SandForce SF-2500 Preview

[Ao1]

couldnt agree more. this looks like another round of SF trickery.....not just ocz but all mfg using the SF ripping people off.
why then is he intentionally not listing the black-eye info of this SSD? why does he seem to avoid it? he doesnt want to get blacklisted by ocz i imagine. im sure hes aware most enthusiasts wouldnt even think to look. they see the big seq and are like WOHOOO!

Also a bit strange that Anand could benchmark the drive yet in a real life application they state: "the drive is basically unusable. I get super long read and write latency. I've already informed OCZ of the problem and I'd expect a solution before we get to final firmware."

Outside of generating market hype I don't get the point of testing unrepresentative hardware that is so underdeveloped the case has not even been designed.

[geo]

Also a bit strange that Anand could benchmark the drive yet in a real life application they state: "the drive is basically unusable. I get super long read and write latency. I've already informed OCZ of the problem and I'd expect a solution before we get to final firmware."

Outside of generating market hype I don't get the point of testing unrepresentative hardware that is so underdeveloped the case has not even been designed.

its a preview?

[Ao1]

A preview of what though? A product that is basically unusable? An early engineering product that will most likely not resemble the finished product? Reviewers keep getting caught out by rushing out reviews just to be first on the block. I wish reviewers would take more time to really understand the drives and find the chinks. I think there should also be a much higher emphasis on reliability, compatibility issues and more realistic assessments of tangible real life benefits. (IMO the AnandTech 2011 light work load bench is not going to give you any indication of tangible real life performance benefits).

Against that we are now seeing rushed reviews of half engineered products using NAND that will not be deployed in the finished product. What is the point?

For me the whole thing is marketing.

Edit: And I could add at a convenient time when OCZ are desperate to deflect attention away from the Vertex 25nm fiasco.

My 2c anyway

[RaV666]

That's not true at all... if you go back to the first gen drives, he did not hesitate to rake them and others over the coals for their rush to market with crappy drives based on early Marvell controllers... and he demonstrated a "fondness" for Intel throughout this period. His recent fondness for Sandforce drives and OCZ is likely due to the fact that they seem to be the only ones really innovating or bringing interesting products to market. Where is Intel?! I think Anand calls em like he sees em. I trust his reviews and recommendations more than most.

Well, yes, and i`ve been anands avid fan from the beginning, but nowadays i see some disturbing things going on in there.
People and sites change, so the fact that he once was great, doesnt mean it still is.
His relationship with ocz`s boss is almost personal now ,if you read all hos ssd articles you know this for sure.
And why he hasnt STILL reacted to OCZ failure regarding 25nm flash ?
Well ,im almost sure hes going to write something in a week or two when ocz already did damage control.
Hes sandy preview was weird too, his change in server hardware at corellating time to intel with corresponding server articles was fuzzy also.
He was asked repeatedly in comments exactly HOW he aquired his new server chips, and while he and his crew were responding to other questions, they plainly ignored this one (asked multiple times by multiple users in multiple ways ;-).
And the using of 480 FTW edition in one of the reviews.
Maybe im paranoid, but its all too fishy hes taken care of in this manner by ocz and intel, and the other sites mostly not.
And going back to the topic, hes previewing new ocz drive just in time when marvell introduces new controller and intel probably also is around the corner.Its marketing.As the product he reviewed isnt really the one were gonna get.

[BigDisk]

Also a bit strange that Anand could benchmark the drive yet in a real life application they state: "the drive is basically unusable. I get super long read and write latency. I've already informed OCZ of the problem and I'd expect a solution before we get to final firmware."

Wow dude, way to get this out of context. At least put the whole paragraph:

I should also note that this is beta hardware running beta firmware. While the beta nature of the drive isn't really visible in any of our tests, I did attempt to use the Vertex 3 Pro as the primary drive in my 15-inch MacBook Pro on my trip to MWC. I did so with hopes of exposing any errors and bugs quicker than normal, and indeed I did. Under OS X on the MBP with a full image of tons of data/apps, the drive is basically unusable. I get super long read and write latency. I've already informed OCZ of the problem and I'd expect a solution before we get to final firmware. Often times actually using these drives is the only way to unmask issues like this.

I guess performance was crappy is OSX due to some early production sample issues. But I can imagine it worked fine on Win7. Do you actually think he pulled all of his number out of thin air?

[Ao1]

How does not quoting the full paragraph take it out of context or change the fact that the drive was not usable on an OS? He did not say crappy performance he stated the drive was unusable. Presumably this was due to "the super long read and write latency".

The benchmarks were most likely done on a drive without an OS. I have no idea why it was possible to benchmark the drives but not possible to run an OS. You would have thought that any performance issues would have been picked up in the benchmarks.

All it proves is that testing an early engineering sample is pointless.

[Computurd]

The benchmarks were most likely done on a drive without an OS.

yes this is unfortunate because it isnt indicative of real performance anyway.

Seems OCZ is doing a PR blitz to wash away the negativity of the 25nnm vertex2 fiasco.

[eva2000]

yes this is unfortunate because it isnt indicative of real performance anyway.

Seems OCZ is doing a PR blitz to wash away the negativity of the 25nnm vertex2 fiasco.

real performance = whatever end users intend to use the ssd for.. not everyone will use ssd for OS disk, so still good to see how it performs in non-OS environment

[Computurd]

real performance = whatever end users intend to use the ssd for

well, you do have a valid point

[Hornet331]

How does not quoting the full paragraph take it out of context or change the fact that the drive was not usable on an OS? He did not say crappy performance he stated the drive was unusable. Presumably this was due to "the super long read and write latency".

The benchmarks were most likely done on a drive without an OS. I have no idea why it was possible to benchmark the drives but not possible to run an OS. You would have thought that any performance issues would have been picked up in the benchmarks.

All it proves is that testing an early engineering sample is pointless.

Because as you put it, the impression is given that the whole drive is unusable under real world scenarios in every instance.

While in reality it only effects a certain os und certain circumstances.

[Ao1]

While in reality it only effects a certain os und certain circumstances.

Says who? Anand did not bother to go in to any details and they did not bother to report if the drive worked with Windows, so who knows? What we do know is that it was unusable under 0S-X.

Perhaps you can clear something up for me. Was this drive not originally slated as the SF-2000 as announced in an article by Anand on 10/7/2010?
This review is of an enterprise SF-2500. So......is the SF-2000 going to be a Vertex 3 and the SF-2500 going to be a Vertex 3 Pro?

Where is the review of the SF-2000 that was announced by Anand on 10/7/2010?

[cx-ray]

Here's another review:
http://pcper.com/article.php?aid=1078&type=expert&pid=1

PROS:
*Really* fast throughput - saturating SATA 6Gb/sec on the first iteration of the tech.
Excellent IOPS scaling at higher queue depths.


CONS:
Odd stutter observed on our batch file write test (possibly due to beta firmware).
Performance at lower queue depths could be improved upon.
Price (see below)

[hyperdoggy]

Jmicron SSD - Shuttered
Intel G1 - no Trim support and when asked about it, Intel basically told you to go yourself
indilinx - poor firmware at launch, and needs to be wiped for many of its update.
Intel G2 - Kills data at launch
C300 - doa, recalled at launch
Sandforce - poor firmware at launch, but not as bad or many as indilinx

Is a Preview, get over it. OCZ always got first dibs on SSD and it shows with its many many firmware updates, at the same time Anandtech and the like is playing with these beta drives. No retail SSD had a perfect launch thus far, with the smoothest being current gen Sandforce. At least we got a look at Sandforce before is launched. Can't say the same for c400/Intel g3...drives that are suppose to be released before Sandforce.

[zalbard]

Heh, X25-M G2 still rules at low QD...

[naokaji]

All it proves is that testing an early engineering sample is pointless.

True, especially if you consider that the ES doesn't even use the same NAND as the retail version will have,

OCZ selected Toshiba 32nm Toggle Mode MLC NAND for these early Vertex 3 Pro samples however final shipping versions might transition to IMFT 25nm. The consumer version (Vertex 3) will use IMFT 25nm for sure.

Besides, with websites testing the higher capacity versions only it remains to be seen if the more affordable smaller versions will perform anywhere near the same or if they will be slower.

[Ao1]

Hard to know how Anand or PC Perspective undertook the benchmarks. Perhaps each bench was done on a clean drive. Maybe the stuttering (or if you prefer "super long read and write latency") occurs when the drive is in a used state with not much free space. Either way presumably it is a firmware bug that will get ironed out.

Anand kept well away from low QD performance. With the PC Perspective IOmeter benches the X25-M is significantly faster in all instances until you get to QD5. The X25-M and Vertex 2 also have much lower latency. Maybe that is something that will be fixed in the final version. If not buying these drives for non enterprise applications makes no sense at all and for enterprise the price is way too high.

Bottom line is that despite the reviews we are none the wiser on how these drive will really perform when they hit the market. That is the only point I try to make. Even when a full review comes out its done so quickly they typically miss the chinks.

Here is my frustration.....What really remains to be seen is the impact of 25nm on low latency and low QD performance. This will soon be clear however when Intel's 320 & 510 come out, which will not be much longer now. The 510 is already listed by some e-retailers

[johnw]

There is something that bothers me about the pcper transactions vs. QD charts. On AS-SSD, the C300 usually beats the X25-M on 4K QD=64 performance. But in the pcper charts, the X25-M is the champ.

I wonder if the pcper test used reads that were not aligned on a 4KB boundary. I know that the X25-M peforms better than most other SSDs for unaligned I/O.

I am anxious to see an AS-SSD benchmark of the Crucial M4 / Micron C400. I hope it maintains the 4KB QD=1 performance that is seen from the C300.

[Hornet331]

The chart is form IOmeter and not AS-SSD, you can't compare different benchmarks directly to each other.

[Ao1]

Jmicron SSD - Shuttered
Intel G1 - no Trim support and when asked about it, Intel basically told you to go yourself
indilinx - poor firmware at launch, and needs to be wiped for many of its update.
Intel G2 - Kills data at launch
C300 - doa, recalled at launch
Sandforce - poor firmware at launch, but not as bad or many as indilinx

Is a Preview, get over it. OCZ always got first dibs on SSD and it shows with its many many firmware updates, at the same time Anandtech and the like is playing with these beta drives. No retail SSD had a perfect launch thus far, with the smoothest being current gen Sandforce. At least we got a look at Sandforce before is launched. Can't say the same for c400/Intel g3...drives that are suppose to be released before Sandforce.

I agree overall with what you say, but not about the smooth SF launch. Intel G2's have been out for nearly two years now. Check out the f/w updates and release notes and compare that to SF drives. It seems even the latest's SF f/w has failed to fix some serious (although maybe rare) bugs that are resulting in data loss.

Intel's "kill data" at launch was an issue only if you had a BIOS password that you then changed. Not really a big deal.

+1 on the G1 no TRIM stitch up.

[johnw]

The chart is form IOmeter and not AS-SSD, you can't compare different benchmarks directly to each other.

I realize that pcper was charting IOmeter performance. That does not change the fact that the C300 generally has better 4KB random read high QD performance than the X25-M in AS-SSD benchmarks. It is difficult to see how IOMeter 4KB random reads at high queue depth should be different. The only reasons I can think of are lack of 4KB alignment, and maybe that the IOmeter setup was mixing reads and writes, which somehow hurt the performance of the C300 relative to the X25-M.

[irev210]

Well, yes, and i`ve been anands avid fan from the beginning, but nowadays i see some disturbing things going on in there.
People and sites change, so the fact that he once was great, doesnt mean it still is.
His relationship with ocz`s boss is almost personal now ,if you read all hos ssd articles you know this for sure.
And why he hasnt STILL reacted to OCZ failure regarding 25nm flash ?
Well ,im almost sure hes going to write something in a week or two when ocz already did damage control.
Hes sandy preview was weird too, his change in server hardware at corellating time to intel with corresponding server articles was fuzzy also.
He was asked repeatedly in comments exactly HOW he aquired his new server chips, and while he and his crew were responding to other questions, they plainly ignored this one (asked multiple times by multiple users in multiple ways ;-).
And the using of 480 FTW edition in one of the reviews.
Maybe im paranoid, but its all too fishy hes taken care of in this manner by ocz and intel, and the other sites mostly not.
And going back to the topic, hes previewing new ocz drive just in time when marvell introduces new controller and intel probably also is around the corner.Its marketing.As the product he reviewed isnt really the one were gonna get.

I agree. Anand has really dropped the ball on his SSD articles.

He made a post about intel's firmware snafu, the micron/crucial C300 snafu, but completely leaves OCZ/sandforce alone despite the fact that there seems to be a trend of higher failures, lots of firmware updates, and the 25nm snafu.

I like how Anand also just focused on peak read/writes using highly compressible benchmarks when sandforce first came out (though he has changed this now and displays both, which is nice).

Glad someone else sees it too.

[trans am]

well, you do have a valid point

paul did ocz ever offer you drives to test on your raid rig? I dont get why these guys are not sending the new hdwr to you. it makes no sense.
If i was a marketing dude I would send the drives to you me and steve ro to test on our raid rigs and we could all hype it up with crazy numbers and help them out with what we know.
they should be sending hardware our way instead of stupid andantech. did you see the andantech dudes at the top of pcmark05/vantage?
at lease send some crap to paul.
paul this is why we need to start our own company. I see gskill, msi and gigabyte getting us contests and goodies to push them. I see it with the vga cards and boards but you guys need to get us samples on drives if you want to push them.
majority of marketing needs to take a close look at how they market new crap. sending us some spam email and a link to andantech is not the answer. get in here and send some of this new crap to really test.
at least send paul some crap. computurd im sorry to put you on the spot but I can think of a crap load of dudes here that feel the same. you are the kingpin of complicated raid hard drive knowledge. dont be shy. you earned it.