RAM size matters, A LOT

[xkm1948]

When i started off my quest of getting 128gb of ram most people were questioning me how the heck would some one use that much amount of ram? Well today i have to use my lab computer for some quick job, and it crashed within 5 minutes. This is with 8gb of ram.

Just some basic workload will eat up to 50GB of RAM. And this is only a tiny sized sample.

Hopefully 256GB ram will be available soon.

 

[ERazer]

do you use 50gb everyday?

 

[ZenZimZaliben]

Very Random Thread...But yeah RAM amount matters, but that amount is all relative to the users environment and what they are doing.

 

[Toothless]

The epeen is strong.

 

[tabascosauz]

While the whole shebang about lab simulations genuinely needing 128GB doesn't apply to most users, it is true that 8GB is no longer the standard for "comfortably multitasking and playing games". Quite a few games in the past few years (GTA V and Far Cry 4 come to mind for me) make poor use of RAM and come with unavoidable memory leaks.

Unless one only wishes to use their PC for accessing the web (disregarding Chrome's RAM habits) and writing word documents, the standard is essentially at 16GB, although it will take most users a few years to catch up.

But seriously, what are you trying to prove? You've already had that lengthy discussion about needing 128GB RAM for your simulations; no one had a problem with your 128GB RAM to begin with, so who are you proving "wrong"?

 

[Naito]

The epeen is strong.

It's about how you use it.

 

[Toothless]

It's about how you use it.

But it's like religion; you don't have to flaunt it.

 

[hat]

Well sure, if you run some shit like that it's going to eat RAM. But, most people aren't running whatever software it is you're running that's using that amount of RAM. 8GB is fine for me at the moment, but if building a new system I'd recommend 16GB, unless, of course, you run some ram-gluttonous software.

 

[xkm1948]

While the whole shebang about lab simulations genuinely needing 128GB doesn't apply to most users, it is true that 8GB is no longer the standard for "comfortably multitasking and playing games". Quite a few games in the past few years (GTA V and Far Cry 4 come to mind for me) make poor use of RAM and come with unavoidable memory leaks.

Unless one only wishes to use their PC for accessing the web (disregarding Chrome's RAM habits) and writing word documents, the standard is essentially at 16GB, although it will take most users a few years to catch up.

But seriously, what are you trying to prove? You've already had that lengthy discussion about needing 128GB RAM for your simulations; no one had a problem with your 128GB RAM to begin with, so who are you proving "wrong"?

Welp, mainly because i never thought it would crash my lab computer. I was expecting it to be extremely slow for a 8gb. But completely crash it really surprised me.

 

[ZenZimZaliben]

Welp, mainly because i never thought it would crash my lab computer. I was expecting it to be extremely slow for a 8gb. But completely crash it really surprised me.

Increase your page file to avoid the crashing..it will be slow as hell though..assuming your running windows or you know how to do that in other os's.

 

[D007]

If you were just playing games, 8 would be fine.. Virtually no on needs 256 gb of ram though..

 

[cdawall]

I have 64GB in my home machine that runs dumb things. My gaming rig has 16 it does fine, my work PC has 32 it also does fine

 

[cadaveca]

If you were just playing games, 8 would be fine.. Virtually no on needs 256 gb of ram though..

8 GB does not meet my needs for gaming. I've been testing a 3866 MHz 8 GB kit and although some things are nice and fast... it's just not enough. 12 GB, sure, but we can't buy 12 GB kits.

 

[cdawall]

8 GB does not meet my needs for gaming. I've been testing a 3866 MHz 8 GB kit and although some things are nice and fast... it's just not enough. 12 GB, sure, but we can't buy 12 GB kits.

Depends how old your rig in plenty of X58 chug along with 12GB kits in them

 

[natr0n]

16gb is a nice sweet spot for me at least.

 

[ZeDestructor]

When i started off my quest of getting 128gb of ram most people were questioning me how the heck would some one use that much amount of ram? Well today i have to use my lab computer for some quick job, and it crashed within 5 minutes. This is with 8gb of ram.

Just some basic workload will eat up to 50GB of RAM. And this is only a tiny sized sample.

Hopefully 256GB ram will be available soon.

You may want to look into E5 Xeons, that support 128GB LRDIMMs (up 12 LRDIMMs per CPU, upto 2 CPUs per machine with the common E5-2600 series, 4 with the 4600) already right now. You could also consider the crazier E7s, but those are way too impractical for workstation use.

 

[RejZoR]

Only time I realistically need 32GB of RAM is when I'm compressing data with 7zip using all 12 threads. And if I happen to run 2 VM's at once and still need plenty of RAM for the host. And that's about it in my case. I just got it because it was rather cheap and I'm future proof because I need 4 sticks for quad channel and it slightly complicates things. Not as much as triple channel on X58 but still.

 

[ViperXTR]

In our case, SpatialNET eats tons of ram and 8GB is chugging and crashing lel,

 

[FordGT90Concept]

I wrote a program that I've seen eat 18 GiB of RAM (could theoretically consume unlimited amounts of RAM given a complex enough problem) and it wasn't done yet and I have 16 GiB. RAM is one of those things that, if you have excess, it is wasted money, but if you don't have enough for what you're doing, it really pays off to upgrade.

 

[Caring1]

I can see the average home user needing and using more ram if they crunch or fold as you can assign how much is used.

 

[FordGT90Concept]

BOINC/F@H don't use much memory. I gave it 2 GiB to play with and it never went over 512 MiB running 8 tasks. They are CPU intensive, not memory intensive.

 

[xorbe]

Multiple vbox sessions can chew through ram in a hurry.

 

[Erazem]

I'm with you op, more ram = better.

 

[Aquinus]

In PostgreSQL, more memory means you can tweak memory usage for certain things or scale out better to more connections. Being able to allocate more to work_mem means fewer chunks to go through when doing JOINs or complex WHEREs or being able to allocate more to maintenance_work_mem which will enable faster building of indexes or imports of data, all of which improve performance without adding CPU power. Also, file caching in Linux will cache database storage files as well so whatever doesn't get used by PostgreSQL itself gets cached in memory.

So from a professional standpoint, I completely agree with you @xkm1948. More memory is almost always a good thing but, that's assuming everything you're working on isn't smaller than the memory you have, at least in my case. The best kind of [server] applications though, I believe, can scale to the amount of available memory in some capacity but, can work around having less. Clearly, on a workstation this dynamic changes a bit but, when it comes to serving, it's almost always the case.

 

[Evo85]

How much one needs to run a computer effectively is situational.

Common sense and some research will guide a person to how much they need.