Great work Ao1!
Will be interesting to see how long it takes for E6 to "recover" or if there are any other movements during the idle time.
Great work Ao1!
Will be interesting to see how long it takes for E6 to "recover" or if there are any other movements during the idle time.
-
Hardware:
Something strange to report. I stopped the endurance test and left the drive to idle. The idea was to let the drive idle for 24hours and then restart the app to see how much could be written before LTT kicked in again.
Once the drive was in idle mode I noticed that the power on hours were not changing from the 166 hours reported at the start. All other drives (including the V2) reported increased hours during this period. (Currently I have 4 SSD's connected to my system).
Whilst the reported hours did not change the raw value for power on hours did change. At the start of the idle period the raw value in 1 byte decimal it was:
52 203 166 0 0 0 166
The values in bold changed every minute. Currently this value is being shown as:
47 116 166 0 0 0 166
Weird. The drive has also remained in a throttled state!E6 has not changed either.
Another surprise is that it has transpired that the V2 40GB drive & the V3 60GB have been set with a 1 year LLT. That makes a huge difference and explains why the V2 was throttled down to 7MB/s.
The V2 was set to:
Life In Seconds: 31536000
Memory Wear Cycle Capability: 3000
For comparison a V2 LE 100GB was set to:
Memory Wear Cycle Capability: 10000
Life In Seconds: 94608000
At first glance it might seem strange that the V240GB was set to 1 year and the V2LE 100GB at three years, but the PE cycle capability holds the answer.
EDIT:
Now that I am writing to the V3 again it seem the power on hours are increasing.
Strange, are you sure that the drive isn't allowed to "power off" during idle time?
-
Hardware:
Dooh.The PC is set to not sleep/ hibernate etc, but the hard disk power setting was set to 20mins.
So much for green computing.
I'm running compression tests currently using your app. (Currently working on 8% fill).
Will revert to the endurance test later.
Working on the thread and I've found someone willing to host the download (TheSSDReview)
-
Hardware:
Drive SE'd before commencing. 8% fill 1GB test file. (Generates random xfers sizes).
25% fill comming up
Drive SE'd again. Avg write speed @ 25% fill = 18.17MB/s, so I already know that 25% is not being compressed very well as the write speed is being throttled. 8% fill was around 50MB/s.
This is going to take some time
Using the V2 for such a task would take time so you're better off waiting for the SF2 series drive.
How is that old V2 drive doing?
-
Hardware:
I'm running this on the V3 due to the 1GB SMART update reporting frequency. The 64GB update reporting frequency on the V2 makes it unsuitable. The throttled state will mean it takes time but it won't effect results.
Currently I use the V2 as a temp download folder.
Will you create a new thread to announce the beta release of your app?
Drive SE'd before starting
Last edited by Ao1; 07-27-2011 at 11:24 AM.
I put the power setting for the hard disk to turn the power off after 5 minutes. I have an X25-M that is idling. The power on count increased for X25-M, so either something unbeknown to me was accessing the X25-M and preventing it from going to sleep, or power on hours are still recorded when the drive is "powered off".
Kind of strange that the V3 has to be fully powered up to count LLT hours and presumably run GC background tasks.
Personally, I'd prefer a disk obey power state requests from the OSOriginally Posted by Ao1
Kind of strange that the V3 has to be fully powered up to count LLT hours and presumably run GC background tasks.
Added NTFS compression and your SF-2200 compression numbers to my compression curves charts. I intend to do the same with the LTT-less SF-1200 when I get it
You obviously can't add my C Volume and D Volume data, so that's been made to a trendline so that it's bridged between NTFS and RAR-fastest. NTFS was incapable of compressing 67%, 67% No Dedup, and 101%; file size didn't even change one byte.
Based on 8% and 25% settings, it looks like the SF-2200 is actually more powerful than NTFS compression
Great graph. 25% fill is now complete and 46% is fill coming up.Personally, I'd prefer a disk obey power state requests from the OS
Regarding the power state I believe that the power setting puts the hard drive into a lower power state as opposed to no power at all. If that is correct it seems strange that SF does not record that the drive is powered on in context of LLT. In a low power state it is still reporting changes to SMART attributes, so why not record power on hours to help alleviate the impact of throttling?
GC could also benefit from a low power idle mode, assuming there is enough power for this to take place.
Last edited by Ao1; 07-27-2011 at 09:18 AM. Reason: typo
While such a state makes sense with a hard drive (spin down the platters), it is less clear with an SSD. I do not remember reading of any low power states with SSDs, and I would be suprised if all SSDs implement a low power state. I suppose it could underclock its controller, or possibly cycle the power to the controller off (and back on periodically). But the gains of such a state would not be great. On SSDs with a large DRAM cache, I suppose there could be moderate savings by clearing and not refreshing the DRAM. But the Intel controllers and Sandforce controllers do not have large DRAM caches. So I think the savings from a low power state would be minimal. May as well just have on and off states.
Agreed on the timers, I'd like for the power on counter to move while in low power, or at least some internal timer to move for LTT....if making a drive last X amount of time is the goal and the drive knows time is passing (i.e., it has some power), it should acknowledge and behave like time is passing.
Last edited by Ao1; 07-27-2011 at 11:36 PM.
Looks like 46% is landing right between the C Volume and D Volume curves so far. And SF-2200 being more effective than NTFS is still in effect too
46% complete now as above. 67% in progress. I ran a quick copy of the Windows and Program folders. The V3 was able to compress a Win 7 & Office install by 50%. Transferring the Windows and Programs folder could "only" achieve 25% compression.
OK something very strange is happening with the power on hours with the V3.
As far as I can see the V2 is counting hours even if the power settings put the hard drive in a low power state.
The V3 does not seem to clock hours unless it is transferring files. (Even when the hard drive is not set to power down).
Maybe this is some sort of work around due to the power related problems with the SF2xxx drives that cause BSOD/ Freezing? I'm using f/w version 2.08.
Has anyone else noticed this on a V3?
Last edited by Ao1; 07-27-2011 at 11:56 PM.
Last edited by Ao1; 07-28-2011 at 06:58 AM.
It's not getting much rest
Great job on the SF2 controller, things are looking better than SF1, as I suspected.
For the lurkers, there is a new firmware for the SF2 Series drives, currently OCZ is the only one that has made it available for download.
-
Hardware:
That's very, very odd....and not a good sign of LTT depends on power-on hours (the only way to allow unthrottled writes is to write?). Do reads up the counter? If they did, they would at least be a stop-gap. I would be surprised if this is related to the BSOD/Freezing, though.
93.4% after 3 runs, still doing better than the 100% that NTFS managed, nice. Obviously won't out-do NTFS with 101%....NTFS doesn't resize it one byte but surely SF-2200 has a WA of greater than 1.00x.
I think I might have the ability to test a SF-2200 with my own C Volume and D Volume data sets to round out the charts, but probably not for a week or so. By then or slightly after that SF-1200 compression testing (taking 64x longer, sigh) will start and after that SF-1200 endurance testing.
Here are a couple of graphs to summarise the findings on compression.
The comparative compression graph shows a fixed level of compression for files generated by Anvils apps. These compression levels are a given and are then compared to what the SF controller can achieve.
Just updated to f/w 2.11 and it seems that the E6 raw 1 byte value for the life curve value is no longer available so I won't be able to monitor the life curve recovering
Power on hours increase now, even after the hard drive power setting is activated.
Bummer that E6 disappeared, was really hoping to see that tick back upward with idle time working.
Not sure if you ever mentioned this, but what was the write speed in Anvil's app with LTT active? If it's anything like the 16-18MB/sec from AS-SSD, then LTT write speeds were almost surely faster than the lifetime slope. 16MiB/s = ~21 disk writes (with parity space, so 64GiB) per day, or 7700 P/E cycles a year, which is way more than what IMFT 25nm is rated for.
Reply With Quote
Bookmarks