Sony Z11 - Long Term SSD Performance - Post your results

Thanks for the reply and for the links to the explanatory information on wear-leveling. It appears that my worry is a non-issue.

 

sorry for this noobish question, after reading thru this thread I am still a lil confused. So does this mean I do not need to do anything to maintain long-term performance of my SSDs?

And are the SSD drives in Z11 considered good SSD drives? Cuz Anandtech mentioned that there are good and bad SSDs.

Thanks in adv!

 

So I randomly decided to run a crystal disk mark bench test and noticed that my random 512K read speed seems really low compared to what I got before... any ideas why this is the case?

btw this is with write-back cache enabled and I have the dual raid setup.

 

Running with write-back cache enabled is good. Testing, however, should be done with caching off, else you also test how much unused RAM you have.
Yes, the speeds will be much lower in the test when you turn off caching. Because you're then testing the actual drive for every write, and not how fast you queue up data in (or for read cache, read from) RAM buffers.

 

Some server-class HP cached RAID controllers apparently do allow for TRIM pass-thru. That's not any help to us.

Also, I think I recall somebody doing some math based on NAND chip counts that indicated that the SSD's aren't over provisioned. Regardless, the maintenance mechanism being used seems to do the trick. I would still love to know what this mechanism is specifically.

 

Is it this lack of being "over provisioned" that leads to the advice of leaving ~15% unallocated space to make sure that the GC has a sufficient scratch pad to do its thing?

 

I think, if you double-check, you will see that your write-back cache is disabled, judging by your numbers.

 

Affirmative. It's also extremely easy to live with. If you need the space at some point, you can simply extend the partition back out. No reboots or anything.

 

I suppose even there's over-provisioning the margin would be slim. Having a buffer zone is always good.

I think that's why it is advised not to filled the disk completely or else performance will take a hit.

 

Thanks. That's what I had assumed from your earlier posts on this topic -- and from the chart(s) in the anandtech article showing the impact of GC overhead as a function of available space. I receive my Z today and intend to leave 15% unallocated, as you previously recommended.

 

Reading what I have just said, I now have a few question.

As it's widely known a 64GB disk is reported as 59.6GB available. The difference would be the spare area for GC to play with. Are I understanding this correctly?

Would the spare area increase or remain the same in a RAID 0 configuration?

BTW, I notice the Sony SSD in my Z has eight 8GB NAND chips per side, so the total capacity would be 8x8x4 (Quad) = 256GB which corresponds to the advertised capacity.

 

Don't think so. I think it's a conversion of bytes to GB.

 

It depends on what you mean by "fill the disk". GC can only do it's job if it *knows* that a certain area is free, so if a part of a raid array is simply not reported to the OS as available, the GC will know that is has this additional space to work with.

But if you mean "fill the disk" as in "don't fill the whole file system with files", then GC has no idea which blocks are free and which aren't, without TRIM that is.

 

I think the following information answered my query regarding spare area

http://forum.notebookreview.com/showthread.php?t=465962&page=28 (from post #277 onward - the discussion among Chirality, arth1, Sunfox and McMagnus)
http://www.anandtech.com/show/2829/7

 

Is there any chance to swap stock 128GB SSD with Intel X25-M ?

 

Zero chance, not possible.

 

You can take out the CD/DVD and put in an Intel X25 there, in addition to or instead of the Sony proprietary drives.
(You also need a hacked BIOS to prevent the X25 from running in RAID mode.)

At present, I wouldn't recommend getting an X25- M, though. It's getting rather long in the tooth, and the newest Sandforce 1200 and 1500 based drives (OCZ Vertex 2, Corsair Force) are faster even for small writes, and with a much larger spare area (which affects not only speed but also longevity). If you already have an X25-M, by all means, use it, but I would not recommend rushing to buy one.
The X25- E is still the champion -- but at $680 for 64 GB usable, it costs about twice as much per gigabyte as the competition. If you need the fastest, or an SSD that will likely still work after 5+ years of hard work, it's still the one to get.

 

thx arth1, for both messages, they were very helpfull. now I'm little bit confused... swapping is little bit risky and it seems that Sony engineers forcing us to do not do.

May be I can go for OCZ Vertex 2 or Corsair Force as you advised.

 

...I tried the search function so I'm forced to ask here; apologies for asking this again (I know this question's been answered, just can't find it again...)

I gonna be setting up my older Z with 2x SSD drives. That involves me getting a flex cable that connects up 2x SSD drives via a standard 2.5 SATA connection, and connecting them to the single original HDD port on the motherboard (my Z26 originally comes with a HDD)

I know that setting up in RAID means there is a greater risk of data loss due to the fact that if one drive fails, then you lose everything on both of them. I also know that RAID lets the OS see the 2x drives are just one. But what are the benefits?

Having the 2x SSDs as just 2x SSDs (without RAID) means I will have C: and D:
What are the benefits of this (I think it is refered to as JBOD here).

Just asking after a long-time member here had her TT damaged after a similar mod; should I risk a BIOS flash to a RAID-capable firmware and go RAID or not; it depends on the benefits...

Cheers in advance.

 

5ushimonster the RAID drives in the VGN are 1.8 drives just like the type i was using.
If you want to get an idea of what performance would be like these benchmarks should give you an idea.

Before with a single drive

http://forum.notebookreview.com/att...-series-owners-thread-crystalmarknewdrive.jpg

After
http://forum.notebookreview.com/att...l-sony-vaio-tt-series-owners-thread-intel.jpg

 

Hmm... That RAID setup; that's quite a good performance increase considering some people I remember said RAID slows some things down...

Ah. Well, your initial (before setup) was an SSD. My current setup is a 7200rpm Hitachi HDD, and I can't even get 100MB/s read for a SEQ... Which is why I'm looking at the SSD mod in the first place ^_^

Strangely enough, I can't find any 256GB 1.8' SSDs in Korea. Might be forced to visit the actual factory to aquire them, a thought I'm actually dreading...

So to answer my own question, RAID allows for faster READ and WRITE speeds... And nothing else in regards to benefits / losses of a RAID vs NORMAL (JBOD) setup?

Have you (or anyone else) tried that Intel RAID Manager 9.6 Zoinks was testing out a few weeks back to bring up the 4k read/write speeds in the older Z?

Otherwise, thanks Rachel for the help...

 

I was using the latest Intel Rapid storage manager for the last benchmark i posted. Doing write back cache did not improve the performance.

I expect that in JBOD mode your result would be like the first one i posted.These were G2 drives as well and some older models shipped with the G1 slower drives.

 

RAID slows things down if you don't have a dedicated hardware RAID controller. RAID 0, on the other hand, isn't really RAID, but plain striping. There's no parity calculations nor mirroring of data, so there's no measurable slowdowns.

The net result for RAID 0 (versus a single drive) is:

Small reads: Same speed
Medium reads: Slightly faster
Large reads: Faster (reading from N drives in parallel instead of 1)

Small writes: Same speed
Medium writes: Much faster (writing to N times the RAM cache)
Large writes: Faster (writing to N drives in parallel instead of 1)

The negative is that you rely on all drives operating at full efficiency. If one drive starts experiencing problems (like frequent bad block remapping), it will effect the whole system, as the other drives will have to wait for that one drive. And if one drive fails, you lose all the data on all the drives. So not only do you get more failures simply because you have more drives that can fail, but the impact is amplified too. This is why RAID 0 is considered the ugly step-child of data storage, or a ricer solution.

 

Remember that with a single drive (HDD or SSD) in a laptop - typical configuration - you also lose all data when single drive fails.
SSD would typically fail on write while HDD may fail anytime. Granted that probability of one of 2/3/4 SSDs in RAID0 to fail is higher than a single SSD drive though. But each individual drive in SSD RAID0 is "used" less (writes and reads are spread over drives in RAID0)) than a single SSD setup. So, in the end, reliability of SSD RAID0 is close to that of a single SSD.

 

I find this article regarding RAID reliability interesting.

PCGuide - Ref - RAID Reliability Issues

I note that the MTBF of an Intel X25-M is 1.2 millions hours. Based on the formula, does it mean that the MTBF of a 4-drive RAID 0 X25-M is close to 34 years ?

 

As is the MTBF for the Samsungs.

With the SSD's, reliability isn't much of a concern for me. I'll take the speed any day and do backups.

A fried display, cpu or, in my case, DVD drive are a more realistic failure candidates.

 

Hello, I wanted to make a clean install of windows 7.

I was a bit curious, I broke and recreated the RAID0, but I didn't know what Strip Size to choose...

I set strip size of 128k

is it good ? or should have I set lower ?

 

128k is the default. Doesn't answer your question but I assume there is a reason Sony set it to that at the factory.

 

I did a few tests with different "Strip" sizes, as the Intel Option ROM puts it.

Stick with 128 and a 4096 sector size when you format.

 

Thank you both!

 

Sorry for double post, but where do I have to set 4096 sector ?
I do not remember noticing it in the RST 9.5 menu I had access just after power on the laptop... but may I'm wrong

 

4096 is the default. You don't have to change anything.

 

After experimenting with two clean installs, and three VAIO Recovery Center restores (one Drive C: restore, a complete system restore with all the factory default sofware, and a complete system restore with Windows operating system, drivers, and basic software), SSD performance degradation became noticeable. After the last restore Windows Experience Index for the SSD decreased from 7.9 to 7.8. I have a VPCZ118GX/S (128 GB SSD x 3).

 

The WEI is a poor benchmark. My quad setup gives me a 7.6.

Run tests with the other tools mentioned in this thread.

I'm about to do another batch of tests since it's been a while.

 

Does anyone know why the "Start-up type" of RST 9.6 service (IAStorDataMgrSvc) is set to "Automatic (Delayed Start)" instead of just "Automatic"? That seems to have caused the RST icon at the tray to display an exclamation mark saying "Intel RST not running" until about 1-2 minutes after boot up.
Would that make any disk i/o during that initial period after boot up unoptimized? Should that service be changed back to start-up automatically?

 

I'm far from certain, but I would guess that this is set to delayed to be able to deal with external eSATA devices, which must be present before the RST driver starts in order to benefit from it.

 

I've changed the start-up type of the service to "Automatic" and it seems to boot up fine and the warning is gone.

 

My Z11 arrives next week (hopefully). I've finished reading this thread and it seems like the consensus is that some kind of GC is occurring that keeps the SSD running fast as long as it has some idle time to recover. Long term performance does not seem to have anyone overly concerned, and it looks like the Samsung SSD should run well and have a long life. Is that a fair summary?

If that is correct, what should an SSD noob like me do when I get my Z next week? Install RST 9.6? Is it necessary given the findings? Disable the defrag schedule? Anything else?

It's been said many times, but thank you to all of the contributors in this thread. It has been very enlightening.

 

Install RST 9.6 and activate Write-back cache in the RST manage menu.

Have a look here, there are some tips.

Windows 7 - Ultimate SSD Speed Tweaks

 

I don't do ANY of those recommendations.

They may tweak things, but not to a noticeable degree. Not worth the effort IMO.

In Windows 7, Indexing is called Windows Search and is integral to Outlook. I use it constantly.

 

Thanks guys. What about the alleged RAID drivers with TRIM support: http://forum.notebookreview.com/6051948-post78.html? Should I download this? My impression from reading the bulk of this thread is that TRIM still does not work for the Sony Z RAID. I'm therefore not sure about whether this download is necessary.

 

Run with RST 9.6. Nothing mind blowing as far as performance, but new is new.

 

Oops, I didn't realize that link WAS the RST 9.6 download (duh). Thanks a lot!

 

umm...wanna know the test result, someone please posts the result if possible? and another question, it would be better if i install a "RST"?

 

I'm rebuilding my machine for what I hope will be the last time, for a long time (yeah, right). Before I get the "keeper OS" installed, I thought I'd do some house cleaning.

Just for kicks, I've set the SATA controller to ACHI and installed Win7 to the 4th drive in the system. To do this and make sure the Windows installer doesn't touch the other three drives, I set them to read only using diskpart.

I just formatted the first three drives from inside Windows using Disk Manager. I started at the top and kicked off the formats one after the other, but all formats ran concurrently.

The drives, which are identical in every way, formatted at different speeds. I normally wouldn't bring this up, but they varied by as much as 5 minutes and the formats did not complete in the order I started them with.

Anybody have insight on this? I don't think this is a problem, just odd.

 

After W7 Clean installed, HD tune shows burst rate about 350-450Mb/s vs 750MB/s on factory system. As a result I have only 7.3 data transfer score on the WIE vs 7.6. Maybe anyone knows how to improve the burst rate.

 

hi may i know which oem brands ssds are being used in vaio z?

any details of the ssd and read/write of 200/200 possible?

 

Samsung 64GB drives.
Samsung MMCRE28GQDXP-MVB to be precise.

 

Not sure if this has been discussed or not...but using AHCI mode on these drives in "other OS's" clearly show NO TRIM support on these drives. I'd attach a screen shot but I don't want to get banned.

Has anyone confirmed this in Windows?

 

XXdeletedXX