OCZ Vertex

[ochadd]

Seems the Vertex's cache eliminated the rest of the technical reasons not to have an SSD as a boot drive. Are there any performance reasons not to upgrade if the size is acceptable?

 

[coolVariable]

long term performance might be an issue with these ... but there will always be an issue.
Otherwise, the Vertex seem to give the intel a run for their money.

 

[Idontcare]

I'm with coolVariable, if the price/performance makes sense to the buyer then the technology itself appears to be order of magnitude superior to prior SSD generations and is in a league second only Intel (for performance) at the moment and probably second to no one for price/performance and price/GB metrics.

 

[Hauk]

Be sure to check OCZ forums, nice tips and tweaks to be found. There's a firmware update that early adopters are doing; new firmware is shipping with fresh Vertex's. May be worth waiting a couple weeks to insure you get one with 1199 firmware.

Faster in many tests than Velociraptors in raid 0. That's solid performance for the money from a single SSD. Despite some reports, all models have 64MB cache. :thumbsup:

 

[ochadd]

Looks like the latest anandtech ssd article confirms my suspicions that the Vertex should rock. Wish I had the $ for an X25-M

 

[magreen]

Yep, today's article by Anand confirms the Vertex is the only other MLC ssd worth buying besides intel's. It's slower than intel's but cheaper, and a good bang for the buck and Anand recommended . It seems to have no scenarios in which it performs worse than a velociraptor, unlike all the other mlc's out there besides intel's.

 

[Andrmgic]

For what it's worth, I ordered a 30gb Vertex this past sunday and it arrived with FW 1199 already on it.

This was from newegg.

I'm quite impressed with the difference over my stock notebook hdd.. sadly, I bought it for my desktop, though.. so my macbook will have to say goodbye to the Vertex for now haha.

 

[EarthwormJim]

Really excited that there is now an MLC alternative to Intel's drive.

I'm also glad that I have held out on getting a substandard ssd.

Now if only the price would get down to $2/gb.

 

[taltamir]

Originally posted by: ochadd
Seems the Vertex's cache eliminated the rest of the technical reasons not to have an SSD as a boot drive. Are there any performance reasons not to upgrade if the size is acceptable?

it wasn't the cache, it was a revised firmware. Read the anandtech article, it explains in depth what happened with the vertex and how OCZ replied to the concerned raised by reviewers by drastically altering the firmware to eliminate the SSD issues at the cost of making it harder to market to the uninformed.

If you don't wanna read all 31 pages of the mega article... you can skip to page 18 where the vertex is discussed:
http://www.anandtech.com/stora...owdoc.aspx?i=3531&p=18

But that will make little sense without a through understanding of SSD and their problems (as explained in pages 1-17)

 

[SickBeast]

OP, if you're grasping at straws, a Raptor has a higher sustained write speed than the Vertex.

I suppose there is also the durability issue, but really, SSDs are more durable in so many other ways.

 

[taltamir]

http://www.anandtech.com/stora...owdoc.aspx?i=3531&p=24

For sequential it is:

vertex:
250 MB/s read
93 MB/s write

Veloci:
118 read
118.9 write.

Please don't forget the random 4K write and read speed, as is listed here, it is MORE important than sequential speed!:

http://www.anandtech.com/stora...owdoc.aspx?i=3531&p=25

Vertes:
32 MB/s read
2.41 MB/s write

veloci:
0.55 MB/s read
0.63 MB/s write

The vertex is the ONLY SSD drive aside from the intel which beats the velociraptor:
1. Across the board
2. In the random write (the most important test)

Every other SSD but those two loses to the raptor in both of these metrics, but has faster sequential writes, which would only matter if you use it NOT as an OS drive / game drive / app drive... but as a bulk storage drive for things like DVD images... due to cost nobody without any exception does that, everyone uses it as OS drive. So random write is the MOST important metric.


PS. that is the info with the new firmware... which is very new... so a little waiting to see if there are any bugs in it are prudent.

 

[alcoholbob]

Originally posted by: ochadd
Seems the Vertex's cache eliminated the rest of the technical reasons not to have an SSD as a boot drive. Are there any performance reasons not to upgrade if the size is acceptable?

Not performance reason but you will need to jump through hoops. The warranty is lower than Intel, and out-the-box performance isn't close.

Originally posted by: taltamir
Vertes:
32 MB/s read
2.41 MB/s write

veloci:
0.55 MB/s read
0.63 MB/s write

The vertex is the ONLY SSD drive aside from the intel which beats the velociraptor:
1. Across the board
2. In the random write (the most important test)

Heh, the X-25M is TEN times faster than the Vertex in the regard of random writes. Looks like Intel got it right. They went for real world performance instead of pure bandwidth for marketing performance.

Out of curiosity, is texture loading sequential read or random read for games? My "guess" is random read although the size of the textures are huge, so it's easy to assume sequential.

 

[taltamir]

yes, intel did get it right, intel is the best... but notice that the vertex is 3.8x faster in random writes than a velociraptor... that is a big deal because the veloci is more than THIRTY times faster in random writes than most SSD drives.

Typical SSD drive:
0.02 MB/s random write... samsung new controller almost 30x faster ... regular drive about 1.2x faster... velociraptor about 1.2x faster still... vertex about 3.8x times faster then veloci, intel about 10x times faster then vertex.

Texture reading is a mix of sequential and random read, but mostly sequential reads... the time it takes for windows to boot or for a game level to load is much faster even on crappy controller SSDs... random writes account for system stutter...
When you send a SINGLE AIM message the drive writes it to your chat log... that is a small write, it takes the minimum size of 4kb... WINDOWS FREEZES FOR THE DURATION OF THE WRITE! causing stutter, you system just "locks" for a second or two all the time during SSD use unless you use the beta firmware on the vertex or the intel X25-M... the samsung might do well but it is slower then the velociraptor so why bother...

Games also often write small data to the drive, so it will also stutter.

 

[ochadd]

The article seems to push the idea that RAID setups can't overcome the inherent deficiencies for the small writes. My take is that as long as they are fast enough to avoid stutters the experience is only going to improve with higher transfer rates. If this is the case it would seem two 30GB Vertex drives in RAID 0 buys you a better experience than a single X25-M?

Does a 430MBps transfer speed from two Vertex drives give you a better experience than an X25-M and it's 10x improvement in small writes? Stutters can't be allowed to happen imho, I would go nuts and destroy a peripheral.

Yes. I am trying to justify buying a couple of these bad mothers. I'm also an OCZ fan as thier support folks are second to none in my experience.

 

[usernamereserved]

the anandtech review used an old Vertex firmware
Vertex seq-write is now 150+MB/s
4K random writes on Vertex is 2500 IOP compared with Intel 10,000 IOP, that does not make the Intel 10x faster.
4K random write speed on Vertex is 12+ MB/s not 2.41MB/s

3x 30GB Vertex is the same price as 1x Intel 80GB and just blows the Intel away
I am planning on 3x Vertex in RAID 0 and will retire my Intel for my laptop.

@taltamir
FYI there is no beta firmware for Vertex but anandtech used an old firmware which makes the results out of date before the review was live

 

[alcoholbob]

What do you mean by old?

Anandtech used the original retail firmware as an example for how bad the Vertex used to be, but used the 1199 for its benchmarking, which is up to date and is the "revised" retail firmware that came out 2 days ago.

Unless you are referring to the 1275 firmware? It's not even posted on any stickied threads, IIRC the forum mods are just emailing it to forum members that request it.

 

[taltamir]

Originally posted by: usernamereserved
the anandtech review used an old Vertex firmware
Vertex seq-write is now 150+MB/s
4K random writes on Vertex is 2500 IOP compared with Intel 10,000 IOP, that does not make the Intel 10x faster.
4K random write speed on Vertex is 12+ MB/s not 2.41MB/s

3x 30GB Vertex is the same price as 1x Intel 80GB and just blows the Intel away
I am planning on 3x Vertex in RAID 0 and will retire my Intel for my laptop.

@taltamir
FYI there is no beta firmware for Vertex but anandtech used an old firmware which makes the results out of date before the review was live

Anandtech SPECIFICALLY told OCZ that their firmware is shit, OCZ went and made a new firmware fixing the problem OCZ said, sacrificing sequential speed, LOWERING it to 90 MB/s but fixing the stutter problem.

The Anything higher then 90 MB/s is using the old firmware, not vice versa. That is because the extra MB/s in sequential writes is not worth the complete and utter crap and unuseability of a stuttering system.

Do yourself a favor and actually READ the article before bashing it.

 

[taltamir]

Originally posted by: Astrallite
What do you mean by old?

Anandtech used the original retail firmware as an example for how bad the Vertex used to be, but used the 1199 for its benchmarking, which is up to date and is the "revised" retail firmware that came out 2 days ago.

Unless you are referring to the 1275 firmware? It's not even posted on any stickied threads, IIRC the forum mods are just emailing it to forum members that request it.

no, he simply did not read the article, he figured that the anandtech second test which I linked (the one with the new firmware that fixes stutters at the cost of sequential read) is older because its sequential read is lower... and if the number is lower is must be older and not as good.

 

[usernamereserved]

RE: Which firmware on Vertex ? by Anand Lal Shimpi, an hour ago
I tested with the shipping firmware for this article (0122). I've been playing around with 1199 in the lab and will most likely have an update in a couple of weeks once I've done a thorough evaluation of it. By then I should also have the final version of the new Samsung drive and maybe even some other interesting things.

For now, I've got to get to work on the new Mac Pro and the updated Ion article I need a small break from SSDs por favor

Take care,
Anand

FYI
1199 firmware is the latest with 150+MB/s and the same IOP performance as the shipping firmware 0122 which had the 90+MB/s seq-write speed.
I DO READ, maybe you should too

 

[Andrmgic]

1275 should be out in a few days, apparently.

 

[IntelUser2000]

Originally posted by: Astrallite

Originally posted by: ochadd
Seems the Vertex's cache eliminated the rest of the technical reasons not to have an SSD as a boot drive. Are there any performance reasons not to upgrade if the size is acceptable?

Not performance reason but you will need to jump through hoops. The warranty is lower than Intel, and out-the-box performance isn't close.

Originally posted by: taltamir
Vertes:
32 MB/s read
2.41 MB/s write

veloci:
0.55 MB/s read
0.63 MB/s write

The vertex is the ONLY SSD drive aside from the intel which beats the velociraptor:
1. Across the board
2. In the random write (the most important test)

Heh, the X-25M is TEN times faster than the Vertex in the regard of random writes. Looks like Intel got it right. They went for real world performance instead of pure bandwidth for marketing performance.

Out of curiosity, is texture loading sequential read or random read for games? My "guess" is random read although the size of the textures are huge, so it's easy to assume sequential.

Sigh, you don't need anything more than 0.5MB/s in random writes for consumers. The problem is that the more demanding workstation users and server users comment on the very same drives meant for consumer sector.

0.5MB/s in random writes is what 4200RPM drives achieve. You don't notice them stuttering do you?

Anandtech is doing a favor saying to us that kind of throughput is fine for the consumer sector. If you want to misinform and say you need 10-20MB/s for home usage, please do not say ANYTHING.

 

[alcoholbob]

Well unlike you, I'm not entirely satisfied running a 4200RPM hard drive. I shouldn't talk? Enjoy your 486.

 

[Idontcare]

Originally posted by: IntelUser2000
Sigh, you don't need anything more than 0.5MB/s in random writes for consumers. The problem is that the more demanding workstation users and server users comment on the very same drives meant for consumer sector.

0.5MB/s in random writes is what 4200RPM drives achieve. You don't notice them stuttering do you?

Anandtech is doing a favor saying to us that kind of throughput is fine for the consumer sector. If you want to misinform and say you need 10-20MB/s for home usage, please do not say ANYTHING.

It's not clear to me what you are basing these statements on.

As Anandtech testing shows it is not solely the bandwidth that determines the user experience (bandwidth is part of it but does not tell the full story) as stuttering takes your 2.4MB/s bandwidth on Vertex and throttles it down to nearly zero for a second while the drive stalls to take care of business with its 0.5-1.0s latency...this you would notice.

However latency does factor into bandwidth, you can't have 2MB/s bandwidth with 4KB random writes if your random write latency is 10ms. The drive simply can't retire enough 4KB random write requests fast enough to hit a 2MB/s bandwidth in that case.

So in a great world where latency was mostly a static attribute of a drive as a function of usage history (which is the case for spindle drives and Intel SSD) then the small file bandwidth tells us all the story we need to see. But if a drive is prone to sporadic latency spikes that reach into the 100ms and higher regime then those bandwidth numbers are meaningless as they fail to capture the typical/worst-case situations where bandwidth is drastically reduced as the latency spikes and the drive stutters.

(this is why standard deviation of the data used to generate an average is usually computed and reported in most technical journals, no surprise it is absent in enthusiast review sites though)

 

[MikhailT]

Originally posted by: Astrallite
What do you mean by old?

Anandtech used the original retail firmware as an example for how bad the Vertex used to be, but used the 1199 for its benchmarking, which is up to date and is the "revised" retail firmware that came out 2 days ago.

Unless you are referring to the 1275 firmware? It's not even posted on any stickied threads, IIRC the forum mods are just emailing it to forum members that request it.

Nope, Anand cleared this up in the comments.

I tested with the shipping firmware for this article (0122). I've been playing around with 1199 in the lab and will most likely have an update in a couple of weeks once I've done a thorough evaluation of it. By then I should also have the final version of the new Samsung drive and maybe even some other interesting things.

For now, I've got to get to work on the new Mac Pro and the updated Ion article I need a small break from SSDs por favor

Take care,
Anand

 

[IntelUser2000]

Originally posted by: Astrallite
Well unlike you, I'm not entirely satisfied running a 4200RPM hard drive. I shouldn't talk? Enjoy your 486.

You really don't get it do you?

Seagate Momentus 5400.6 0.81 MB/s
Western Digital Caviar SE16 1.26 MB/s
Western Digital VelociRaptor 1.63 MB/s

The Samsung SLC most people don't notice stutter gets 0.5MB/s random writes in that "used" portion.

And I use the X25-M, thank you. Used to use the first generation Raptor 36GB. IMO they are still not fast enough.

So in a great world where latency was mostly a static attribute of a drive as a function of usage history (which is the case for spindle drives and Intel SSD) then the small file bandwidth tells us all the story we need to see.

Yea, people are concentrating too much on IOPS and bandwidth, which are the same thing. Bandwidth-wise you don't need anything more than 0.5 MB/s. But say your random writes are 100MB/s but it hits a brick wall, for example runs out of buffer, you'll still stutter.

You need balance of everything. The unbalanced read-writes are what makes SSDs stutter and the newer ones are trying to mitigate. Focusing on IOPS/bandwidth alone is merely moving your problem from one place to another.