Paging File on boot SSD or second HDD?

[jkauff]

I have 8GB of RAM, and I've yet to find a program that breaks the 5GB usage mark (no Photoshop for me). Despite that, I still see activity on my page file. I've got it on a SATA 3 7200RPM HDD, because it was on my SSD and I saw no performance difference. Samsung recommended not putting the page file on the SSD, so that's what I did.

 

[tweakboy]

My only experience with an SSD is in my work notebook. My usage is very multitasking and memory intensive, and the SSD has improved things greatly for me. Context switching between programs just happens, where before upgrading to the SSD, it could take up to ten seconds thanks to the slow 5400 rpm notebook drive.

So, I know that a paging file on the SSD makes my setup fly, but I was recently reading this article ( http://www.zdnet.com/blog/bott/windows-7-and-ssds-cutting-your-system-drive-down-to-size/2941?pg=1 ) and on the third page, he talks about moving the paging file to the secondary hard drive, as if it was a valid choice for some setups.

For those of you with SSD boot/HDD data setups, can you notice the difference if you put the paging file only on the SSD or only on the HDD?

Thanks,
SnakeJG

If you have 8GB or more you dont need a page file my friend.

 

[birthdaymonkey]

Unless you need the space, moving your swap file to an HDD is silly. Current gen SSDs will endure HUNDREDS of TB of writes.

Check out he xtremesystems thread where members have been torture testing their drives to death. The current leader is the Samsung 830, which endured 6000 TB of writes before giving up the ghost. I believe their 64GB m4 is still going strong at over 2000TB.

http://www.xtremesystems.org/forums/showthread.php?271063-SSD-Write-Endurance-25nm-Vs-34nm

Not all drives fare this well. But even the controversial TLC-based 840 endured more than 400TB before failing (and this was the 120GB - the 240 should last a lot longer).

 

[sub.mesa]

Yes, but what they do not test is retention. Because it is a public secret that not the actual write cycles is a major problem, but rather the fact that the retention drops significantly when the NAND cells are showing wear. This process starts at about 65% of rated NAND writes so already well before the official lifetime of the NAND.

What some of the tests in XtremeSystems show, is that during all this writing the number of replaced NAND cells will go up. But what remains hidden is that due to all the writing the data doesn't get a chance to remain there for a couple of months. When your flash is worn and you do realistic tests (instead of overwriting all day long) you will find out that the number of read errors is staggering. In many cases, the SSD will start being unusable if the NAND is severely degraded.

One good protection is the RAID4 or RAID5 scheme that many SSDs use to facilitate some kind of bitcorrection. This means that if a cell cannot be read, the redundancy still might. This softens the problem of lower retention, but not enough.

If you believe your SSD will work perfectly until it is worn down, then unfortunately I will have to disappoint you. The SSD starts being much less reliable as it approaches its last successful write and even well before this point the reliability and retention is far lower than required for reliable operation.

 

[birthdaymonkey]

Yes, but what they do not test is retention. Because it is a public secret that not the actual write cycles is a major problem, but rather the fact that the retention drops significantly when the NAND cells are showing wear. This process starts at about 65% of rated NAND writes so already well before the official lifetime of the NAND.

What some of the tests in XtremeSystems show, is that during all this writing the number of replaced NAND cells will go up. But what remains hidden is that due to all the writing the data doesn't get a chance to remain there for a couple of months. When your flash is worn and you do realistic tests (instead of overwriting all day long) you will find out that the number of read errors is staggering. In many cases, the SSD will start being unusable if the NAND is severely degraded.

One good protection is the RAID4 or RAID5 scheme that many SSDs use to facilitate some kind of bitcorrection. This means that if a cell cannot be read, the redundancy still might. This softens the problem of lower retention, but not enough.

If you believe your SSD will work perfectly until it is worn down, then unfortunately I will have to disappoint you. The SSD starts being much less reliable as it approaches its last successful write and even well before this point the reliability and retention is far lower than required for reliable operation.

Even so, well before the writes reach 65% of the Xtremesystem test figures, most users will have mothballed the computers or replaced the SSDs with new ones.

Even 100TB is a lot of writes over the course of several years. A user could write 10GB a day (which is actually a bit more than I average myself for my system drive) for 27 years before accumulating that much wear on the drive.

For very heavy users (e.g. a video editing scratch drive) endurance could be an issue, but not for most people--and certainly not enough to make putting the page file on a spinning disk necessary.

 

[sub.mesa]

True, write endurance is not a crucial factor for ordinary users. With SSDs getting larger and larger this problem also largely disappears since the wear can be spread over more NAND cells. With SSDs of 1TB soon available and used in the wild, you have to do some serious punishment to bring such an SSD down to its knees.

Nevertheless, people expect the drive to function perfectly until it reaches 0% and have learned it will continue to function even long after that. What they don't know is that the SSD will not provide the same reliability as when it had only light usage (>65% of rated cycles).

 

[j3SeCh6d]

I've been using a RAM drive for almost 3 years now to store all temp files of Windows, IE & Mozilla Firefox for my 3 computers. This avoids unneccessary write wear on all my SSD [4]. I have not used Windows Pagefile/Virtual memory for a long time now, since I have lots of RAM. Even my 7 years old laptop with 2gb ram with a 512mb RAMdrive & no pagefile has been performing well.

It's simple to setup & free, too.

 

[GAO]

True, write endurance is not a crucial factor for ordinary users. With SSDs getting larger and larger this problem also largely disappears since the wear can be spread over more NAND cells. With SSDs of 1TB soon available and used in the wild, you have to do some serious punishment to bring such an SSD down to its knees.

Nevertheless, people expect the drive to function perfectly until it reaches 0% and have learned it will continue to function even long after that. What they don't know is that the SSD will not provide the same reliability as when it had only light usage (>65% of rated cycles).

But for most users we are talking 5-10 years of wear cycle here. Even with page file, all caches, temp files and everything else on the SSD.

RAMdisk only lasts for a boot. if you want to get any utility out of disk cache for a browser, it has to span boots. temp files for windows isn't very much load either. Now if you were doing a lot of production like processing that required temporary space, then RAMdisk might be useful - like Linux /tmp RAMdisk.