• We’re currently investigating an issue related to the forum theme and styling that is impacting page layout and visual formatting. The problem has been identified, and we are actively working on a resolution. There is no impact to user data or functionality, this is strictly a front-end display issue. We’ll post an update once the fix has been deployed. Thanks for your patience while we get this sorted.

OCZ Vertex EX

A

alcoholbob

Diamond Member
http://benchmarkreviews.com/in...=view&id=186&Itemid=60

The first review is pretty bad and uses benchmarks that are rarely used for SSD benchmarking (other than ATTO).

The popular ones are Crystal Disk Mark, HD Tune 3.5 Pro, IOMeter. The Vantage HDD suite is fairly decent as well.

A couple of review samples are out in the wild, hopefully more reviews come in soon. OCZ wants to sell this as the X25-E "alternative." First week there will be price gouging (right now its 95% of the Intel X25-E according to Mwave) but their plan is for the 120G to be ~$1300 instead of the $1460 it's currently pegged at, and less than mwave's $740 for the 60GB.

http://www.ocztechnologyforum....768&stc=1&d=1240523903

Sequential read/write, Random 512K Read/Write are neck and neck with the X25-E (within error margin). Where the differences lie is in the 4K performance (X25-E numbers are from my own benchmarking with a 64G X25-E)

Intel: 23MB/s read, 65MB/s write
EX: 35MB/s read, 25MB/s write

Based on the Microsoft article below, the EX might have a slight advantage as a Windows-based SSD. It states that normal pagefile characteristics are a 40 to 1 ratio of reads to writes, with 90% of reads under 16K, and 70% 4K or under. The Intel drive overcompensated with writes but reads are much lower.

http://blogs.msdn.com/e7/archi...-state-drives-and.aspx

The Vertex EX is kind of intriguing...I'm considering jumping in for the 60G. $1.3k (and that's a "non-gouge" price) for a 120G is a little *too* much.


 
I like the numbers, well except the price of course :laugh:

Seeing how quickly (trend-wise) the MLC-based SSD's have ramped up performance in the past 12 months, I'm really curious to see whether the SLC vs. MLC performance differentiation will continue to support the parallel market segments of MLC and SLC or if SLC will simply cease to be viable at some point in the near future.

At this time it reminds me of the (then) competing memory standards for the technically superior but more costly rambus RDRAM versus the lower-performing (at the time) but far less costly DDR SDRAM.

SLC's performance attributes simply cannot be contested by an MLC device, but the market opportunity for SLC-based SSD's seems to be dissipating as MLC-based SSD's continue to reduce the magnitude of those advantages.
 
Originally posted by: Astrallite
Intel: 23MB/s read, 65MB/s write
EX: 35MB/s read, 25MB/s write
Click to expand...

Even with slc it can't catch up to the mlc intel drive in random writes? The intel drive seems like a steal now, they're pricing the EX too high.
 
Originally posted by: shabby
Originally posted by: Astrallite
Intel: 23MB/s read, 65MB/s write
EX: 35MB/s read, 25MB/s write
Click to expand...

Even with slc it can't catch up to the mlc intel drive in random writes? The intel drive seems like a steal now, they're pricing the EX too high.
Click to expand...

It's only a steal if you believe you'll expect a need for some sort of constant logging that would require such high writes. Per Microsoft though, 4K read and sequential write performance are optimal for virtual memory. The MLC Vertex actually has 1/2 of the write speed of the EX (12MB/s), whereas the X25-M and the X25-E both share very similar random write performance.

That indicates
1) a high chance that Intel is intentionally reducing the sequential write performance on the X25-M in order to differentiate it from the X25-E
2) they have an excellent controller that has near identical random write (and read) performance between the SLC and MLC models
3) (and if the previous points bear any similarity) Indilinix might be intentionally reducing random access performance on the MLC controller in order to differentiate it from the SLC variant

But I agree the pricing is a bit high, most people in the OCZ forums were hoping for a $600 price point rather than $670-700 range it's at right now.

Here's the benchmark on my X25-E

http://i40.tinypic.com/2cf422r.jpg

When I first got the X25-E, I was getting random 4K reads of 33MB/s although it quickly dropped to the mid 20s until settling at 21-23. I found my 120GB Vertex to maintain consistent performance since I first purchased it. Here's hoping for an X25-E firmware update I guess...
 
RAMBUS Was different, there was a ram cartel out to kill it with illegal market behavior(according to the court verdict).

No matter how much better your controller gets, SLC is inherantly faster, a lot faster. So I see a permanent parallel market there... its just that pretty soon MLC drives will be "fast enough" for most users, who could not justify the price premium of SLC...

Although, maybe later the overall prices will be so low that paying twice the cost for SLC would make it more appealing to many (EX: 80$ in stead of 40$ for a fast small OS drive)
 
Doesn't the SLC version use the same memory chips with the same controller(with different firmware) as the MLC version? So why the heck is the price significantly higher than the 120gb MLC vertex? Someone correct me if I'm wrong.
 
Because when you modify the firmware to make it MLC, you double the space (at the cost of a lot of speed)... so a 120GB SLC vertex has the same hardware as a 240GB (well, 250GB in that case because there is less "hidden" space used for performance increasing) MLC vertex only with a different firmware.

That is assuming that your controller can handle both with just a firmware change (which I believe is the case with the vertex)

http://www.newegg.com/Product/...x?Item=N82E16820227396

However, there is a good chance that there is some slight design differences in the control board or power circuitry between an SLC and MLC version (which could prevent cross flashing, should not affect price)
 
It may seem odd that the Vertex EX is 3 1/2 times more expensive then the Vertex (MLC), but OCZ is trying to price themselves against the X25-E. Sometimes "improving your price" like in the DirectTV commercial, is a viable way to advertise that you are a luxury product--while at the same time, being cheaper than the competition.
 
Originally posted by: Hacp
Doesn't the SLC version use the same memory chips with the same controller(with different firmware) as the MLC version? So why the heck is the price significantly higher than the 120gb MLC vertex? Someone correct me if I'm wrong.
Click to expand...

I could be completely wrong here, but it is my understanding that MLC chips are not just SLC chips configured to operate as MLC chips, or vice versa.

They are fabbed differently (specifically the floating gate process tech is different for MLC vs SLC chips, for MLC tha floating gate stack is customized specifically for higher yielding chips and lower error rates, better performance, etc), binned and validated differently (different operating voltages, acceptable error rates, etc), and sold as distinctly different IC's...in a lot of ways you could think of it by way of analogy between DDR2 vs. DDR3 in that the end products are not the same, nor are the underlying individual IC's the same, but they do function in similar fashion and to a novice could be mistakenly simplified as "DDR3 is just a double-pumped DDR2"...which would be way oversimplified and untrue of course but it is easy to see where such a simplification could be generated.

http://www.supertalent.com/dat...s_MLC%20whitepaper.pdf

Checkout this one highlighting the divergence in process tech for Intel/Micron's IM flash with SLC to MLC:
http://www.semiconductor.com/m...ex.asp?destination=710
 
Well OCZ support reps states that the price of SLC NAND is completely dependent on worldwide fab prices and right now production is something like 70/30 in favor of MLC NAND, which is why costs are so high. Demand for SLC is rising and fabs are retooling from MLC for SLC production. Indilinix sells a hybrid controller, so I would be surprised if a significant amount of the price hike involves the controller firmware alone.
 
The write speed issues are overstated. When you're writing files you typically don't need the performance and expect it to be slow. I'd say that 90% of the time I'm reading files.
 
Originally posted by: Astrallite
Well OCZ support reps states that the price of SLC NAND is completely dependent on worldwide fab prices and right now production is something like 70/30 in favor of MLC NAND, which is why costs are so high. Demand for SLC is rising and fabs are retooling from MLC for SLC production. Indilinix sells a hybrid controller, so I would be surprised if a significant amount of the price hike involves the controller firmware alone.
Click to expand...

Well, the same guy wasn't sure if SLC drives would need TRIMM(that is suffer from write performance after filling up the drive), so I wouldn't take his word for anything.
 
The Vertex (MLC) has K9HCG08UIX Samsung NAND.
The Vertex EX (SLC) has K9NCG08U5M Samsung NAND.

The Vertex (MLC) has an Indilinx Barefoot IDX110M00-FC controller.
The Vertex EX (SLC) has an Indilinx Barefoot IDX110M00-LC controller
A crossflash will almost certainly not be possible.
 
Originally posted by: Hacp
Originally posted by: Astrallite
Well OCZ support reps states that the price of SLC NAND is completely dependent on worldwide fab prices and right now production is something like 70/30 in favor of MLC NAND, which is why costs are so high. Demand for SLC is rising and fabs are retooling from MLC for SLC production. Indilinix sells a hybrid controller, so I would be surprised if a significant amount of the price hike involves the controller firmware alone.
Click to expand...

Well, the same guy wasn't sure if SLC drives would need TRIMM(that is suffer from write performance after filling up the drive), so I wouldn't take his word for anything.
Click to expand...

You mean Tony? He's one of the QA guys at OCZ.
 
Originally posted by: SickBeast
The write speed issues are overstated. When you're writing files you typically don't need the performance and expect it to be slow. I'd say that 90% of the time I'm reading files.
Click to expand...

when i read from a drive, its usually an ongoing operation needing a specific bandwidth and doesn't matter if it goes over, be it to burn a disk or watch a movie or play a song or a game, it needs to be "Fast enough"

when i write something i initiate a write and then i have to WAIT on it to finish, and until it does i am prevented from doing other operations (or get a lot of slowdown doing other operations)...

So I would say in reality its the exact OPPOSITE of what you said.
 
Well his comment really lacks context since it's basically an open-ended statement, but he might be talking about the excessive 4k write performance on the X25 series, compared to the less than stellar 4k read performance, which in fact, both the Indilinix and Samsung SSD controllers actually manage to surpass.

I can understand why you might need very high random writes for enterprise applications (X25-E), but the M has the same random write/read performance, which is odd for mainstream use (4k reads and sequential writes being the typical behavior of Windows). Having this much headroom is NICE, but merely the ratio is seems out of place (I imagine it would be much better if the numbers for read/write were reversed).
 
Originally posted by: Astrallite
Originally posted by: shabby
Originally posted by: Astrallite
Intel: 23MB/s read, 65MB/s write
EX: 35MB/s read, 25MB/s write
Click to expand...

Even with slc it can't catch up to the mlc intel drive in random writes? The intel drive seems like a steal now, they're pricing the EX too high.
Click to expand...

It's only a steal if you believe you'll expect a need for some sort of constant logging that would require such high writes. Per Microsoft though, 4K read and sequential write performance are optimal for virtual memory. The MLC Vertex actually has 1/2 of the write speed of the EX (12MB/s), whereas the X25-M and the X25-E both share very similar random write performance.

That indicates
1) a high chance that Intel is intentionally reducing the sequential write performance on the X25-M in order to differentiate it from the X25-E
2) they have an excellent controller that has near identical random write (and read) performance between the SLC and MLC models
3) (and if the previous points bear any similarity) Indilinix might be intentionally reducing random access performance on the MLC controller in order to differentiate it from the SLC variant

But I agree the pricing is a bit high, most people in the OCZ forums were hoping for a $600 price point rather than $670-700 range it's at right now.

Here's the benchmark on my X25-E

http://i40.tinypic.com/2cf422r.jpg

When I first got the X25-E, I was getting random 4K reads of 33MB/s although it quickly dropped to the mid 20s until settling at 21-23. I found my 120GB Vertex to maintain consistent performance since I first purchased it. Here's hoping for an X25-E firmware update I guess...
Click to expand...



Shoot, I thought the X25E was the ultimate solution. I guess not.
 
Originally posted by: Astrallite

When I first got the X25-E, I was getting random 4K reads of 33MB/s although it quickly dropped to the mid 20s until settling at 21-23. I found my 120GB Vertex to maintain consistent performance since I first purchased it. Here's hoping for an X25-E firmware update I guess...
Click to expand...

No... You have it wrong. The random reads were always low-20s for both X25-E and X25-M.
http://www.tomshardware.com/re...ance-power,2279-6.html

http://images.google.ca/imgres...3Dko%26sa%3DN%26um%3D1

The fluctuations in scores is CrystalDiskMark's fault because it doesn't average the result but only show the highest of the results. You might get 25MB/s in the first run, and 20MB/s in the rest but the result would still end up to be 25MB/s. The reason they did that I suspect is because the benchmark program dealt with platter HDDs until very recently and for those showing max speed isn't really a bad thing as hard drive benchmarking shows relatively stable scores.

Look everywhere, it never reached anywhere above 24MB/s.

1) a high chance that Intel is intentionally reducing the sequential write performance on the X25-M in order to differentiate it from the X25-E
2) they have an excellent controller that has near identical random write (and read) performance between the SLC and MLC models
3) (and if the previous points bear any similarity) Indilinix might be intentionally reducing random access performance on the MLC controller in order to differentiate it from the SLC variant
Click to expand...

Do YOU REALIZE that SLC is faster at the chip-level than MLC?

-Yes, its true X25-M and X25-E has similar random write performance. However, their sequential write performance is not.

-And its true that Vertex SLC and MLC versions have similar sequential write performance. But random writes are not.

No, its just a different "adjustment" depending on what the manufacturer thought was optimal for the market segment. X25-M document states that while similar IOPS to X25-E are possible, the performance will not be robust over the whole drive as the X25-E.

And for the Vertex, the initial firmware versions lowered sequential writes for random writes! Nobody's getting anything free.

Intel sacrificed sequential write speeds on the X25-M because reaching both the extraordinary random write level and high sequential speeds might not be possible in MLC. OCZ sacrificed random writes to get higher sequential writes for same reason as Intel. MLC is slower than SLC by nature!!
 
Originally posted by: IntelUser2000
Originally posted by: Astrallite

When I first got the X25-E, I was getting random 4K reads of 33MB/s although it quickly dropped to the mid 20s until settling at 21-23. I found my 120GB Vertex to maintain consistent performance since I first purchased it. Here's hoping for an X25-E firmware update I guess...
Click to expand...

No... You have it wrong. The random reads were always low-20s for both X25-E and X25-M.
http://www.tomshardware.com/re...ance-power,2279-6.html

http://images.google.ca/imgres...3Dko%26sa%3DN%26um%3D1
Click to expand...


Crystal Disk Mark is *A* benchmark. It's not the end-all-be-all. HD Tune Pro 3.5 gives me different numbers with a reverse 4k read order: X25-E, X25-M, Vertex. And I don't know if you typed in the wrong link, but your Tomshardware link has nothing to do with what you are talking about.

http://img25.imageshack.us/img...epro35randomaccess.png
http://img8.imageshack.us/img8.../x25mhdtuneproread.png

Do YOU REALIZE that SLC is faster at the chip-level than MLC?
Click to expand...

Of course I do.

Yes, its true X25-M and X25-E has similar random write performance. However, their sequential write performance is not.
Click to expand...
The X25-M has unusually low write speed. The write speed is almost a complete straight line at 70MB/s, which online reviewers have postulated that Intel is intentionally limiting X25-M performance.

And its true that Vertex SLC and MLC versions have similar sequential write performance. But random writes are not.
Click to expand...
No they don't have similar sequential write performance. ATTO shows a 50MB/s seq write differential. Crystal Disk Mark shows 60MB/s.

Intel sacrificed sequential write speeds on the X25-M because reaching both the extraordinary random write level and high sequential speeds might not be possible in MLC. OCZ sacrificed random writes to get higher sequential writes for same reason as Intel. MLC is slower than SLC by nature!!
Click to expand...

I have no comment here. This is borderline non sequitur.
 
Originally posted by: sxr7171
What makes the Intel X25E so slow in sequential 4K reads?
Click to expand...

CrystalDiskMark indicates that Vertex (and EX) are faster than X25 series in 4k random reads. ATTO shows that the Intel drives actually have about a 7% advantage in 4k performance. And my HD Tune Pro 3.5 numbers seem to show a reverse order too.

However, assuming CrystalDiskMark is the only legit 4k benchmark, then I'd say that Intel designed their random writes for a very specific enterprise application (maybe lots and lots of data logging) and where generally reads are sequential in nature (like opening a file on the server).
 
Originally posted by: Astrallite

Crystal Disk Mark is *A* benchmark. It's not the end-all-be-all. HD Tune Pro 3.5 gives me different numbers with a reverse 4k read order: X25-E, X25-M, Vertex. And I don't know if you typed in the wrong link, but your Tomshardware link has nothing to do with what you are talking about.

Of course I do.

The X25-M has unusually low write speed. The write speed is almost a complete straight line at 70MB/s, which online reviewers have postulated that Intel is intentionally limiting X25-M performance.

No they don't have similar sequential write performance. ATTO shows a 50MB/s seq write differential. Crystal Disk Mark shows 60MB/s.

I have no comment here. This is borderline non sequitur.
Click to expand...

Tomshardware review has to do with the fact that degradation is minimal. Which is the whole point this started.

The Vertex MLC gets 150MB/s and SLC gets 200MB/s writes. That's only a 33% difference, not a whole 200MB vs. 80MB(measured) for Intel, which is 150% difference.

Go back and look at what Intel X25-M datasheet says.

4K Read IOPS 35,000
4K Write IOPS 3,300

"Measurements are performed on 8GB span"

You look at the X25-E datasheet it says same numbers for IOPS but...

"Random IOPS cover the entire range of legal logical block addresses(LBAs). Measurements are performed on a full drive(all LBAs have valid content).

The X25-M has unusually low write speed. The write speed is almost a complete straight line at 70MB/s, which online reviewers have postulated that Intel is intentionally limiting X25-M performance.
Click to expand...

I tell you, there are reasons its done and its a good thing. Whatever reasons Intel put high random writes as important(and according to benchmarks X25-M performs better even compared to the Vertex which has 2x sequential writes), and that super-high random writes is achieved as a result of sacrificing sequential writes.

Whatever the drive, it uses a buffer for scratch space and when the scratch space runs out, the drive LBAs are full, but if the speed of the random writes are too high it'll possibly flood the process.

Slowdowns:
X25-E: http://forum.ssdworld.ch/viewtopic.php?f=1&t=82
FusionIO: http://forum.ssdworld.ch/viewtopic.php?f=4&t=59
Vertex with 1199 firmware(1199 has problems, but not related to performance and they have it working fine at least here): http://forum.ssdworld.ch/viewtopic.php?f=4&t=95

This is probably the reason its put as a server drive:

OCZ Vertex MLC IOPS: http://forum.ssdworld.ch/viewtopic.php?f=4&t=95
Intel X25-E IOPS: http://forum.ssdworld.ch/viewtopic.php?f=4&t=81

Go look back at the X25-E review in SSDWorld. This review was done back in mid-December of last year, when barely anybody knew about the firmware(probably no one outside Intel). See on HDTach R/W benchmark how it recovers quickly, a similar behavior to the 8820 firmware on the X25-M.

"In order to resetting the drive we don't needed to format it, just while we made the bench tests the Intel X-25 E have calibrated himself to the right performance, you can see it in this HDTach Bench, please note how at the end of the test the performances will regain the normal rate (200MB/s with fast slow down to 150MB/s)"

"If stressed in a normal server use the ioDrive is quick KO, after only 2 min of Iometer you can register a slow performances. It was the same for the MLC model, and was the same for the ioDrive of FusionIO Company.
This happens because there are many "refuse" files after session with Iometer, this drive can't handle so many fast, so the performance will drop."

Yea whatever. So even the Vertex, can be screwed up with enough I/O testing. However, most PC users probably won't see the problem. It's true even with the older firmware and in rare cases even the JMicron. Some people know how to degrade it to the level where it looks like its "dying" almost.
 
However, assuming CrystalDiskMark is the only legit 4k benchmark, then I'd say that Intel designed their random writes for a very specific enterprise application (maybe lots and lots of data logging) and where generally reads are sequential in nature (like opening a file on the server).
Click to expand...

Yea according to

OCZ Vertex MLC IOPS: http://forum.ssdworld.ch/viewtopic.php?f=4&t=95
Intel X25-E IOPS: http://forum.ssdworld.ch/viewtopic.php?f=4&t=81

X25-E achieves a magnitude higher IOPS.

Probably what X25-E is doing is very relevant unlike your claim.
 
You must log in or register to reply here.

TRENDING THREADS

Back
Top