Is the Windows cache off when Primo Cache is on? Topic is solved

[Violator]

What about them? If you don't know for sure you need them, I would recommend not using them. Same goes for most of the performance tweaks you can find on the internet.
See http://www.tweakhound.com/2009/12/11/nt ... ge2-tweak/ for NTFSMemoryUsage.

Seems like the author didn't read the pasted link..
http://technet.microsoft.com/en-us/libr ... 85435.aspx
"Applies To: Windows 7, Windows 8, Windows Server 2000, Windows Server 2003, Windows Server 2003 R2, Windows Server 2008, Windows Server 2008 R2, Windows Server 2012, Windows Vista"

Controls the amount of paged memory used for file operations. The default value of 1 uses the default amount of paged memory for file operations. Setting the value to 2 increases the amount of paged memory available for file operations.
However, this feature can have unwanted side effects. The added file operation memory only nets increased performance when you don't take paged memory away from other areas of the operation system.
Generally, you'll need to experiment with this setting and monitor its effect on overall system performance before you make it permanent.

[GunnzAkimbo]

Found another "File Cache" tweak tool called CacheSet by SysInternals.

This is the system wide - all drive - file cache tweaking tool.

http://technet.microsoft.com/en-au/sysi ... 97561.aspx

I set it to max = 960MB, and hit a peak of 7xxMB the other day.

[InquiringMind]

Found another "File Cache" tweak tool called CacheSet by SysInternals...

This looked interesting, but when I tested on my (XP SP3) system, it didn't actually do anything. It didn't change any Registry entries relating to cache settings - the only change it made was adding a flag to indicate whether the EULA had been accepted or not! Task Manager did not indicate any change in the system cache either when it was supposedly cleared.

I guess this was designed for much older versions of Windows (pre-2000) which used different settings for their file cache, and therefore won't function on anything newer.

If anyone is interested, I've included a zipped .csv file showing registry activity reported by ProcMon (a SysInternals utility that *does* work well). In this, I tried changing the maximum cache size and clearing the cache - but the only registry activity reported is on process start and when I close it a few seconds later.

[idefix44]

I tried it too and had no change. Like InquiringMind.

[Davey126]

Although I have not experimented this utility recently it never seemed to produce the desired results. At some point I saw an explanation about the original intent (believe it was coded for NT) and how Windows has changed since - but that was a looooong time ago ....

For those unfamiliar the tools produced by Mark Russinovich (now employed by Microsoft) and his team are simply outstanding; many are essential for advanced Windows debugging/management. Most can be freely downloaded at http://www.sysinternals.com.

[InquiringMind]

A deleted post here suggested the AnalogX CacheBoost utility - I just tested that and it also does nothing for current Windows versions (it adds [vcache] entries to the system.ini file, which dates back to Windows for Workgroups 3.11) and made no perceptible different to cache size in WinXP.

AnalogX's other software generally does the job advertised so this seems to be very much an exception.

[AikenDrum]

I'm pretty sure it's impossible to disable the Windows read cache. It's used for memory-mapped file I/O, which is a way of saying, "Make this RAM address range mirror what's in this part of the disk," kind of like a reversed virtual memory. Many programs use this, so you simply can't turn the underlying mechanism off.

I'd say PrimoCache should be considered more of an internally-emulated external cache, as if it were, say, a dedicated ram cache attached to a disk controller. It makes the hardware appear to be very fast, but otherwise the system behaves as normal.

[InquiringMind]

I'm pretty sure it's impossible to disable the Windows read cache...

It would be nice to be able to reduce it to minimum size though, rather than having it grow to 2GB+. Aside from data duplication when running PrimoCache and similar software, a large cache means more time taken to check whether requested data is present or not.

[Davey126]

I'm pretty sure it's impossible to disable the Windows read cache. It's used for memory-mapped file I/O, which is a way of saying, "Make this RAM address range mirror what's in this part of the disk," kind of like a reversed virtual memory. Many programs use this, so you simply can't turn the underlying mechanism off.

Agreed. In my experience claims of disabling the Windows read cache really boil down to killing Superfetch and/or the Windows file cache both of which are separate/distinct from the system cache. I keep all native caches enabled as recent versions of Windows do a fine job managing this space without adversely impacting performance. Sustained free memory serves no useful purpose IMO.

As for PrimoCache, I find a small L1 read/write cache yields a smoother running system even when the source disk is a SSD. Such claims are hard to demonstrate through benchmarking; it's more of a perception and certainly something I wouldn't loose sleep over if proven wrong. One item I can measure: a 10s write delay does reduce hard writes by 5-15% as outlined in other threads.