Why does Windows 7 disable Superfetch?

[FishAk]

superfetch is on, but it ignores SSD drives.

If this is true, how can you explain this?

PS: And no, the service isn't just running because I've got HDDs in the PC as well.. or maybe it is, but it's still caching stuff from the SSD - according to the stuff in Windows/Prefetch at least (actually it's almost only caching stuff from the SSD since that's where the apps are )

 

[=Wendy=]

If you wish to check how much data is being written or read by superfetch, you can monitor it in Windows 7 task manager.

Start task manager.
Select the processes tab
Click on the view menu and select "select columns" then tick the I/O read bytes, and the I/O write bytes.
Now find the superfetch entry in the "processes" list, and you can monitor how much is being read/write during a user session.

PS.
It's one of the svchost processes, so you will need to dig around and find which one is running superfetch.

 

[postmortemIA]

If this is true, how can you explain this?

Windows 7 superfetch can prefetch app data and boot files. I also have stuff in Prefetch folder, but most appears to be loaded every time, like window manager, svchost, etc. and is fairly outdated.

However, to see if superfetch is activelly loading data/programs after boot, take a look into amount of free RAM after logging in with Task Manager. If you don't open anything, and Free Physical Memory remains the same - then superfetch is not caching data/apps. On regular mech. drive, superfetch will load data/apps into free ram so "Free" column will show about 0MB after few minutes.

Mine still has 1790MB after 1/2hour of use.

EDIT. Mark explains that .db file in Prefetch folder is for database that stores what data will be cached.
http://technet.microsoft.com/en-us/magazine/2007.03.vistakernel.aspx

 

[Voo]

Windows 7 superfetch can prefetch app data and boot files. I also have stuff in Prefetch folder, but most appears to be loaded every time, like window manager, svchost, etc. and is fairly outdated.

But you said it wouldn't fetch anything from the SSD, which it does
Also the free memory is <300mb usually AND stuff from the SSD is loaded into RAM (photoshop is surely no default file that is always loaded on every windows install And yes Photoshop is installed on the SSD), so that theory just doesn't hold up.

 

[tweakboy]

This is a Microsoft feature , just as how they improved their benchmarking to suite the newer hottest hardware to 7.9 , soo if in vista it says your hd is 5.9 that can be a regular HD, but vista doesnt know any better by that score. Soo W7 is accurate and understands what hardware you have. It sees a SSD and loves it takes score up to 7.7 or 7.9

I think for me size is the issue. I rather have size over an extra 150mbps . Its still a drive and will be no where near as fast as RAM,,, anyhow gg and tc my brothers and sisters of anand in faith.

 

[coolVariable]

This is a Microsoft feature , just as how they improved their benchmarking to suite the newer hottest hardware to 7.9 , soo if in vista it says your hd is 5.9 that can be a regular HD, but vista doesnt know any better by that score. Soo W7 is accurate and understands what hardware you have. It sees a SSD and loves it takes score up to 7.7 or 7.9

I think for me size is the issue. I rather have size over an extra 150mbps . Its still a drive and will be no where near as fast as RAM,,, anyhow gg and tc my brothers and sisters of anand in faith.

What BS is this? Is this even supposed to make any sense?

 

[FishAk]

On regular mech. drive, superfetch will load data/apps into free ram so "Free" column will show about 0MB after few minutes.

Usually, when I look at Task Manager, I have between 0 and 30Mb free on this system. However, after cranking my system up this morning, and looking at this forum- and then looking at the free space after seeing your post- I have 8.5Gb free, with only 2Gb cached from a total of 12Gb. Superfetch is set to start automatically, says it's running in the Services tab, but I don't see in in the Processes tab.

The funny thing is I don't have an SSD on this system. It's completely mechanical.

I don't have a set schedule, but I do have several programs I run most days, or at least very frequently. I don't know why these are not loaded into RAM at this moment. This system has been on for over an hour.

Usually, when I look at Task Manager, it's after working with the machine for some time. Perhaps the stuff cached is from programs I have opened, and not Prefetch for the most part.

I don't see any of the big programs I run, like Photoshop, Photoshop Elements, or Poser in the Prefetch folder. I have one 10Mb file, one 4.5Mb, one 3Mb, 7 in the 1Mb range, and the rest of the files are fairly small. the total size of the Prefetch folder is only 53Mb. One would think that with 12Gb memory, more would be loaded from Prefetch. Perhaps I'm doing something wrong.

 

[postmortemIA]

Superfetch loads different data on different time of day as well ... that is a just a rumor

 

[taltamir]

If we look at MLC NAND
50nm = 10,000 write cycles
34nm = 5,000 cycles
25nm = reported to be around 3,000 write cycles for good quality stuff.

actually:
50nm = 10k write cycles
34nm = I don't know
25nm = 35k for the good stuff.

 

[=Wendy=]

35K for enterprise grade, and that won't find its way into the "M" product line.

 

[taltamir]

35K for enterprise grade, and that won't find its way into the "M" product line.

you said "for good quality stuff"... "enterprise grade MLC" is "good quality stuff".
M product line is "cheap stuff" not "good quality stuff".
The MLC for the M grade stuff is still rated at 10,000 writes btw... do you have any evidence that it went down to 3000?

 

[=Wendy=]

The enterprise grade stuff hadn't been announced when i made my post in #27.
OCZ use Intel 34nm NAND in their Sandforce series of drives, they quote 5,000 cycles.
For the 25nm stuff they expect 3,000 cycles at best.

 

[bad_monkey]

using that logic, if intel told you to jump off bridge, would your response be the same??

yes!

 

[sub.mesa]

25nm NAND is 5000-6000 cycles, the "25nm Enterprise-MLC NAND" has 6x higher write endurance; around 30k cycles.

The 34nm NAND would be around 10.000 cycles.

But cycles per cell is NOT IMPORTANT.
What is important is TOTAL WRITE ENDURANCE.
What depends on that? Yes max write cycle count multiplied by number of cells. In other words; double the capacity of your SSD and you automatically double the write endurance.

So what we are doing now, producing NAND at smaller process technologies, means the total write endurance pretty much stays the same: the number of writes per cell is lower, but with more cells available we can write to more and use wear-leveling to get a much higher total write endurance.

So the low write cycle count should not be a problem; the higher cell count will compensate for that.

 

[=Wendy=]

25nm NAND is 5000-6000 cycles, the "25nm Enterprise-MLC NAND" has 6x higher write endurance; around 30k cycles.

The 34nm NAND would be around 10.000 cycles.

But cycles per cell is NOT IMPORTANT.
What is important is TOTAL WRITE ENDURANCE.
What depends on that? Yes max write cycle count multiplied by number of cells. In other words; double the capacity of your SSD and you automatically double the write endurance.

So what we are doing now, producing NAND at smaller process technologies, means the total write endurance pretty much stays the same: the number of writes per cell is lower, but with more cells available we can write to more and use wear-leveling to get a much higher total write endurance.

So the low write cycle count should not be a problem; the higher cell count will compensate for that.

Are you not mixing up write cycles with P/E cycles?
It was my understanding that 25nm had 5000 to 6000 P/E cycles.
P/E = program or erase
Given that a cell has been written too, it has to be erased before it can written again, that's 2 cycles per write = 2500 - 3000 write cycles.

 

[Rubycon]

Does not matter to desktop users.
You will have forgotten about the drive long before it reaches EOL.
If you use a smart host with lots of cache life is even longer but like I said it does not matter.

 

[taltamir]

Are you not mixing up write cycles with P/E cycles?
It was my understanding that 25nm had 5000 to 6000 P/E cycles.
P/E = program or erase
Given that a cell has been written too, it has to be erased before it can written again, that's 2 cycles per write = 2500 - 3000 write cycles.

my understanding is that it is program AND erase... that is, 10,000 writes = 10,000 erases + 10,000 writes... not 5,000 erases +5,000 writes.