Discussion on:

How do I use a RAM disk to help speed up disk-intensive applications?

RAM disks used to be a common power user device for increasing performance; do you still use RAM disks for this purpose? What application do you recommend for RAM disks?

ramdisk.sys.

I use ramdisk all the time. When we edit audio with goldwave, we have wav files expanded in the ram disk. in this way, non-linear edits is very fast!. with regular disk (even raid0), the speed is very slow, even with coreI7 and new cpus.

I use ramdisk in rendering sets in truespace as well.

In 1990/91 I run AutoCAD 10 from RAMDrive. I run it the way what was later known as "DOS mode".

I copied WHOLE program to RAMDrive. It was around 3Mb. Yes, it was *three MEGs!* - complete design/draughting software which today Autodesk delivers on 1x DVD 32bit version and another DVD for 64bit to be installed.

It easily beat AutoCAD 10 DOS Extender version when it became available. And we compared 2 PCs:

- my 286 PC running at 11MHz (4Mb RAM, 3Mb used as RAMDrive; part of DOS and some drivers loaded HIGH, which is between 640 and 1024kb for those who are too young to know). I also run some *memory resident* programs: Sidekick, plot and print spooler (2 different queues!).

- competing machine was 16MHz 386 with the same 4Mb RAM "virgin", only AutoCAD and DOS with EMM (Expanded Memory Manager).

We timed opening the same drawing from the same floppy (1.2Mb, 5.25 inch).
Difference was around 25%. Editing the drawing (i.e. using various AutoCAD features) was much, much faster on my PC. Starting the program was much faster as well - obviousy copying it from Hard disk to RAMDrive I put in autoexec.bat ......

Today I use AutoCAD 2011 and AutoCAD Electrical 2011 on multicore, multi-GHz PC with 2x 22" screens, 5-button (+2 rollers) optical mouse, listen to the radio on the same PC etc. But I still cannot forget excitement and the pleasure of those days. Single 14" VGA screen didn't look bad; nobody had more! Cleaning mouse's ball didn't bother us either.

I have to try it again sometime...

32-bit OS (Vista), 4GB RAM. The OS can only address 3 GB. Will a 1GB RAMdisk use the part of RAM that the OS can't access? or will it take that GB out of the 3 that the OS is using, leaving only 2 for the OS?

Not. It's an architecture issue.

VSuite Ramdisk (Free, Server edition etc) does exactly that. Especially helpful in 32-bit systems. In the setup menu there's an option to "Use OS Invisible Memory" for 32-bit systems with non-addressable RAM, and you may use a large portion of that RAM in such a system with this RAM Disk application, both effective and smart to do, since it's sitting there non-utilized.

Personally all my systems here are 64-bit and don't have that issue, but it IS an option with VSuite Ramdisk, both free, and the various paid versions. so there you go!

Here's the URL of the company whose RAM disk software we use, that is VSuite Ramdisk from Romex software:

http://www.romexsoftware.com/en-us/vsuite-ramdisk/index.html

I have a PC OEM desktop with 24GB physical RAM and I have a grand total of 6 working RAM disks with the Server Edition of the program...most are non-volatile but two keys are a 10GB scratch disk for audio, video and PS work; 512MB Ramdisk for my Chrome browser is the 2nd Ramdisk that I use with the Image Option, where it's saved during restarts, shutdowns, etc to cache memory on my main RAID 0's HD space.

The application uses Dynamic RAM Allocation also, ie it's not addressed until used; the 10GB scratch disk is set up where I can actually write and read fastdata to it in NTFS format, with NTFS compression for indexing for example.

The Chrome browser Ramdisk, by comparison, also uses the Image directory option, is not SCSI but rather Direct I/O Disk in options, which is faster than the SCSI option but not recognized by Disk Management or any other storage utility, so you cannot use it like an SCSI non-Direct I/O Disk.

The SCSI Ramdisks we use with DDR3 RAM in a Core i7 environment have read and write speeds up in the 5,000MB/sec range (five thousand MB/sec!), which compares very favorably with normal read/write to disk speeds of 400-600MB/sec for our RAID 0 boot volume in this particular computer.

You can see where that is a huge advantage for work I mentioned above, on the 10GB scratch disk for example.

Since we started using the Ramdisk software my personal productivity is up perhaps about 25-40% overall on big I/O projects, and simple things like the Chrome browser action is very fast, web pages are rocket-like launching, recurring memory functions are also rapid.

Overall we are pleased with the software! (there's a 15-day trial period for the paid versions of the program where everything works 100%, and you can see where it benefits you or not.)

I haven't used a RAM disk in years, but back when I had a 286 with a whole megabyte of RAM, DOS could only use 640K. I don't recall what application I was using (probably RAMDISK.SYS), but it used the upper 360K for a RAM disk, which allowed me to move things around a lot faster. I haven't checked yet, but I'm sure there are applications that will do the same for today's hardware.

however, the truly unreachable RAM for the 32-bit Desktop Windows OS is above the 4GB line,

which is not necessarily a limitation for a 32-bit server OS starting with Server 2003 certain versions are capable of using much more RAM
see the MSDN article here:

http://msdn.microsoft.com/en-us/library/aa366778%28VS.85%29.aspx

SuperSpeed's RAMDisk and RAMDisk Plus will use the RAM above 4GB on a Desktop OS
- when you use RAMDisk Plus you can also shove the pagefile in the RAM above 4GB

It would be great if the software worked...

I used to use a RAM disk. I think I used Himem.sys in the config.sys file to set it up, but I forget the exact syntax.
. . and was that in Expanded or Extended memory?

use:
- a system with 8GB RAM
- a 32-bit OS
- stuff the pagefile in the upper 4GB of RAM using: Gavotte's RAMDisk or SuperSpeed's RAMDisk Plus

That's a great idea!

How well does it work? Obviously the pages will be stored in physical RAM, but the OS won't know that so will continue to have page faults and will happily swap things out of RAM to your page file to make space to fault the old page back in. Does this constant paging slow things down signifiantly, or does it run pretty similar to a 64-bit OS with 8GB RAM?

I've been thinking about this a bit. The setup looks good but does it not require a 64bit OS? The ramdisk utility is running through the OS's memory addressing isn't it? Or, does ramdisk somehow address and manage RAM chips beyond the 3.5'ish gigs a 32bit OS is physically able to see?

I do see how this would benefit 32bit apps on top of a 64bit OS though. Win64bit can see your 8 gig of ram but you have Photoshop32bit which itself can only make use of 3.5'ish gigs. Stick Photoshop's swap file on a ramdisk and now the 32bit Photoshop is addressing 3.5'ish of RAM with the rest in a swap file running at the speed of the ram.

I've personally got an 8 gig system with 4.5 gigs wasted when I boot over to WinXP for a bit of gaming so I'm honestly asking if I've missed something. I'd love break out of the 3.5 gig limit even if it was using a swapfile trick.

the RAMDisk program(s) does that

with Gavotte's and SuperSpeed's RAMDisk Plus you get to use "Un-managed RAM"

and the 4GB limitation in 32-bit Desktop OS is a programmed and hardware limitation

as PAE in certain versions / editions of 32-bit Server OSes on server boards can address and use 8GB - 128GB RAM

this capability was available as far back as win2000:
- Server 4GB
- Advanced Server 8GB
- Data Center Server 32GB

(even the 64-bit desktop editions have programmed limitations)

see MSDN article for limits:
http://msdn.microsoft.com/en-us/library/aa366778%28v=vs.85%29.aspx

cheers for that clarification. So Ramdisk actually addresses and manages the extended RAM beyond 32bit's 3.5.

This makes no sense at all to me. A page file gets used when there's no memory. By limiting the amount of memory an application has by having a ram disk you are increasing the chances of swapping to the pagefile.
If you have large memory needs, it's best to leave your system alone. Programs that perform a lot of HD access would make a lot of sense to use a ram disk.

If you have over 3.5 gigs of ram and 32bit WindowsXP/Vista/Win7 anything beyond that amount is extra RAM the 32bit OS can't make use of.

This ramdisk utility, on a 32bit Windows, is able to make use of that extra RAM above what the OS can see. It can turn the extra RAM into a "hard drive" until next reboot.

8.0 gigs RAM installed
3.5 gigs RAM visible to the 32bit OS
4.5 gigs RAM used as "hard drive R:\"

Since R:\ read/writes at the speed of ram instead of the much slower physical hard drives, placing your swap file here gives you 3.5 gigs RAM and 4.5 gigs Swap running at the same speed. "Swap" behaves like extended RAM instead of a cache file on slow storage.

Never really had enough RAM to use one with Windows already running.

But the configuration and diagnostic software I use at work uses RAM disks extensively, primarily because the register PCs don't have internal hard drives.

Does installing RAM disck create a new partition on the hard drive? If I uninstall at a later time will my disk space be restored?

RAM disk do have performance gain. Especially, the application must work with temp file, i.e. office (cache file), crystal report (generate report). But, in fact, the performance gain only valid if your OS don't use swap disk. Therefore, you must balance the ram disk against swap disk.

I had setup a ram disk as temp drive, performance gain is noticeable cause my XP installed with 4G ram but XP only use 3.2G. I reserve 1 G ram as temp as well as eclipse workspace.

A swap disk creates extra "ram" by borrowing hard drive space to "swap" out part of what is stored in ram. A ram disk creates extra hard drive by borrowing ram space to present as storage. The two provide different functions which are closer to a mirror of each other.

Since a swap disk is a blob of file on a hard drive and a ramdisk is a simulated hard drive using part of the system ram, you can host your swap file on your ram drive. I'm not sure what benefit there would be over simply using the full amount of ram as physical ram rather than swap ram on a ram hosted virtual drive.

I think I'd do what he did. XP only recognizes 3.2 GB, so I'd give up a couple hundred meg and configure a 1G RAM drive, then use that as the location for the swap file.

All the benefits of swap space without the wait for the hard drive.

I used it before. Though it speed things up quite a bit, but it has a serious flaw: scheduled windows backup will not be successful. A less serious flaw is, you must clear the temp folder or some programs, like a print spooler, will stop working due to "not enough memory".

Will this applicatiob eb suitable for Win Vista or 2007 running at 64bits?

How is this different/better than ReadyBoost ?

With ReadyBoost, the data has to pass through the drive controller and the USB hub, then leave the mainboard, and the maximum speed of the transfer is usually at USB 2.0 speeds.

With a RAM drive, data remains in memory and all data transfers at memory bus speed.

Some say ReadyBoost isn't that useful - gives you that "extra" memory but does it really improve things?

I added a 1GB SD card [which I configured with ReadyBoost] on my netbook which already has 2GB of RAM. I can't see any improvement. Maybe I needed a larger card.

Between RAM and the CPU things are stinking fast.

Between RAM and IDE/SATA, things are fast'ish.. but not stinking fast. This is your swap speed slow down; you get more memory but running at the speed of the storage not memory.

Between RAM and USB, thinks really slow down. Like 0.680Gps versus SATA at 3.0GPS.

I just don't see how ReadyBoost using USB bus speeds is any different from simply using a swap file on the SATA bus which would be over three times faster. I'm open to the idea that I'm not understanding something inbetween though.

Your 3.0 Gps is the data rate once the sector is found. But your disk has to first seek to the sector, which takes several millisecs each time. The USB has no seek time.

I'd still rather just use SSD on my 3 Gig SATA bus. I've never seen the same overall performance come out of usb attached storage even with the lesser seek time.

Cheers though. I thought it was the speed of the bus more than the seek times.

I see huge speed improvement in LOAD Speeds with software like Adobe Photoshop, Premiere etc when I have readyboost enabled with a 4Gb usb fast drive.

just run on the command line,

mount -t tmpfs -o size=2gb tmpfs /myramdisk

Note that the software mentioned in this blog is a release candidate. Seems Dataram is using people as guinea pigs. Should offer a stable version and optionally this RC. It's an RC - it's in the version number from their web site.

I don't think your system will speed up by using the RAM disk to store the temporary internet files.

If anything, moving the temp folder [or if possible, the swap file] into the RAM disk.

"If anything, moving the temp folder [or if possible, the swap file] into the RAM disk."

That doesn't make sense. If I have free RAM, I shouldn't put the swap file on a RAM disk, I should configure my computer to use RAM before swap file... a much better (and cheaper) choice on a Windows PC is "Ready Boost", which moves the swap file to a flash drive.

J.Ja

causes the PageFile to operate at the speed of RAM with no Disk subsystem I/O delays

... it still doesn't make sense. Look at it like this:

System RAM: 12 GB
Size of swap file: 4 GB
Total RAM used by applications: 11 GB

Now, if I am running like this, I should be using 11 GB of RAM, all in physical RAM. If I push the swap file to RAM disk, what happens? The system using 8 GB of physical RAM for apps, then goes to swap, and pushes it all... guess where... into RAM.

It's totally illogical.

If you want to force 100% of your RAM usage into physical RAM, the solution is simple. Disable swap file. Why have the system jump through the hoops and layers of putting a swap file on RAM disk? I guarantee you that pushing the swap file onto RAM disk will be slower than not using a RAM disk for it, and in terms of the total RAM available, it is *identical*.

J.Ja

an x86 32-bit windows desktop OS system won't use more than the first 3.xGB with the last part of it going to managing system devices etc.
leaving you with 8GB of un-managed RAM that can't be used for anything but a RAMDisk

disabling pagefile in a 32bit desktop OS doesn't always produce the desired result either, especially for huge apps using huge files

I run my XP 32 4GB RAM system with a 2GB pagefile and it always has at least a 40 - 50% VMem / pagefile allocation for system processes and apps.
ie.
if Mem Usage for any one process is 500MB, then VMem / pagefile usage is around 400 - 500MB

I've never had any decent level of success with setting:
"No PageFile / 0byte pagefile"

shutting if off makes the system crawl especially when editing huge 1.8GB 24-bit audio files

I've had better success with creating a pagefile at the very beginning of a second HDD than setting it to 0Bytes

I've heard that Windows is actually pretty good at managing the swap file. If you create a system managed swap on multiple drives; it'll use the apropriate drive/swap. I'm guessing this means it'll swap to D when your making heavy use of C for active files or it'll track which drive provides the best performance?

I believe it was an interview with the guys behind Sysinternals but it's a long while back now and the details are faint.

Agreed. I always set my swap file to system-managed. I've always considered "tweaking" the swap file to be stupid. If you're hitting the swap file a lot, don't "tweak" the thing that is an order of magnitude slower than your actual RAM. Get more RAM instead to make sure you never hit the swap file hard. My second thought on the matter is that some really smart person at Microsoft has spent a lot of time writing code to manage the swap file. Chances are they have thought of all sorts of things I haven't, and will therefore handle the swap file far better than any setup I come up with.

Anyway, I haven't seen the interview you refer to but it will no doubt be by Mark Russinovich. And if he says Windows is good at handling the swap file, you can take that as a fact.

I don't doubt that Microsoft can hire some of the smartest developers out there. In past though, it seems the quality of the developers work gets degraded by the OS design choices and general inter-office politics. Most of the time I'd consider if a tweak does refine something towards my uses or beyond the factory defaults.

I believe it was Mr Russonivich (sp?) long before MS bought Sysinternals and hired him on. He did state pretty clearly that one of the things Windows does far better than tweaked hard setting is swap management. If I can track down my PDF or the original URL I'll post it for those interested.

a static pagefile will not fragment

a system managed page file will eventually be all over the drive

especially if you frequently get the
"windows is running low on virtual memory ..."
popup dialog

I have and always will set a static pagefile of a min 2x the installed RAM to a max of 2GB

ie.
- a system with 512MB gets a static 1GB page file
- a system with only 128MB RAM gets a 512MB page file
- a system with 2GB and up gets a 2GB page file

doing this I have never received the
"windows is running low on virtual memory ..."
popup dialog


etu

@Who Am I Really

If you see the "Windows is running low on virtual memory" then there's a great chance you would benefit from more physical RAM. Seeing that message means you're out of physical RAM, and just about out of swap file too. If you're hitting the swap file so much that it's running low, then performance has likely already taken a massive swan dive. The answer to that is more RAM.

Setting a static page file isn't actually that useful. Firstly, setting system managed you basically get a static page file anyway. It can increase (and later decrease again) if needed, but in normal circumstances this doesn't happen much. It's not like your swap file is constantly shrinking and growing as your memory usage changes. Think of system managed as "a static swap file, but with the ability to grow if it has to". It's able to avoid the "Windows is getting low..." message in situations where a static swap file wouldn't be able to.

Secondly, it doesn't matter if your swapfile is fragmented. Pages get swapped back in to physical RAM on demand. There's no way for an application to tell Windows "hey, I'm about to use a whole bunch of pages now that I haven't used in a while... can you go ahead and load them all into physical RAM". Paging is transparent to applications - they aren't even aware that their pages have been swapped. So as pages are needed, Windows swaps them in. This is adhoc, there are never streams of pages being paged back in in a single hit, therefore sustained transfer speed isn't important. It's seek time that matters, and fragmentation doesn't affect this.

I also question the wisdom of setting your page file based on some multiple of your amount of RAM. Soooo many sites recommend this (all with different multiples!). It really makes me laugh. If the "perfect" page file size happens to be some neat multiple of your RAM then that's purely coincidence. In reality, the appropriate page file depends on what you are doing. No multiple or any other formula will be able to tell you what *your* memory requirements are. Analyse your needs and set the page file accordingly. Or, given the small cost and large capacity of hard drives, just set it high and forget about it. Then again, if you're going to do that you may as well set it to system managed. That gives you the benefit of being able to go very large if needed, but without actually wasting that space if it's not needed.

Of course, all of the above is over-analysing something that just isn't that important. My advice has always been that if you see *any* benefit from tweaking page file settings, you should instead go out and get more RAM. You will see far more improvement.

Yeah, in that scenario, it actually makes sense (at a technical level). That being said, it points out that you need to install an x64 bit OS, badly. And probably upgrade the RAM a bit too, because for that you are doing, 4 GB of RAM really isn't going to cut it one way or the other. When working with a file that large, in my experience (having written applications which work with large multimedia files in the past), there is a multiplicative effect on RAM usage. For each byte of file, the application will need, say, 1 byte for the file data plus 3 bytes of application data. For example, if you are working with a 100 pixel image, not only does it have each pixel in memory, but it has data about each pixel too (like its position). If the original file had any compression in it, what the application brings into memory will be the size after decompression, naturally.

So, if you are working file files like that, coming from someone who used to do the same on a Vista machine with 4 GB of RAM, I can tell you... upgrade. No tweak you make with RAM disk will have the same benefit of upgrading.

J.Ja

system is not optimized for x64

plus, the only way I can use the DDR2 pc6400 800MHz RAM is with 4GB as 2x 2GB or 4x 1GB modules

if I want to use 6 - 8GB I have to drop back to PC5300 667Mhz or PC4200 533Mhz
(MB/chipset limitation)

and really the only time it's "slow"
is if I try to set it to no pagefile

I have the 320GB C:\ short stroked to 39GB where the OS /Programs an page file are located, and all user files are stored elsewhere

System: (custom build: 2007)
OS:
XP-Pro SP3
MB:
- Intel DG965RY
Proc:
- Core2Duo e6600 @ 2.4GHz
RAM:
4GB OCZ PC6400 800MHz 2x 2GB (5.5.5.15)
Optical:
- 3x CD/DVD DL +-RW multi-RW DVD RAM
HDD:
- HDD0 320GB Seagate (C:\ 39GB, F:\ 254GB, L:\4.8GB)
- HDD1 1TB Hitachi (D:\)
- HDD2 1.5TB WD Green (E:\)
- HDD3 1TB WD Green AFD (J:\)
- HDD4 1TB WD Green AFD (K:\)

I use XP x64 at the studio
on a Dual Socket Dual Quad Xeon 2.33GHz with 8GB RAM PC4200 533MHz FB DIMM

and there's not much difference in performance between these systems

I'm running the same Audio Apps on both,
the differences appear when running FFT processes against the audio and file saving
- FFT Processes run about 20-25% quicker
- saving a file is about 1/4 less time (mostly due to a larger File cache)
- the difference in loading a file is indiscernible

What you are describing is the textbook example of a company not spending $500 or $1000 and costing themselves a fortune. They could get you a fully tricked out W7 x64 machine for $1,000. My work machine is a custom spec'ed box with RAID 1 drives (I will *never* use a box without RAID 1 as my primary work machine again!), 12 GB RAM, Intel 920 CPU, and a gorgeous aluminum case, and it ran around $1,200. I could step up to SSDs for a few hundred more.

From the sound of what you are doing, and a rough estimate of what it costs to keep someone at your level on payroll, they are losing money by keeping you on this 5 year old PC.

J.Ja

first and foremost, win vista / win7 doesn't support Kernel Audio drivers,
it's user mode audio only
which doesn't work with my hardware or apps

I'm not replacing my hardware and applications until absolutely necessary

plus it's way more than $500.00 - $1000.00 to upgrade

the XP x64 Dual Socket Dual Quad Xeon system came in @ just over $5000.00
- not including apps

the next time I drop $5000.00 on a system
I expect to get:
at least a Dual Socket, Dual Octo-core Xeon running @ over 3GHz with 100GB RAM
and that's not including replacing the audio hardware and all the apps with real x64 apps / drivers
which will probably come in around an extra $1200.00 - $1600.00

so we're talking about $6000.00 - $7000.00 to upgrade

that's the current repair / upgrade budget for a whole year
_

here's my task manager on the main workstation

http://i832.photobucket.com/albums/zz249/WhatNameShoudIUse/XP-ProTM.png

24-processes with one 436MB 44,100 16-bit stereo file loaded in the audio editor
_

edit: add

It's my minimum for trusting storage; wouldn't touch a NAS (home or business) without minimum two drives mirrored. Hardware RAID where possible.

(more lenient with my workstations though as they mirror critical data to trusted storage or don't have it locally at all. Nature of the work though.)

I've heard of RAID5 failing to rebuild (software RAID) even when using the device specific "reconstruct RAID disk". I've yanked a drive out of a RAID1 without the machine so much as hickuping or failing to re-mirror when a second drive was stuffed back in.

After working for some time for the help desk for [name of NAS maker who was acquired a few years ago], I will NEVER touch RAID 5. It was so sad, enraging, and embarrassing to get calls day in and day out from users who had RAID 5 installed, and when it had to rebuild, it blew up. After all, there is no guarantee that when it's doing the rebuilding that the disk it is moving data to is healthy! All it takes is a minor disruption during that rebuild process (Murphy's Law, folks) and you can kiss it all goodbye.

There are three common RAID levels that I trust, in this order: 1, 10 (aka 1 + 0), and 6. And even 6 I'm not a huge fan of. Given the cost of drives, and the value of data, if you have enough data to justify more than one drive, you can spend the extra cash on RAID 10. If you don't have enough data to fill one drive, RAID 1 is an extra $80 or so to add (cost of a second drive).

I experience one drive failure every 9 months, on average. Even though I perform nightly backups, the time lost due to a drive failure with a single disk is unacceptable and costs me a LOT more than what a second drive costs. If I'm going to keep a spare disk on the shelf, I might as well have it in the box in a RAID 1. I buy Western Digital disks lately, and one reason why I do is that every time I have a failure, their RMA process is painless. I'm rarely more than 3 days without 2 healthy drives in my system.

J.Ja

The customers have all chosen RAID 10. I've said it before and I'll keep saying it: None of my servers has ever crashed due to a drive failure in a RAID array, and the only time I had a disk fail to rebuild, the replacement was DOA. I always carry two...

One of our customers is upgrading their servers and eliminating up to 5 separate boxes per store by going virtual. They're using IBM's DS3200 System Storage units for the new servers, configured with multiple RAID 10 arrays hosting logical drives for each virtual server.

You are of course correct about using all the RAM. I assume the other poster was talking about having a 32-bit OS with >4GB physical RAM.

With the swap file - it's a bad idea to disable it completely. Some programs (poorly written ones generally) request large amounts of RAM when they are launched, but never use a lot of it. Windows is smart enough to service these RAM requests with the pagefile, and only actually swapthose pages into physical RAM if the application actually uses them. If you disable the swap file, the OS has no choice but to waste physical RAM with all these unused pages.

Would I have to re-install the game on the RAM disk for it to work? Or is there some other way to get the game to utilize the RAM disk?

Using Junction Link Magic [http://www.rekenwonder.com/linkmagic.htm] you can link any files or folders onto your RAM Disk, and link them to the "new" location (being the RAM Disk).

Hi,

The content of the virtual drive can be saved on exit (optional feature). Restarting the computer the data is loaded again, so you can continue where you leave it. RAM drives are great to play games.

I too am dubious that putting Temporary Internet Files on a RAM disk would provide any perceivable speed increase.

Speaking personally, I can't perceive any difference in speed once we get below about 200ms. Anything below that speed seems "instant" to me. 200ms is plenty of time to yank a bunch of html and images from a hard drive, so I would be amazed if I could tell the difference between temp internet files being on a drive and being on a RAM disk.

And that's ignoring that most of the time I am looking at new content, that has to be pulled down the wire, often from the other side of the world.

RAM disks definitely have their place, but it seems to me trying to optimise the lesser common scenario (looking at content you have already viewed recently), when it is already fast enough to begin with is pointless.

Then there's the issue that you lose the contents of the folder when you restart the computer. Next time you revisit a web site after a restart you will again have to download the content over the Internet, instead of viewing it from your local cache. That speed difference I definitely would notice.

My guess is that a browser isn't going to have any benefit from it's cache directory being on a ramdisk. The speed of the network will determine it's responsiveness more than what cached files it has from recent browsing; it's still going out to the DNS and grabbing the base html that the cached images and such may fit into.

is to keep it from otherwise plugging up the system with multiple thousands of useless files and MFT Bloat

the more you surf the more the MFT Bloats
the more you surf the more your disk gets fragments
etc.

plus the benefits of:
- at boot the T.I.F. is always empty
- at reboot / shutdown the T.I.F. is dumped including those pesky .dat files
- persistent critters that hide in T.I.F. are gone with every boot / restart / shutdown
etc.

Yeah, I can see benefits of doing this for some people. I would add privacy to your list as one of the key benefits. Not only is your browsing cache "deleted" automatically for you on restart, it is near impossible to retrieve (as opposed to deleting something from an actual hard drive, which can be recovered).

I can see different circumstances where a RAM disk can speed things up. That's why I find the author's choice of example as a weird one.

Is a ramdisk effective on x64 systems or is it better suited for x32 to utilize the "lost" ram?

Is there a yardstick for how much ram should be allocated to ramdisk?

Does this improve efficiency where the amount of onboard ram is limited?

It's basically a hard disk in ram; some of your ram is allocated to be used as standard storage mounted with a drive letter (R:\ if you like).

It would be better to use a 64bit system since you can drop in far more ram than the OS needs giving you that extra room for R:\directories\fi.les. The measuring stick would be how much "storage" you want to apear out of the free RAM not used.

The one interesting use mentioned was for 32bit systems where the amount of physical RAM is limited by 32bit memory addressing. In that case, it may be possible to add far more than 3.5 gig of ram but use that additional ram to host the R:\ which in turn hosts your swap file. Theoretically, this should bring your swap file speed up to RAM speeds rather than running at slower IDE/SATA/USB speeds.

I question this a little based on the 32bit OS not being able to see more than 3.5 gig of RAM so anything running on the OS is limited to that total sum. With a 64bit system, the OS can address the more than 3.5 gigs though a 32bit program running on top of that 64bit OS is still limited to 3.5 gigs it can make use of. Now, what you can do is use multiple 32bit programs each limited to 3.5 gigs. They don't have to share the same sum total 3.5 gigs so your two 32bit programs would potentially use 7.0 Gigs total.

I am just getting into Photo Shop so I will give this a try, plus I just like tweaking. Good way to learn.

USE the RAM Disk "better" than just Temp Internet Explorer files... (are you still on a 500 Mhz Celeron?)

With Junction Link Magic [http://www.rekenwonder.com/linkmagic.htm] you can link any files or folders onto your RAM Disk, and link them to the "new" location (being the RAM Disk) just like how symlinks work in Linux.

Also, with better RAM Disk applications, you can have specific data copied to the RAM Disk upon startup. See SuperSpeed RamDisk Plus [http://www.superspeed.com/desktop/ramdisk.php] and Romex's VSuite Ramdisk [http://www.romexsoftware.com/en-us/].

Thanks for these tips, Jack. I agree that it?s important to squeeze out as much performance from your PC as possible and RAM disks seems like one great way to do this. I?m curious?do you know how long the Dataram RAMDisk?s registration license is valid for?

Can i use Dataram RamDisk on a computer formatted with NTFS, windows XP Pro Version 2002 service pack 3,
Intel pentium 111 processor 730MHz 512MB of Ram.

But with only 512MB RAM in system, you won't gain anything in speed. In fact, you will probably notice your system slows down significantly.

I had been advised to set a static pagefile as well long ago.
The reasoning was so Windows will not be busy changing the size of the pagefile.
I believe this helps and I don't see the low on pagefile prompts.

i just found out about it and i can't wait to use it, it will centainly make my life easier.

How to go about making Chrome to use the Ram memory ?

@ Ichene: For Chrome apparently no GUI is available to set the cache directory, but it can be done through switches, see the following link: http://www.steelfrog.com/speed-up-and-secure-google-chromefirefox-by-using-a-ram-disk/

The page file (aka swap file) is used by the system to free up RAM by moving infrequently accessed RAM data to disk. Putting a page file on a RAMDisk not only entirely defeats the purpose of the page file, but adds unnecessary overhead and increases the margin of error. You're better off just disabling your page files entirely! I have 8GB RAM in Windows 7 (64-bit) with a 256MB RAMDisk for TMP, TEMP, browser cache, etc., and absolutely no page file set up on any drive. I do a large amount of 3D modeling, graphics work, and programming, not to mention some pretty heavy gaming (e.g., Battlefield 3) and I never have a problem and my system flies

I have, for quite some time, been perplexed with a particular application I use daily. Be advised, this is only a hobby for me, and it's not so terribly important that I want to take up time and effort from anyone that doesn't wish to be bothered.
Here is the problem:
1. I use a fantastic ebook management program, Calibre. This program was, for Windoze, a 32-bit program. It has recently been updated to a 64-bit program for Windoze. It is written using Python scripting language. From the beginning, there was a 64-bit program written for Linux. It has a feature for converting ebook file formats, ie, from *.pdf or *.epub and, in my case, turning them into a *.mobi file that is friendly to my Kindle products. Whenever I perform conversion operations in bulk, the program is very CPU intensive, boosting CPU usage to 100%, but, with the 64-bit program, only maximum usage of about 4.2-3 of the 16 GB RAM I have installed, 14+ or so usable.
My system: HP p6677c
Motherboard Manufacturer: Foxconn
Motherboard Name: H-ALVORIX_HF-RS880-uATX
HP/Compaq motherboard name: Alvorix-GL8E FN-Alvorix-RS880-uATX (Alvorix)
Foxconn Model 2AB1 (1.00) Motherboard
Chipset AMD 785G(RS880/SB710), Alvorix Board: FOXCONN 2AB1 1.00
Bus Clock: 200 megahertz
BIOS: American Megatrends Inc. 6.04 09/07/2010
OS Version: Microsoft Windows 7 Professional, Service Pack 1, 64 bit
Processor: AMD Athlon(tm) II X4 635 Processor, AMD64 Family 16 Model 5 Stepping 3; Processor Count: 4
RAM: 16127 Mb
Graphics Card: ATI Radeon HD 4200, 256 Mb
Hard Drives: C: Total - 941411 MB, Free - 851600 MB; D: Total - 12354 MB, Free - 145 MB; (external) J: Total - 476937 MB, Free - 276144 MB

2. I'm wondering whether creating a RAM disk to crunch a 15 GB database of ebooks, 10,300 individual books - each with supporting metadata, would be more efficient?

3. Can anyone briefly tell me how I can do this? The library is on my C:\ drive and it is backed up on the G:\ drive external. The Calibre program is located, also on my C:\ drive.

Understand, my system isn't a bulldozer, but it isn't a piece of junk, either. My CPU fan, presently, when rendering this database, sounds like a "C-130, rollin' down the strip." I don't have a bios that supports heat sensor information. Core Temp reported 23 degree C at idle and off the chart under 100% load. I'm adding an aftermarket cooler when it arrives, but I'd like to ease that processor load with this idea if it is feasible.

Thanks for your time.