Solved Is A RAM Disk Effective With Just 2 GB Of RAM?

Magicmicro / CUSTOM BUILD
September 4, 2014 at 04:20:03
Specs: Win 7, SP1, AMD 4130 FX+ 8/12 RAM
Hello All,
I've been using a RAMDisk software on my mainframe computer (AMD 4100FX, Quad Core with 12 GB RAM); of the 12 GB RAM, 4 GB is dedicated to the RAMDisk which handles all the Windows TEMP files. Because of the base amount of RAM, I know its fine to split it 1/3 | 2/3, and the computer certainly seems faster, to say nothing of the break my hybrid hard drive gets from not having to thrash files all the time.

I've installed this same RAMDisk software on my laptop and netbook. Each has 2 GB of RAM, of which about 1000 MB is free. I've installed a 512 MB RAM Disk with compression, and pointed the temp files there. Now the question: is there a way to measure whether a 512 MB RAM Disk on a 2 GB machine is both effective and helpful to speed up the system? Something similar to the way that ReadyBoost can be measured to see if its effective.

All help appreciated.

Regards,
BearPup


See More: Is A RAM Disk Effective With Just 2 GB Of RAM?

Report •

✔ Best Answer
September 4, 2014 at 09:44:41
Set up performance logging. Keep track of Memory -> Page Reads/sec and Memory -> Page Faults/sec. For every page fault that generates a page read, you're reading from the page file. This probably means your disk is thrashing. Compare the results with/without the RAM disk. If your computer has more reads per page fault with the RAM disk, you're causing more harm than good.

More in depth: http://technet.microsoft.com/en-us/...

How To Ask Questions The Smart Way

message edited by Razor2.3



#1
September 4, 2014 at 06:36:52
AMD 4100FX, is not a mainframe, it is a desktop. Mainframe computers cost millions of dollars and fill whole rooms.

RAM disks on modern computers are a waste of time, regardless of how much RAM you have available, even more so on a computer with only 2Gbs of RAM.

RAM disks were devised at a time when hard disk were painfully slow. That is not the case any more and using RAM as a RAMDisk is a waste of resources that could be better employed elsewhere.

Stuart


Report •

#2
September 4, 2014 at 06:46:33
Measuring the effectiveness of a RAMDisk is not a simple thing but is far more complex than most people imagine. But in this situation I doubt it is necessary.

I am not a big believer in the use of a RAMDisk with a modern OS. Modern systems have a complex and sophisticated caching system that provides most of the benefits of a RAMDisk with fewer of the problems. The system cache is very good with a 32 bit OS, even better with 64 bits. The big problem is that the RAMDisk dedicates a substantial portion of RAM to one specific purpose, even when it may not be in use, even when the memory may be put to better use elsewhere. The size and contents of the system cache varies dynamically according to the situation.

In most cases a RAMDisk is an inefficient use of memory. There are situations where it is justified but they are unusual. With 12 GB RAM you may have such a situation. I would tend to doubt it but I grant the possibility. With only 2 GB RAM the possibility of that being true becomes very small indeed. With a 64 bit OS that possibility would be vanishingly small.

Don't be misled by the memory usage graph. Most memory that is shown as available is not free but is in fact providing substantial performance benefits. Only memory that is labeled as "Free" is completely unused.

Dedicating 512 MB RAM of 2 GB to a RAMDisk is almost certainly going to impair performance, not enhance it.


Report •

#3
September 4, 2014 at 08:35:47
@ StuartS: Excuse me. I call it my mainframe. If you want to be a snob about it then technically you are correct. However, I will continue to call it my mainframe.

Report •

Related Solutions

#4
September 4, 2014 at 08:40:11
@ LMiller7: Thank you for responding. And not to take this personally, but I'm looking for more than opinion. I'd like to find a way to measure the effectiveness myself. The parallel with ReadyBoost is on point. Many people dismiss ReadyBoost as a gimmick, not worth it, etc. . I on the other hand learned how to measure its effectiveness and find it can help in certain situations. I'd like to do the same with RAM Disks, even if I have to learn new things!

Report •

#5
September 4, 2014 at 09:39:31
@ StuartS: Excuse me. I call it my mainframe.

It is not a matter about being a snob, it is a matter of using the correct terminology so everyone understands what everyone else is saying. It is like calling a desktop computer a CPU or formatting when you mean install an operating system. It marks you out as a newbie that has heard a word that sounds good but and has little understanding of what it means.

If you want to continue to confuse people you communicate with then carry one.

Stuart


Report •

#6
September 4, 2014 at 09:44:41
✔ Best Answer
Set up performance logging. Keep track of Memory -> Page Reads/sec and Memory -> Page Faults/sec. For every page fault that generates a page read, you're reading from the page file. This probably means your disk is thrashing. Compare the results with/without the RAM disk. If your computer has more reads per page fault with the RAM disk, you're causing more harm than good.

More in depth: http://technet.microsoft.com/en-us/...

How To Ask Questions The Smart Way

message edited by Razor2.3


Report •

#7
September 4, 2014 at 10:35:30
What I hear you saying is that labels are more important than what is being said. If you think my talking about a mainframe marks me as a newbie that is your perception - label me if it makes you feel better, and in doing so, you don't hear a word I'm actually saying. That is your loss.

BTW, I also have a device that I call "mini", even though it really isn't a DEC PdP8. I bet that makes me look like a real newbie, especially as I carry it with me all over the place. It must just drive you crazy that when people see my mini they've exclaimed to a one, "I'm jealous!" and haven't flown off in fear of the newbie.


Report •

#8
September 4, 2014 at 10:37:16
@ Razor 2.3: Thank you, Thank you. Exactly the info I was looking for.
I very much appreciate it.

Regards,
BearPup


Report •

#9
September 4, 2014 at 11:50:40
@ Razor 2.3: Well I did as you said and got the following results: I'm better off with the RAM Disk!

With the RAM Disk enabled Page Faults/sec stayed constant at 0 to 100 over a 1 minute period. During the same period, Page Reads/sec stayed constant at around 15 per second. However when I unmounted the RAM Disk, Page Reads went to spikes of 50, 94 and 100 during the same 1 minute interval. Also observed were fewer Faults being manipulated some how.

Again, my thanks for your help.
Regards,
BearPup


Report •

#10
September 4, 2014 at 12:51:49
What I hear you saying is that labels are more important than what is being said.

Yes they are if you call an motor car an aeroplane then wonder why it wont fly.

Stuart


Report •

#11
September 4, 2014 at 12:55:07
You'll need a longer testing interval than that. Windows' state can change suddenly without any prompting from a user, so you need an extended test. At a minimum, two days of normal usage, one day with the RAM disk, and the other without. You should be able to create a data collector set to save the data to a file. After your tests, you can compare the averages for both sets of data.

Remember, you're looking at the ratio of page reads to page faults. Every fault can generate a read, and that ratio is what you're watching. An increase in page reads means nothing if page faults increase proportionally, as page faults are one measure of system activity.

Instructions on making a data collector: http://technet.microsoft.com/en-us/...
Not mentioned in that article is that data collectors do not start with the OS. You'll have to do so yourself after any reboots, or schedule it to do so: http://blogs.technet.com/b/jeff_sto...

How To Ask Questions The Smart Way


Report •

#12
September 4, 2014 at 13:19:38
@ Razor2.3: Again, many thanks. Guess I thought of it as an Instamatic and not a long-term exposure. However, as its the right way to do the test, that's how we'll do the tests.

Thanks too for the heads up on Data Collection Sets. I'll simply start them when I first boot-up, and make sure any manual reboots cover that base. Its different than when I was calibrating my hard drive temps, where I was looking for an instant snapshot at boot-up and then calculate it from there.


Report •

#13
September 4, 2014 at 13:30:54
@ StuartS: It is sad, really sad. You actually, really think that anyone would entertain the thought that I was trying to fly my automobile? Really?!

Report •

#14
September 4, 2014 at 14:15:12
Well if you were to call it an aeroplane it would be reasonable to assume that it could fly because that is what aeroplanes do.

Stuart

message edited by StuartS


Report •

#15
September 4, 2014 at 20:13:37
@ BearPup, are you still running DDR3-1333 RAM on your "mainframe"? You can tout the use of a RAMDisk all you want but the RAMDisk doesn't make up for the performance hit your system is taking by running the wrong RAM. I couldn't find a memory scaling article specific to the FX-4130, but this should give you a general idea:

http://www.pcper.com/reviews/Proces...


Report •

#16
September 4, 2014 at 21:16:45
@riider: Thanks for your response. To answer your question, I'm running a Magic Micro Memory Kit spec'd as DDR3 PC10600 / 1333MHz.

And to be honest, I'm not understanding whatever point that you are trying to make. suppose you start there: so what I'd suggest is for you get any and all information onantimax


Report •

#17
September 5, 2014 at 08:21:13
Let me try again.

The memory controller built into the FX-series of CPUs supports up to DDR3-1866, so ideally, that's what you should be running for optimal performance. By running DDR3-1333, the memory bandwidth is 19% less than optimal (as shown in the link in response #15), therefore your system is theoretically taking a performance hit.

You seem to think your RAMDisk config is a wonderful thing. I'm not going to argue the fact. But the added performance you may be getting from it does NOT make up for the performance loss due to running DDR3-1333.

A Question of RAM

Also, we touched on ReadyBoost in the following thread, responses 6 thru 9:

Improved Video Quality Needed (Not Gaming)

message edited by riider


Report •

#18
September 5, 2014 at 08:30:42
@ Riider: Thank you, I now understand. The memory kit was a sale item when I was specing out my machine, and it saved me about $100 (special + cost of better memory); pure economics. And yes, I do believe using the results I see and am now testing for, that there is a performance gain, and on my mainframe (not a server), a saving of hits to my hybrid SSD / HDD drive (more economics, about $75).

My question really goes to the use of a RAM Disk on a 2GB machine. And now I have the tools to measure it. I'll wait and see what the results are before passing judgment.


Report •

#19
September 6, 2014 at 09:27:17
Please don't call your computer a mainframe. You have a desktop computer. Click on the highlighted word in response #15, THAT is a mainframe. And yes, terminology does make a difference. A couple of other incorrect terms I've seen for a desktop/tower are CPU & hard drive.

"My question really goes to the use of a RAM Disk on a 2GB machine"

IMO. your question was answered at the bottom of response #2 - "Dedicating 512 MB RAM of 2 GB to a RAMDisk is almost certainly going to impair performance, not enhance it."


Report •

#20
September 6, 2014 at 09:47:18
Once and for all, if I want to call my computer Bill, or Mary, or Mainframe, I will call it that. As to your picture, sure looked like a server to me. Guess you just can't rely on pictures or words. Despite it all, everyone certainly understood that I wasn't talking IBM 370/70.

As for comment #2, people are free to express their opinions all they want. And I got opinions from "Of course it'll help, just don't use compression or you'll lose your gains" to "RAMDisks never help (regardless of configuration)". What I was looking for was a specific, objective means of measuring whether or not the RAM Disk was helpful or not.

So now that everyone has chimed in with their opinions - including dumping on the OP - as far as I'm concerned the only reply that actually addresses the asked question is the one I marked as the Answer, #6. And as far as whether or not the RAM Disk is helping or not, preliminary numbers suggest that it is helping, with lots more testing still to be done.

That's the nice thing about objectivity, it doesn't care what the outcome is, only in how the outcome is determined.


Report •

Ask Question