Vertex 4 Reviews and Thoughts

[Christopher]

OCZ is saying it doesn't happen on AMD systems... that they think its an Intel problem.

[Cyber-Mav]

so regardless of new 1.4 firmware there is still issues with this drive?

[Christopher]

so regardless of new 1.4 firmware there is still issues with this drive?

The 128GB will still have the write speed reduction at 50%, but the 512GB and 256GB don't seem to have the issue. But OCZ support staff hasn't tested it with a 128GB or on anything but AMD systems (if I understand the post I'm referencing in the V4 1.4 final thread at OCZ's forums).

I am aware of some reviewers getting twitchy about this and they need to take a step back and use the 1.4 final and retest.

I cannot reproduce any such Drive Fill scenario drops (on 1.4 Final) in Linear Read or Write with my 256 GiB drives across any of my AMD platforms.

I won't finish the Intel platform testing until late tomorrow (GMT).

Regards,

[AMD_Freak]

I’m going to guess it’s a form of throttling then. Makes sense looking at that graph; a sudden drop and then more or less flat line performance.

Maybe it is also why you can see less of an impact with smaller sized xfers if the throttling is based on a bandwidth restriction. If I remember correctly the SF drive also throttled reads, which was presumably linked in some way to the mechanism that throttled writes. Is read performance also impacted when write speeds drop?

It’s not actually that unreasonable if you need to protect a 5 year warranty against server type “abuse”, but “if” true $OCZ should be upfront about it.

Attachment 126944

Yea like they are upfront about everything even the controller the Vertex 4 had in it

[canthearu]

The 128GB will still have the write speed reduction at 50%, but the 512GB and 256GB don't seem to have the issue. But OCZ support staff hasn't tested it with a 128GB or on anything but AMD systems (if I understand the post I'm referencing in the V4 1.4 final thread at OCZ's forums).

So basicly the last 2 years have taught OCZ absolutely nothing about good QA testing ....

[Anvil]

Hopefully I'll be getting the 256GB by Wednesday.

Looks like they (OCZ) are going to post some info about what they have found on the performance drop, maybe today.

[Cyber-Mav]

went with another plextor m3 in the end. cant risk using vertex 4 with so many issues with it.

[AMD_Freak]

So basicly the last 2 years have taught OCZ absolutely nothing about good QA testing ....

because they know people will buy into the line of BS they spew and sell

[johnw]

because they know people will buy into the line of BS they spew and sell

It amazes me the people still buy OCZ SSDs. Heck, some people are actually OCZ fanboys. Weird.

[Anvil]

I got the 256GB drive today, will check for the fill issue sometime tomorrow.

A quick test using LBAs > 128GB did not show any issues.
(I copied ~25GB of data to LBAs past that point and speed was as expected)

[Ao1]

OCZ is saying it doesn't happen on AMD systems... that they think its an Intel problem.

lol, should have guessed it was not an $OCZ problem and nothing to do with the 128Gb drives being throttled.

[Nizzen]

Same story about the sandforce ssd's.... Blame intel and endusers :p

[JR.]

Hi,

It is somewhat dissappointing that people post comments in this thread without having any first hand experience of the Vertex 4.

My testing shows that with a 256GB V4 the drop off in speed only occurs when one writes a massive amount of data continuosly to an unformatted drive in one windows session. I hit the drop after writing 4 test files (80GB + 80GB + 40GB + 20GB), full of random data, at this point I deleted the 20GB file and performance recovered immediately and in a second test I rebooted at the 220GB level and performance recovered immediately. I also observed the drop off in a linear write test against a raw, unformatted drive, at 78% full - however when I formatted a drive prior to the linear write test I observed no drop off.

I understand that OCZ may soon publish definitive test results to remove the myths surrounding the drop off in speed.

Meanwhile, here are some performance benchmarks using the new 1.4 firmware.

First, a single 256GB Vertex 4:



and secondly 2 x256GB in Raid 0:



Remarkably, the Raid 0 shows no performance degradation even when well and truely pummelled with massive amounts of data. Indeed I see no difference between the persistence of performance with one V4 (with trim) and two in Raid 0 - it's as if the Raid 0 benefits from some form of pseudo trim functionality.

Regds, JR

[SteveRo]

^^ If no other problems are discovered I should start selling my stack of acards!!

[Christopher]

The 128GB is the only one affected significantly. The 256GB and 512GB don't really have the issue. But it's not a bug, at least I don't think. It's too controlled and prescribed.

[johnw]

I understand that OCZ may soon publish definitive test results to remove the myths surrounding the drop off in speed.

Heh, you used "OCZ" and "definitive" in the same sentence! Of course, it is rather balanced by the words "may" and "myths", which go well with "OCZ".

[Zaxx]

I'm betting they'll get it sorted before long. From what I've read, difficulty flashing the fw is the biggest complaint so far...also read (here ofc) that the overclocked controller heats up during endurance testing (which didn't go to well). Too bad we can't just crack it open, slap a cooler on the controller and jack it up...lol. Can't wait to see what nand Ocz goes with for the Vertex 4 MaxIOPs...24nm Type B or maybe the upcoming 20nm IM?

[JR.]

Hi,

For a 256GB V4 you will only experience a marked drop off in speed if you write more than 185Gb of data continuously (in one Windows session) to an unformatted, raw drive. Performance recovers immediately following a reboot. If one runs a linear write test against a raw, unformatted 256GB V4, you will see a drop in speed at 78% full. If you format or write at least once to all blocks and then run a linearwrite test you will see no drop off.

To make things clear, I am pro OCZ, which seems to be a sin to end all sins around here

Placing a careful guard on my objectivity, I don't think the V4 has a problem. I feel that reviewers should step their game up a bit as the running of a synthetic bench against a raw unformatted drive has thrown up a fog of misleading vibes, which have then mindlessly been picked up and compounded by the popular anti-OCZ front - Much ado about nothing, i.m.o.

I know - I must immediately go and wash my mouth out with soapy water for saying something pro OCZ.

Regds, JR

[Ao1]

Prospective buyers can only ascertain if it’s not a problem if the know what triggers the reduction in read and write speed. The simple thing would be for $OCZ to provide a statement on why this occurs and under what circumstances. They have avoided providing any detail to date despite the fact that they know exactly why it is occurring, so it’s not that surprising that everyone is speculating.

I believe the issue occurs on formatted drives (not just on unformatted drives) and that the 128GB drive is more susceptible to whatever is occurring. My guess is that it is a form of throttling, but no one seems to be saying anything and I’m not about to buy one to find out.

[Anvil]

Welcome JR

There are strong opinions on some brands, it is expected as some people have different experiences with different brands.
(some have had bad experiences with SSDs in general)

I've got both drives and the issue is easily reproducible in real-life situations on the 128GB, I deal mostly with large files (VMs) and so the initial "install" triggers the issue for me.
The 256GB is different in the way that it recovers more of less w/o "help" and like you say, the drop in speed is at about 80% which makes it unlikely to surface during a single "session".

[Anvil]

^^ If no other problems are discovered I should start selling my stack of acards!!

Hi Steve, looking for something new to play with

[SteveRo]

Hi Steve, looking for something new to play with

Good morning Mr. Anvil, I took the plunge on a pair of V4 128s. A new toy!
Do and compare benches of old firmware vs. new - on ich10r, pch, 1261 and then 1880 - should be fun!

[Anvil]

Steve

They should be great fun!
I've got one of each and so I need another one to start playing.

Bring us the results, in the Xtreme section

[SteveRo]

^^ very good will do

[JR.]

I believe the issue occurs on formatted drives (not just on unformatted drives) and that the 128GB drive is more susceptible to whatever is occurring. My guess is that it is a form of throttling, but no one seems to be saying anything and I’m not about to buy one to find out.

Hi Ao1,

My personal experience is with the 256GB drives (both as a singleton and 2 in R0). I have only seen the increase in latency on a 256GB drive in the circumstances I described in my previous post. I have seen some interesting results from the testing of a 128GB drive but I am not in a position to refer to this.

I hope OCZ will take steps to remove the confusion sometime soon - I get the impression that the forefront of SSD technology is kept strictly under one layer of NDA after another, which makes things difficult.

Meanwhile, I find that the V4 is an incredible bit of technology. See the results I've posted for 2 x 256s in R0 above - 2125 in AS SSD was until recently the stuff of only extreme multi-drive raiders and now it is 'easy peasy' for the average Joe and a relatively modest investment - I feel folks should applaud OCZ for taking the lead in driving the market forwards over the last few years.

By the way my 2 x 256GB V4 array just simply refuses to drop in performance and I have pummelled it relentlessly with Anvil's Endurance Test and large file tests of my own. I dont know how the V4 achieves this - but heck who needs trim?

Regds, JR

@Anvil - thanks for the welcome m8 - much appreciated

p.s. when I have previously refered to a 'formatted drive' I actually mean a full format. I know this goes against the grain of the normal advice for an SSD, i.e. one should preform a quick format only - it seems to me that the V4 controller prefers to manage blocks it has previously written to.