Overclock.net banner

1 - 20 of 37 Posts

·
Registered
Joined
·
563 Posts
Discussion Starter #1
As someone who works with NAND technology, I know that having the page-file on an SSD is one of the worst things you can do to it. Not sure if this has been posted before, but here is a link to a very simple explanation of MLC NAND flash and correctly talks about the impact of the quality of the controller which does all the workarounds.

http://www.tomsguide.com/us/ssd-value-performance,review-1455-5.html
 

·
Registered
Joined
·
563 Posts
Discussion Starter #3
Thanks! All hope is not lost however, in the article it correctly shows that as the capacity increases, life of the SSD increases, and by virtue of the total size, the effect of something like a page file on the capacity degradation (due to non-correctable blocks) will decrease as a percentage of the total capacity.

I think that SSDs are the best thing since sliced bread and will be ubiquitous as system/OS drives, but people need to understand that the technology needs to catch up with this new use-case which is very different than the Cell-phone PMP media storage (mostly reads) use-case, and that all controllers are not equal. The cheap guys will not implement the best-in-class wear-leveling methods (allowing them to use a cheaper controller), but what you get is a drive that is going to deteriorate rather quickly.
 

·
Registered
Joined
·
640 Posts
Bingo, it makes perfect sense after all, same as avoiding to defrag your SSD.

Any knowledge about making my parts last longer is great knowledge indeed, hence the rep

 

·
Premium Member
Joined
·
9,511 Posts
Quote:

Originally Posted by sti_boy View Post
As someone who works with NAND technology, I know that having the page-file on an SSD is one of the worst things you can do to it. Not sure if this has been posted before, but here is a link to a very simple explanation of MLC NAND flash and correctly talks about the impact of the quality of the controller which does all the workarounds.

http://www.tomsguide.com/us/ssd-value-performance,review-1455-5.html

Sorry OP, but why is it the worst thing you can do to them!
As having such a fast page file has got to be good
 

·
Not new to Overclock.net
Joined
·
77,827 Posts
Quote:


Originally Posted by ACHILEE5
View Post

Sorry OP, but why is it the worst thing you can do to them!
As having such a fast page file has got to be good


Pages 5 and 6 explain it. It doesn't say "so this is why you shouldn't have the Paging File on your SSD", but it does explain why it's a bad idea even though it isn't talking about the Paging File.
 

·
Registered
Joined
·
1,822 Posts
Quote:


Originally Posted by TwoCables
View Post

Pages 5 and 6 explain it. It doesn't say "so this is why you shouldn't have the Paging File on your SSD", but it does explain why it's a bad idea even though it isn't talking about the Paging File.

That's another reason I went with RAID0 on my two 80s.

But here's my take on the page file thing....

Quote:


Already, the average SSD endurance will likely far outstrip its useful life in an everyday PC.

 

·
Premium Member
Joined
·
5,330 Posts
OP - what do you do with NAND technology exactly? I work with Excel everyday, but it don't make me a programmer...

Modern, decent SSDs have sufficient write-cycles to last significantly longer than their useful lifespan. The correct advice is to buy a good SSD, not to tell people to switch off their pagefiles. Although the article is accurate in the way it describes the differences between MLC and SLC, it does not actually state not to use high write cycles on MLC. In fact it is a little out of date - the latest enterprise SSD devices are now going towards MLC because it is so much cheaper - add a wear-leveling controller & the life and speed goes up to that of SLC because you can buy more cells for your money. Seen the Fusion-IO cards?

Also before you berate pagefiles on SSD it is also useful to understand exactly what kind of I/O you get when paging. Actually most pagefile activity are reads - as explained by Microsoft here (scroll down a little to the question about whether your pagefile should be on your SSD - for those that can't be bothered to read it - they say it should be there).

Basically - if you buy an SSD, make sure it is a good one, otherwise you are wasting your money. Then use your purchase to its full potential - otherwise again you are wasting money. Sure - you might reduce its life a little. But not so much as will actually make a difference. How many people decide not to overclock because to do so will make their cpu die in 10 years rather than last for 100? Intel rate their SSDs for 5 year minimum life, when used for 20GB/day writes. Still think your drive is going to die before you want to replace it?
 

·
Registered
Joined
·
563 Posts
Discussion Starter #10
Quote:


Originally Posted by the_beast
View Post

OP - what do you do with NAND technology exactly? I work with Excel everyday, but it don't make me a programmer...

I am a chip-designer and architect working on embedded SoCs. These SoCs have raw NAND controllers so I am quite familiar with the technology including error-correction and have been working with it for the last 5 years or so. Having said that, one does not need to have this level of understanding to see that large writes/erases are going to reduce the capacity and performance of the drive. As the errors increase, the correction load increases and slows things down. So, yes, I believe its a good idea to reduce the page file (turn it off if you have the RAM), or perhaps re-target it to another drive.

Its tempting to think that NAND being solid state is such a robust technology, but its constantly changing and is one of the most difficult technologies to deal with and keep up with (in terms of the controller HW and associate SW stack). You'll never really know if you're buying a "good" one because write/read performance tests are done on "fresh" drives so the degradation aspect does not really kick in until you start accumulating bad blocks.
 

·
Not new to Overclock.net
Joined
·
77,827 Posts
Quote:


Originally Posted by the_beast
View Post

OP - what do you do with NAND technology exactly? I work with Excel everyday, but it don't make me a programmer...

Modern, decent SSDs have sufficient write-cycles to last significantly longer than their useful lifespan. The correct advice is to buy a good SSD, not to tell people to switch off their pagefiles. Although the article is accurate in the way it describes the differences between MLC and SLC, it does not actually state not to use high write cycles on MLC. In fact it is a little out of date - the latest enterprise SSD devices are now going towards MLC because it is so much cheaper - add a wear-leveling controller & the life and speed goes up to that of SLC because you can buy more cells for your money. Seen the Fusion-IO cards?

Also before you berate pagefiles on SSD it is also useful to understand exactly what kind of I/O you get when paging. Actually most pagefile activity are reads - as explained by Microsoft here (scroll down a little to the question about whether your pagefile should be on your SSD - for those that can't be bothered to read it - they say it should be there).

Basically - if you buy an SSD, make sure it is a good one, otherwise you are wasting your money. Then use your purchase to its full potential - otherwise again you are wasting money. Sure - you might reduce its life a little. But not so much as will actually make a difference. How many people decide not to overclock because to do so will make their cpu die in 10 years rather than last for 100? Intel rate their SSDs for 5 year minimum life, when used for 20GB/day writes. Still think your drive is going to die before you want to replace it?

It is in Microsoft's best interest to say that it's a bad idea to disable the Paging File. Why? Because they know that by saying "We strongly recommend that you do not disable the Paging File", they are killing 2 birds with 1 stone: they're hopefully preventing users from disabling it who just don't know any better (or who don't know what they are doing), and they are also enabling users like us who know better to think for ourselves - to do the research and to see whether or not we want to disable it. They effectively leave that choice up to us. They know that people like us ignore the recommendation. But they also know that it doesn't mean that they can complicate it for less-experienced users by including the exceptions that we know about. That would mean that they'd have to hope that they read more information on it, and we all know that they won't.

So with that said: I have had my Paging File disabled for about a year now, and I've also had my SSD (the G.SKILL Falcon) since the first week this past August. Did I notice a difference when I disabled it? No. Absolutely nothing changed. Will I ever enable it in the future? It's unlikely.

Oh, and by the way: the article does not have to come right out and say say "try to minimize write cycles to the drive". Why? Because it would be redundant.
 

·
Registered
Joined
·
563 Posts
Discussion Starter #12
Quote:


Originally Posted by the_beast
View Post

Also before you berate pagefiles on SSD it is also useful to understand exactly what kind of I/O you get when paging. Actually most pagefile activity are reads - as explained by Microsoft here (scroll down a little to the question about whether your pagefile should be on your SSD - for those that can't be bothered to read it - they say it should be there).

This is a good article and to some degree backs my "concern". Its clear that MS has been working with the SSD makers on this issue. I find some of the statements in the article a little unconvincing and its clear that both NAND flash makers, controller makers and MSFT have work to do in order to make SSDs have longevity (in terms of capacity and performance). What promises does intel make on their 5yr claim? is it of complete malfunction? What I am referring to is a slow but steady decline in performance and capacity, not a hard failure like with a mechanical HDD. It appears from this article that there are in fact other things that are worse on the SSD than the pagefile (but they dont say what exactly).

Another thought - for those that like to re-install their OS very often : it may not be a great idea on an SSD.

Again, I'm all for this technology and I believe its the future. I will stress that this particular application of the technology is *very* new and quite immature, but I think the advancements will be quick.
 

·
Not new to Overclock.net
Joined
·
77,827 Posts
Quote:

Originally Posted by sti_boy View Post
This is a good article and to some degree backs my "concern". Its clear that MS has been working with the SSD makers on this issue. I find some of the statements in the article a little unconvincing and its clear that both NAND flash makers, controller makers and MSFT have work to do in order to make SSDs have longevity (in terms of capacity and performance). What promises does intel make on their 5yr claim? is it of complete malfunction? What I am referring to is a slow but steady decline in performance and capacity, not a hard failure like with a mechanical HDD. It appears from this article that there are in fact other things that are worse on the SSD than the pagefile (but they dont say what exactly).

Another thought - for those that like to re-install their OS very often : it may not be a great idea on an SSD.

Again, I'm all for this technology and I believe its the future. I will stress that this particular application of the technology is *very* new and quite immature, but I think the advancements will be quick.
I think the part that I bolded needs to be in a thread all by itself and stickied. This makes absolutely perfect sense.
 

·
Premium Member
Joined
·
9,511 Posts
Quote:

Originally Posted by sti_boy View Post
Another thought - for those that like to re-install their OS very often : it may not be a great idea on an SSD.

Quote:

Originally Posted by TwoCables View Post
I think the part that I bolded needs to be in a thread all by itself and stickied. This makes absolutely perfect sense.
I do think we need a post with the Definitive SSD guide, maybe with the OP edited to reflect what is in the post

Like the "Do's and don'ts of SSD" Or SSD's Urban legends!

But am not so sure that writes are all that a problem


As with 10,000 write/erase, you could install every day for the next 27 years
And that's without "ware leveling"
If we assume ware leveling works.
And my OS takes up 1/3 of my drive. So now times 27 by 3= 81 years


In fact, the more I read the more I think these things were made to be abused


But having said all that, "I have moved my IE Temp file to a HDD"
 

·
Premium Member
Joined
·
5,330 Posts
Quote:


Originally Posted by TwoCables
View Post

It is in Microsoft's best interest to say that it's a bad idea to disable the Paging File. Why? Because they know that by saying "We strongly recommend that you do not disable the Paging File", they are killing 2 birds with 1 stone: they're hopefully preventing users from disabling it who just don't know any better (or who don't know what they are doing), and they are also enabling users like us who know better to think for ourselves - to do the research and to see whether or not we want to disable it. They effectively leave that choice up to us. They know that people like us ignore the recommendation. But they also know that it doesn't mean that they can complicate it for less-experienced users by including the exceptions that we know about. That would mean that they'd have to hope that they read more information on it, and we all know that they won't.

So with that said: I have had my Paging File disabled for about a year now, and I've also had my SSD (the G.SKILL Falcon) since the first week this past August. Did I notice a difference when I disabled it? No. Absolutely nothing changed. Will I ever enable it in the future? It's unlikely.

So, you go against the advice of those who wrote the software you use because you know better, but gain nothing for doing so? Why would you do that?

If the pagefile was not a 'good thing', why would MS persist with it? Surely it would be the work of a few moments to code "if RAM>3GB then pagefile=off" if such a thing was actually a better idea?

Quote:


Originally Posted by TwoCables
View Post

Oh, and by the way: the article does not have to come right out and say say "try to minimize write cycles to the drive". Why? Because it would be redundant.

Why would it be redundant? Because you knew that minimising writes is necessary? If you already knew everything about SSDs, why read the article? A flash cell may have limited write cycles, but the life of your SSD is not necessarily limited in the same way.

Assumption is the mother of all f*#~ ups (Under Seige 2, for those who don't know their Segal movies).

Quote:


Originally Posted by ACHILEE5
View Post

I do think we need a post with the Definitive SSD guide, maybe with the OP edited to reflect what is in the post

Like the "Do's and don'ts of SSD" Or SSD's Urban legends!

But am not so sure that writes are all that a problem


Such a thread would be a good idea. But writing it in a balanced, objective way would be hard, and there would always be people who would dismiss everything.

There is also the problem of different brands/types of SSD - what is true of current-gen, cache-equipped, wear-levelling enabled MLC drives was not so true with older drives. Such an article would have to deal with different types, or make it abundantly clear that all SSDs are not created equal.

Quote:


Originally Posted by ACHILEE5
View Post

As with 10,000 write/erase, you could install every day for the next 27 years
And that's without "ware leveling"
If we assume ware leveling works.
And my OS takes up 1/3 of my drive. So now times 27 by 3= 81 years


Or to put it another, slightly more realistic way - your SSD will last longer than it remains useful for. Honestly, how many people that care about performance still use HDDs over 5 years old for their OS? Even 3 years old is pushing it - around half the minimum useful life that Intel quote, even with the somewhat high 20GB/day usage figures.
 
  • Rep+
Reactions: ACHILEE5

·
Not new to Overclock.net
Joined
·
77,827 Posts
Quote:

Originally Posted by the_beast View Post
So, you go against the advice of those who wrote the software you use because you know better, but gain nothing for doing so? Why would you do that?
Because the Paging File proved to be unnecessary in my system.

Quote:

Originally Posted by the_beast View Post
If the pagefile was not a 'good thing', why would MS persist with it? Surely it would be the work of a few moments to code "if RAM>3GB then pagefile=off" if such a thing was actually a better idea?
If they did that, then their OS would be less universally-compatible than it is now. So, by having the Paging File enabled by default, they are covering all of their bases. But since I proved that the Paging File is not necessary on my particular system, I just leave it disabled. I mean, what's the point in using something that makes no difference with or without it? And what's the point of leaving it enabled when I can further streamline my system by disabling it?

Quote:

Originally Posted by the_beast View Post
Why would it be redundant? Because you knew that minimising writes is necessary?
No. It is because the article makes it obvious to anyone who's paying attention to what they're reading. If after reading that article you don't get the sense that you should try to gain more control over your SSD writes, then you weren't paying attention to the article.

Quote:

Originally Posted by the_beast View Post
If you already knew everything about SSDs, why read the article? A flash cell may have limited write cycles, but the life of your SSD is not necessarily limited in the same way.
I'd rather go by all of the research I've done than take your word for it here. That article isn't the first little bit of information I've read about SSDs.

Quote:

Originally Posted by the_beast View Post
Assumption is the mother of all f*#~ ups (Under Seige 2, for those who don't know their Segal movies).
You are assuming that I am making an assumption.

Quote:

Originally Posted by the_beast View Post
Such a thread would be a good idea. But writing it in a balanced, objective way would be hard, and there would always be people who would dismiss everything.
Yeah, kinda like how you're trying to dismiss everything that article said and also like how you're trying to dismiss everything I said. You're right. We have way too many people like that.

Quote:

Originally Posted by the_beast View Post
There is also the problem of different brands/types of SSD - what is true of current-gen, cache-equipped, wear-levelling enabled MLC drives was not so true with older drives. Such an article would have to deal with different types, or make it abundantly clear that all SSDs are not created equal.
Yeah, you're right. It's not worth it.


Yes, that was sarcasm. Of course it's worth it! You have to start somewhere! Why do you have to come up with all of the reasons why it won't work? With an attitude like that, of course it won't work. But if you ignore those possible reasons why it won't work, then you'll get started on it and you'll eventually finish it. But once it's finish, you don't have to stop there: you can just continue perfecting it.

Quote:

Originally Posted by the_beast View Post
Or to put it another, slightly more realistic way - your SSD will last longer than it remains useful for. Honestly, how many people that care about performance still use HDDs over 5 years old for their OS? Even 3 years old is pushing it - around half the minimum useful life that Intel quote, even with the somewhat high 20GB/day usage figures.
With technology this immature, it is impossible to be as certain as you are trying to be. I mean, can you prove it? Hey, we can prove this kind of stuff with hard drives, but how about with SSDs? We can't. Therefore, time will tell.
 
  • Rep+
Reactions: sti_boy

·
Premium Member
Joined
·
5,330 Posts
Quote:

Originally Posted by TwoCables View Post
Because the Paging File proved to be unnecessary in my system.

If they did that, then their OS would be less universally-compatible than it is now. So, by having the Paging File enabled by default, they are covering all of their bases. But since I proved that the Paging File is not necessary on my particular system, I just leave it disabled. I mean, what's the point in using something that makes no difference with or without it? And what's the point of leaving it enabled when I can further streamline my system by disabling it?
How would automatically turning off the pagefile if it were not needed make anything less universally compatible? If it is uneeded with lots of memory, why not switch it off? Surely that would be making it more universally compatible? Unless of course it is needed for some other reason - for example with the many legacy apps that require a pagefile to be present in order to run? (Before you ask, versions of Photoshop and Call of Duty (IIRC) will not run without some pagefile - and I will bet there are others). Or for times when your installed RAM just ain't enough - a few years ago, 4GB would be laughable overkill in a system. Now it is a sensible minimum. How long before it is not sufficient?

If it makes no difference to your system, you have not streamlined anything. You have just made your system less capable.

Quote:

Originally Posted by TwoCables View Post
No. It is because the article makes it obvious to anyone who's paying attention to what they're reading. If after reading that article you don't get the sense that you should try to gain more control over your SSD writes, then you weren't paying attention to the article.

...

You are assuming that I am making an assumption.
You are making an assumption - gain more control over SSD writes = minimise SSD writes as far as possible. This is not necessarily true - but you assumed it is. There is no point making unneccesary writes (such as defragging). But that does not mean that all writess should be avoided. If this is what you are aiming for, why not put your SSD in a cabinet and look at it? That way it can last for ever...

Quote:

Originally Posted by TwoCables View Post
I'd rather go by all of the research I've done than take your word for it here. That article isn't the first little bit of information I've read about SSDs.

...

Yeah, kinda like how you're trying to dismiss everything that article said and also like how you're trying to dismiss everything I said. You're right. We have way too many people like that.
I am not dismissing everything that article said. In fact I am not dismissing anything that article said. I am dismissing the assumption that you made from the above article and presented as fact from said article (see above).

And what other research have you done? Both Intel and Microsoft, the 2 people responsible for the technologies involved, say that the technologies are compatible, and nothing you have posted other than your own opinion shows otherwise.

Quote:

Originally Posted by TwoCables View Post
Yeah, you're right. It's not worth it.


Yes, that was sarcasm. Of course it's worth it! You have to start somewhere! Why do you have to come up with all of the reasons why it won't work? With an attitude like that, of course it won't work. But if you ignore those possible reasons why it won't work, then you'll get started on it and you'll eventually finish it. But once it's finish, you don't have to stop there: you can just continue perfecting it.
Why is something not worth doing because it is hard? With that kind of attitude your boss must love you.

It is common sense to get a feel for the magnitude of any task before you start. If you intend to create an article to aid people set up their systems, you need to be aware of the potential pitfalls and put in place a strategy to deal with them. To do otherwise will likely end up with a piece that makes matters worse rather than better, and merely serves to confuse further.

Quote:

Originally Posted by TwoCables View Post
With technology this immature, it is impossible to be as certain as you are trying to be. I mean, can you prove it? Hey, we can prove this kind of stuff with hard drives, but how about with SSDs? We can't. Therefore, time will tell.
Likewise. All I am suggesting is that SSDs are used to maximise their potential. They are spec'ed to handle the load and are ideally suited for pagefile use, which does not hammer disks nearly as hard as is hinted at here in a properly set up system. I would not buy expensive cpus or RAM, then underclock them just in case the extra heat will shorten their life. Why would I take the same kid-gloves attitude with an SSD?
 

·
Premium Member
Joined
·
17,110 Posts
According to Intel AND Microsoft a page file is perfectly fine on an SSD.

Most MLC SSDs are rated at 100GB/day for FIVE years. Unless you want your SSD for 20 years, than a page file is fine. I, for one, plan to upgrade to newer and better SSDs at least 3 years down the road. So I could honestly care if these die in 8 years with a page file. Not only that, windows is much more snappier with it enabled. I disabled it, and certain programs take almost twice as long to load.

Not only that but I have programs AND games that will not launch without the page file. It doesn't give you a clear "you need a page file error either". It took me days to figure out the issue.
 

·
how two spel
Joined
·
3,036 Posts
Quote:


Originally Posted by Murlocke
View Post

According to Intel AND Microsoft a page file is perfectly fine on an SSD.

Most MLC SSDs are rated at 100GB/day for FIVE years. Unless you want your SSD for 20 years, than a page file is fine. I, for one, plan to upgrade to newer and better SSDs at least 3 years down the road. So I could honestly care if these die in 8 years with a page file. Not only that, windows is much more snappier with it enabled. I disabled it, and certain programs take almost twice as long to load.

Not only that but I have programs AND games that will not launch without the page file. It doesn't give you a clear "you need a page file error either". It took me days to figure out the issue.

I'm going to test this theory, if some of my apps don't launch I will def. put it back on and redirect it to the caviar black.
 
1 - 20 of 37 Posts
Top