Anyone running 128GB ram?

[Shawn]

I have 128GB of ram on one of the work PCs and twice now I have run into issues with installers crashing with cryptic error messages because of too much ram. One was an old version of Matlab that refused to install (2017 version installed fine) and another was an old version of Sigmaplot. I was able to get Sigmaplot installed by pulling all but one 16GB stick. After it installed though it worked fine with all 128GB.

Has anyone else run into this issue. Are there any workarounds? It seems if I could limit the amount of memory the installer sees it should work fine, but that's not an option in Windows 10's compatibility mode.

 

[Elixer]

Well, with that amount of RAM, I would make a VM, and run the crappy programs on that.
If that isn't an option, then you could play with Application verifier, and limit the amount of RAM that, that crappy program can see.
HEAP_ALLOC is the one you would play around with.
http://gusclass.com/blog/2011/10/31/application-verifier-save-your-memory-debug-your-apps-also-hax/

 

[VirtualLarry]

Yeah, most installers, even for 64-bit enabled applications, are themselves 32-bit.

 

[XavierMace]

I don't see how having 128Gb of RAM is going to cause something to crash. I also don't see what the installer being 32-bit has to do with anything.

 

[bbhaag]

I don't see how having 128Gb of RAM is going to cause something to crash. I also don't see what the installer being 32-bit has to do with anything.

I had it happen to me last year when I was trying to install IE5 on my Win3.11 system. Because I had 68mb of ram is wouldn't install. I had to enter setup.exe /F:16 at the command line to bypass the memory checker. So it does happen.

 

[XavierMace]

I had it happen to me last year when I was trying to install IE5 on my Win3.11 system. Because I had 68mb of ram is wouldn't install. I had to enter setup.exe /F:16 at the command line to bypass the memory checker. So it does happen.

Not sure if serious.

 

[Elixer]

I don't see how having 128Gb of RAM is going to cause something to crash. I also don't see what the installer being 32-bit has to do with anything.

If the installer used a 32bit unsigned int, then 128GB = 137438953472 bytes, and that is too big to fit. Which means, they overflowed the (unsigned) int which can only hold 4294967295. Which can lead to crashes, or other odd things.

And yeah, that happens all the time on badly programed code.

 

[XavierMace]

If the installer used a 32bit unsigned int, then 128GB = 137438953472 bytes, and that is too big to fit. Which means, they overflowed the (unsigned) int which can only hold 4294967295. Which can lead to crashes, or other odd things.

And yeah, that happens all the time on badly programed code.

And if that was the case, it still wouldn't have worked with a single 16Gb stick as that still would have overflowed the unsigned int.

 

[Elixer]

And if that was the case, it still wouldn't have worked with a single 16Gb stick as that still would have overflowed the unsigned int.

Yeah, thought of that later, but I was only guessing on one possible thing that can go wrong... They really did a bad job coding this installer it seems.

 

[XavierMace]

What I was getting at is how was the conclusion reached that the issue is too much RAM?

 

[mikeymikec]

The older installer's function may have been intentional to try and ensure that far more modern computers have to use a newer version of the software.

 

[Shawn]

I figured it out because it worked on an identical system with less ram. Had it not been for that I probably would not have figured it out. The installer gave some non-informative error like "The installer has encountered a fatal error" and that was it. The program runs fine after being installed.

 

[Shawn]

Well, with that amount of RAM, I would make a VM, and run the crappy programs on that.
If that isn't an option, then you could play with Application verifier, and limit the amount of RAM that, that crappy program can see.
HEAP_ALLOC is the one you would play around with.
http://gusclass.com/blog/2011/10/31/application-verifier-save-your-memory-debug-your-apps-also-hax/

Thanks. I will try that if I encounter another poorly written installer.

 

[mikeymikec]

I figured it out because it worked on an identical system with less ram. Had it not been for that I probably would not have figured it out. The installer gave some non-informative error like "The installer has encountered a fatal error" and that was it. The program runs fine after being installed.

Out of curiosity:

That still left quite a wide margin of possibilities. Did you find some further bit of evidence to point you to the cause or did you just think "maybe it's because of too much RAM" and tried removing some?

 

[XavierMace]

So you're assuming the memory is fine?

 

[bbhaag]

Not sure if serious.

No I'm being 100% serious. That happened to me last year. I guess installers have or at least had parameters that you can change using the CLI(dos prompt). Without the /F:16 parameter the install never would have happened. Microsoft used to have an article about it kb239550 to be exact but they have removed it from their support page.
But my original point was that certain installers hang and then crash if you have to much memory. I've seen it happen first hand.

 

[XavierMace]

That fact that something dumb happens on Windows 3.11 isn't really relevant to a modern OS and application. Not to mention 16-bit/32-bit vs 32-bit/64-bit.

 

[Shawn]

So you're assuming the memory is fine?

I believe so. I ran the built in Win10 memory check a while back and it passed. The only other reasons I can think of are 1) the motherboard only officially supports 8x8GB sticks, but we put in 8x16GB sticks (because we like to live dangerously) or 2) Windows was installed using legacy boot and not UEFI. I don't know if that matters though.