8 GB versus 16 GB

heedaf

Distinguished
May 19, 2010
24
0
18,510
Is an increase from 8 GB to 16 GB a huge increase in speed to justify the $$$? I've got and Ivy Bridge MB and I will be using it for all types of applications (Xilinx being the big one) and running on Win 7.
 
Probably the best indicaror for addi9ng memory is, How often Does the operating system have to utilize the page File??

You could do a Ctrl alt Del and start select task manager, then select performance at top and look at memory usage. can also click on resourse monitor at bottom then expand the memory down arror.
 
If there's any 'Free' listed in the Resource Monitor then there's no reason to add more RAM, but if there's no Free RAM listed then some applications won't be 100% optimized -- maybe 99% to 50% optimized -- it all depends on how your apps utilized Free & Shared memory <or> if it's a leaky app.

Resource-Monitor-32GB.jpg
 
Memory is one of those things that "more" does not translate into "faster", unless you need it. If you are running apps and programs that are using up all available memory and the OS has to start treating the hard drive like memory, then yes adding more memory will help tremendously. If you are not running out of memory, adding more will do absolutely nothing for you.
 

nazgron

Honorable
Apr 12, 2012
39
0
10,540
Not sure if there're some programs so big that would drain out 8GB RAM tho...

My friend is studying atomic physics and he uses programs that stress his phenom 1055t like raping it yet his 8GB RAM is fine :| (sry for my bad English)
 

rajyohanson

Honorable
Apr 25, 2012
81
0
10,640
Haha well you got the raping part correct and that's the funny part. :) Depending on the program it's going to be more GPU intensive than CPU intensive like in the case of your buddies program . In all reality, should get a tesla card and cover both!


http://www.nvidia.com/object/tesla-servers.html


just a standard PCIE reference card designed to do exactly what he's doing. :D


Not that they are practical but they are cool.



 

InvalidError

Titan
Moderator

No single program I am using uses more than 2GB of RAM on its own but the sum of multiple programs, multiple instances of some programs and the file cache do end up using more than 8GB RAM and performance ends up bogged down due to disk IO if I start actively using more than one large application at a time. To make multitasking more bearable, I ended up spreading applications across multiple HDDs so they don't get in each other's way when simultaneously loading/writing stuff from/to drive, that is already 3-4X faster than running everything off a single HDD.

I have a C2D-E8400 and my 8GB RAM currently is my biggest bottleneck, I need at least 12GB to eliminate most swapping/reloading and 16GB if I wanted to be able to turn swapfile off.
 
Yea my son has a GPU that cost as much as my whole computer - Not good for gaming.
He does electronic design an simulation.

have to remember the vast majority of common programs (applications) are still x86 programs and as such will only use up to 2 gigs, so with them it's more a question of how heavy is multitasking. There are a few programs that will take as much memory as you can throw at it, but these are generally the more expensive photo/video editing programs and some cad/cam drawing packages (ie hugh data bases).
 


And then you got applications like Lightroom 3.0 that eats up 2GB on its own ;)
 

InvalidError

Titan
Moderator

While 32bits x86 processes can only "own" 2GB of RAM directly (3GB with registry tweak), many games and some applications will access a much larger on-disk data set during a typical session. All that data passes through the OS' disk cache and with most modern OS, that cache will grow to consume whatever spare RAM is available which can translate into much faster reloading times if there is enough RAM for the cache to grow big enough to fit all frequently accessed data when reloading levels, reloading dungeons, teleporting, switching accounts/characters, etc. You do not see those quite tangible gains in FPS benchmarks but they do contribute to user experience.
 
Some data doesn't have to reside full-time in addressed memory space. It's inefficient and just a waste.

What the guy said about disk I/O, and kinda what the guy said about hitting the page file. Page faults would be a killer, too.

And my understanding is Xilinx is big on OpenCL. It may be so efficient your CPU, RAMs and disks are kicked back with their feet on the desk.