Ocz's refurbs not close to spec in speed.

[formulav8]

I've bought approx. 7 or more Ocz refurbs the past month or 2. I've bought 60gb up to 240gb. Agility 2's to Vertex2's, and Agility 3's. I had to send one back for replacement.

Anyways, not a single 1 of these drives gets even close to the theoretical speeds OCZ is claiming.

I've tried many different computers, Oses, and chipsets, and still the same.

All of them average around 140-150-160gb mb/s reads and 60-70mb/s writes.

I could see getting 1 drive or so not up to spec but not All of them. I saw a guy earlier on here who had the exact performance problems with his.

I'm wondering if they're purposely putting lower spec chips on these drives and still claiming the same speed as the non-refurbs knowingly?

I've tried every tweak, alignment, latest firmware, and everything but still at the performance area I mentioned above.

Is there something I might have overlooked though? Or someone know what is going on with these drives?


Jason

 

[cheez]

Have you tried other brand SSD in the same system before? What score did you get?

Did you set to AHCI mode for the storage controller in BIOS prior to installing Windows? If not, that could be the problem.

Also the scores will vary depending on the size of the SSD according to OCZ site.


cheez

 

[dave_the_nerd]

What benchmark utility are you using? Sandforce controllers use compression, and some benchmark utilities use incompressible data, so you end up with the worst-case scenario instead of the best-case.

Also, file copies on Explorer are way slower IRL than the benchmark scores.

 

[VirtualLarry]

What benchmark utility are you using? Sandforce controllers use compression, and some benchmark utilities use incompressible data, so you end up with the worst-case scenario instead of the best-case.

This. SandForce benchmarks with the zero-filled ATTO benchmark, because it is the best-case for SF's compression feature. Other benchmarks will not show nearly as high a score.

 

[formulav8]

Hi, I tried a Kingston V100 (not + series). and it is close to spec with 240-250 mb/s reads. I also tried a Corsair Neutron 240gb and it basically pegged the Sata2 controller at 270mb/s read/writes.

I even tried testing on my main computer that has the Kingston v100 128gb drive I mentioned above and its still low reads/writes. On an Intel P35 board.

They are still faster than hd's (seek times are fine on them). But definitely want them to be up to spec.

I tested as add-on drives and tried them as OS drives and still the low performance. One of them I remember got almost 200mb/s read but writes are still bad. If they would perform just 200mb/s or so I would be fine.


Edit: I also tried reading using HD Tach and AS SSD to test the reads and they both show similar reads at only 130mb/s or so on many of them.


Jason

 

[formulav8]

I will test with atto again tonite and see if it still shows the similar reads/writes that I tested before.

Thanks a lot for your help

 

[Hellhammer]

Have you secure erased them? It's possible that OCZ's validation has degraded the performance to a state where even TRIM doesn't help (incompressible data and SandForce don't like each other). If the speeds are still bad after a proper secure erase, then I would contact OCZ.

 

[razel]

On some models the smaller capacity drives won't have as fast sequential writes as the larger to largest capacities which is often what's only stated when rated. To top it off some of the Vertex 2 E models do have slower performance that the standard Vertex 2's.

When I saw the refurbed OCZ Vertex 2 E 240GB-ish model on sale at newegg for I think $95, I nearly bought it but wanted to double check it's benchmarked specs on ocz's site which they used to list for every capacity in a nice PDF. Unfortunately OCZ updated their site and I can't find it anymore.

Regardless, don't get too caught up if your sequential reads are 85% as rated. That 15% difference over your ownership of the drive will probably amount to about a few minutes difference.... that's all the time wasted constantly benching it.

 

[formulav8]

Hi, I did fully wipe them using a tool. But I will try Ocz own eraser this weekend and see what happens.

If the reads were 200-220 or so that would be fine. But 110-150 reads just seems really low for a product that's optimally supposed to nearly peak out a sata2 controller (up to 280 or something). Especially when your using a bench that they apparently use for rating their drives.

But anyways, I do strongly appreciate all your info! If anything changes I will post more info.

 

[Hellhammer]

Hi, I did fully wipe them using a tool. But I will try Ocz own eraser this weekend and see what happens.

What tool did you use? Most tools have been designed for hard drives so what they do is write 0's to all LBAs which just degrades performance. OCZ's tool should work (though my experience with Windows tools has been mixed) but I would use e.g. Parted Magic instead.

 

[Emulex]

it's well known that even samsung 830/840/840 pro's will fail to a state where so much error correction is going on that they perform forever even after secure erase at 30-50% original speed.

I'd bet a $20 that you got a return and reformat special. oops accidently reset the firmware counters to zero.

 

[Zap]

Didn't OCZ silently switch NAND during the 2 series that caused a large drop in performance and even some capacity loss?

 

[formulav8]

That's what I thought I heard as well about them changing the nand chips. That's one reason I posted this to see if anyone had some info on Ocz doing that.

I plan on using Ocz own tool to erase the next day or 2 and that just may help. At least it won't hurt I don't think.

 

[Ao1]

Where are you getting the refurbs from? Direct from OCZ? Have you opened up one of the SSD’s to see what is inside?

 

[Zap]

That's what I thought I heard as well about them changing the nand chips. That's one reason I posted this to see if anyone had some info on Ocz doing that.

Found this article with a simple web search.

Several readers contacted me in the past two weeks, complaining about OCZ's recent adoption of 25 nm NAND and its effect on the capacity and performance of certain SSDs that they expected to be both larger and faster.
...
There are two fundamental issues to explore here: lower capacity and reduced performance.
...

There's more to the story because it apparently is not always slower, and can be faster in certain benchmarks. However, the crux of it is that in some benchmarks the performance is extremely sub-par.

 

[Ao1]

lol they are all over the net. Goes to show how many RMA’s OCZ get. They should perform the same as a new drive however as the specs have not been changed. If they don’t you could open one up to see what NAND is being used, although you will invalidate the warranty by doing so.

OCZ Refurbished SSD

Q: What is a Refurbished unit?

A: A Refurbished unit is one that is "Newly Overhauled". At some point they have been returned to OCZ from either an end user or a dealer, and may or may not have been used.

OCZ take all these units that come back, rigorously test them, and repairs them if necessary. Then before they are repackaged, OCZ ensure they have the latest firmware updates installed.

OCZ then offer these units out to their dealers at a discounted price. Ebuyer then pass this saving on to our valued customers. Ebuyer provides a 1 Year warranty on these Units so you can buy with confidence

IN SUMMARY - Since these units have been previously released to the market they cannot “officially” be sold as new, but are as good as new
http://www.ebuyer.com/455032-ocz-240gb-agility-3-ssd-agt3-25sat3-240g-rf

 

[Zap]

they cannot “officially” be sold as new, but are as good as new

The "problem" is that performance of the 25nm versions are not the same as the 34nm versions, even when new.

 

[Ao1]

Forulav8 says he has the speed issue on all drives including an Agility 3, which uses a SF-2281 controller and 25nm Micron NAND. As NAND is the most expensive part of a SSD it would not make sense to swap NAND over and in this case it’s 25nm anyway.

It’s much more likely that a SSD is returned due to a controller failure, so I’m wondering if it is the controller that they have swapped over as presumably an Indilinx controller would be cheaper than a SF controller.

The only way to tell is open one up, but even then it’s touch an go as they may have rebranded the controller or removed any marking.

Reading buyer reviews it seems they are having a lot of failures on these reburb drives so something is dodgy.

 

[formulav8]

I finally got around to retesting a couple drives. The 2 120GB 3.5" versions are showing much better reads now. Maybe erasing helped? About 225-ish. The uncompressed writes stink at 50-60. But compressed are 170-ish which is much more acceptable. Still not close to the 285mb/s read-writes or whatever they advertise but better than I originally tested.

Anyways, I plan on testing the Agility 3 especially to see what going on with it being as slow as it showed the first time.

Thanks for all of the great info and such as well!


Jason

 

[Hellhammer]

I finally got around to retesting a couple drives. The 2 120GB 3.5" versions are showing much better reads now. Maybe erasing helped? About 225-ish. The uncompressed writes stink at 50-60. But compressed are 170-ish which is much more acceptable. Still not close to the 285mb/s read-writes or whatever they advertise but better than I originally tested.

Anyways, I plan on testing the Agility 3 especially to see what going on with it being as slow as it showed the first time.

Thanks for all of the great info and such as well!


Jason

Those scores are normal for Vertex 2 with 25nm NAND. If you want to "max out" the drive, you can try ATTO as it uses a higher QD by default and the data is highly compressible (i.e. gives great numbers for SandForce drives).