Intel SSD 520 Review: Taking Back The High-End With SandForce

[bildo123]

And with an SSD, your computer comes out of standby faster than your monitors do. Not kidding.

I know. I USED to be able to blast my password in and press enter just in time for my desktop to pop up as my monitor would catch the signal. That however was with my 74GB raptor with nearly half the access time of my current Samsung F3. I think the delay I have now is more mobo related though since I did a new rebuild. Sometimes my G15 and Mouse don't "come alive" for a bit.

 

[ctv]

What did Toms use to measure boot times with? Anyone please?

 

[nikorr]

As far as Anand can tell... no.

http://www.anandtech.com/show/5508/intel-ssd-520-review-cherryville-brings-reliability-to-sandforce

Intel's SF2281 implementation seems to be stable, and BSOD free.

That is pretty much the reason that I still use the Vertex2 : )) No complains there!

 

[Guest]

I am also very curious how does the encryption option work in Intel SSD 520? Does it use ATA password? When I connect it to another computer or via USB the data cannot be read?

 

[zerapio]

[citation][nom]borynek[/nom]I am also very curious how does the encryption option work in Intel SSD 520? Does it use ATA password? When I connect it to another computer or via USB the data cannot be read?[/citation]
Encryption is enabled using the user password in the BIOS. Because the encryption/decryption is done in HW performance is identical whether you enable it or not. Although I haven't tried it yet I'd imagine if you take the drive to another computer it can't be read until you enter the correct password.

 

[schouwla]

I am not very excite about the new SSD from Intel. How does it stack up against A-DATA AS510S3-120GM-C that has reads at 550 MB/S and writes at 510 MB/S. A 120G disk is 11500 yen in Japan.

 

[michelrochette]

Three drives in RAID 0 give you 1.5 GB/s with an Highpoint 2720SGL (160$)

 

[noahgoodnik]

I read articles about SSDs and whether I should "upgrade" to one. But, how do I get my existing hard drive installation on the SSD. Installing everything from ground zero would be very painful.

 

[meradz]

Very well thought-out and written article, thank you very much!

Something else I would take into consideration when deciding on a brand is how well and quickly the manufacturer responds to issues with firmware updates. SSD performance and reliability are still a rapidly evolving technology. We would love to hear about your experiences when you bring up issues to the attention of the manufacturer.

 

[josejones]

Where is the very best place to purchase that Intel 520, 120G "Cherryville" from in order to get the best price? Is it Newegg or that ssdtracker website?

I have to play it safe and go with an Intel SSD for our small business. I'd rather get a cheaper one but, I can't afford to get cheap on reliability and lifespan.

I've never had an SSD before so I'm curious about basic maintenance. What all do I need to know before getting my first SSD?

 

[Combat Wombat]

Awesome Review!

 

[rabinnh]

Your chosen hard drive for comparison is a 5400 RPM budget drive? A better choice would be the Western Digital WD6000BLHX. It's a 600GB 2.5" 10,000 RPM SATA 6.0GB notebook drive for around $350. Still cheaper than most SSDs, 2.5x the capacity, and a huge performance difference from the hard drive that you chose for comparison. That would give the reader a choice between a reasonable performance/storage tradeoff at a similar price point.

If you want to move to the desktop, for the last 3 years I have used an LSI MegaRAID SCSI controller with 2 Seagate 146GB 15,000 RPM drives in a RAID 0 (striped) configuration. At the time that I built my system, there was no other way to get thte performance that I needed at my price, which at the time was over $900.

Well, times have changed. My research indicates that an OCX Vertex 3 240 or Intel 520 240 will (roughly) improve my performance.

Cached reads: +278%
Buffered reads: +96%
Sequential writes: +151%
Non-cached sequential reads: +96%
Cached sequential reads: +348%

Improvement enough for me to drop $340 on an SSD.

 

[beavermml]

i have intel 320 80GB.. how much is the different between 330 120GB? i want to use it as my boot drive.. is it wise to upgrade?

 

[w_barath]

The read speed drop during the torture test is not surprising at all - Sandforce is doing foreground garbage collection, which consumes its read bandwidth, leaving reduced bandwidth for the host reads under a mixed read/write load. Maintaining 400M/s with random 4k writes on a full drive - that is real progress. If a log-structured filesystem were in use then the read performance would have stayed at 50% instead of 25%.

 

[rjranay]

These drives are prone to get bricked. I know I had one, it got replaced for free by Intel, though, but still you have that lingering insecurity at the back of your mind.

 

[leonfeldman89]

Interesting article seeing as it's completely irrelevant to anybody.
The chance that someone will be writting 100% compressible data with at Q1 in the real world is exactly 0%.
In fact, you will never see ayone anywhere writing real data with a compressible to incompressible ration which yield write amplification below 1.

Furhter, just accessing(reading) any single bit of data means you have to at the very least write 1 full block if not a full page and more than once for sub 30nm die sizes.

Finnaly, rating the enurance of an ssd based on 7GB of writes per day is like rating the gas mileage of a car based on pulling it out of the garage.

Every time you boot windows you're putting more than 28 billion nand cells(7GB) through 1P/E cycle worth of wear.
45nm mlc endurance is 3k, below 32 nm you'd be lucky to see 1k in the real world.

The real failure rate for sandforce driven mlc ssd's is between 3-5% over the warranty period.
OCZ is on the high end because they deviate more from the reference firmwares than most other LSI partners.

Either way a manufacturer will typically decide on how much they want to allocate to servicing a competitive warranty, then they produce drives with failure rates that don't exceed the capital outlay.

Once the drives are ready for production they design compatibility and bug testing parameters to fit a competative stated mtbf and performance in addition to a generic set of tests for internal purposes(predicting warranty service costs).

The set of tests you see published and released to reviewers(mouthpieces) are for public consumption only and shouldn't be taken seriously over the predictable decay of endurance corralated to smaller die sizes.

 

[leonfeldman89]

[citation][nom]ctv[/nom]What did Toms use to measure boot times with?[/citation]
I'd say you'd be wrong. Sanforce driven is boot or cache drive only and OCZ is the worst offender.
Unless you're okay with a 3-5% chance for data loss, then go ahead and store your files there.

 

[Rumboogy]

How do you know it is over-provisioned by 7%? I can't find this in the Intel spec-sheet. Are you just going by the capacity of chips vs the capacity of drive? What about storage that is used for ECC etc?