SSD Update: Vertex gets faster

[Denithor]

AT article

In short -

4k random writes go up nearly 3x (2.4mb/s to 6.5mb/s)
4k random reads go up 9% (32mb/s to 35mb/s)
2m sequential writes go up 45% (93mb/s to 135mb/s - nearly 2x the Intel speed)

Also, SuperTalent is launching an SSD drive featuring the Indilinx controller (UltraDrive ME) that appears to use the same 1275 firmware as the improved OCZ Vertex (based on very similar 4k random writes). This trend should continue, according to Anand:

I spoke to three more manufacturers evaluating the Indilinx Barefoot controller during my time out in the bay area. The only complaints I?ve heard are that Indilinx is a very small company and the Barefoot controller is perhaps priced a little too high. With pressure from additional manufacturers, and as volumes go up, we should see these drives get even more affordable.

 

[n7]

Good news overall.

 

[amddude]

I want to see 2 vertex in raid 0 against the x25-e.

 

[magreen]

Great news that this 1275 firmware release candidate is so promising. Now let's just give 3-6 months time to OCZ to validate this latest firmware... oops, I mean to the consumer guinea pigs to discover all the bugs... and it will be a great product!

Seriously, I am excited about this, but would never be an early adopter of a product with such a wing-and-a-prayer development cycle.

 

[Denithor]

I'm seriously considering becoming a "consumer guinea pig" as you so quaintly phrased it.



I think I can string together an eBay/PayPal/mwave deal to get the 60GB model for about $160, making it about $2.67/GB (reasonable for a fast SSD).

 

[taltamir]

i agree with, well, all the sentiments... it is very impressive speed wise and price wise, and very scary at the same time, it is tempting, and yet it is worrisome that the customers are guinea pigs and that there are already bugs and issues being found...

 

[Cookie Monster]

Its impressive indeed, but cant they have done this without claiming "retail" status of the product??

 

[Denithor]

Originally posted by: Cookie Monster
Its impressive indeed, but cant they have done this without claiming "retail" status of the product??

How - exactly - would you sell a beta SSD to consumers?

Good luck with that... :roll:

 

[usernamereserved]

Originally posted by: Denithor

Originally posted by: Cookie Monster
Its impressive indeed, but cant they have done this without claiming "retail" status of the product??

How - exactly - would you sell a beta SSD to consumers?

Good luck with that... :roll:

It is just some folks find it hard to admit that something is good and have to find another way of knocking something.

noone would buy a motherboard if it did not have an upgradeable BIOS
noone would buy a graphics card if it did not have upgradeable drivers
noone would buy a BD writer if it did not have upgradeable firmware

So anyone that buys a retail product is a guinea pig using this analogy.

 

[Cookie Monster]

Originally posted by: Denithor

Originally posted by: Cookie Monster
Its impressive indeed, but cant they have done this without claiming "retail" status of the product??

How - exactly - would you sell a beta SSD to consumers?

Good luck with that... :roll:

Hm you got my point wrong. I never said that a "beta" SSDs would be sold to consumers have I?

Instead they should have done more thorough inhouse testings before releasing these SSDs as "final" retail product that you dont have to read through multiple "tips and tricks" and upgrade the firmware every now and then to squeeze the performance out of a product that wasnt performing to its full potential during release.

This is what sets intel apart from the rest in this market. The former knows what they are doing when it comes to SSDs, while the rest is beginning to do their homework.

 

[imported_Scoop]

Originally posted by: usernamereserved

Originally posted by: Denithor

Originally posted by: Cookie Monster
Its impressive indeed, but cant they have done this without claiming "retail" status of the product??

How - exactly - would you sell a beta SSD to consumers?

Good luck with that... :roll:

It is just some folks find it hard to admit that something is good and have to find another way of knocking something.

noone would buy a motherboard if it did not have an upgradeable BIOS
noone would buy a graphics card if it did not have upgradeable drivers
noone would buy a BD writer if it did not have upgradeable firmware

So anyone that buys a retail product is a guinea pig using this analogy.

Well I sure as hell don't buy an MB without a proven BIOS. And I'm still using catalyst 8.2 drivers

 

[alcoholbob]

Technically it is a beta...

The current "e-tail" firmware in majority of stock is 1199, which is faster than 1275 but has bugs, which is why indilinix rushed out 1275.

It's not an easy drive to work with...it doesn't like pagefile (will stutter in some cases, recommended you turn pagefile off since writes can saturate its bandwidth), OS installation can be a real chore depending on choice of OS and installation method.

Flashing the drive is a bit of a chore too, its easy to do but annoying given the juggling thats required...I have the Vertex in a 3.5" bay...had to open the case from the far side, pull it out, uncrew/unmount...put in a jumper...boot into bios, flash into windows...remove jumper, remount...then the joy of deleting the partition, re-aligning the drive, creating new partition at 4096k, and then re-installing the OS.

Then reinstalling EVERYTHING else you need to get up and running. Just this task alone will take you 3+ hours. I spent the better part of 18+ hours of my weekend juggling the drives ideosyncracies before it finally worked without noticeable hiccups. And its sad that this drive can't even handle a pagefile properly when the random writes start to pile up.

The best part is, the drive (120G variant) dropped in price by about $100 one week into retail availability, and it had nothing to do with any incoming competition. This is why you realize there are major problems with your product.

 

[taltamir]

Originally posted by: Cookie Monster

Originally posted by: Denithor

Originally posted by: Cookie Monster
Its impressive indeed, but cant they have done this without claiming "retail" status of the product??

How - exactly - would you sell a beta SSD to consumers?

Good luck with that... :roll:

Hm you got my point wrong. I never said that a "beta" SSDs would be sold to consumers have I?

Instead they should have done more thorough inhouse testings before releasing these SSDs as "final" retail product that you dont have to read through multiple "tips and tricks" and upgrade the firmware every now and then to squeeze the performance out of a product that wasnt performing to its full potential during release.

This is what sets intel apart from the rest in this market. The former knows what they are doing when it comes to SSDs, while the rest is beginning to do their homework.

correction, the rest let consumers and real experts like anand do their homework for them...and then they "correct" their answers to match.

anandtech should charge companies for their services

 

[Idontcare]

Originally posted by: amddude
I want to see 2 vertex in raid 0 against the x25-e.

Me likes raid-0 anything but why do it for comparison to the SLC silly spendy X25-E?

Comparing a raid-0 vertex to the MLC X25-M makes much more sense IMO, similiar price/GB marketspace and total GB marketspace.

 

[sgrinavi]

I just ordered three of vertex 120's - lets see how these things run!

 

[Hacp]

Interesting. With new competition, I hope to see indilinux drives to reach jdmicron prices!

 

[alcoholbob]

Originally posted by: Idontcare

Originally posted by: amddude
I want to see 2 vertex in raid 0 against the x25-e.

Me likes raid-0 anything but why do it for comparison to the SLC silly spendy X25-E?

Comparing a raid-0 vertex to the MLC X25-M makes much more sense IMO, similiar price/GB marketspace and total GB marketspace.

Gah, I don't think its a good comparison at all. Vertex already wins half the synthetic benchmarks with the X25-M, in RAID it will run all over it. RAID will simply increase throughput whereas the real reason why Anand calls it (Intel drives) the head of the pack is real-world performance. It's random I/O performance is off the charts. For example, the Vertex I own, depending on the app, can stutter with a pagefile enabled since the random writes are just totally saturating the Vertex. A regular mechanical HDD, even 4200RPM, would easily be able to shrug this off. I feel *safer* with a pagefile enabled, but the Vertex just can't handle it. I may end up upgrading to 12GB of RAM later to offset this.

 

[taltamir]

so what? you compare things based on COST not based on performance... and the x25-M is similar in cost...
you THINK two vertex in raid will run all over the intel x25-M... a benchmark will verify if your hunch is correct or not.

 

[alcoholbob]

What are you comparing? 2x60s vs the 160?

Does Raid really improve random writes?

 

[Idontcare]

Originally posted by: Astrallite
Does Raid really improve random writes?

http://it.anandtech.com/IT/showdoc.aspx?i=3532&p=6

 

[alcoholbob]

Assuming equal RAID (and LINEAR) scaling you'd need 12 Vertex drives in RAID to match a single X25-M in random writes. You'd get about 6 x 30GB Vertexes (180GB) at the same price of the new X25-M (160GB). Of course I'm pretty sure there are drawbacks to so much parallel processing; just look at multi-GPU.

 

[Idontcare]

Which delivers better value to the consumer - (1) $600 graphics card that generates 300fps min, or (2) $200 graphics card that generates 100fps min?

(assuming all other things are equal, AA, screen resolution, etc)

No one is arguing (that I can tell) that raid-0 is the solution to creating equal 4KB random write bandwidth given enough vertex drives merely to have equivalency to a X25-M.

But we do enjoy humoring ourselves with discussions on the merits of using raid-0 with vertex drives to create a disk subsystem that might have enough random write bandwidth to meet a typical user's needs.

It might not be feasible once $/GB, etc, are all factored in, but the comparison itself is not without justification or merit.

 

[taltamir]

Originally posted by: Astrallite
What are you comparing? 2x60s vs the 160?

Does Raid really improve random writes?

no, 2x60GB vertex (215$ each) or maybe 3x30GB would be better (100$ each) vs a single intel X25-M 80GB (360$ each). the 160GB one is 700$.

3x30GB vertex is about the same GB and should perform a lot better than a single intel X25 80GB.

 

[taltamir]

Originally posted by: Idontcare
Which delivers better value to the consumer - (1) $600 graphics card that generates 300fps min, or (2) $200 graphics card that generates 100fps min?

(assuming all other things are equal, AA, screen resolution, etc)

No one is arguing (that I can tell) that raid-0 is the solution to creating equal 4KB random write bandwidth given enough vertex drives merely to have equivalency to a X25-M.

But we do enjoy humoring ourselves with discussions on the merits of using raid-0 with vertex drives to create a disk subsystem that might have enough random write bandwidth to meet a typical user's needs.

It might not be feasible once $/GB, etc, are all factored in, but the comparison itself is not without justification or merit.

yap, we need enough to prevent stutter.

 

[alcoholbob]

Originally posted by: Idontcare
Which delivers better value to the consumer - (1) $600 graphics card that generates 300fps min, or (2) $200 graphics card that generates 100fps min?

Playing Quake 3?

The answer is 100fps if only because at 300fps the capacitors are going to whine so much you'll want to throw your case in a closet and dynamat it.