Sony Z11 - Long Term SSD Performance - Post your results

Just got a call from E-Link who verified my Z11 did indeed come with 256GB. And I was all prepared to wage war when took it back too. I have to wait until the SSD's come in.

IMO the problem is that a CTO system when scanned @ E-Link does not show anything. The invoice like I stated is similar to Linear A and can't be read by anyone. My brother ordered a Macbook at the same time I got the Z11. I could easily tell you EXACTLY what's inside. What a joke, on me it seems.....

It seems ridiculous that my only real proof is my email which could easily be edited to show anything, not a tamper proof PDF.

 

I think the pdf can be hacked too, just open it with a text editor

Anyway, I'm glad you managed to convince them!

 

I thought PDF's could be locked.

I basically got a lazy tech, the "local" E-Link isn't exactly close either....

 

So did Adobe.

 

Hi Mikey,

I have had 5 VAIOs including this particular new 128GG/Z. Their service isn't terrific but I've never had any complaints about them.

What has happened to you is indeed terrible!

I really hope whoever screwed it up for you makes up for it in some way or apologizes sincerely...

 

it is performance degradation. for vpcz129gg
if you dont beleive it is 129gg can see the image on the link below

http://a.imageshack.us/img708/5894/ramstealed.jpg

 

Hi Andrew,

I totally believe you're using the top of the end, premium model 129.

But how is the screenshot of windows experience telling us that your SSD has degraded?

 

the score of Disk Data Transfer Rate decreased from 7.9 to 7.6

 

I believe Zoinks or Arth has given a lot more data on how you should be benchmarking your system before coming to that conclusion.

BTW, I certainly believe there is a performance degradation, I was the moron who started the thread about SSD TRIM. Nonetheless, you'd be helping everyone here if you were more extensive in your testing as the windows experience bench tells very little.

 

I am sorry because i was too shock when my computer faces a lot of 5 seconds freezes,
but maybe the degradation is too little as the Crystaldiskmark 3.0's software report. but why WEI scores' can change as it like?

note: i will be away until 7/8 August 2010

 

Crystalmarks is very popular, but it only reports average times, not workst case. And it's worst case that hurts you.
Or, to put it another way, if you have 9999 writes that take 2 ms and one that takes 2 s, your average as reported by Crystalmarks will be just above 2 ms. But that won't help, because what you'll notice are the stutters. Someone with a drive that does 3 ms without the stutters will feel much faster, despite what Crystalmarks says.

Again, what you should be measuring is worst case timings for the operations you do the most of.
For most people, the most common operation would (by far) be small random reads, followed by sequential reads or random writes, and sequential writes all the way at the bottom.

 

Heyas Arth,

Maybe you can help me with this.

Have you noticed that Andrew's machine stutters?
The benchmark he was running was showing awesome throughput with the exception of some areas where it dropped to nada/zilch.

I thought stuttering were issues to do with very early controllers?

The reason I'm asking this is because I'm experiencing the same thing with no need for benchmarks. It's almost blood boiling to be zipping through a large copy and then the whole computer suddenly...PAUSES.

Is it due to the Intel controller driver? I didn't update mine since [As you have pointed out before] Vaio UPDATE is F'ed.

 

You can pull the updates directly from the Sony support site and install them. That includes Vaio Update , but I think that one requires you to uninstall the old version first. Vaio Update works for me now, after getting the new version.

 

Hi Gerg,

I'd like to welcome you on both counts.

To Notebook Review AND joining us as a happy and proud VAIO owner.

 

Okay,

Maybe not happy.
My machine STUTTERS!!!

 

Thanks! Hopefully we can find the source of these hiccups.

 

Well, that is symptomatic of oversaturated drives without TRIM. Every so often, the drive can't find any pre-erased sectors to write to, and has to do a copy-erase-write operation, an operation which can take as much as a second. That gives you preformatted blocks to write to for a short while, but soon enough, the drive has to repeat the process.
If the writes you do aren't cacheable, but committed writes, i.e. the OS waits until it gets a response back from the drive saying the writes are actually done, the net effect is freezing stutters.

 

I know something definitely wrong with my vaio z129 and i called sony 2 days ago, they asked me to send my laptop there, and i am going to do so this weekend...

so my problem is not only because of the laptop freeze, the WEI hardisk transfer rate score lowered to 7.6 from 7.9 but also there is some freeze when running the HDTUNE...

edit: new photo uploaded 06/08/2010

 

To get a reliable test result with any benchmark tool you need to runn the test a while (or several tests) and use the largest work/test area possible.
The HD tune read test are only have of the medal. The most important tests are the write tests.

Here are some benchmarks. 2x Toshiba SSD's (THNS064GG2BNAA with FW AGSA0203) in RAID 0 vs 1x Intel SSD Gen1 (SSDSAMH080G1HP with FW 8820). Keep in mind the two disks in RAID0 should be close to twice as fast as a single drive.

CrystalDiskMark (left the Toshiba's, right the Intel):



IOmeter with disk size= 10000000 sectors (5GB), 1Worker, test time 4 minutes (first result of the Toshiba's and 2nd the Intel):
tests with 1 Outstanding IO:
64KB sequential read: 289MB/s vs 161MB/s
64KB sequential write: 146MB/s vs 50MB/s
4KB random read: 12.25MB/s (3136 IOps) vs 8.5MB/s (2177IOps)
4KB random write: 1.2MB/s (310 IOps) vs 15.5MB/s (3969IOps)

tests with 32 Outstanding IO's:
64KB sequential read: 288MB/s vs 153MB/s
64KB sequential write: 110MB/s vs 57MB/s
4KB random read: 29.8MB/s (7630 IOps) vs 17MB/s (4368IOps)
4KB random write: 1.12MB/s (287 IOps) vs 19.4MB/s (4962IOps)

I also run a test with enabled write back cache but that almost did not change anything on the results. I guess the write back cache is very small.

regards

Enrico

 

It's been two months and my array has gotten much slower. It was a bit over 300mb/sec. I tested it again yesterday and now it's 200mb/sec at best.

 

How full is the drive? Have you let it idle for a good chunk of time?

 

21GB free out of 119GB.

 

you should let the laptop just be on and idle for a hour or so. This will give the disk some time to do some garbage collection and all. Had slower performance a while ago, but let the laptop just run the whole night and during the day, and speeds are even higher than they we're before.

All speeds are still fairly the same. With 4k speed even higher than it was @ 22,5 read and 20,1 write.

 

I think you're getting pretty close to the threshold though for where garbage collection will begin to have trouble. If I remember right, you want to have some free space to allow the GC to do its work. I think you're still okay with 21GB free, but something to keep in mind.

 

I think 21GB is too low, try to get it above 25GB. Clean the crap and put less important stuff on an external drive. And let it idle over night like the other said and Garbage collection should go to work.

Also, if you don't know something, ask before making statements, as it might give the wrong impression of the product.

 

This is my new primary drive in my VPCZ12:


IOmeter with disk size= 10000000 sectors (5GB), 1Worker, test time 4 minutes :
tests with 1 Outstanding IO:
64KB sequential read: 163MB/s
64KB sequential write: 117MB/s
4KB random read: 17MB/s (4355IOps)
4KB random write: 6.5MB/s (1680IOps)

tests with 32 Outstanding IO's:
64KB sequential read:237MB/s
64KB sequential write: 131MB/s
4KB random read: 139MB/s (35750IOps)
4KB random write: 31MB/s (7982IOps)

 

at last i fixed the spike on my computer it is windows fault, believe me?

It is "SUPERFETCH", it make the whole computer unusable...

you must disable "SUPERFETCH"!!!!!!!!!!!!!!!

even though i can't make the windows score back to 7.9 at least the ssd is alive...

 

The SSD Review: Windows 7 SSD Performance Optimization Guide

The Windows score (experience index) is moody, especially at the disk drive score. Run it a few times until you have the score you like

 

hmm thanks Enny02 but i prefer something to be on...

 

Just uninstall everything except notepad, wordpad and calc
I agree, not all suggestions in the article make sense.

 

but my SSD performance can't get back to 7.9 sadly

how WEI works:
WEI not only benchmarking HD TUNE 4.50 with Full San 64 Kb, but it try every single Kb starting from the lowest to the Largest...

 

For the WEI is only a "guess-o-meter" to check what the performance bottleneck in your system could be and is not really meant to compare your performance with other systems.

I trust more in the tools from Futuremark and HDD tools like Crystal Mark and IOmeter (with the proper test settings).
HD Tune and other physical disk benchmark tools are useless for your system drive because you cannot do write tests without removing the partition/system.
Screw WEI . Run your benchmarks with other tools and compare.

 

true enough but i am too scared to replace ssd inside my notebook
is that true that i can put 2 micron 256 1.8" RealSSD C300 inside part code MTFDDAA256MAG-1G1 or 2 Intel X18-M inside my vpcz1?

 

Well you don't have to replace your SSD's just to get your 7.9 score back at WEI, even on my notebook with RealSSD I still have only 7.6. That is why I don't really care about WEI.

I'm not sure if all VPCZ12 are the same regarding SSD connectors (whether they are mSATA or LIF/ZIF). If you have Toshiba SSDs then you can put other 1.8" mSATA SSD's in your notebook and 2 micron 256GB 1.8" RealSSD C300 should fit in there. But to be sure about the connectors you have to open your notebook.

 

thanks a lot Enny02

 

Decided to benchmark my recently bought vaio z.

i7
256gb (2x128) raid 0
8gb ram
write cache enabled

Do these results look out of the ordinary?

 

Are that Toshiba or Samsung SSD's ?

 

Been a while since I've been around these parts (forums). I see that Sony released a Intel 9.6 driver kit and a SATA patch. I've installed both and haven't seen any differences.

Am I alone in this observation?

Also, my SSD's are still performing well. No marked decreases in write performance.

 

I may as well post this now. I've been running CrystalDisk benchmarks periodically, and saving the results. Here are my last seven runs since April (close to once a month). I'm a moderate to heavy user, though my drives are only about 33% full. I do software development almost all day, along with normal gaming and browsing tasks. As you can see, no big change in numbers overall except for sequential write between points 2 and 3 (I am not sure, but I think that is when I turned on write cache. Can't think of anything else that would cause the jump). I'm running 4x128 in raid0:

Code:

 seqrd   seqwt  rr512k  rw512k  rr4kq1  rw4kq1  rr4kq32 rw4kq32
457.228	209.946	162.543	282.051	14.235	51.929	92.341	60.106
485.152	170.945	208.568	275.394	13.835	53.12	91.938	60.494
460.44	359.964	184.817	277.355	15.152	52.394	91.516	61.041
483.958	376.96	205.706	299.05	19.82	49.376	95.156	61.135
455.771	333.517	183.9	269.708	13.969	53.598	87.59	58.828
405.848	341.778	187.587	262.619	12.974	47.465	92.419	60.407
446.393	361.204	202.984	264.086	13.326	48.697	93.926	61.603

 

Awesome statistics, thanks for compiling!

You are probably right about the write cache being enabled.

Other than the dip in reads at point 6, that looks great. Nice and flat numbers showing a pretty consistent trend. Blindly, I'd say this dip is more of CrystalMark problem than the drives doing something funky. Run it several more times over a day and you'd likely see it even out.

I wouldn't be too concerned about the random 512 write numbers trending down for the same reason. Looks like you had a really good benchmark pass at point 4. Take it out and it is nice and flat.

 

Yeah, I agree - CrystalMark is inconsistent between subsequent runs. The numbers I posted were the last 7. I actually had 11 runs, but the first 4 were all done on the first day I got the laptop (I did 4 runs in 30 minutes), and they showed comparable variance. For instance, seq reads varied between 480 and 457, and seq writes varied between 293 and 226.

I sort of wished I was consistent in when I did the runs, but it was mostly whenever I thought of it, so it could have been after a reboot, or after it had been idle for a while, or after doing a bunch of work. I'm happy with the results so far.

-Peter

 

Not completely scientific, but still reasurring and interesting. Thanks for taking the time to compile that and share it.

 

Since you are running it on the same disk from where you booted from.
Run it in safe mode that'll give you maybe more consistent results.

 

Those are the toshibas in my z12. Anything fishy?

 

Even if the drives results are not very regular, it is what science is made of, and why science use statistics...

Maybe multiple iterations would be nice in crystalmark.

 

I wrote up an iterative testing model in the opening post, but CrystalMark will still give you some level of variance. CrystalMark does run multiple passes of each test type, but having an option to run the "test suite" a few times would be nice.

 

Thanks for the testing!

 

As mentioned above the other apps/OS affect benchmark results. CDM is an easy to use tool but not really reliable. Also it has by default a small work area. It should be selected 2000MB and 4000MB, results with smaller work areas are useless. Long Term Performance, you'll most likely see differences only at write speeds.
Also keep in mind you should not constantly run write benchmarks because that could harm your SSD over time, a few test per month are certainly okay.

 

I am in 7th Heaven,.... Almost. I just upgraded my Vaio SZ340p to a 1290X and I am just blown away with how fast it is. With a 512Gig SSD and 8GB RAM it really cooks. After a week of using it for work, my HD Tune Read specs are below:

Min 462 MB/s
Max 576 MB/s
Av 544 MB/s
Access Time 0.165 ms
CPU Utilization 19.1%

The only problem I have with it is the screen size. The 340P was 13.3" and was a perfect size. The 1290 is 13.1" which seemed almost the same, but I didnt realize that the height to width ratio would be so VERY different. The 1290X only has a 6.5" screen height, which makes report writing rather difficult, on the verge of being annoying. No matter what work you do, sooner or later you have to write up the results.

The 14" Vaios might have been better, but I needed the speed for my natural resource modeling work, and the 14's don't have the same CPU horsepower as the 1290. I have to admit, I love the Vaio, but why did they have to make this into a short stubby wide screen movie box. There are still a few people that do work on their computers.

Sorry for getting off topic, I just had to vent my joy and sorrow.

 

Is the 1290 using SATA 3 for the RAID? It seem that SATA 2 would max out at 375MB/s.