[SSD] Vertex 2 EX or Intel X25-E

G

Gagnon

Weaksauce
Joined
Oct 28, 2009
Messages
93
I'm looking to get an SLC SSD for my new laptop and I am having a hard time finding decent reviews of these two drives (ideally pitted against each other).

Should I assume the performance / value (or lack thereof) are pretty much equivalent and just go with a brand? If so which brand would you recommend?

Options:
Intel X25-E 64GB
Vertex 2 EX 60GB
Vertex 2 EX 120GB

I am not familiar with any Intel options > 64GB so I guess if I want 120GB I'd stick with OCZ. Although with a 500GB hybrid as a second drive I think the money spent to go to 120GB would not be warranted.

Is there any truth to what I read about larger SSD's having a longer lifespan?
 
It is still a month or two from release, but the Micron P300 looks like it will take the performance crown for SSDs. It has SLC flash and will come in 50, 100, and 200GB capacities. I suspect it will be expensive -- $4 to $8 per GB would not surprise me.

http://www.micronblogs.com/2010/08/setting-a-new-benchmark-for-enterprise-ssds/

As for lifespan of the SSD, the fundamental things to keep in mind are that the NAND flash are specified to be erased/written a certain number of times, and beyond that they wear out. For MLC SSDs, that number might be 5K (up to 30K for "enterprise" MLC). For SLC, it can be much larger, even 100K, but for the P300 it looks like they are assuming 17.5K. So, that spec, combined with the wear-leveling algorithm, and the write-amplification, determines the longevity of the SSD. Also, keep in mind that SSDs have "spare area" which is not counted in the specified data capacity. The larger the spare area, the more efficient is the wear-leveling and lower the write amplification. Bottom line is that for a given amount and pattern of data writing, a larger SSD will last longer because there is more NAND to distribute the writes over and/or more spare area.
 
Last edited:
john4200 said:
It is still a month or two from release, but the Micron P300 looks like it will take the performance crown for SSDs. It has SLC flash and will come in 50, 100, and 200GB capacities. I suspect it will be expensive -- $4 to $8 per GB would not surprise me.

http://www.micronblogs.com/2010/08/setting-a-new-benchmark-for-enterprise-ssds/

As for lifespan of the SSD, the fundamental things to keep in mind are that the NAND flash are specified to be erased/written a certain number of times, and beyond that they wear out. For MLC SSDs, that number might be 5K (up to 30K for "enterprise" MLC). For SLC, it can be much larger, even 100K, but for the P300 it looks like they are assuming 17.5K. So, that spec, combined with the wear-leveling algorithm, and the write-amplification, determines the longevity of the SSD. Also, keep in mind that SSDs have "spare area" which is not counted in the specified data capacity. The larger the spare area, the more efficient is the wear-leveling and lower the write amplification. Bottom line is that for a given amount and pattern of data writing, a larger SSD will last longer because there is more NAND to distribute the writes over and/or more spare area.
Click to expand...

Hi John - If those benchmarks are to be believed it does look pretty nice, but won't help my cause for notebooks since it is a 6GB/s drive. The price even at $8/GB would be a bargain compared to Vertex 2 EX prices which are in the $12-14 / GB range.
 
I guess there is not a lot of interest here in the high performance (enterprise) SLC drives. Still looking for benchmarks/comparisons between the two.
 
Performance? Performance is MLC nowadays.

SLC still is useful in excessive write situations where you cannot replace the SSD easily when it exceeded its write cycles.

For most users, MLC is the answer. Especially if you want performance. The Intel X25-E is still at 50nm, almost 2 generations back. Yes it has high write endurance, but at what cost?
Interesting that the new X25-E will be 25nm "enterprise" MLC NAND; which does have more cycles than normal MLC but still much less than SLC. They likely also have a lot more reserved spare space (invisible space) than the X25-M counterparts, which only have 6.8%.

I say, wait until Christmas when you get third generation Intel (IMFT) 25nm MLC NAND. These SSDs should be much faster. And cheaper too, so you can buy multiple ones and replace any that exceeded write cycles. But normal usage that doesn't happen too often, really.
 
Gagnon said:
Hi John - If those benchmarks are to be believed it does look pretty nice, but won't help my cause for notebooks since it is a 6GB/s drive. The price even at $8/GB would be a bargain compared to Vertex 2 EX prices which are in the $12-14 / GB range.
Click to expand...

What do you mean "won't help your cause for notebooks"? It will work in your notebook computer, but if your computer only has SATA 3Gb/s, then the performance would be capped at 300 MB/s or less. It would still be the fastest SSD you could put in your notebook.
 
sub.mesa said:
Performance? Performance is MLC nowadays.
Click to expand...

Wrong. SLC is higher performance than MLC. That has not changed.

The MLC controllers had been getting ahead of the controllers in the SLC SSDs, but the P300 will change that.
 
Interesting; would love to see more thorough benchmarks on that upcoming P300!

But it is also interesting to note that X25-M G2 (34nm) has lower read latency than X25-E SLC (50nm). And it also appears that Intel is moving away from SLC altogether, using Enterprise MLC for their X25-E series.

Even though SLC on the same manufacturing generation (xxnm) is faster than MLC, i would argue that performance potential lies in the controller chips instead. More channels, more parallel I/O, more mature firmware, better chip designs. So instead of using SLC and letting your price-per-GB crash, you can also look at the option of better controller chips which simply use more parallel I/O to sustain higher speeds. I think that's more worthwhile trade-off.

But designing a NAND controller chip that scales with channels isn't that easy, and the market is quite young. So SLC can be an easy way to add some write performance mostly. But is that worth double the cost? It's not like SLC is double as fast.

Nonetheless interesting note! Does it have supercapacitor for safe writing? Then it would be ideal for database servers and as log devices. That's still a thing many current SSDs miss: they have unsafe writes. Though i think some Sandforce models also have ability to use supercapacitor/BBU.
 
sub.mesa said:
Interesting; would love to see more thorough benchmarks on that upcoming P300!

But it is also interesting to note that X25-M G2 (34nm) has lower read latency than X25-E SLC (50nm). And it also appears that Intel is moving away from SLC altogether, using Enterprise MLC for their X25-E series.

Even though SLC on the same manufacturing generation (xxnm) is faster than MLC, i would argue that performance potential lies in the controller chips instead. More channels, more parallel I/O, more mature firmware, better chip designs. So instead of using SLC and letting your price-per-GB crash, you can also look at the option of better controller chips which simply use more parallel I/O to sustain higher speeds. I think that's more worthwhile trade-off.

But designing a NAND controller chip that scales with channels isn't that easy, and the market is quite young. So SLC can be an easy way to add some write performance mostly. But is that worth double the cost? It's not like SLC is double as fast.

Nonetheless interesting note! Does it have supercapacitor for safe writing? Then it would be ideal for database servers and as log devices. That's still a thing many current SSDs miss: they have unsafe writes. Though i think some Sandforce models also have ability to use supercapacitor/BBU.
Click to expand...

Are you joking? The benchmarks I linked to are the best I have seen anywhere. They are done using Calypso's SSD tester.

And I agree about controllers. As I already wrote, the controllers are why it looked like MLC was gaining on SLC. In reality, it was just the SLC controllers lagging the MLC controllers. But Micron probably used similar quality controllers for the C300 and P300, and the P300 is about twice as fast as the C300. Yes, comparable process generation SLC is about twice as fast as MLC.

No supercapacitor. But the Micron literature states that the controller is designed to have good performance with write cache disabled, which is how enterprise applications that require write security would operate.
 
Sorry i hadn't reviewed your link yet. It does look good, but 480MB/s reads and 320MB/s writes isn't really double the performance. Random read IOps the 50nm SLC Intel does pretty good against such a new product. Is that really worth double the money? I still do not think so. Unless you have very specific requirements and going to use it in such an environment where the benefits make sense. But i think SLC has most of its benefits in its longer write endurance, not really performance. It does look like it can write twice as fast sequentially, though.

But i'll reserve my judgement for when i see better benchmarks, with good comparisons of all parties involved. In particular, it would be more interesting to put this upcoming product against the upcoming third generation Intel SSDs, especially the newer 25nm Intel X25-E, based on 'enterprise' MLC instead. I think it's too early to call P300 a winner or SLC to taken back the performance crown. But i agree SLC appears not to be dead either, yet. ;-)
 
The P300 definitely has the performance crown, there is no doubt. Nothing on the market (or coming on the market) this year can beat those benchmarks. Sequential and IOPS, the P300 is the highest.

The really impressive thing about the P300 is how it performs in the steady state. All of those benchmarks, except the write saturation, were performed on drives that had been heavily and fully written to. The Sandforce MLC people like to throw around 270 MB/s numbers, but in the steady state, with compressed data, the Sandforce numbers are typically in the mid 100's.

And that write saturation benchmark shows how impressive the P300 is. Being able to sustain 20,000 random write IOPS after hours of continuous 4KB random writes is incredible. Additionally, the P300 performance degrades smoothly as it is exercised, starting high (and holding high for longer than the others), then dropping smoothly to the steady-state level. In comparison, notice how the Intel overshoots into a "pit" before recovering to a steady state IOPS of less than one third of the P300.
 
Last edited:
Nothing on the market (or coming on the market) this year can beat those benchmarks.
Click to expand...
How can you tell? I think it's too early to make that kind of conclusions. I surely haven't seen any benchmarks of the upcoming Intel SSDs. I think they would at least be competitive to this expensive P300 controller with SLC NAND. And then price will determine final judgement.

One question i still have: are the writes done in that article compressible or not? If they are, this could amplify it's real underlying speeds and the compression would start to take artficial forms. But that random write graph is indeed impressive. But not if that was all due to compression giving the SSD much more spare space and doing much less work under the hood.

I think you may be jumping the gun a little on benchmarks results released by Micron itself; why wouldn't you wait for at least a few impartial reviewsites to test this thing with real competition? Then you would have more facts to base your opinion on, instead of these Micron PR-approved benchmarks. Yes they look plausible, but you don't know if this is the complete picture.
 
The 25nm NAND flash does not have any performance advantage over 34nm. The Intel G3s and the Micron P300 will both use IMFT flash, the G3s 25nm MLC, and the P300s 34nm SLC. SLC NAND flash DOES have performance advantages over MLC, particularly with erase/write speed. Therefore, a good controller paired with good SLC flash will beat a good controller with good MLC flash. And Micron has proven they can make a high performance controller in the C300. The P300 controller will be at least as good as the C300 controller. Furthermore, the Intel G3 SSDs are widely expected to have similar performance to the G2s, except perhaps somewhat higher write performance. I expect the G3 performance to be similar to the C300 performance. The P300 has higher performance than the C300, and the C300 has about the best performance of any MLC SSD on compressed data.

G3 = C300 < P300

The Micron controller does not do compression. Sandforce is the only one I am aware of that does compression.

Those benchmarks are from an impartial third party, Calypso Systems, which makes industrial SSD testers. That is a more reputable source than review websites, which often do a half-assed job, not following industry standards.
 
john4200 said:
The 25nm NAND flash does not have any performance advantage over 34nm.
Click to expand...
Any source? At least on Intel going from 50nm to 34nm caused lower read access latencies. The new 25nm NAND will also be ONFi 2.2 compliant. But most importantly, the next-gen 25nm Intel SSDs will likely get a new controller and that's where most performance gains could be expected.

Perhaps the next-gen Intel isn't as great as some hoped, leaving the Micron P300 as superior performance SSD product by a good margin. But it's just a little bit too early to make any such conclusions, at least to me.

SLC NAND flash DOES have performance advantages over MLC, particularly with erase/write speed. Therefore, a good controller paired with good SLC flash will beat a good controller with good MLC flash.
Click to expand...
I agree. But differences in controllers could yield far more extreme performance differences than differences in NAND speeds.

Furthermore, the Intel G3 SSDs are widely expected to have similar performance to the G2s, except perhaps somewhat higher write performance.
Click to expand...
G3 = C300 < P300
Click to expand...
Any source on that? I did not read anything that was leaked about Intels upcoming SSD line regarding performance numbers. Not even FUDzilla.

If you suggest that Intel would quit its own NAND controller and switch to Micron's controller instead for the upcoming third generation SSDs based on 25nm NAND, then i would think this is highly disappointing. It would effectively mean there isn't going to be a third generation Intel controller, and there's only Micron-Sandforce-Indilinx-JMicron-Samsung now.

So i hope you'll forgive me when i say i hope you're wrong. ;)

The Micron controller does not do compression. Sandforce is the only one I am aware of that does compression.
Click to expand...
Appears you're right. That does give more validity to your claims.

Those benchmarks are from an impartial third party, Calypso Systems, which makes industrial SSD testers. That is a more reputable source than review websites, which often do a half-assed job, not following industry standards.
Click to expand...
Yes but its still Micron who picked things out of that source that it liked to release, it may have chosen not to include some benchmarks. The basic things like random writes appear to be covered though, and especially the random writing looks exceptionally good.

It appears the tests themselves are conducted properly and you're right that review sites often miss the mark as they got no clue what they are actually testing. But at least it would truly be impartial and can uncover undisclosed (performance) issues with the new Micron P300 controller. So i'd still like to see more data before i make any conclusions about 2011 SSD performance lineup.
 
I cannot provide a definite source for the performance of the Intel G3s, but rather have heard "off the record" from people who I think have a good idea. Additionally, it just makes sense that there is not a lot more that can be done with MLC flash -- the controllers are already quite good. About the only significant gains left are to use huge amounts of overprovisioned flash...perhaps the new Intel X25E's will use 100% overprovisioning. But even that would only increase performance modestly over the current best MLC SSDs; mostly it would improve the endurance.

I did not suggest that Intel G3s would use the same controller as the C300 (they will not). What I meant was that the controller in the C300 is currently the best performing controller for MLC flash, at least for already compressed data. There is no reason to believe that the Intel G3 controller will be significantly better than the C300 controller. Perhaps it will have a slight edge in some areas, but trail in others. Overall, I expect comparable performance in the G3 and C300 controllers. But the P300 controller will be at least as good as the C300 controller -- basically an improved and "enterprised" version of it. Pair that with good SLC flash, and the P300 takes the crown. No doubt.
 
Last edited:
You must log in or register to reply here.
Back
Top