Cebit 2012: OCZ Vertex 4

[Gilgamesh]

Tests OCZ Vertex 4 (Indilinx Everest 2 + nand MLC Intel 25nm sync) 256GB
http://www.computerbase.de/news/2012...-ocz-vertex-4/

Atto:
http://pics.computerbase.de/3/9/9/6/0/2.jpg

Iometer:
http://pics.computerbase.de/3/9/9/6/0/3.jpg

Asssd:
http://pics.computerbase.de/3/9/9/6/0/1.jpg

[johnw]

What bus/controller was the SSD connected to?

AS-SSD says it is using an "IDE" driver. I wonder if that is why the sequential speeds look so bad.

[Ao1]

I thought that strange as well. The article does not say. Hard to image that OCZ would present the drive in a sub optimal system though. Seems Indilinx remain behind the curve.

[Kain665]

I'd be willing to bet that it's a configuration issue. There's no plausible reason that ANY SSD will perform better under IDE mode rather than AHCI mode. Without AHCI, you don't have NCQ...

[canthearu]

What bus/controller was the SSD connected to?

AS-SSD says it is using an "IDE" driver. I wonder if that is why the sequential speeds look so bad.

No, if it was using the IDE driver, the 4k-64q figure would be the same as the 4k figure. (No NCQ in IDE mode)

[Anvil]

Not sure why they are using V1.2.n of AS SSD, could be a bug in detecting storage in early versions?

The latest fw for Octane resulted in higher random reads/writes while sequential suffered, might be the same that's happening here -> tuned for max iops at small blocks.

[Tony]

http://www.ocztechnologyforum.com/fo...l=1#post716980

the version of AS-SSD on the show machine is old and misreporting ahci mode, I had Nico test on his X79 with RSTe drivers ahci mode, his AS-SSD is in the post I linked.

Remember, this is on a daily FW build, no where near final

[Computurd]

TBH i rather like it with the great low QD numbers, especially the 4k write is making me feel randy...

none of the incompressible/compressible mumbo jumbo, and great performance at low QD is what we really need. I just hope it isnt tuned up too much higher on the seq at the sake of the low qd.

as it stands, I like it, and think it would make a great OS drive for 'normal' users. Seriously, your never gonna hit those tall sequentials and high qd random. looks to be a screamer for desktop as-is.

we need more education for end users so that they will see that a drive like this is actually the ideal !

[johnw]

It is interesting to compare AS-SSD benchmarks of the Plextor M3P and the OCZ Vertex 4.

The Plextor wins on QD1 4KiB random read, sequential read, and sequential write. The V4 wins on QD1 4KiB random write, and QD64 4KiB random read and write.

In my opinion, the importance of these figures for most desktop power users, from most important to least important, is:

QD1 4KiB random read
sequential read
QD1 4KiB random write
sequential write
.
.
.
QD64 4KiB random read
QD64 4KiB random write

The high QD stats (beyond about 80MB/s) are distant last, since they almost never come into play for most desktop users. They are only important for situations such as a heavily loaded database server. (But AS-SSD places a very high weight on those stats in its "score", which is misleading for most desktop users)

Of the other four stats, the Plextor wins three and the V4 takes one. I think Plextor has the performance lead by a wide margin. And with a 5 year warranty and no history of mistakes and BSODs, the Plextor also wins for reliability.

256GB Plextor M3P
http://www.xtremesystems.org/forums/...6&d=1331062263
m3p.jpg

240?GB OCZ Vertex 4
http://www.ocztechnologyforum.com/fo...1&d=1331125325

[Anvil]

I wouldn't use the C600 with the 3.0.x drivers for promoting products just yet, for most drives it cuts on sequentials and that could be the case here as well. (to some extent)
(11.x drivers brings more speed to C600/X79)

I wouldn't mind a faster "C300" and that is what this sort of resembles.

Looks like I need to have a look at one of these drives , now if they could make a 360-400GB edition at an affordable price.

[Ao1]

Taking into account lazy writes I’m not so sure how important 4K write performance is anymore. I’m going to look into it in more detail but I’d rate sequential write higher than 4K QD1 write performance, which with current drives is already way above anything required in a client environment.

Too early to say as the V4 is an engineering sample and there is not a direct performance comparison available between either drive, but it looks like the Plextor comes out on top, plus it’s on the market now.

[canthearu]

Taking into account lazy writes I’m not so sure how important 4K write performance is anymore. I’m going to look into it in more detail but I’d rate sequential write higher than 4K QD1 write performance, which with current drives is already way above anything required in a client environment.

My take on 4K write is that it is reasonably important for a drive to be able to quickly complete write when it is required to. When software issues a write to the SSD, and then instructs the SSD that all writes must be flushed to permanent storage, it cannot hang the system for seconds at a time like early SSDs did. However, beyond this, 4k write performance is not important at all.

If 4k write performance really was important, hard drives simply wouldn't work at all, as they benchmark at only a few meg per second best case for 4k write performance.

The system write back cache is able to handle most burst writing, and will gracefully degrade as long as the ssd controller responds reasonably quickly to commands.

[Gilgamesh]

Ocz Vertex 4 will debut with nand Intel 25nm next june or with nand IMFT 20nm?

[RyderOCZ]

I wouldn't use the C600 with the 3.0.x drivers for promoting products just yet, for most drives it cuts on sequentials and that could be the case here as well. (to some extent)
(11.x drivers brings more speed to C600/X79)

I don't have info at the moment on whether this is with P67 or Z68, but here is a bench with the "Non-Enterprise" RST drivers.

[Wendy]

Taking into account lazy writes I’m not so sure how important 4K write performance is anymore. I’m going to look into it in more detail but I’d rate sequential write higher than 4K QD1 write performance, which with current drives is already way above anything required in a client environment.

That's my feeling on the subject as well, and has been for quite some time.
I would be very interested in reading your findings, should you decide to look into this in more detail.

[Kain665]

For the normal user, anything over 200mb/s sequential read/write is overkill. You're only going to see real "snappiness" with small block transfer speeds at low QDs - where the V4 shines. 4KRW performance scales extremely fast with QD, reaching almost 80k at just QD3. Combine that with the excellent internal GC of the Everest 2, most users will not see IOPS performance degradation as with SF controllers of this generation.

[Computurd]

and thats the elephant in the room. what is the REAL performance over a period of time with all devices. FOB testing is just meh.

[Tony]

Systems at cebit are X79

[Computurd]

usually fast write speeds @ low QD equate to faster reads, as you are freeing the controller. This is why i still love the low QD write just as much. We all know the punishment mixed read/write workloads can put on any device.
also, just a good measuring stick of overall latency/snappiness. either read or write IMO.