After many test, here are the programs with most bytes written:
- Antivirus Software
- Firewall log file
- Some files belonging to my Firefox Profile are constantly written to
- Windows Temp Directory
After sym-linking some of those directories mentioned above to another hard drive, I did another test. In this test, I let Process Monitor record for 12 hours. I was not using the computer at all. Total Written Bytes where approx. 1.5GiB, Intel Toolbox recorded 4,44GiB. This might be due to physical disk sector size.
Now, the two Folders taking most written bytes to are Virus Scanner and Windows Temp directory. I thought about redirecting the Virus Scanner folder but need to measure the system impact before I'll do that.
Original Post for reference:
Owning two 128GB SSDs, both suffer the same phenomenon: Even if I do nothing, the host writes value increases very fast.Both SSDs are in different computers (both Win 7 x64). One computer is switched on 24x7. After 150 hours my host writes value is at about 0.5 TB. What I tried to track the issue down:
- relocate database writes to another harddrive (with junction points)
- ended all running programs
- ended some services (like firewall and virus scanning) With Process Monitor and a filter for "Operation contains Write" and "Path start with C:\" many events are captured, but data written from all events is not even close to some megabytes per minute. Do I have to worry about this? Warranty is for about 35TB I read, so I might be there very soon.... Can someone help me, please? Jens Edit: Host Writes seems to increase in minimum steps of 60MB. And only once a minute.
Very strange of course. What are you using to check the amount of Host Writes? I imagine the SMART data, but what is interpreting it?
I've read about a similar phenomenon in another manufactures forum regarding one of their SSDs, but the issue was not resolved.
Just some suggestions, did you try running Windows in Safe mode, and see if the same thing happens? Have you tried using Resource and Performance monitors, to possibly find what is causing this?
Windows Task Manager, in the Processes tab, allows you to watch the I/O read and writes at the byte level per process. Those columns are not displayed by default, but can be selected with the View button, and Select Options. Scroll down to find six I/O options, among others that can be displayed. You can also change the update speed of the display with the View button. If you've never looked at this data before, I suggest that you do, it just might surprise you.
SMART write increments are updated every 32MB. 0.5TiB over 150 hours = AVG ~1MiB/s.
An AV scan is primarily a read operation. Do you by any chance have a lot of RAM and do you frequently use hibernation/sleep? If so that can potentiality generate a lot of writes, as any info in memory is transferred to the disk/SSD. I'm not 100% sure if the Resource Monitor, or the Intel Toolbox for that matter, are able to record the writes that occur as the PC goes into hibernation mode.
If host writes are being recorded by SMART/ CDI the Resource Monitor should be able to see the correlating write activity.
Another tool you coudl try would be Process Explorer. You can use the View>Select Columns to select Disk Writes.
I have 8GB RAM, but don't use any kind of sleep or hibernation. As I said before: summing up all written bytes per second for a minute (roughly estimate) I get a value around one megabyte.. Over the last hour, I run my home PC in Safe Mode (without touching a key), stats gone up approx. 70MiB. That's much better. But:Just opening Excel and updating my chart with this gave an instant increase for another 60MiB. So two increment step seems to be 130MiB. This without any virus scanner etc. in safe mode. Is that usual? I don't know. As a test I copied a 10GB file to see the host writes counter. But that did increment just as is should. Going on that fast, I reach 35TB "limit" soon. I'll gladly test any new idea! :-) Jens
I just checked the 510 Product Spec and the update frequency is 32 MB. I write around 0.65 GB of host writes per power on hour. (E1 Host writes / 09 Power-On Hours Count). I don't often copy large files, so that is mostly OS and app generated writes.
You are writing 0.5 TB/ 150 hours = 3.41 GB per power on hour, which seems quite high.
If you are feeling adventurous you could try hIOmon to find out exactly where the writes are being generated.
I belive this is in fact an issue within the SMART's Host Writes algoritm because I also noticed host writes value being to high compared to the drive usage. When I first received my 510 120 GB drive , and written the first 60-70 GB of data the Intel SSD toolbox reported something above 500GB. After it went through the 1 TB, I decided to disable page file, but despite very few writes which mostly consists of browser cache I am now at almost 2TB.
It could be an error in the reporting mecanism, especially knowing Intel's late track with their products.
Carved, I have a similar experience with my 510, which is noticeable in contrast to my G2 80GB SSDs, which I have used for a much longer time. The 510 shows nearly twice the amount of writes than any of my G2's do, which are more than a year older than the 510. My usage of each should be virtually identical, except for the length of time. Of course, the rest of the PCs they've been used on are not identical, but all use Window 7. The software used on the PCs has evolved over time, and newer versions could be contributing to the difference. I've wondered about the SMART data reporting on the 510 as you mentioned, which could be what we are seeing here.
My 510 seemed to be writing quite a few GB per hour as well. I have the same OS image on several drives, and when I swapped the 510 for an X25-E I dropped to only about .9GB a day of host writes. The good news is you can use the Windows task manager and add host writes (In bytes) so you can figure out what's writing the most. My worst offenders were google chrome and windows live mesh. If you look at the first entry in my taskmgr shot, you'll see that Anvil Storage Utilities had written over 4TB since I opened the program 10 hours ago... but it's writing to a second SSD, and I wanted it to do that.
All you have to do is add the colums you want to see, and pick the IO's.
You can use this to track how many writes and reads each program makes since your windows session started.
Just for your information, "Anvil Storage Utilities" has written in 10 hours 43 MB. (Not 44 TB)
The write load is negligible so.
On my picture you see posted 31 GB of "Symatec" ("Norton Internet Security"). This corresponds to 0.03 TB. However, this was written over several days and is not a problem.
I observe this problem for quite some time. Unfortunately I did not have time to test in this matter.
I have since 07/2010 Intel G2 160 GB. Start by cloning (07/17/2010) with 194 GB. Until today (10/30/2011) 2480 GB.
That was exactly 471 days and 2286 GB. Average per day 4.85 GB.
Since 06/2011 I have an additional series 510 250GB. Start by cloning (6/18/2011) with 119 GB. Until today (10/30/2011) 3900 GB.
That was exactly 3781 MB in 135 days. Average per day 28 GB.
Both SSD work in "100% identical" conditions. Same hardware "HDX Dragon" notebook.
Operating system 100% identical as clones. (Dual Boot Vista 32 and Win7Prof.64)
There are no differences of the systems and software.
Systems are fully "optimized SSD." (And I know exactly what I do)
8GB of RAM per system. Page File on the Vista32 bit Ramdisc. Win764 bit without page file.
Temporary files on all systems on ramdisk.
Hibernate is not used.
Systems are more on than off. (Average about 13 hours per day)
Of course, sometimes off, sometimes turned on for "24 / 7 hours.
Main system is using Win7 64 bit.
Both hardware will be used equally. Main use: Internet browsing
No software to write as much data. The operating system is not guilty.
Largest software write load for me is "Norton Internet Security." Average of 1GB per 24 hours.
This program works, however, has always been on "Intel 160GB G2".
So the hardware is to blame "Intel SSD 510".
There are two possibilities:
1.Intel calculated from the operating time of a "Host Writes value" of the warranty is adapted.
Warranty: 20 GB per day for 5 years.
This option is no physical "host writes," and would not affect the service life.
(Unless Intel has even programmed a "35TB forced shutdown.")
2.The write load (host writes) is physically generated by the controller. That would be a nightmare.
But even with this possibility, the SSD is working to secure more than 35TB.
(Unless Intel has even programmed a "35TB forced shutdown.")
http://www.xtremesystems.org/forums/showthread.php?271063-SSD-Write-Endurance-25nm-Vs-34nm Here is a link to some long-term tests. (Sorry, no 510 here)
With the enormous write load the 35 TB limit warranty in 3.5 years would be achieved. Calculated from the installation of the SSD. That would be a great impertinence.
Whether the firmware plays a role I do not know. I have "PWG4".
Who among "PWG2" (or PPG2 at 120 GB SSD) also has a write load of about 20 GB in 24 hours, please post!
I pursue this issue further. With new knowledge, I'll write here.
An Intel employee:
I ask for clarification of this issue! The Series 510 is acquired because of higher durability.
This uncertainty in the consumer may not be in the interest of Intel.
I see this problem as "not solved" on "!!!!!!!!!
Ps Sorry, this is Google translator!
Jenson, Just so I understand, what is your main concern, that your warranty will be void due to the write threshold being exceeded before the warranty period ends? That certainly is a valid concern, I would be unhappy about that as well. I've never heard of anyone complaining about being refused warranty service due to excessive use of a SSD, but who knows?
IMO, the so-called SMART data and it's interpretation is one of the dumbest things I've ever seen in the PC world. For example, the SSD Toolbox 3.0 (which IMO is excellent) reports in the SMART data that a conventional HDD I have, about a year old, has a "Total LBAs Written" of 127.86 PB. Yes, PB, PetaBytes. I know that is a misinterpretation, but just an example of the free-for-all that SMART data is. Not to mention another attribute that was unknown to me, "SATA Downshift Count", which I can imagine would cause users that experience "downshifts" much grief as well as the manufactures of their drives and mother boards.
The Media Wearout Indicator on my 510 is reported as 0 100 0 (Raw, Normalized, Threshold), and I have Host Writes of a bit under 0.5TB. Two Intel G2 80GB SSDs that I can check at the moment both report 0 100 0 as well, one at almost 0.75TB, the other approaching 0.25TB. A new Crucial M4 SSD I have does not even report an E9 number in the Toolbox, or Host Writes, but does show other attributes.
Did you mention earlier what your E9/Media Wearout Indicator shows in the Toolbox? If others could post their values, I would appreciate that. What is the value for E9 that would indicate worn out/no warrantee?
Just so you know, the SSD Toolbox doesn't calculate host writes on HDDs correctly, so it will always be 127PB which is what I believe the maximum value is. Furthermore, the M4 has no host write attribute, so you can't ever know how much has been written to it. Also keep in mind that some utilites, like CrystalDiskInfo, will report Intel SSD life as the spare area attribute instead of MWI.
Basically, even if your drives had 50TB and 75TB of host writes, that's still nothing. Intel will guarantee that a drive will write at least 75TB for the 300GB 320 model. The reason it's so low is those are full span 4K random writes, which dramatically increase write amplification.
I've been taking part in an SSD endurance test on the XtremeSystems forums. 1G, 2G, and 3G Intel drives are being tested as well as a couple SandForce drives, an M4, and a few others. My Mushkin Chronos Deluxe 60GB has 460TB of host writes on it and is increasing by 10.8TB a day -- that's a lot. The M4 had 750TB on it when it died a few days ago, and the X25-V and 320 40GB (with over 500TB on it) are still in the test. So if you're only writing between 1GB and 12GB a day to your drive, it's not going to die before you will.
The SMART data is important, but there is no standard across manufacturers. Some drives report host writes in terms of sectors written, while others use straight GB, and Intel uses 32MiB chunks. Some attributes have the same label, but report different values. Every drive basically has some "Life Left" attribute, and they're super conservative. On most drives, the MWI/life attribute will go from 100 to 0 and then back to 100, then back towards zero. SandForce drives go from 100 to 10 where they stay pretty much forever. So that's not even a good indicator of actual life, but JEDEC specification that state the drive must retain data for 1 year without power, and if the drive gets too worn it can't do that. What's supposed to happen is the drive ends up becoming read only, and then your data should be accessible for a year while the drive is disconnected from power. It's unclear that this will always happen, but so far every drive should be able to put down hundreds of terrabytes with no ill effect.
Jenson, I recently read that Win 7 will create a Restore point every time the PC is booted. I don't know if that is accurate, but that may be contributing to your Host Writes. Then there is Pagefile, that is equal to the amount of DRAM you have, but is that written on each boot?
Ckryan, perfect examples of why SMART data can be totally random. When firmware can be programmed to write values to the SMART data whenever it is so inclined, where is the consistency? I am well aware that SMART data is totally inconsistent between manufactures and devices, which is why I don't pay much attention to it.
The Pagefile resides on another drive. This 'habit' is something I'm used to since like Windows 2000? Though I don't know if this is still some good advice.