Prefetch and SSDs

[carwiz]

My apologies. Superfetch DOES create indexes on disk that allows it to carry over between boots. But I have not been able to locate these files. I've read so many expansions of "how it works" that I got my own rope in a knot. The link I posted above is totally wrong. I actually posted it before reading it. The author starts out talking about Superfetch then himself changes over to Prefetch and identifies files that don't exist in Windows 7.

Yes, Superfetch does preload programs into memory. It also releases the memory based on current memory requirements. The preloaded programs have the lowest priority for memory space. Superfetch is supposed to be move "intelligent" than earlier version and does support faster startups. The intelligence is how it determines what should be included in the memory preload. So that I don't confuse me, you or anyone else, I'll just keep my mouth shut. But from all the different explanations, the memory can contain preloaded programs that may be used at startup.

 

[GeneO]

The files are the .pf in the Prefetch folder and there is a database (.db) file in there with a hash appended to the file name for each user.

I have been reading up some more. I still would like to know what distinguishes the prefetch from superfetch registry settings. For now I assume prefetch will load some programs (not boot related) at boot time and superfetch will load them after login with some delay, and at different time in the day according to how it profiled your daily routine.

Prefetch will also arrange to have the disk defragmented for optimal boot. It depends on the scheduled defrag to do this. Since SSD are excluded from defrag scheduling, there is no chance that prefetch will result in your SSD being defragmented

I don't believe superfetch is involved in releasing memory, it just loads programs in the standby cache in memory and the windows OS paging system treats them like any other pages in standby and will evict them if needed by its normal means. There may be some logic to prevent thrashing, I don't know.

 

[essenbe]

The path to Superfetch exe. is C:\Windows\system32\svchost.exe -k LocalSystemNetworkRestricted

I think it also depends on how it is set. It can be set to automatic, automatic(delayed) or disabled.

 

[carwiz]

Here's what I HAD in the Prefetch folder. There were many .db and .fx files but not a single .pf file. The .ini file contains a list of just about every file on the system. As you can see from the date modified, not much was added from the time I built this system. (2/10/2011). It seems the last change to the data base files was on 2/22, or 12 days after the AppLaunch and Cx-SC4 were created.

I deleted these files for a test. After deleting, I changed the Registry entry for Superfetch to "3" in both control areas (users) and rebooted several times. In between each boot, I ran some of the programs I normally use and and created some dummy files for Word, Excel then recorded a short portion of a TV program with WMC. After subsequent boots, I went back to these dummy files and read them back in. I let the system sit idle for about 15-20 minutes then went back to the Prefetch folder. The folder was still empty! So, I don't know what those files really do and no new files were created. Especially no .pf file.

 

[carwiz]

Update - I was looking around for my WEI logs (Winsat) and discovered the first time I ran the index was on 2/22/2011. Is this a coincidence or does Superfetch use the results to turn itself off? Or it could be that by 2/22, I had used every file I was going to use. I don't know.

Also found three errors in my Admin log. They occured after each boot described in my previous post. These errors have prompted me to re-instate the Prefetch files above. :shock: But, again, there's no .pf file.

 

[GeneO]

I have read that it will turn superfetch off if it detects the boot drive is a high performance SSD. It doesn't on my 'puter even though my SSD is high performance (SATA 3.0 m4 with WEI of 7.9).

 

[carwiz]

I just booted up and came online and much to my amazement, .pf files suddenly started appearing. The problem is I don't know if it's because of Superfetch or Prefetch. They are both on at the moment.

 

[GeneO]

They may be written at shutdown after analysis so they would not have appeared on your first boot.

 

[carwiz]

They appear to be written in real-time too. More appear every time I go look. As various services are used, it looks like an entry is made. I re-ran WEI and picked up 0.3 on my Gaming Graphics but that was because I was tweaking settings. I want to find out if Windows will turn off the Fetching with a new assessment so I'm just going to monitor it for a few days.

 

[essenbe]

I have read that it will turn superfetch off if it detects the boot drive is a high performance SSD. It doesn't on my 'puter even though my SSD is high performance (SATA 3.0 m4 with WEI of 7.9).

GeneO, that was the point of my post about 1st generation drives. Eveb though it says that, after the 1st generation experience, it dosen't do it. I have an OCZ and an Intel and it has never turned them off. They may not be as gast as yours, but they are not low performance drives. Microsoft saig that the threashold was being able to read at 8MB/s, which I believe all SSDs can do today.

 

[GeneO]

You are right. Micro$oft must have realised that there is not much reason to have specialized code and that there is not any reason to disable suprferch anyhow. So I say follow ms lead.

 

[essenbe]

You are right. Micro$oft must have realised that there is not much reason to have specialized code and that there is not any reason to disable suprferch anyhow. So I say follow ms lead.

I really don't know which is best, but Microsoft's reasoning was on 1st GEN SSDs, we are now in the 3rd GEN. I would think that the SSD manufacturers would want their SSDs to run as well as possible and they all recommend disabling it. What I would really like to know is why SSD manufacturers recommend that. They must have some reason. They just won't tell what it is.

 

[cluberti]

Given it's not for performance (pro or detriment), I'd wager a guess it's for extending the life of the drives. Given a desktop/laptop or even workstation with a 2nd or 3rd gen SSD moves the resource bottleneck to the CPU, it's not really necessary to have any of that hitting the "disk" anymore. Most of the superfetch and prefetching is meant to alleviate the performance hit to a mechanical disk by Windows, and given an SSD does not have those limitations those technologies really aren't useful anymore.

 

[stereo]

What about to move the pagefile on the HDD ( as D partition ) ?

 

[cluberti]

The paging file is mostly read and rare writes (unless under incredible load that passes RAM capacity), an SSD (even Gen1) is a very good place for a paging file (given the nature of how the paging file is used).

 

[essenbe]

In an interview Microsoft said that pagefile reads outnumber writes 40-1, and that the majority of writes were something like 60kb or less. Considering the write tests they are doing at xtremesystems, I don't think a few writes will hurt. These are the latest numbers released

xtremesystems write endurance test

Post #1228, #1230

Kingston SSDNow 40GB- Intel 25 rebrand
220.19 TiB writes
MWI- 1 July 25th
RaS-6

Intel 320 40GB No update since 8/4/11
271.0 TiB writes.
MWI- 1 - Since July 9th. @ 190.5 TB
RaS-39

Samsung 470
368.289 TiB writes
MWI-1 (approx. 5 weeks ago)

They also have a C300 , an M4 and a couple of others that have just started.

 

[GeneO]

In an interview Microsoft said that pagefile reads outnumber writes 40-1, and that the majority of writes were something like 60kb or less. Considering the write tests they are doing at xtremesystems, I don't think a few writes will hurt. These are the latest numbers released

xtremesystems write endurance test

Post #1228, #1230

Kingston SSDNow 40GB- Intel 25 rebrand
220.19 TiB writes
MWI- 1 July 25th
RaS-6

Intel 320 40GB No update since 8/4/11
271.0 TiB writes.
MWI- 1 - Since July 9th. @ 190.5 TB
RaS-39

Samsung 470
368.289 TiB writes
MWI-1 (approx. 5 weeks ago)

They also have a C300 , an M4 and a couple of others that have just started.

I have been following that forum for a while. Their m4 is still running strong at > 150% of its rated life with no errors or reallocated sectors. Looking at the number of erase cycles on my m4 and the number of power on hours, if mine does as well as theirs up to now, then my m4 SSD should last over 17 power on years, which is more than 25 of my years

 

[essenbe]

In an interview Microsoft said that pagefile reads outnumber writes 40-1, and that the majority of writes were something like 60kb or less. Considering the write tests they are doing at xtremesystems, I don't think a few writes will hurt. These are the latest numbers released

xtremesystems write endurance test

Post #1228, #1230

Kingston SSDNow 40GB- Intel 25 rebrand
220.19 TiB writes
MWI- 1 July 25th
RaS-6

Intel 320 40GB No update since 8/4/11
271.0 TiB writes.
MWI- 1 - Since July 9th. @ 190.5 TB
RaS-39

Samsung 470
368.289 TiB writes
MWI-1 (approx. 5 weeks ago)

They also have a C300 , an M4 and a couple of others that have just started.

I have been following that forum for a while. Their m4 is still running strong at > 150% of its rated life with no errors or reallocated sectors. Looking at the number of erase cycles on my m4 and the number of power on hours, if mine does as well as theirs up to now, then my m4 SSD should last over 17 power on years, which is more than 25 of my years

Just be careful, don't do too many writes to it.:roflmao:

 

[carwiz]

It won't matter by then. We'll all have computers in our brain and SSDs will be the size of a postage stamp.

 

[essenbe]

It won't matter by then. We'll all have computers in our brain and SSDs will be the size of a postage stamp.

Right, I have 2 M4's, an X25 and a Vertex2. I started to say that my 9 year old Granddaughter will be able to use them. But, by then they will be too large and too slow.