Second SSD for the page file?

[Doomer]

Would I gain anything by adding a 32gb SSD for the sole purpose of holding the page file?

Thanks

 

[BrightCandle]

If you are using your page file a lot then I suspect quite a lot. But of course it might be better to add performance by adding more RAM instead, it will likely be cheaper and faster.

However if you aren't paging often then basically nothing, the number of hard faults on a modern machine is minimal.

 

[UaVaj]

just get more physical ram and totally disabled the page file.

you will thank me later.

 

[Cerb]

Would I gain anything by adding a 32gb SSD for the sole purpose of holding the page file?

You should get more RAM, so you won't notice it as much and/or can disable it (until you start using that RAM, because there's never permanently enough ).

Page file access is an ideal case for a fast SSD, and won't be substantially limited by anything but controller/NAND, in most cases. It'll all be random access, and usually small. A newer/faster SSD replacing your main SSD might be better, but only if you could really take advantage of that potential speed, which would generally mean no less than 120GB--and today, that's already starting to become 250GB. But, RAM, while costlier per GB, is still far and away faster, generally by several thousand times (200-300us v. <100ns). So, nothing beats more RAM.

 

[mango123]

I have wondered this myself personally, and I have found contradicting information on page files vs. no page files from both people who seem to know what theyre talking about.


:hmm:

I too agree that if you're worried about performance enough to think about adding a SSD just for your paging file that adding ram would probably be the best thing to try first.

 

[goobee]

Just make sure you are running 64bit Windows. Adding more than 4gbs to 32bit Windows is useless as it can not access that additional memory.

 

[UaVaj]

I have wondered this myself personally, and I have found contradicting information on page files vs. no page files from both people who seem to know what theyre talking about.

there is no contraction information. there is only how big is your wallet?

if your wallet is big enough for physical ram. skip page file.

if your wallet is too small. then you do not have a choice but to use page file via ssd or via hd.

here is the calcalus. hd = 130mb/s. fastest ssd = 500mb/s. ddr3 1600 = 12800mb/s. any questions?

 

[Deders]

In the old days when HDD's had trouble handling more than one request at once than yes having a separate disk for your pagefile unclogged the I/O significantly. SSD's handle multiple instructions at once so well that I highly doubt you would get any noticeable performance benefit from adding a 2nd for a pagefile.

It's also been proven that completely disabling the pagefile completely does have a negative impact on performance but whether it is generally noticeable in anything other than benchmarks is still up for debate.

 

[Zaxx]

Another point to consider is how much paging does your rig do. Some apps can/will use and actually benefit from a lot of pagefile space if available to them...mostly professional software like Adobe and CAD type apps. In this case an extra large pagefile kept either on a sizable boot ssd or a separate, smaller ssd may be the best way to go. Most of us don't fit this model tho. Imo, if you have 8GB or more RAM then you're fine just keeping it in memory...4GB or less and I'd use a drive. For a professional/heavy user (or those using a workstation level rig), I think a minimum of 16GB min. would be a good idea if you want to keep it in ram. Just my 2 cents ofc.

 

[Cerb]

I have wondered this myself personally, and I have found contradicting information on page files vs. no page files from both people who seem to know what theyre talking about.

That's because different people have different priorities. Even with an SSD and Superfetch combined, that post-AV post-backup time in the morning will still suck. Some people also have this idea that asking for race conditions is a good thing, v. getting more work done in that time by letting whatever screwed up crash like it should have in the first place (biased? Me? ). If you spend $500 on a mobo+CPU and can't afford another $50-120 for enough RAM to prevent that, most of the time, you're looney, IMO. But, you have to have some confidence that nothing you will need to do will exceed, in commit, your RAM, and don't do things like not save often, or it could blow up on you.

Also, averages of loopy benchmark results are often seen as useful evidence, despite modern PCs being faster than the human in front of them, if well-configured, for most typical work, except when bottle-necked by local or remote storage (and there's not a lot you can do about remote storage--if you think SSDs are expensive...).

IoW, the corner cases can be 99% of what matters. In typical non-gaming use, I can only barely tell a difference between a new high-end CPU and a fast Core 2/Phenom II, if the PC has an SSD. What matters most is how often I have to wait on the computer, when I am not expecting to (some things, like installing updates, take a long time for no apparent reason, FI). Even with an SSD, not having enough RAM, and/or Windows having no concept of important v. unimportant data or IO requests (and I think still being stuck with lots of big locks--though that's never been confirmed, Linux with meager RAM is a totally different animal, and old-style locking+exceptions would explain it), is quite noticeable, IME.

 

[Fred B]

I always have the page on a second hd/ssd , basicly it sits on the fastest disk availeble . With ssd you do not notice when it is paging it is happening in ms . There is headroom enough for paging with one ssd for normal pc usage.

Also the pagefile can pe put on several ssd/hd and the page start writing 64K in raid 0

 

[Deders]

Does cluster size actually affect SSD performance these days?

 

[taq8ojh]

just get more physical ram and totally disabled the page file.
you will thank me later.

Please go away.

 

[Cerb]

Does cluster size actually affect SSD performance these days?

Not in any positive way. LBAs are now 4K, so anything else serves little to no purpose. Potential reduction in fragmentation from bigger clusters will be made up for by the greater IO load and reduction in free space.

 

[Doomer]

Thanks all. I'll just stick with my current setup which is 8gig of ram and a 240gig SSD. Performance is fine but low capacity SSD's are dirt cheap so i was just wondering.....

 

[UaVaj]

Please go away.

don't trash this thread because you too broke to buy enough physical ram.

now get lost.

 

[Emulex]

paging still occurs even if you have too much ram. It is unwise to reduce the page size since some applications may do memory mapped IO (basically loads the whole file into ram, paging ram too).

If the C drive has any contention (database) then moving the swapfile to another lun will allow concurrent access.

most folks just install OS to C:, page to C:, apps to e:, database to F, log to G, tempdb to H, scratch to I:

you want to equalize the # of cores to concurrency so each core can dispatch i/o threads to maximize queue depth.

Some folks are less nerdy and some are more nerdy so you can take this advice with a grain of salt if you don't find it fun to install sql server or VS on your home pc

 

[hanspeter]

paging still occurs even if you have too much ram. It is unwise to reduce the page size since some applications may do memory mapped IO (basically loads the whole file into ram, paging ram too).

Mapping a file into the address space doesn't use the page file.

You can create a memory mapped file that points only to memory. That will use the page file, if you are low on physical ram.

 

[zir_blazer]

If you can spend money on a SSD for the sole purpose of having a Pagefile on it, you may as well purchase more RAM to reduce the chances you actually need to use it.



I seriously hate the Pagefile. Its purpose was to take advantage of the Virtual Memory feature introduced on 386 Protected Mode to let you use Hard Disk space to simulate RAM at a time when that RAM was both scarse and very expensive. It made the difference between being able to use an application at speeds slow as a crawl, or not using it at all. But 20 years later, we're STILL using it in an era where RAM is usually cheap and very abundant to not having any need to do that.
I'm not sure how the Pagefile works on W7 and W8 because I'm still a WXP user, but when I was experimenting with this a decade ago, it always seemed that WXP sends to the Pagefile applications that are left minimized and idle for too long. If I left the machine turned on during the night, when I wake up again, restoring minimized applications could cause ton of HD activity plus a very apparent lag/delay as it seems that it was reading on demand all that data to put it back on RAM. I started disabling the Pagefile back when I had "only" 640 MB RAM (But was a ton at that time, 256 MB was standard) and it feeled like multitasking was snappier due to the fact that WXP was forced to hold everything on RAM instead of sending data that wasn't used often to the Pagefile even if there was enough unused RAM.
This aggressive behaviator of sending unused applications to the Pagefile was fine for RAM limited computers, but the more RAM you have, the more retarded it gets. In an era where its easy to have enough RAM to cover a worst case scenario (A healthy marging above your highest record peak commit), it should be even more eaiser to not have any need to use a Pagefile at all. It makes absolutely no sense to annoy the HD/SSD with read/writes (Specially with all the sleep states and such as there is an extra delay to wake up an HD if it spinned down, etc) during any normal usage scenario.

 

[Fred B]

If you can spend money on a SSD for the sole purpose of having a Pagefile on it, you may as well purchase more RAM to reduce the chances you actually need to use it.



I seriously hate the Pagefile. Its purpose was to take advantage of the Virtual Memory feature introduced on 386 Protected Mode to let you use Hard Disk space to simulate RAM at a time when that RAM was both scarse and very expensive. It made the difference between being able to use an application at speeds slow as a crawl, or not using it at all. But 20 years later, we're STILL using it in an era where RAM is usually cheap and very abundant to not having any need to do that.
I'm not sure how the Pagefile works on W7 and W8 because I'm still a WXP user, but when I was experimenting with this a decade ago, it always seemed that WXP sends to the Pagefile applications that are left minimized and idle for too long. If I left the machine turned on during the night, when I wake up again, restoring minimized applications could cause ton of HD activity plus a very apparent lag/delay as it seems that it was reading on demand all that data to put it back on RAM. I started disabling the Pagefile back when I had "only" 640 MB RAM (But was a ton at that time, 256 MB was standard) and it feeled like multitasking was snappier due to the fact that WXP was forced to hold everything on RAM instead of sending data that wasn't used often to the Pagefile even if there was enough unused RAM.
This aggressive behaviator of sending unused applications to the Pagefile was fine for RAM limited computers, but the more RAM you have, the more retarded it gets. In an era where its easy to have enough RAM to cover a worst case scenario (A healthy marging above your highest record peak commit), it should be even more eaiser to not have any need to use a Pagefile at all. It makes absolutely no sense to annoy the HD/SSD with read/writes (Specially with all the sleep states and such as there is an extra delay to wake up an HD if it spinned down, etc) during any normal usage scenario.

Basicly that is a worst case scenario, when the page is set to a slow hd with the os also it is a punish . When downloading from the web the page is used in the way a cache is used . It writes 64K to the page and writes
the file to the same disk. When using 1 slow PATA hd it is a worst case scenario . Now whe have NCQ that realy improves hd disk speed with pagefiles . For example , when having a ssd for the os and a hd for the page it is a very smel diference in read time from ssd or hd because it is seqentieel with basicly no seektime

I tryed to stop the pagefile from w7 and it automatic turned on again , saying that w7 misses the page and will make one .
I think the best thing is making the page as small as can be , set it to a little size and when it is to small windows automatic resizes the size of the pagefile

 

[Cerb]

I tryed to stop the pagefile from w7 and it automatic turned on again , saying that w7 misses the page and will make one .

All you did was delete the one that was there, then. It's easy to turn off, and no, it won't make a new one.

Uncheck automatically manage, then select "No paging file," and it will go away, and stay gone.

I think the best thing is making the page as small as can be , set it to a little size and when it is to small windows automatic resizes the size of the pagefile

If you make it really small, you can get a mini-dump, but then why bother? If you're going to have one, let Windows use it as a backing cache, like it wants to. Now, if you have 32GB RAM, but aren't doing serious CAD work or something, you probably don't need a 32GB page file, so resetting the base size should be fine, letting Windows grow it a bit as desired; but capping it at a low size is gaining you very little over not having it, except for a place to put dumps.

 

[Fred B]

All you did was delete the one that was there, then. It's easy to turn off, and no, it won't make a new one.

Uncheck automatically manage, then select "No paging file," and it will go away, and stay gone.

.

With xp i can turn it on and of ,so i think there is something on this W7 install that made it depend on the page file . I made this W7 to perform with less memory and made it a fast booting w7 and it is updated . There must be a service stopped that is essential to work withouth a pagefile.

 

[aigomorla]

just get more physical ram and totally disabled the page file.

you will thank me later.

not unless he runs out of ram and requires a page file.

then he will be posting in general help asking why the program keeps running out of memory.

 

[UaVaj]

not unless he runs out of ram and requires a page file.

then he will be posting in general help asking why the program keeps running out of memory.

original post is asking about performance associated with page file. having more than enough physical ram for the given task does not increase performance, however it does "prevent" unnecessary slow down due to paging.

-----

the point here is: to have enough physical ram so one does not run out of memory for the given task, much less needs a page file.

if you position is to run windows with "lesser" physical ram and to allow page file on slower ssd/hd to make up the short coming and willing to accept the associated slow down. more power to you.

some of us just choose to run with max/enough physical ram and negate the slow page file. if such motherboard already maxed out does not hold enough physical ram for the given task. we simply buy the next better motherboard capable of accommodating more physical ram.

-----

to get "more" memory performance, faster memory needs to be consider.

 

[Nec_V20]

Would I gain anything by adding a 32gb SSD for the sole purpose of holding the page file?

Thanks

Don't be silly, you will gain a lot more if you create a RAMdisk http://www.softperfect.com/products/ramdisk/ and move your virtual memory there.

With every reboot you will then get a completely clean pagefile.sys which will prevent the mysterious faults associated with a corrupted vitrual memory.