Cherryville - SSD 520

[Ao1]

Thanks for posting. Did you realise you left the serial number on the last post?

[johnw]

Thanks for posting. Did you realise you left the serial number on the last post?

Why do people worry about that? It is not like he obtained the product illegally (I hope).

[Anvil]

...can find no reference to controller

Gone for supper.

Could you run a few more test using i.e. 46% or 25% compression.

Not that we really need it but still it would be great to know how they perform.

[DooRules]

with the 11 series driver



with 25%compression

[DooRules]

46% comp


100% comp


safe to say drives are hammered pretty good, have done no SE's yet

[johnw]

If you get a chance to take them out of RAID for a minute, I'd like to see a single-SSD benchmark with ASU, at 100% (incompressible).

[some_one]

Thanks DooRules. With my own RAID0 Sandforce based Agility drives both reads and writes take a huge hit with compression. Unlike the writes your reads seem little affected with compression so whether that means non-SF controller or caching of the drive IDK. To check the controller would probably mean opening the drive case which IIRC voids warranty so probably not too many people going to be doing that.

SF Agility 60GB RAID0 results 0-fill and 46% for comparison.

[Computurd]

secret firmware sauce. Guys cmon.... all should know what this processor is.

[Computurd]

@Doorules- Kudos, you managed to get these drives before the NDA from Intel has expired... unbelievable
Im glad that it ended up in your hands though, instead of some idiot who had no idea what they are doing....nice to see it land in one of the 'guys' hands

[felix_w]

Well, the AS-SSD could match 2xM4's, so maybe it's still Marvell with xtra treated fw, or revised controller ?

[johnw]

With my own RAID0 Sandforce based Agility drives both reads and writes take a huge hit with compression. Unlike the writes your reads seem little affected with compression so whether that means non-SF controller or caching of the drive IDK.

The Sandforce SSDs with synchronous flash exhibit little or no decrease in read speed for incompressible data when tested with ASU or AS-SSD (except in certain heavily-used states). But the Agility uses asynchronous flash, which has much worse performance than synchronous flash. I never recommend an SSD with async flash, because the sync flash is only a little more expensive but peforms a lot better.

I'm sure Intel is using sync flash in the 520.

[johnw]

Well, the AS-SSD could match 2xM4's, so maybe it's still Marvell with xtra treated fw, or revised controller ?

Very unlikely it is Marvell, since the SSD is clearly compressing the 0-stream data, and Marvell does not have built-in support for compression (yes, the data could be compressed with another chip before being sent to the Marvell, but that would be a mess that I do not believe Intel would make)

Now that we know the 520 is using compression, I think it is 99% certain that the controller is a Sandforce.

[some_one]

The Sandforce SSDs with synchronous flash exhibit little or no decrease in read speed for incompressible data when tested with ASU or AS-SSD (write speed is another story). But the Agility uses asynchronous flash, which has much worse performance than synchronous flash.

Thanks johnw, I realize synchronous offers better performance, just wasn't aware it was that great. Thanks for the heads up, something I will certainly consider next time purchasing a SSD.

[DooRules]

no comp single drive

[Anvil]

Superb!
Thanks again DooRules

@johnw

There's no point in arguing about it being SandForce or not, it's more a question of what have they been able to tweak.

To me it looks like a standard Sync SF2281 with negligible adjustments, it's really close to i.e. the Corsair Force GT 120GB.

I'd say they are late to the party though, if the prices are favorable and they have managed to make a "better" firmware then they have a winner.
I've been using the Vertex 3 240GB for some time now on the LSI-9265 and they are just as stable as the C300's I've been using on the Areca 1880, meaning no issues at all.

[DooRules]

There are still driver issues so I would only expect it to get better. With the enterprise drivers Intel Toolbox is pretty much completely dysfunctional. With the 11 series drivers I have complete use of the toolbox, SSD optimizing works, forced trim, SE works a charm, shows all drive data for drives in R0 and single drives as well. Toolbox also shows all smart data for M4's and C300's.

[DooRules]

OS installed, drives were SE'ed prior to install thru Intel Toolbox, this is with E drivers, in benching they are a fair bit slower than the 11 series.

[johnw]

@johnw

There's no point in arguing about it being SandForce or not, it's more a question of what have they been able to tweak.

Huh? Who's arguing? I said I was 99% certain it is Sandforce.

One thing that could make the 520 potentially better than most other Sandforce SSDs is if Intel included power-loss-protection capacitor(s), and did not screw up the implementation like they did with the 320. Any idea if that is the case?

[Anvil]

It looks like there are some differences to the SMART attributes vs the standard SandForce one, will compare to one running the latest official firmware.

The links to the 520 are nowhere to be found here in Norway, looks like they have been told to clean up all traces to the 520 both on the webshops and on "prisjakt".

[Anvil]

01 Raw read Error rate is gone from the Intel

AA Available Reserved Space is Added on the Intel

B1 Wear Range Delta is removed from the Intel

B8 End-to-End Error detection is Added on the Intel

B5 Program fail count
B6 Erase fail count are both removed from the Intel

C0 Unsafe shutdown is Added on the Intel

C2 Temperature
C3 Hardware ECC Recovered
C4 Reallocated Event count
C9 Soft Read Error rate
CC Soft ECC correction are all gone from the Intel


E1 Host writes
E2 Timed Workload Media Wear
E3 Timed Workload Host Read/Write Ratio
E4 Timed Workload Timer Added on the Intel


E6 GMR Head Amplitude
E7 SSD Life Left are both removed

E8 Available Reserved space
F9 'Unknown' are both Added

From what I can see, the Raw "NAND" Writes are missing from the Intel implementation

[Ao1]

Hey DooRules, how easy do those case screws come out? It would be great to see the SF controller and if it has a power cap.

[Ao1]

From what I can see, the Raw "NAND" Writes are missing from the Intel implementation

E1 to E4 will makes things interesting though for calculating wear using different levels of compression

[Anvil]

It certainly adds something so it might look like they are aiming at using the 520 more as a semi-professional drive.
(targeted at SMBs)

[DooRules]

Hey DooRules, how easy do those case screws come out? It would be great to see the SF controller and if it has a power cap.

Very easy in fact. I can say with 100% conviction that it is a SF 2281 VB1-SDC

[Ao1]

Thanks. Is there a power cap