such as raising physical time scale.

///////

I really read everything in this forum

Sans is right. All you need to do is increase the memory allocation factor. However, your case is probably too big to run on a single 32 bit processor.

Hi,

In Windows 32 bit while the theoretical maximum process size is 2GB the memory is always fragmented so a process cannot grab a 2GB contiguous chunk of memory. So the actual maximum size of memory you can grab depends on whatever other stuff is loaded on the machine. For instance you will get a bit bigger chunk on a freshly booted machine compared to one which has checked email, done some web surfing etc - even if the software has subsequently been closed.

The maximum size of memory a process can grab is 1.2-1.7GB. So your 1.5GB job is right in this range, so sometimes it will work and sometimes it won't. You need to go parallel or go to 64 bit.

Glenn Horrocks

Thank you! I realized you were right then i switched on /3GB on boot.ini (win32) and now it seems i'm not getting problems anymore.

Now i'm planning try 64bit for better memory usage. I feel relieved now, thank you again.

R.

Hi Reeko,

Would you let me know how you got over the memory problem?

As like your case, my mesh is composed of 1845334 elemments(hex).

my system has 3G memory but sovler want to more.

the description of my problem is at the top of the threads.

Thanks

Hi,

Don't do big serial simulations in 32 bit. They are unreliable. Either go to a 64 bit OS or run it parallel. Otherwise you are wasting your time. You can do tricks to squeeze a little more space (run on a freshly booted machine, memory defrag programs) but I would not count on them working all the time, and they only give you a little bit anyway.

If you go 64 bit or parallel then you massively increase the size simulation you can do.

Glenn Horrocks