SSD Tweaks and Tips

Here is a collection of Tweaks and Tips for Solid State Disks (SSDs).

These are gathered together from discussions folks have had on this Forum.
It centers around the VPC-Z with SSD's and RAID0, but should be generally applicable.

As suggestions, and corrections, are made I will update the OP.

Enable Write Caching: This provides faster writeback of related small disk writes, at the expense of less safety in the case of a power loss. Not so much of a risk on a laptop (which has batteries) compared to a desktop (which can just have a power cord kicked out).
On the Z use the Intel Rapid Storage Technology app: " Manage | Advanced | Write-back cache: Enabled"
Otherwise can use: " Device Manager | Disk drives | Double click on Volume 0 | Polices". Select " Enable write caching on the device", unselect " Turn off Windows write-cache buffer flushing on the device"

Disable indexing: This imposes some overhead every time files are written, modified, or deleted. With lots of small files it can have a noticeable impact. Even with it off, searching happens lightning fast.
In Windows Explorer, or My Computer: Right click on C: Drive | Properties. Uncheck " Allow files on this drive to have contents indexed..."

Turn Off the Disk Defragmenter Schedule: It's Ok to occasionally defrag an SSD, and a small benefit will result. But there is no need to do so on a daily or weekly basis (and it increases wear).
Run Disk Defrag, turn off regular schedule

Disable Superfetch: This preloads programs into memory that Windows thinks you may use. Some people like to turn it off since programs load so quickly anyway off the ssd.
In RegEdit: " HKEY_LOCAL_MACHINE\SYSTEM\CurrentControlSet\Control\Session Manager\Memory Management\PrefetchParameters". Change " EnableSuperFetch" value from 3 to 0.

Disable Prefetch: This imposes some overhead, increases boot time, and seems to provide negligible benefit on an ssd.
In RegEdit: " HKEY_LOCAL_MACHINE\SYSTEM\CurrentControlSet\Control\Session Manager\Memory Management\PrefetchParameters". Change " EnablePrefetcher" value from 3 to 0.

Firefox - Use memory cache instead of disk cache: Small block disk writes are the bane of SSD's everywhere , and Web browsers are one of the worst perpetrators of small block disk writes.
Either create a RAM disk and point the browser cache to it, or set Firefox to use memory cache:

1. open up about:config (type it into the url bar)
2. type browser.cache into the filter bar at the top.
3. Find browser.cache.disk.enable and set it to false (by double clicking on it).
4. set browser.cache.memory.enable to true
5. create a new preference by right clicking anywhere, hit New, and choose Integer.
6. Call the new preference browser.cache.memory.capacity and hit OK.
7. In the next window, where it asks for the number of kilobytes you want to assign to the cache, just enter -1 to tell Firefox to dynamically determine the cache size.

See: Speed Up Firefox by Moving Your Cache to RAM, No RAM Disk Required

The downsides are:

-Disk cache, unlike the memory cache, persists across restarts.
-Size of memory cache is capped at a lower number than size of disk cache.
-Possible slight slowdown for some plug-ins such as Adobe reader, which get to their data as a disk file.

See: moving cache to RAM instructions making the rounds - mozilla.dev.apps.firefox | Google Groups

Future Firefox may be able to have both disk and memory cache and so have the best of both worlds.

Set up a RAM disk: This can be used to speed up many types of programs, also save power and increase drive life, whether used with an SSD or a mechanical hard drive. A side benefit can be security if properly set up, as temporary files containing sensitive information go away instead of sticking around forever, when the the machine is shut off. Downside is it uses up precious RAM which might otherwise be put to better use.

From psyang: DataRAM provides a RamDisk with persistence between reboots and is free for up to 4GB. See: RAMDisk Software - Dataram
Also: Set up a RAM disk

Disable Hibernation: While useful for speeding up boot times on a mechanical drive, is less useful with SSDs as they boot more quickly (plus there's always the option of Sleep/Standby mode). Disabling Hibernation can return 1/2GB - 4GB disk space. From start menu type “ cmd” and run as Administrator. In the command line type “ powercfg -h off”.
From anseio: SSD Optimization Guide

Turn off Pagefile: If you have 4GB+ of RAM, consider turning off or reducing the pagefile size. On an 8GB system the pagefile size will default to 8GB. The irony is that the more physical memory is available, the less pagefile is needed. Nevertheless this one is more controversial than the others as it can cause problems when all available physical memory is in use and there is no pagefile to turn to for virtual memory. If you know your memory usage and feel so inclined:
1. My Computer | Properties | Advanced System Settings | Performance tab | Advanced tab
2. Select the virtual memory change box.
3. Unselect automatically manage pagefile size for all drives.
4. Select no paging file, or reduce it. and apply (then restart as required). Fwiw: Microsoft indicates debug info doesn't get written out if pagefile <400MB.

Other Links:

Jim_Kirk listed this useful link with SSD tips: Guide * Windows 7 Ultimate Tweaks & Utilities *

anseio listed: SSD Optimization Guide, and http://forum.notebookreview.com/hardware-components-aftermarket-upgrades/429972-ssd-thread-benchmarks-brands-news-advice.html

sugarkang suggests that less is more when it comes to optimization: ssd optimizations

 

Nice compilation Oscar. Rep added.

 

About the regular defrag issue; I thought I read somewhere that Windows 7 will automatically disable regular defragging if you have an SSD. Is that true?

 

The defrag schedule was on on my Sony Z until I disabled it. I don't think W7 cared that I had an SSD. Then again, I never gave it a chance to run on its normal schedule.

 

Check out this link on the OCZ forum. I've used several of these and they are very well explained at OCZ.
Guide * Windows 7 Ultimate Tweaks & Utilities *

 

Yes indeed, that's why I had included your link in the original post. The only thing is, you have to read through to the end of the post to be able to see it.

 

SSD Tweaks are always welcome!

Thanks Oscar2 (^_^).

 

Hmmm... a bug man... in a rainy city... with a Z12. I know you!

 

Win7 is supposed to, but I don't think it's consistent. My first Intel X25-M that I got in August was never seen by Windows as SSD. I replaced it a month later, for other reasons, and the new one is recognized by Windows.

Can't say why.

 

Good Post Mate!

 

Oscar2==>Glad the OCZ link helped to compliment your post.

 

Thanks for the post. I've disabled the 2 fetchings and windows indexing, and enabled the write-back caching.

Would love to hear any dissenting opinions to doing any of these, in terms of my 256gb Z12.

 

Oscar2 did a nice job of providing some opposing views on some of the recommendations, like the increased risk of data loss with enabling the write-back cache and the advantages/disadvantages to defragmenting the SSD. Those are a few controversies I've seen sparked by some of these recommendations in the past.

 

I don't really see the benefit of that one. Is it just for saving on disk space? (I'll add it if some feel it useful).

 

Disc space being one thing.

More important is that every time your lappy hibernates, the contents of your RAM are written to the SSD. Found in The SSD Optimization Guide | The SSD Review

If you're looking to decrease unnecessary writes to your SSD, hibernation and pagefile are two options. Page file being a little more controversial. Opinions for and against it. I've got 4GB of RAM, so have no need for a page file. The exception is that one of my games requires it, so I have it located on my 2ndary HDD. Pagefile is also a space saver move.

 

Deleting the hibernation file just to save extra write cycles seems ridiculous. It only does a single write to that file when you hibernate the machine. That's what, 1-2 times per day max for most users? So over 2 years you'd have 400-700 writes to that area? I can see trying to free up disk space, but seems otherwise pointless to disable.

 

I have added it to the OP.

 

There are other tweaks that are much more ridiculous than this one. We take the ones we like and discard the rest. I, personally, do not like the "disable drive indexing" tip.

4. DISABLE DRIVE INDEXING
The purpose of drive indexing on a hard drive was to allow quicker access to a file. As access times on a ssd are almost instantaneous (.1ms), there is a common belief that indexing does nothing more than increase the total number of writes to the ssd which results in a lesser life cycle. Having stated this, there is no confirmed performance increase by disabling indexing and the chances of wearing out ones ssd is somewhere in the area of impossible to one in a million.

I like drive indexing, it's necessary for searching from the start menu to function well. It also does not write repeatedly and incessantly like web browsing can. My indexing folder is 360MB and contains information for 95,000 files.

It's very easy, with SSD's, to become obsessed with its lifespan and to try to decrease writes as much as possible. That must be tempered with what we'd like out of the SSD in regard to everyday usage.

I've only got 80GB and like to have my photos on the SSD, since they load MUCH faster into Picture Motion Browser than when they're on my HDD. Since I need the space, I prefer to disable hibernation and have relocated my pagefile to my HDD.

It's the nature of the drive to degrade as it is filled to capacity, but it's hard to really drive those numbers down. TRIM and Garbage Collection are supposed to be very helpul in keeping performace up.

http://forum.notebookreview.com/har...ssd-thread-benchmarks-brands-news-advice.html

 

^Agreed. I agree on indexing being useful as well, but I'm not sure any of us know how many writes the indexes makes. I rely on indexing a lot in outlook.

My point with hibernation is that it's one write each time the computer hibernates, which the user controls, and isn't very often. I can't see any logical case for disabling it in effort to decrease writes. I do agree with disabling it to free up disk space, which happens to be why mine is disabled

 

Your math is flawed. Each "write" consists of a lot of writes. The file system's block size is 4k, and the disk's block size is 512 bytes. So if you have 4 GB RAM, you write 8388608 blocks to the disk. Multiply that with 400-700.

 

My math is fine. You are only writing to each block one time each time you hibernate. The issue with SSD's manifests itself in a single block being written to too many times and causing wear. And due to wear leveling they won't even be the same block each time.

 

It will only write 4GB of RAM if 4GB is actually what's in use at the time of hibernation. I'd love to have mine on. Will give it a try, since sleep has BSOD issues if left sleeping for more than a few hours.

 

To put into more perspective:

If you have a 4GB hibernation file on a 128gb (really 120) ssd that's 3.3% of the space on the drive. With programs, windows install etc. it's 4.4% of the free-write space. Most mlc cells are rated for about 10,000 writes. Each time you hibernate it wears 4.4% of the drive by a factor of .0001. So.....you will need to hibernate your drive about 250,000 times to kill it.

I'm pretty worried myself

 

End of the world!!!

 

Wrong again. You assume that the wear leveling is 100% efficient and doesn't have to copy blocks.
In reality, wear leveling causes blocks to be moved around on the drive, which causes a minimum of two writes for each write that has to be wear leveled, rising towards four writes as the drive approaches full or doesn't have TRIM.

 

Out of curiosity anyone know if the hibernation reserved size increases when going from a 4 to 8GB system? I suppose I can live with the current level which appears to lock yet another 2 gig of my 192 (down to 140 after my cleaning out most of the junk and installing the stuff I need). But when I go to 8, will the reserved size climb?

 

Not wrong again. The block that is being freed up due to wear leveling isn't written to or erased again until data is reallocated to that location (when the drive needs it again).

Even if you were right (you're not), then you'd have to hibernate 125,000 times to trash the drive.

Let's look at it this way. If your drive doesn't wear level at all, and all the data keeps getting written to that same 4gb over and over....you can hibernate around 10,000 times before that 4gb of the drive is "shot"...or more likely replaced by spare cells.

 

Pretty much it was the primary inspiration for this thread, as far as expanding beyond the previous thread's more singular topic of using memory cache for Firefox.

 

You guys can keep hashing it out but I just wanted to say thanks for including me in the thread eheh.

 

Wow! 5300+ posts and with the title "SSD guy"

Welcome to our humble thread. We are not worthy...

 

You might be interested in the very first internet ssd review done by 'The SSD Guy' here at NBR which is ar0und here buried somewhere eheheh.

 

all the dispute about hibernation -- I don't really see the point of it. Wasn't it mainly introduced for more rapid startups? A non-issue on the Z. Or are people really working with so many windows open that it's more desirable to hibernate over shutdown? meh I disabled it first thing, and just use sleep.

 

It's not just the number of windows, but the state of what you're running.
If I'm working at compiling something, I like to get back to where I were, with the same files open, looking at the same line numbers, with the same history buffer for each console window. It can easily take 10 minutes or more to get back to where you were, if you can even remember it.

That said, I much prefer standby over hibernation. I never leave the machine disconnected from the mains for long enough for standby to deplete the batteries.

 

So it turns out that as you might suspect, adding 4GB of RAM winds up stealing another 8GB of disk space. Didn't look at my page size prior to upping the memory, but afterward it was at nearly 8GB. Idiotic. Set it to a fixed 1GB which is still overkill (largely as if they're this stupid to begin with who knows why they might be swapping), but at least its a little more reasonable.

Most of the other 4GB seems to have gone to hibernate. I've left it on for the moment, but I frankly dont think the cost in SSD space is worth it and likely will shut that down as well. There seems to have been a further 1/2-1 GB unaccounted for when I shut all this down, but likely that was just a loss related to foolishly doing a series of Windows updates. When someone gets a solution for the SxS debacle, I'd love to hear about it.

 

Well, this solution works well for me

 

Unfortunately, I bought this particular box as a highly portable workstation for handling video and data analysis which is sadly all Windows based. The whole point of buying the fastest, okay nearly, porta-box available is to squeeze every last bit out of it while on the road, so I'm not, in this case, predisposed to virtualize. 'sides, I'm a Suse guy

 

Added entry for disable Pagefile to OP.

 

I enabled hibernate today to see just how much it would write to my SSD. I don't leave a lot of programs running, but today's hibernate used 500MB. Nowhere near as high as I thought it would be.

 

I'll throw in a tip, though I haven't tried it yet myself. Maybe other people will guinea pig it But I trust Lifehacker's recommendations. Here it is

SSDLife Keeps You Informed of Your Solid-State Drive's Health and Lifespan

 

Hmm... Well I downloaded the free version to try it, but it failed to detect my SSD drives, saved a log and exited. Unfortunately, the log seems to a binary format so no assessment as to why it could see the drives, but perhaps iits not happy with raid arrays.

 

As far as I can tell, this app only pretty-prints a subset of the S.M.A.R.T. values.

For S.M.A.R.T. to work, it needs to be pointed to the individual physical disks, not the logical disks (in addition to being enabled in the BIOS setup, and the drives actually supporting S.M.A.R.T.)

To get the actual data, you can either boot a Linux liveCD (pretty much any one will do) and run "smartctl -a /dev/sda" (and sdb, sdc, sdd), or install cygwin and the "smartmontools" package (under "utils"), and do the same.

If so, note that the S.M.A.R.T. counters count down from a set value towards an alert value, not up. So if the Power_On_Hours report 94 with a raw value of 3000, it means that for this drive, 3000 hours is considered equal to 100-94 = 6% of how long the drive can be powered on before it's expected to fail. Of course, that doesn't mean the drive will last that long -- it's just one of many parameters.

 

Most of the tips on ssd optimization are contentious at best and completely false at worst. There's was a big debate about this already over at anandtech with the SSD optimize site owner. He lost the debate.

I'm not going to say that he is definitely wrong, but before you guys start tweaking your system, it would be prudent to have all the information.

 

I looked through that thread, it contains a lot of thoughtful discussion (with some entertaining trolling thrown in for good measure )
But if what you got out of all that discussion was that "most ssd optimization is contentious or false" then you should go back and re-read the thread. What they are saying there is not so different than what has been discussed here.

Here are the suggestions that are in the OP for this thread:
-Enable Write Caching
-Disable indexing
-Turn Off the Disk Defragmenter Schedule
-Disable Superfetch
-Disable Prefetch
-Firefox - Use memory cache instead of disk cache
-Set up a RAM disk
-Disable Hibernation
-Turn off Pagefile

-Enable Write caching has a little bit of contention but (for laptops) generally falls into the category of 'why not?', for desktops it's a little more serious of a decision due to the potential for data loss.
-Disable indexing always stirs up some discussion, same everywhere.
-Disk defrag: Everyone seems to agree that its not necessary to do a regular defrag on sdd's, and that Windows cannot reliably be depended on to automatically do the right thing here (hence the tip to turn off the auto defrag schedule).
SuperFetch and Prefetch: It can slow boot times, no real reason not to disable these on an ssd.
Firefox memory cache/ramdisk: This is just a good idea in general.
Hibernation: after all the screaming about this one settled down, the conclusion was: If you don't really use it turn it off, it saves disk space. Otherwise leave it on as it doesn't impact anything performance wise.
Disable Pagefile: Everyone always mentions the dangers inherent in turning off pagefile, just a duplication of the same discussion that happens on every thread that mentions it (including this one) nothing new there.

So, in conclusion, it's not very informative to say: " Most of the tips on ssd optimization are contentious at best and completely false at worst".

 

Before you suggest someone re-read something, you should probably make sure that you didn't overlook something yourself. You missed OP's critical line "IF YOU ARE READING BELOW THIS LINE BE AWARE THAT THE FOLLOWING IS MOSTLY WRONG ADVICE WHICH I AM CRITIQUING!"

If you look at the very first post, the OP goes to debunk all of these you have listed. His advice consists of:
1. enable AHCI; and
2. install Windows 7

The list you have up is the subject of debate over at anandtech. That debate is long and detailed and if you want to skip to the conclusion, the general consensus was that these "optimizations" are indeed contentious at best and harmful at worst. Unanimous? No.

Don't misunderstand me. I'm not saying I'm 100% right and you are 100% wrong. I'm saying what the consensus of the thread was from very knowledgeable people.

Again, I don't care what you decide you want to do with your particular system. I'm merely presenting alternative, credible information that people on this board might be interested in before making a bunch of changes to their machines.

 

Yes, I did read the first post. I ALSO read past the first post. Just because the first post says only two things are required. the next 280+ posts make it clear there is more to it than just those first two lines.

What I summarized, just above, is in agreement with what people on there and elsewhere generally agree with. Instead of speaking in generalities, can you instead point to the particular ones you have issue with? (that's why I listed them and there aren't that many).

 

I would caution you to check more into things. Just because the SSD is installed, doesn't mean that Win 7 will recognize it. This I have confirmed with firsthand experience. Also, Les (The SSD Guy) isn't just some person with thoughts on SSD performance. He's a "go-to" in the industry and has a ton of knowledge. If you read into the SSD Review, you'll notice that at the very beginning he writes "There is no need to Optimize Win7 or your SSD whatsoever for it to function correctly." All things that follow are things that can be explored, by by no means are necessary.

I've accepted about half of the optimizations as very useful. Proper chipset drivers being one of them. The anandtech forum post was written quite a while ago. Intel's latest chipset drivers are very useful and 10.0.0.1046 is actually better than msahci (as far as boot times are concerned).

 

True. I have the same experience because I'm on VAIO TZ with ZIF connector which doesn't disable superfetch by default. Maybe because I don't have an AHCI option in my bios. No matter. The options that Windows ends up choosing for me end up conforming to the suggestions listed anyway, which is detailed in the debate.

I didn't say that he was some unknowledgeable noob spreading disinformation. He knows a lot more about computers than I do. He also had some good reasons for posting info as he did. He referred back to the advice of Intel and Microsoft engineers as well. And yet the other guy was even more knowledgeable. I boot times/performance are better with prefetch and superfetch on.

Yeah. I presented the anandtech thread as another source of information, not to end or win a debate. You are all free to do whatever you wish.

 

btw, I have added your recommendation, not to get too carried away with the optimizations, and the link, to the OP.
I still disagree with your phrasing however: "Most of the tips on ssd optimization are contentious at best and completely false at worst", but I figure everyone can make up their own mind on that one.

 

EXACTLY. And it's also clear that OP was not suggesting they follow your list above.

Uhh. Perhaps you missed the entire superfetch debate? SSD guy gives good reasons as to why he thinks they should be off, but OP gives better reasons for why they should be on. Same with almost all of the other stuff.

I'm not going to go into detail about what's the best procedure. I'm not as knowledgeable as EITHER of those two. And I'm not trying to be. I'm not here to win some debate.

Like I said before, and will not say again. The tips listed here are the same as the tips listed in the anandtech forum debate. Those tips are not universally accepted good tips. There are credible, knowledgeable people that say that these tips will either do nothing or will harm performance. I am just providing a link to info that nobody else was providing on this forum. I am not trying to relive the arguments in that forum on here.

For potential tweakers, my suggestion is you read that forum thread. And if not, caveat emptor.

 

Well, I just went back and looked at that debate. Taltamir vs Flamenko (i.e. Les the SSD guy) Seems to me Les had the final word on that one when he pointed out that the advice of leaving it on really only benefits older SSD's, that it's recommeded by Intel (remember them?) to leave it off.
I think you are confusing "better reasons" with "more emphatic", there is no doubt that taltamir is the "more emphatic" in that debate.

In any case, like I pointed out above, your suggestion of using reason and understanding when applying any optimizations, is now added to the OP.

I still see that we are talking about two separate things however. My main issue is your statement that ssd optimizations fall into the the range of: contentious to false, as per your comment "Most of the tips on ssd optimization are contentious at best and completely false at worst"
It is that comment, not the underlying notion that we should all become better informed, the way you just throw that out there, that makes one take anything else you say with a grain of salt.