VincentKeyboard
After ~6 months, my ssh health suddenly dropped from 100% to 99.0% according to hwinfo.
1) What was it not gradual?
2) I presume even 1% per month is not "bad" as it means a 100 months. But what exactly does this mean? Did I just lose 1% of the storage capacity?
nuclearcat
VincentKeyboard wroteAfter ~6 months, my ssh health suddenly dropped from 100% to 99.0% according to hwinfo.
1) What was it not gradual?
2) I presume even 1% per month is not "bad" as it means a 100 months. But what exaclty does this mean? Did I just lose 1% of the storage capacity?
Approx meaning: possible write cycles count.
https://www.ontrack.com/blog/2016/10/25/wear-leveling/
VincentKeyboard
Oh. I see. That explains a lot. Thank you for the link.
rolf
Your SSD resale value just dropped.
VincentKeyboard
What happens when the number reaches 0?
nuclearcat
In best case it will become read only. In worst - you will lose all data on it.
I suggest to not reach lower than 15%, experienced earlier failures on some extreme case usage (but more likely due too high DWPD).
VincentKeyboard
I just saw the prices. pcandparts are selling the 1TB Intel nvme for 139 USD. I presume that one has quite a bit more write cycles?
nuclearcat
Not necessary. Newer SSD mostly TLC and have _less_ cycles, if we talk write per cell. But because capacity higher, amount of TB you can write per day and per lifetime is higher.
Check this:
https://www.mydigitaldiscount.com/everything-you-need-to-know-about-slc-mlc-and-tlc-nand-flash.html#eMLC
anayman_k7
I am currently using 1TB Intel 660p Nvme SSD which is based on TLC, I watched once a video about the longevity of this drive and as far as I remember it needs like 40 years to die based on write cycles and an average write values from users, but I am sure it will die much more earlier than this, hopefully it will give me signs that it is going to die so I can backup stuff, based on my experience the main sign is freezing with high disk percentage usage while reading/writing a very low amount of MBs per second
VincentKeyboard
This is partially Window's fault. It takes me three years to write 15TB on a hdd on Linux.
Windows, however, won't keep the disk activity down.
Even at idle, the System process, To-Do, and Skype are doing a combined 0.5MB / second.
DNA
VincentKeyboard wroteThis is partially Window's fault. It takes me three years to write 15TB on a hdd on Linux.
Windows, however, won't keep the disk activity down.
Even at idle, the System process, To-Do, and Skype are doing a combined 0.5MB / second.
change pagefile location to the HDD as it writes a lot to disk and you will be good, most system processes mostly do read operations and small writes for logs and such
nuclearcat
DNA wroteVincentKeyboard wroteThis is partially Window's fault. It takes me three years to write 15TB on a hdd on Linux.
Windows, however, won't keep the disk activity down.
Even at idle, the System process, To-Do, and Skype are doing a combined 0.5MB / second.
change pagefile location to the HDD as it writes a lot to disk and u will be good, most system processes mostly do read operations and small writes for logs and such
Wont help much, as windows is pile of crap in this matters. I wonder how much write amplification is happening because of similar issues on write:
https://twitter.com/BruceDawson0xB/status/1170924849004367873
DNA
Well on my windows pc my ssd went down like 90% when i changed pagefile location its now negligible for the ssd's health.
Well this can't be the case for writes as it doesn't work this way and actually if that's the case with writes it is better for SSDs health to write 68bytes than 4K blocks each time for example although it is much slower.
I see what this guy means i agree with him but i don't see that being an issue as most explorer reads are from memory and not the disk so latency is minimal the only things i guess that are read from disk every single time are things that can be changed like history or things like that. i will test it out on my pc 700ms seems exaggerated if it takes that much every single time.
nuclearcat
Exactly opposite, 61x68 bytes much worse than 1x4K blocks.
https://en.wikipedia.org/wiki/Write_amplification
It is explained somewhere in other tweets, each syscall have overhead cost. At the end we get 700ms delay, while it should be near instant operation.
DNA
nuclearcat wroteExactly opposite, 61x68 bytes much worse than 1x4K blocks.
https://en.wikipedia.org/wiki/Write_amplification
It is explained somewhere in other tweets, each syscall have overhead cost. At the end we get 700ms delay, while it should be near instant operation.
No dude you didn't get what i am saying, i know 61x68 is much much worse than 1 4k.. what I mean is that if windows for example needs to update a file with just 1bytes of data for example and can do that directly without having to write 4k blocks it would be much more efficient for SSDs health but it can't do that nor can any OS. for sure if you need to write large chunks 4k is better.
VincentKeyboard
I symlinked firefox's cache dir to drive D. I will change pagefile location now.
I also noticed that every hibernate is a 3GB disk write... I am not sure how to change the hibernate image location on this stupid OS.
On Linux, it's just a matter of creating a swap partition on the hdd and booting with resume=/dev/sdxy (where sdxy is the swap partition on the hdd).
Edit: Windows doesn't wish to use the pagefile on the hdd and says it created a temporary pagefile on c:\ the ssd so the pagefile trick doesn't help.
DNA
how come it don't wish to create a pagefile on your HDd there is definitely something wrong from your end. if you can provide more info...
well windows isn't stupid there is something wrong for it to create a temporary pagefile. linux being the open source OS it is things like that are simpler. but overall windows is much better
VincentKeyboard
@DNA I disabled the pagefile on drive C, created one on drive D and rebooted. Windows complained with the following warning "Windows created a temporary paging file on your computer because" and the pagefile on C was back. I reverted my changes and rebooted.
There was a unused paging file on D:\
Regardless, my machine doesn't use half its ram so paging isn't an issue.
Anyway, what would have helped was if I could move hibernation image to D:\ but windows needs it at boot before mounting D:\ (according to google) so this approach won't work.
VincentKeyboard
Ok, I did some investigation (resource monitor) and most of the disk activity is in two locations:
1) The hibernate image. I solved this by disabling hibernate. I only lose 2% battery per hour on standby anyway. This was 3 to 4GB of writes every hibernation. Pretty bad.
2) heavy writing to c:\users\account\Appdata\
Is there any way to move my profile to the HDD?
@DNA There are a lot of shortcoming in the Windows architecture and ecosystem. One of the main reasons why Windows applications start up slower than Linux is the excessive internal library shipping. I have transmission, onedrive, cmake, etc.. all using their own copies of Qt. If they used a systemwide installed copy instead, startup would be faster as the copy would be already cached in memory with the first application that used it.
The same concept applies to disk writes. There is likely too much duplication and inefficiency.
DNA
There are lots of ways to fix the unmovable pagefile do ur research am sure you will fix it, personally most of the usage is from pagefile i use up lots of ram.
for the appdata yea you can move it the using the symbolik link you mive the folder and link the original directory to the new one there is a tutorial u can easily find it on google but please be warned if done incorrectly it could corrupt your profile and apps.
I know windows is bad compared with linux in the way it runs things but it is bery complex and runs on older artitecture it needs an overhaul for sure but as it is and ith my PC i face 0 issues and no lag whatsoever the context menu mentioned above opens instantaneously am not sure where the 700ms can be felt.i have zero issues with windows really.
