vergilangleos
Posts: 11 +6
Nice job .
what about ram frequency ? 1600 , 1800 , 2100 , 2400 ? any differences ?!
what about ram frequency ? 1600 , 1800 , 2100 , 2400 ? any differences ?!
How Much RAM Should You Get? 4GB vs. 8GB vs. 16GB Tested
- Thread starter Steve
- Start date
cliffordcooley
Posts: 13,141 +6,441
Just because you waited too long to upgrade doesn't make the article false. Price increases doesn't negate the potential for upgrading later, even after managing to operate several year without upgrading. Setting yourself up to where you can upgrade is "Future Proofing" whether you want to see it that way or not.
"Once you have 'enough' memory for all your applications to run, having more memory won't increase performance any further."
That is not entirely true. Windows keeps all sorts of things in your memory if it has the memory to do so. Not only that, but when more stuff needs to go in it will usually just append it instead of replace it in your standby memory. This is why, if you are not the kind of person who just uses one time of memory intensive program on your machine, would benefit from setting your memory profile to "Background Services" over "Programs" to get these ongoing background performance kicks.
That is not entirely true. Windows keeps all sorts of things in your memory if it has the memory to do so. Not only that, but when more stuff needs to go in it will usually just append it instead of replace it in your standby memory. This is why, if you are not the kind of person who just uses one time of memory intensive program on your machine, would benefit from setting your memory profile to "Background Services" over "Programs" to get these ongoing background performance kicks.
infiltrator
Posts: 171 +31
Having more RAM is always better than having no sufficient RAM at all.
cliffordcooley
Posts: 13,141 +6,441
Windows usage is covered under what was stated. The amount of unused memory is irrelevant to system performance. The application either loads or it does not.
On side note I truly believe testing memory should be done with paging turned off as well as any other test. This way you get a true idea of when more memory is actually needed.
Paging should be a safe guard not a function to depend on. If paging was a function that was absolutely needed, there wouldn't be an option to turn it off. God forbid Microsoft to decide removing even more options of user control such as this one. I've enjoyed (for several years) not wasting 20GB of my disk space with only two files.
Burty117
Posts: 5,108 +3,689
Well no, you've got that wrong.
1) Disabling the Page File means if your machine ever Blue Screens, you will never be able to find out what caused it as the PC won't do a memory dump.
2) It doesn't increase performance at all and can actually negatively affect performance as certain applications actively look for the Page File and when it's not there it screws them up.
3) Since Windows 8.1 on wards, you only need 2048-4096MB free for a Page file to do a decent memory dump, SSD's have got cheaper, this isn't a problem any more.
4) Turning the page file off when you've got 8-16GB of RAM installed will barely change the amount of RAM actually in use, I'm talking around 12MB or so, Windows isn't stupid.
Really, the best way to get a performance boost is modifying the Page File settings, instead of being on "Automatic" make sure the page file is only on the system drive (otherwise it still won't memory dump on a blue screen) and as long as you're running a modern version of Windows (8.1 / 10) just set the RAM to start at 2048MB and Max of 4096MB. The reason this will help performance is that Windows won't grow the Page File on the fly, stopping quite large writes to your SSD usually when you're doing something quite intensive.
cliffordcooley
Posts: 13,141 +6,441
What part of I have enjoyed my PC with paging turned off, did you not understand? I've loaded two MMO's (Neverwinter & Skyforge), two browser windows, Steam, Origin, Popcorn Time, and uTorrent with only 8GB of memory.
Never have I had an application that required a page file. So don't even try pushing this page file crap is needed on me. If you want to post it is recommended then fine, but it is not I repeat not absolutely needed for everyone. I'd be willing to bet you don't have an application that requires a page file and you are so sure that you do.
Ohh and memory dumps mean nothing to me. I've already been down that road and feel they are quite often misleading in what is actually wrong with the system. Way too often memory dump readers want to analyze system file, when hardware is faulty everytime.
Windows manages the use of paging not the application. Applications create their own files to manage.
Burty117
Posts: 5,108 +3,689
I used to have a program that refused to load without the Page File, that was a long time ago I'll grant you that, but still, I've experienced it first hand. And it doesn't come up with an error that says "I need a Page File" it just comes up with what looks like a bunch of memory errors (like your RAM is failing).
Point is, it only takes up 2-4GB, is that really too much to ask? You are recommending to average users to turn off the Page File which has been known to cause instability when you start to hit your memory limit and is generally there for when things go south? ok...
So because you can't read memory dumps, recommending to average users to disable the page file is a good idea? So when a techie comes over to find out why it's blue screening he'll just have to turn the page file back on to get a memory dump?
You need to chill out a bit though, fact of the matter is that the Page File is useful, doesn't decrease performance and increases stability. Sure, you haven't run into any issues but many people have so recommending something that won't help and will probably cause issues isn't great.
On a side note, the issues I had with Battlefield 4 when I was using lots of RAM (first post), my PC probably would have blue screened or Battlefield would have crashed regularly without the Page File...
Only if you're running an older Operating System, you can use 2-4GB from windows 8.1 on wards and it will still do what it's supposed to do without eating loads of space.
Skidmarksdeluxe
Posts: 8,645 +3,290
Possibly. The 16 GB kit was so closely priced to the exact same 8 GB kit that it would've been silly to buy it in order to save about 15 bucks.
Skidmarksdeluxe
Posts: 8,645 +3,290
I dunno so much. Windows sets my swap file space at about 16 GB irrespective of how much memory I have installed. On both my 8 GB & 16 GB kits the swap file was about the same size. I know I can manually reconfigure it but I'm not bothered.
D
DelJo63
You confirm precisely normal memory operations and what occurs when real ram is consumed - - heavy paging.
Solutions have always been:
*) reduce the number of programs running
*) add more memory
Removing the pagefile is NOT a solution as that will lead to programs beginning to fail.
However Windows will always require a small pagfile on the boot partition. The custom pagefile size ccan then be mapped to the SSD.
The SSD is ideal for the pagefile - - EXCEPT, they degrade in performance over time and heavy paging to it will surely accellerate the demise
Spot on the money. 32bit users who add large amounts of RAM usually discover this or complain when they see <3gb in actual use.
Hopefully, all the advocates of large RAM knew that 64bit systems were required.
Good point. Those with heavy multitasking or
allowing many browser tabs or 'standby' open programs would need the Background profile, while the typical gamer would benefit from the Programs profile
Kind'a true, but we need to understand the implementation of Virtual Memory in modern systems.
1) VM is not a feature nor a 'safe guard' - - it's a basic architecture and is operative EVEN without a pagefile!
2) The only intertface to memory for every program (including the OS) is *through* the VM system.
When the pagefile exists, the system is running in V=V mode (ie all virtual) and
when the pagefile is non-existing, it operates in V=R mode (virtual is all real).
Running V=R avoids paging at the risk of programs asking for memory and there's none to give.
Applications have no interface to 'conditionally request' memory - - is a boolean success vs fail.
The intent of the V=R option is to support "real time" systems which control special use devices
(think manufacturing or satellites[don't go ape on this one, DOD does not operate Windows in spacecraft!]).
These systems have no multitasking or background jobs - - they are 'dedicated' to a few specific task and thus have zero chance of memory request failures.
Several members posting to this thread have discovered the marginal performance boost of running V=R and avoiding paging.
That's fine, that's their choice. Personally, I've always opted for reliability.
:Grin: I opt for Min=Max to stop expansion altogether
Expansion is a big time degredation!Historically There have been two major designs for program startup:
1) load every page into memory before giving control to the program
2) load only the page that contains the 'entry point' of the program and then give it control.
The thinking on (1) was to preload the pagefile and never access the application image again.
On option(2), the program could be operational asap and if the confg was in error, could quickly say so and let the program terminate.
The downsize however, is the system must be able to access the applications image at all time while it is running.
(which explans why you can rename, delete or move a program which is currently active).
"Once you have 'enough' memory for all your applications to run, having more memory won't increase performance any further."
This isn't generally true. The reason: Disk Caching. Any modern operating system will cache data read from the hard disk into RAM, as long as there is empty space (windows still shows it as free even when there is cached data in it!), because why not...it doesn't hurt anything and can make a difference if your application decides it needs to read a file again.
I can see this being most useful for web browsers, that often cache (on disk) graphics and content that websites include. The first time you visit, it downloads it and saves it to disk. Then you go back a few minutes later (or to another site that uses the same graphics...like the facebook like button for instance) and the web browser grabs the image from disk instead. Except if you have lots of memory available the operating system won't actually have to go and read the disk, it will know that the file kept at the requested location on disk is already stored in the memory cache so will simply return that instead...in about 100ns instead of 100,000ns.
This isn't generally true. The reason: Disk Caching. Any modern operating system will cache data read from the hard disk into RAM, as long as there is empty space (windows still shows it as free even when there is cached data in it!), because why not...it doesn't hurt anything and can make a difference if your application decides it needs to read a file again.
I can see this being most useful for web browsers, that often cache (on disk) graphics and content that websites include. The first time you visit, it downloads it and saves it to disk. Then you go back a few minutes later (or to another site that uses the same graphics...like the facebook like button for instance) and the web browser grabs the image from disk instead. Except if you have lots of memory available the operating system won't actually have to go and read the disk, it will know that the file kept at the requested location on disk is already stored in the memory cache so will simply return that instead...in about 100ns instead of 100,000ns.
codgerface
Posts: 44 +69
Fantastic article! Great experiment design and conclusions.
My hope is that this reduces so much of the misinformation and bad advice given out in forums to people looking for help with their builds. Unfortunately, it's probably an unwinnable debate when you are dealing with arguments manifested out of opinion.
My hope is that this reduces so much of the misinformation and bad advice given out in forums to people looking for help with their builds. Unfortunately, it's probably an unwinnable debate when you are dealing with arguments manifested out of opinion.
cliffordcooley
Posts: 13,141 +6,441
Well for one caching from disk to disk with a paging system doesn't increase any performance. And with SSD (which I have) that is one feature I want turned off regardless. Caching from disk to memory could help, but with paging I'm not so sure that would be the case for all memory caches.
Are we actually trying to hold on to the dark ages where paging and disk caching was needed? We are now living in a time where paging and disk caching can be outdated and move on to bigger and better things. Memory capacity and disk read/write speeds are no longer a major bottleneck.
D
DelJo63
The whole point of caching is to avoid I/O. The virtual mapping is:
Code:
Application <=> IO driver <=> cache <=> physical deviceHave you noticed that MS has implemented a Font Cache? Once the application has reference Font Times New Romain for example, the cache can avoid the performance hit of
- open font x
- read {bold,italics,normal}
- close font x
tipstir
Posts: 2,873 +206
I've tried 32 GB and didn't seem much gain. So I had gone to 16 GB very happy using it. The rest of the systems are 4 GB and 8 GB. Seems to be okay running at that much RAM since they maxed out. Test will show how much the RAM installed will do base on the PC used in the test. So test results will very. Now I use No Page File on 8 GB and 16 GB PC since there is enough Real RAM to use instead of HD RAM which tends to be slower in results.
Actually I ran for years with 6GB on my previous Win XP(32) system and it made a huge difference from previous 2 and 4GB.
The trick was simply to put everything above 3-4GB into a RAMdisk with the Gavotte RAMdisk software(the only I found to work perfectly for doing this), then I put a big chunk of a preset swapfile on that, and suddenly XP had fullsize "normal" RAM, and a few GB extra of superspeed swapfile.
With some tweaking, while having a SSD or more useful RAM total makes a bigger difference, the difference to having only the previous 2 or 4GB was BIG. And the change going from that layered approach to having a SSD and 32GB RAM on 64 bit W7 that I have now, is definitely much smaller than the difference was from not using the RAMdisk trick.
Even now, I am considering finding a 2*2GB DDR2 set to replace the 2*1GB that was my original RAM, because I might still want to use my older computer, with how much incompatibilities I keep finding W7 to have with my older games.
D
DelJo63
Right, it was a trick and it needed the Gavotte RAMdisk to make it work. From the view point of the Win/32 system, memory was still limited to 2^32-1
Paging is different. That is where there isn't enough physical RAM for all the requests that programs have made of it, instead the operating system writes some of it (hopefully the less-used stuff) to disk. In this day and age where RAM is plentiful, you should definitely invest in more if you find that your computer is paging to disk! This will give you a huge speedup any time that data is used.
Caching is actually the opposite of paging. Instead of writing RAM to the disk, it keeps data from the disk in RAM. This is how modern systems with tons of RAM can put it to use to actually speed things up (unfortunately the article didn't address this possibility). As more people get more RAM, this will be more and more useful.
I have seen this make a *huge* difference for me when compiling large bodies of C++ code. Without changing anything, but adding a bunch more memory suddenly made my compilation time drop to 20% of what it previously was. When I investigated, I figured out that for each (of the thousands of) .cpp file that I had to compile, it was including several headers (and almost all CPP files included some common ones, like <list> <string> etc). With the extra RAM, once the files were read from the disk the first time the OS kept them in the cache. Then when the compiler went to the next .cpp file and issued a read request to the OS for the same header file again, the OS was able to return in just a few nanoseconds instead of a few microseconds with the result, because it was sitting there in the disk cache. Multiply that by the thousands of .cpp files that needed to be compiled and there was some serious time savings.
tl;dr - Paging RAM to Disk is old and outdated, we really shouldn't be doing it anymore. Caching Disk to RAM is awesome and getting more awesome...this is a great use for all that extra RAM.
Here's a great illustration of the speed of different levels of the IO:
http://blog.codinghorror.com/the-infinite-space-between-words/
When you're getting stuff that has been cached off of the disk, you get the speed of the main memory 120ns. Otherwise you're stuck going to SSD 50-150 microseconds, or HDD 1-10 milliseconds.
Jad Chaar
Posts: 6,481 +976
Power users will definitely benefit since it allows them to run intensive programs + a lot of browser windows. This is a great article though for the average user, like you said.The article was aimed at the "average user" I assume... and is therefore correct in stating that there is little difference between 8 and 16gb...
But for some people, the more the merrier... I have 64gb - I generally use 32gb as a ramdisk, and it makes everything far speedier
I used to play a certain browser game (poorly coded flash, but it was bloody addictive) that I needed insane amounts of RAM to run "bots" to play for me while I was asleep/working (don't ask, it's just that addictive!!). I used to run out of RAM with 16gb...
I understand I'm in the minority - but there are others like me!
Similar threads
