>SSD fucks up because I stupidly put a swap partition on it
>have a 300GB backup 10k RPM raptor
>plug it in
>it fucking sucks
IT'S JUST NOT THE SAME
FUCK
So... no swap partitions on SSD's then?
>>51881996
>putting wrote intensive partition on ssd
How dumb r u?
>>51882013
Swap Partitions rack up the R/W cicles on ssd's like there is no tomorow.
Besides ram is faster you don't need swap when you have ram.
Unless this story takes place over several years OP is lying
SSDs don't literally die that quickly retard
>>51881996
It's simple OP you just need....
AN ALL NEW SCION TC
>>51881996
You should know better than to use a raptor m8.
>create swap
>set swappiness to 10
how acoustic are you
>>51882167
IS THAT THE NEW SCION TC?
SCIRON TEE SHEE
>almost 2016
>still using a swap partition
do you seriously have less than 8gb of RAM?
>mfw I had a page file on my SSD the first couple months because I forgot that was a thing
>3TB of data written to a 256GB SSD
>>51884716
Try hibernating without one.
>>51881996
>have 32GB of RAM
>nothing ever has to be paged
i use ssds in my laptops and a raptor in my desktop. the desktop runs just fine
>>51884716
>>51884802
>what is hibernate
ITT: tards and memesters.
>>51882013
People saying that are making shit up, don't know how to configure a system, of have greatly undersized their RAM.
You keep swap, but you turn down the swappiness so it only uses is when it's actually needed to keep the box from invoking the OOM killer.
Swap will get used from time to time, but nowhere near enough to wear out the ssd. We use ssds for journals at work, and they get constantly, and completely rewritten. The write cycles on them dwarf those on the ssds we have swap on... and at two years in, show no sign of wearing out.
My 4 year old Samsung 830 has been in 5 different computers, had dozens of OS and game installs and always had a swap file or partition on there. It's still at 100% life.
The mistake you made was buying a shit SSD.
>>51884832
Redundant when your computer boots in 3 seconds.
>>51886294
You don't like having all your work exactly like you left it? The current state of a dozen web apps, office, cad and usually an ide are a hell of a lot easier to restore from hibernation then by opening each one by one, logging into each one, and getting to where it needs to be.
Then again, I have a job, so little stuff like that is important to me.
>>51881996
>muh swap meme
I don't know what you niggers are doing wrong.
>>51887320
>4chan
>>51886361
>dozen of web "apps".
>office
>cad (in linux?)
then install everything into a VM and pause it.
also, web apps and hibernation?
total kek
>>51884559
ITS WHAT MOVES YOU
>>51887585
>then install everything into a VM and pause it.
"Just make it more hassle for yourself"
>>51886361
you should be using sleep mode instead of hibernate you dumb shit. everyone knows that hibernation is frowned upon with ssds because it causes unnecessary wear. its not going to make the drive fail unless its just bad luck
>>51887619
>you can virtualize your machine in very few steps.
>you can pause your work anytime.
>do backups of the entire system.
hassle? why?
>>51886234
>>51882166
>>51882141
This. Here's a wonderful article, it took the eighteen MONTHS of constant writing to kill off all the SSDs in this test.
http://techreport.com/review/27909/the-ssd-endurance-experiment-theyre-all-dead
>>51884750
I had a pagefile for years on a 2009 Sata 2 SSD.
Accumulated 50000 power on hours and 80 TiB of writes, and I'm finally putting it out of service this month after an upgrade. Not a single issue with it.
I just let windows manage swap on an HDD. You still need swap, even if you never use 100% of your ram. Otherwise, Windows under automatic swap handling would almost never write to the drives when you have a ton of excess RAM.
Also, not having swap or page filing increases disk thrashing. I guess that isn't too important if you're using an SSD though.
>>51888182
SSDs last a lot longer than most naysayer act like, but that's kind of a easiest-cast endurance test method.
TR's test spread the writes across the whole disk, whereas you could kill one faster by keeping the bulk of the disk full of long-term static data and churning writes over the smaller remainder.
SSD longevity is proportional to the degree of empty space available during when writes are happening, and you can kill one rather fast if you have no idea what you're doing.
>>51881996
>SSD fucks up because I stupidly put a swap partition on it
I'm running swap on my SSDs. It's my understanding that modern SSDs will outlast their HDD contemporaries.
>>51889003
That's not entirely true. An SSDs controller will move static data around to free up block which have more cycles left to wear the drive evenly.
That's an issue only 2008 and earlier SSDs had. Any SSD build in the last 6 years will not have this problem.
>>51886234
>how to turn down swappiness?
>inb4 /flt/
>>51889054
the problem is definitely vastly mitigated, but the relocation still ages disks a bit faster than repeatedly filling and emptying the disk as a whole.
>>51887636
except sleep uses power, which is not always available, dumbass
>>51889046
Each bit has a finite number of write operations that can be done to it in SSDs. Swap partition will constantly be overwriting bits. Not a good idea. It might be true that an SSD under normal usage will outlive an HDD but in theory an HDD has no finite life span. All ssd's do.
its a bad idea
>>51887606
PUT IT ALL TOGETHER!
>>51889097
It doesn't cost more than 1% of the drives life, which is a small sacrifice for the benefits.
>>51889262
you have any sauce on that?
I'd like to see some proof that SSD writes are really fungible to that level before treating them that way potentially foolishly.
Whats with all these annoying SSD threads. SSDs last long enough and are faster than mechanical drives as well as being silent. Any drive can die so have backups. If your data is all on your ssd and it dies. Your an idiot. SSDs are cheap now. Whats the debate. Stop complaining.
>>51881996
>"you're doom"
>>51882104
No, you need SWAP.
Oses aren't yet designed for no swap, not having any swap will fuck with them.
Anyway, just get a cheap ssd for a swap, I'm in the exact situation, used to use a 10k raptor as swap to protect my 2nd gen ssd, just recently got a new ssd, and got rid of the raptor with a storage ssd, and a dedicated swap ssd.
It's wonderful.
No more whirring at all, my fans are silent from outside the case, and when I do open a shit ton of shit, and start using swap, the swapping happens in a few seconds, rather than over a minute or so.
Anyway, just get one of these for a swap SSD.
http://www.amazon.com/gp/product/B003YMJPE8
>>51891043
That's actually a good idea, might get one just to toss my programs on
>>51888182
Fuck that article, according the article I should be seeing 750TB writes before failing but my MX100 won't even make it to 200TB at the rate it's health is dropping.
The swap partition did not kill the SSD.
The SSD died because SSDs die sometimes.
>>51891043
I have 32GB RAM, no swap, and zero fucks to give.
I have my crucial M4 for less than 4 years and I have a 16gb pagefile. My SSD is still at 98% health according to SSDlife
>>51891325
>assuming linear health to life relationship
top kek bro
There's also the issue of there being a range of durabilities, but they're all high.
But you are pretty stupid though. Linear relationships, kek.
>>51881996
>he fell for the SSD meme
Is this some forced meme about how linux is bad at paging?
As far as windows goes, if you have an SSD, just stick a page file on your OS partition. Dedicated swap partitions only make sense for actual spinning rust.
>>51891325
p sure half those drives hit '0 health' long before they actually started to fail
>>51891652
I'd really like to take that blonde for a test drive, if you know what I mean.
>>51891325
are you sure health=0 isn't just set at 200TB and it'll probably keep working past that? assuming it's linear, anyway.
>>51891325
>>51887320
Mine has been on 96% for fucking years m8
>>51881996
>>SSD fucks up because I stupidly put a swap partition on it
stupid motherfucker
>>51891934
>forced meme about how linux is bad at paging?
It's more the forced meme that a swap file will kill your SSD over the weekend.
I think it's mostly down to luck. As long às it has a half decent firmware and you're not unlucky with getting bad flash these things will last a long time.
>>51889729
My source?
a) Hardware engineers are not foolish enough to cause a lot of wear on an SSD in an attempt to reduce wear, despite all the scammy stuff their marketing departments pull.
b) How would moving static data every couple hundred cycles even cause that kind of wear?
>>51889095
https://en.m.wikipedia.org/wiki/Swappiness
Actually covers it better than the Ubuntu pages that often come up in the searches. There are a whole bunch of tuneables in the kernel that you access in /proc and /sys, and all of them can be changed on the fly. This, straight from the kernel docs, is just the memory specific ones. There are many many more.
You can learn a lot by digging into th, and what you can learn about a running system & change with them.
>>51895031
And I managed not to include the link....
https://git.kernel.org/cgit/linux/kernel/git/torvalds/linux.git/tree/Documentation/sysctl/vm.txt
>>51889169
Sleep mode is good enough, even my oldass t60 could live on sleep for a week with an old ass battery.
>swap partition
Wait, doesn't the OS just use the available disk for virtual addresses? What's the point? Saving everything on the RAM to the drive?
>>51887636
My PC annoyingly blinks with a led when in sleep mode. I stupidly bought a case with a huge-ass led.
>SSD fucks up because I stupidly put a swap partition on it
Ignore the troll.
>>51895891
When I built my desktop, I made sure my case didn't have any annoying LEDs on it that would light my room up at night.
Then my dumbass bought a pair of bookshelf speakers with the brightest blue power indicator LED I've ever seen. I don't know how many times I've had to get back out of bed just to go turn my speakers off because the power LED on them was keeping me awake.
>>51887320
>almost same POH
>twice the wearout
>>51887320
>>51891325
How do you get crystaldisk to show the health percentage, for me it just says "good".
>>51896037
sho, power cycle count is the same, POH is 1/5th, but double the wearout!!
WORRY^2!
>>51893987
the point is that there is a 2x difference in longevity between best and worst cases even with perfect wear-level remapping.
your 1% number seemed too good to be true and to be pulled out of your ass, and now I can rest easy knowing it was.
>>51896037
Mine is a LOT more basic m8. They are renowned for failing, I was just super lucky.
>>51896049
No clue, might depend on brand/firmware.
>>51896236
Yeah it might be the brand, I dont see the amount of data written either. But if it's "good" it's probably just fine, SSDs don't fail as easily as many people think.
>>51882013
Aside from Windows, there's very little need to use swap on modern desktops. If you use hibernate/sleep, or need to do a lot of debugging you'll want swap.
Manufacturers claim that modern SSDs don't degrade from writing like older ones. They also want to sell as many SSDs as possible, so I don't trust their word. I don't take chances. /, /usr, /usr/local, / boot, and if used, /boot/efi go on the SSD, and /var, / home, /altroot (for OpenBSD), and /pkg-cache (OpenBSD again) go on a HD. /tmp goes in tmpfs if possible, otherwise it goes on the HD.
>>51882167
BOOSH!
>>51888858
It is used on Windows. It is used routinely on Windows, for no good reason than just to be using it.
>>51896222
Explain to me how moving static data every few hundred cycles would cause as much wear as you claim.
>>51896732
>Aside from Windows, there's very little need to use swap on modern desktops.
that's some high-quality delusion.
>>51893987
Good source, glad you know what that question means
>>51896791
>doesn't know what it's used for
>thinks it must be used for nothing
Look at this retard.
>>51881996
Frisky Dingo was a a great show.
>>51889180
>Swap partition will constantly be overwriting bits
What the fuck kind of shitty operating system are you using? Not even fucking Windows constantly writes shit to the swapfile.
>dur I have an entire gigabyte of RAM xd
>>51888182
>constant writing
if you passed math in elementary school then you should know its total bullshit.
you can wear out SSD in 1 day of constant full speed writing.
>>51887585
>then install everything into a VM and pause it.
what the fuck do you think "pausing" a VM does, you idiot?
it saves the VM RAM contents to disk, same as hibernating.
>>51897365
I think I will trust the well respected reviewing site over a random poster on 4chan.
>>51882013
Have swap on chinese ssd for last 2 years on machine with 4gb ram so swap is used a lot and ssd still works like new.
>>51896882
the rate of relocation-per-overwrite is obviously near 0% when the writes are spread perfectly or at least pretty evenly, but it will reach 100% is the disk is full and the rewrites are focused on a small set of logical blocks.
if the controller didn't operate this way, the SSD could be killed trivially.
> e.g., write 249,999MB to a 250GB disk, then overwrite the same last 1MB only a couple hundred times, and you're done. (reserve blocks defer death, but it would still be soon)
as the disk's nominal size fills up and the reserve blocks die off invisibly in the background, you have to do a relocation for essentially every write to spread the pain around, not every "few hundred cycles", whatever that means.
>>51896955
>Nuh uh!
>Cuz I sed so.
Do you actually have any other reasoning for using it besides this?
>>51895891
you could just disconnect the LED if it annoys you that much. I typically find no use for the power and activity LEDs on a computer case anyway.
>>51897365
>well reviewing site can defy math
>10k RPM
Where do I get a 10000RPM HDD?
>>51897980
>10k RPM raptor
If only there was a way to search online for such a phrase
>>51897641
In that extreme example, it will wear the free 1 MB of space by a few hundred cycles, then move some static data there, and wear another 1 MB for a few hundred cycles. Rinse and repeat ad infinitum.
Even when your drive is filled 99.99%, you still will have to wear down the entire drive before it fails.
No single SSD will only write to that free 1 MB space until it dies, then reallocate a reserve block, and kill that one, and so forth.
>you have to do a relocation for essentially every write to spread the pain around
No. The controller does not forcibly keep all cells autistically at the exact same write count.
Meh.
>>51898400
>37TB written
>100%
samsung-quality
>>51898511
It's MLC so it's bound to be robust.
TLC isn't as durable now, but it's only gonna be better.
The G:\ SSD is solely for games so it's not been written to as much.
I have swap and full disk encryption on all my SSDs. Both Samsung, no problem yet.
>>51898566
I remember when nobody would touch MLC with a 10-foot pole and SLC was the only way for reliable SSDs.
The most expensive "enterprise" SSDs are probably still SLC
The smaller the tech the easier it is for it to break.
I made copper parts for machinery where the circuits where 18x2mm big because it was made to last.
Set vm.swappiness=0 you mongoloid
>>51897365
>you can wear out SSD in 1 day of constant full speed writing.
Citations etc
>>51888182
while its true it would be very hard to actually hit the write limit of the drive.
what is really interesting from that is how everyone loves to brag about how 'when it hits the limit, its just read-only!'
well, that isnt fucking the case, the drive either nukes itself, or it is read-only just until it reboots, and then the data is lost.
the problem is ssd reliability isnt really the write number, it is when the drive fails whether it be the flash or the controller you lose all of your data, while with spinning disks you can almost always get your data back.
i use ssds in my laptops, but i use a raptor and other spinning disks in my desktops
>>51898658
Copper won't break as easily as hardened steel, because of how soft it is. I've had larger parts made with hardened steel just shatter at seemingly nothing. I mean you aren't going to bust it up with your bare hands, but a press can fuck it up a lot easier than most people would think.
>>51899288
>while with spinning disks you can almost always get your data back.
After one of the mechanical parts fail?
>>51891043
>Oses aren't yet designed for no swap, not having any swap will fuck with them.
Retarded the post
absolute nonsense
swap is rarely used unless you have 512 MB ram
and faggot with ssd has at least 2GB
>>51881996
swap partitions are still a thing? Jesus christ. This is what you get for using linux.
>>51899425
head failures are not nearly as common as they used to be, especially since these things have been in production for decades. even with a head failure if you pay the money you can usually get something back.
if the controller fails you can simply swap to a new one and be on your way like nothing happens. if a controller fails on the ssd even if the flash was in 100% good condition, the data is lost
the reason ssds are good for laptops is you dont have to set the drive to power off during movement to prevent failures, but in a desktop as far as head failures go it is just the luck of the draw
>>51899288
>while with spinning disks you can almost always get your data back.
For several times the cost of the drive.
>>51882013
Only if you have an older model.
SSDs made after 2012, even with swap, have enough cycles to last over 20 hours of high performance usage.
It's the older models (especially ones that are only SATA 2) that die very quickly.
>>51898566
>It's MLC so it's bound to be robust.
>wear leveling count is at 0x1a6 (362 dec)
>meaning the drive has written 363 (sic!) times to each cell
>830 series is rated at 3000 p/e cycles
ergo: 88%
your welcome.
>>51899519
>over 20 hours of high performance usage.
That doesn't sound very flattering
>>51899519
>20 hours
Kek
>>51899288
>it is when the drive fails whether it be the flash or the controller you lose all of your data, while with spinning disks you can almost always get your data back.
I think an important difference is that you can very easily disassemble an SSD and recover data with home tools, while data recovery when a controller fails on an HDD is almost impossible without professional equipment.
>>51899499
What do you suggest as an alternative?
>>51899550
>>51899549
But those 20 hours of high IO will be worth it.
>>51899563
Where do you get those 20 hours from senpai
>>51899551
id like to see these home tools, because the controller maps the actual blocks of data, if it fails you are in a bad place.
if the controller fails on a hard drive you literally just buy another one and attach it
>>51899621
Note the 4 pins on the top left. Power, ground, input, output. Don't need to buy anything or even replace the controller to talk to the device.
>>51899684
you are missing the point. if the controller is lost then when you put a new one in you have lost the mapping of blocks, which means all of your data is gone.
>>51891362
>>51899451
It's true though.
OSes expect swap, an won't perform optimally without it.
http://www.alexonlinux.com/swap-vs-no-swap
Your performance loss.
>>51900490
Windows has similar issues.
http://lifehacker.com/5426041/understanding-the-windows-pagefile-and-why-you-shouldnt-disable-it
Disabling SWAP is a retard move.
Configure your system to use it less, and if you're worried about drive life, spend the 10 dollars on a dedicated swap drive.
>>51900490
>http://www.alexonlinux.com/swap-vs-no-swap
Holy shit, this is painful to read.
>I never heart of a program
>yet it significantly prologues live of the individual processes
I know he claims his name is Alexander Sandler but I smell curry and poo in the loo.
>>51900724
Then google other articles/read
>>51900555
OSes are optimized with the expectation swap exists, since historically it was very important.
That hasn't changed yet.
>>51895891
>>51897839
>>51896006
>only 1 room
>being this poor
>>51889180
>>51881996
I have swap on ssd and even with 8gb ram only it is empty 98%of time. Op fucked it up somewhere. I dunno where you take constant writes linux manages swap bretty well
>>51900868
I wasn't questioning whether or not disabling swap has a negative impact on Linux, I'm just saying that you picked one of the worst possible articles on the net. He's a clueless muppet, making baseless claims and is totally incapable of writing a paragraph without a significant grammatical error.
>>51901309
Meh, blame google, he's the first result for
linux no swap issues.
http://superuser.com/questions/639618/linux-dont-create-swap-partition
i was going to say something but then i realised i've only turned this computer off 42 times since 2012
i should probably go outside
>>51881996
>10k RPM raptor
>it fucking sucks
Really, OP? That is right about the best that a consumer HDD gets. You need to lay off the 1337 "gamer-grade" components for a bit and come back to the real world.
>>51901491
And it's pretty bad these days.
It was nice in it's day, but compared to an SSD, it's incredibly slow, and very noisy.
>>51882104
>>51881996
>>Swap Partitions rack up the R/W cicles on ssd's like there is no tomorow.
Uhm not sure if this is a bait thread.
Even with 2gigs of ram years ago I had only a minor amount of data on my swap partition.
Now with current hardware, even with default 'swappiness', I barely have _any_ data on them. I mean, I do have Swap, but because I have more memory, the OS does not use it.
And well, the first thing you do, is adjust swappiness.
Right, OP?
I mean, we are on /g/....
>>51901627
>And it's pretty bad these days.
It depends on how much ram you have.
How much ram you have anyway?
4gb? lol.
>>51902271
What?
First off, you need swap regardless of ram amount.
See
>>51900555
>>51900490
Second off, a normal 7200 RPM drive is what you get if you want storage, a ssd if you want speed.
A 10k drive is faster than and HDD sure, but doesn't compare to ssd, and with their market vanishing, you can't even get them in large sizes.
>>51902344
I never said that you don't need one...
My point was, that if you have enough ram, swap rarely - if ever gets used.
And everyone should modify the vm.swappiness value on Linux anyway.
By the way 10K drives were like SSD these days - it does not make sense to buy them anymore. OP just had one lying around.
>>51902360
Yeah, and that's exactly what I said.
They were great in their day.
I actually just replaced my old 10k drive a few weeks ago, I moved it to storage when I got an ssd for my main drive.
(And a network storage off in the closet for large storage.)
Anyway, I'm saying that 10ks were good, aren't anymore, and you need a swap partition, not how much you need.
What re you arguing with me about, we seem to be saying the same thing?
>>51900555
>2009
Do you actually have something recent?
>>51896741
>and/or Kakow
>>51896037
>almost same POH
But his has 25000 hours and hours likes 5000.
>>51887619
Only tards and GAYmers use Windows outside of a VM in this day and age.
