• Welcome to Overclockers Forums! Join us to reply in threads, receive reduced ads, and to customize your site experience!

Is there any advantage to buying more RAM than you need?

Overclockers is supported by our readers. When you click a link to make a purchase, we may earn a commission. Learn More.

SPL Tech

Member
Joined
Nov 28, 2006
Often I'll see people build ridiculous systems with like 64 or 128GB of RAM, which is clearly unnecessary for gaming. However, I did wonder if there would be any speed benefit to buying said more RAM? The main theoretical benefit I could see would be that prefetch could store more information in the RAM and thus there is a higher chance that when something in the OS or an app needs to load, it's prefetched into the RAM and will be accessed from RAM instead of your SSD. Then, if you had say 128GB of RAM, you could theoretically prefetch an entire game into RAM and thus load every file related to a game directly from RAM. That would obviously be far faster than accessing it from an SSD, but in reality does this actually happen?
 
Ram disks are nice. Moving browser caches and such as well as applications that utilize a faster disk to a ramdisk can make quite a difference. Photoshop, imaging software, etc. I dont have any clue about prefetch though. I am curious how you would set that up?
 
Depends on how much ram you will use. I keep 50 webpages open and play BF1 at the same time so I use 11 GB. If I only had 8GB of ram it would take more time when going through the web browsers tabs as they would be swapping from the storage drive with a swap file.
 
Ram disk maybe but... no. Your theory in the first post would he nice, but it doesn't work that way. I have 32gb (was sent for review, not purchased..I'd have 16 if I bought it) and it never comes close to fidling it 'just because it can'.


Depends on how much ram you will use. I keep 50 webpages open and play BF1 at the same time so I use 11 GB. If I only had 8GB of ram it would take more time when going through the web browsers tabs as they would be swapping from the storage drive with a swap file.
Hopefully, wingy, your storage drive is an ssd...or put it back on you OS drive, which I presume is an ssd.
 
Yes my OS is on my SSD storage drive and is fairly fast when swapping files when using more than 8GB of memory for the time I did not have 16GB.
 
Well I did not know what the OP had for a drive, so in my example I referred to my drive as a storage drive.
 
Agreed, outside of server/virtualization or very specific applications that will take advantage of as much ram as you can throw at it (I imagine very very heavily duty animation/3D modeling/effects/etc)

 
Having more ram than you need allows the OS to use it as a disk cache. I've used this as a workaround to slow disks before SSDs were available. As it is built into the OS, you don't generally need to do anything to gain the benefit. I believe optimisations are different for SSDs than HDs in Windows, so only HDs might gain a benefit and you may be limited to SSD speeds on SSDs. Note this only applies to reads, I don't believe there is any significant write caching as there is a risk of data loss should something happen before the data is committed.
 
Simply if you can't use RAM capacity then additional RAM is useless. Only RAM which can be used has any point. You won't see any difference above ~32GB if applications can use ~30GB but you have 128GB.
Disk cache isn't using more than ~2GB ( or you have special soft which can use more ).

You can compare it to using 300GB on your HDD which is 3TB. You will have a lot of free space which won't be used but you actually paid for that. That's why for OS drive most users pick small SSD so they trade capacity to speed at similar price.

However there are some exceptions. Full SQL will use more RAM than it actually needs ( I doubt you are using it at home ).
There are also architecture matters. The fastest memory modules are 8GB so users who have X99 motherboards may need ~10-12GB but they won't buy 16GB kit ( 4x4GB ). Instead they get 4x8GB which let them run at tighter timings and higher frequency. The same, in some applications dual rank memory is faster. The best dual rank memory modules are 16GB so if you really need it then you get 2x16GB or 4x16GB when you really could live with 16GB total RAM.
 
There are also architecture matters. The fastest memory modules are 8GB so users who have X99 motherboards may need ~10-12GB but they won't buy 16GB kit ( 4x4GB ). Instead they get 4x8GB which let them run at tighter timings and higher frequency. The same, in some applications dual rank memory is faster. The best dual rank memory modules are 16GB so if you really need it then you get 2x16GB or 4x16GB when you really could live with 16GB total RAM.

I have this problem on a couple of systems.

My single socket Xeon system runs 4x4GB single rank, although I don't need 16GB in it at all. I would like dual rank in this application, but decided against throwing yet more cash at it as it meant going to 8GB modules.

My dual socket Xeon is similar, currently running 8x4GB dual rank for 32GB. I don't need more than 8GB of total ram on the system, but I do need the bandwidth. I actually bought the system configured with 64GB (16x4gb) as I feared the modules may be single rank, and as used kit, the price differential wasn't significant in this case. I never did get around to testing if going beyond dual rank per channel impacts performance in any way. If only it wasn't registered ECC so I could swap it with consumer systems. Feels wrong to have 32GB of ram sitting idle in a bag. Note it isn't in the server as I had stability problems with it fitted, and haven't had time to diagnose further.
 
Woomack's post jogged my memory a bit and I remembered something about Linux (donk ask me what about that post made me remember but it did :))

According to this page : http://www.linuxatemyram.com Linux does exactly what the OP was suggesting.
 
So does Chrome and Firefox, but it looks like Linux does it a little better.
Well Im not sure they are the same? To my knowledge the web browsers create a cache directory on the hard disk. To see benefit there you want to put your cache directory on your fastest disk - making your browser faster.

Linux on the other hand does this system wide into available ram. Ive run across this phrase alot: "Unused ram is wasted ram". I have also read vague articles that say Linux has moved beyond that statement somehow, but my savvy doesnt extend to comprehending the inner workings. Not in my attention span anyway. I certainly plan to revisit the topic someday.
 
Linux or some applications like web browsers are locking some RAM in case they will need it. The same is doing MS SQL. The more RAM you have, the more RAM SQL will use but it doesn't mean it will make it faster. There is some limit which is helping in performance. Like when all applications need 20GB RAM, lock 30GB RAM but you have 48GB in total then still 18GB won't be used by anything and/or RAM in use will refresh fast enough to not use what's left above that 30GB. It's only an example but I see that in business environment where I work for some long years.
 
Linux or some applications like web browsers are locking some RAM in case they will need it. The same is doing MS SQL. The more RAM you have, the more RAM SQL will use but it doesn't mean it will make it faster. There is some limit which is helping in performance. Like when all applications need 20GB RAM, lock 30GB RAM but you have 48GB in total then still 18GB won't be used by anything and/or RAM in use will refresh fast enough to not use what's left above that 30GB. It's only an example but I see that in business environment where I work for some long years.

How do you know web browsers are locking some Ram?
 
Back