something SSD you didn't know about from OCZ

[flopper]

My boss brought this drive to me a short while ago, they shipped an ES Apex sample for me to evaluate and for me to decide if it launches or not...OCZ are acutely aware V2 and V1 core was over hyped and peoples expectations were a little to high..hence the person who ended up supporting SSD the most for the company gets to decide what route future releases will take.

Vertex is a new drive, new controller, new rules (to OCZ) and its positioning will be marketed as a performance option...Apex (coreV3) had been dropped altogether till we decided to try internal raid.

Times are hard, the economy is getting worse, I need to sell drives, that is blatantly obvious...people have less expendable cash now hence you have to rethink your products. We could have followed Kingston and brought an Intel clone SDD to market costing $600 for MLC...but how many would we sell, that is the big question you have to ask?
Take Asus and ROG boards as an example, they sell 60k mid range boards compared to 400 or so ROG boards similar SKU, I know what is earning Asus the most money here

Overall we are taking a chance revisiting the JM602 controller, but we are doing it with a twist. We know anandtech will put it thru IOmeter and pronounce is crap...thing is i do NOT trust that bench or care what the review will say as you can have a good experience with a product even though the numbers from the bench say you can not. I am starting to think the whole benchmarking frenzy we see with SSD's is not really taking and showing SSD's for what they are or what they can do.

I know some will argue with me here, and for those that do come see me at CES, I may have my DIY with me which has a V1 SSD, i will let you play with it...but i will ask you to say what the experience was like, not how it bench marked.

Benchmarks dont tell the whole story.
Its easy to belive you need to run using LN2 to get the fastest computer in the world, well benchmarks say so, right?
Its just damn expensive to do so and also hire one to just stay there all day pouring ln2 cost a bunch. My computer is faster than those in a 24/7 use, my computer will run after those ln2 rigs blows up or runs out of ln2.
So, my computer then is actually faster than Fuggers and the others so called speedy system. since it be running along fine

Running benchmarks, is just fun sport, to actually feel a difference, in your daily desktop use, that is where its get tricky, does the benchmark reflect your actual use of your computer?

I run cheap ass ram, kingston value, 3gb on an i7, bad timings cl9, and my desktop use is smooth as baby ass, and much in my experience faster than my dualcore 4ghz system.
I want my desktop to be snappy and smooth.
If I press the button, I dont want to wait.
If I load up, install, and so on, it has to be snappy and fast.
And to add to this, I am getting a soundcard a essence stx for my headphones which I will also replace for a more high end quality experience which I also can measure in my use.
I am a headphone nut, not at a audiophile level but a nut.

For me, the accesstimes is the first criteria, then, how well does it run the rest, and then to add to this with cost and price/performance then SSD is the way to go.
Vertex seems for me to be a good balance between price/performance if it outperforms my Vraptor and adds a more snappy desktp use.
I just are sensitive to my desktop use, much more than the average Joe user.

Benchmarks are theoretical values, if the system as I use it, runs faster and I can subjectivly feel it, then it adds for me value to my use.
Ram dont cut it in that category.
Cheap ass ram give me the same experience as more expensive?
Sure, if I run memtest benchmarks, I can see a difference, but can I feel it?

Benchmark programs have been done for harddrives that runs in different ways than ssd does.
And I guess, someone will make a program soon that show us the difference in opening, closing applications, games and use between ssd and mech drives that more show the real usage.
I once bought the hype with ram timings, and frankly I am not impressed with theoretical data.

How fast is the data avaible and how fast can the data be read, and then if the write random are solved, the disk then is much better than the mech drives out there.
No matter the benchmarks.

I am however suprised that m$ dont have a better ssd driver out.
Usually they are in the forefront of tech.

[Lestat]

ok so now NO Apex?
cripes here i thought i was going to be able to afford an SSD finally

[Tony]

I saw a small delay in drive recognition on the ICH8m controller which I do not have on the NVidia controller. The office has not reported any raid issues to me yet also.
It took 5 secs or so for the ICh8m to see the drive, on the NV controller there is no delay at all.


The drive is seen as 1 drive, IE its not seen as a raid array and works much like the XFX raid card that needed no drivers.

To be more specific at this time is hard...

[One_Hertz]

The point you are missing is the small file write speeds on this Apex, its very fast. Now the controllers are not IO monsters but the raid controller does help. I have a feeling 2 or more of these on SB raid may be killer. i wish I had another here as i would have shown it for sure.

you guys need to get away from needing 5 gazzilion IO/s 4k random writes, you don't need that to have a good experience...Im trying to find just what you DO need to have stutter free untweaked and tweaked.
Going by what i feel this drive is performing like its not that many..maybe just 1 gazzilion...

Most of what I was talking about are the real world benchies, not necessarily IOPS. If you clicked on my link you would see that simple copy/paste tests get 3-4x faster on hardware raid versus software raid when using your SSDs. It doesn't get any more relevant than that, hence I said your SSDs perform horribly on software raid and well on hardware + cache.

Edit: don't get me wrong, I think V2s are a great product, it just has to be set up correctly. Hey Tony, if you can tell me, are V2s going to drop in price anytime soon as they are being phased out? I seriously want two more.

[cadaveca]

What sort of warranty is going to come with these, Tony? MTBF?

[Tony]

Regards price drops...V2 is not being phased out straight away but will be once volume comes to us on Apex.

warranty etc will be released on official launch, what you had here was an early leak...a holidays present if you will

[Tony]

ok so now NO Apex?
cripes here i thought i was going to be able to afford an SSD finally

big post edit there man...you ask a question then remove it once i answered you..makes my post look kinda strange.

[roofsniper]

Anyway, honest posting does not count for some it seems...maybe i should leak info like this else where and let you guys get it second hand next time?

although some don't show it i believe mostly everyone here is happy for the benefit you have made to this forum. i know i am.

im new to this whole ssd and raid business as i haven't done either of them yet . i know that ssds are really fast for small file transfers but for large file transfers hdds are actually better. what is an estimate of when an hdd would actually perform better than a ssd for large file transfers? also i read on wikipedia that ssds have a lower density than hdds. what is this supposed to mean? that lets say i have 30gb on info on my hdd but if i put that 30gb on a ssd it will actually take up more room?

[alfaunits]

you seem to be a person who needs benchmarks to prove something is good or bad, I used to be like this, i even used to play superpi things change though.

Unlike SuperPI I know the IOMeter test actually does sum up the speed for me. These are 2 totally different tests.

Average hard drive seek is 11 to 13ms, add in other latency and you can ofter see 14ms in a bench. SSD even crappy JM602 ones seek in 0.2 to 0.3ms

That's right for reads - but not writes. If the drive does 100 IOPs, that makes the seek time = 10ms.


For the Apex vs. Vertex, I was only mentioning the prices of the two are about the same in Europe for the same capacity - so that makes no sale sense. Am I missing something?

[Lestat]

sheesh i had that thing there for like 45 seconds....

Apex (coreV3) had been dropped altogether till we decided to try internal raid.

either way did i see your earlier post correctly that Apex has been put on the back burner for now, or did i misunderstand?

[Gunslinger]

Tony, with the single drive and internal raid 0, what kind of cold boot time can we expect? (power on > desktop)

[RyderOCZ]

either way did i see your earlier post correctly that Apex has been put on the back burner for now, or did i misunderstand?

You misunderstood. Until we raided the controllers, it was not going to happen (meaning we would not produce the drive)

Now we are going to produce it under the name Apex. This is what Core V3 "was" going to be, now it is not called Core V3, it is called Apex.

[NapalmV5]

alfaunits

my 4x ssd more than matches 8x adfd raptor read/write and im not referring to benches

yup 1:2

[Tony]

Tony, with the single drive and internal raid 0, what kind of cold boot time can we expect? (power on > desktop)

depends on controller and how you have the system setup...overall from boot selection screen to OS can be like 10 secs, me with tweaks 30 secs.

its all in how its set up

[Tony]

Unlike SuperPI I know the IOMeter test actually does sum up the speed for me. These are 2 totally different tests.


That's right for reads - but not writes. If the drive does 100 IOPs, that makes the seek time = 10ms.


For the Apex vs. Vertex, I was only mentioning the prices of the two are about the same in Europe for the same capacity - so that makes no sale sense. Am I missing something?

That means resellers are price gouging..Apex should be cheaper

[DFI pit bull]

Thanks Tony for the heads up on the new drives, there are very few people as helpfull as you too. I think i'll buy one when they come out. The ocz ssd's are such great value, fcuk the other ssd makers, especialy intel with their rip off prices.

[ecidious]

That just shows me the following:

1 even though i have run this drive for 1 full week in this AMD laptop you don't believe the ATTO i posted is real.

2 you don't understand how we are using a byproduct of steadystate to enhance the drives. If I could extract the caching feature and release it as a stand alone SSD enhancement i would, and i bet the world and his dog would use it with any brand and any type of drive.

I don't care about benchmarks; disk benchmarks can vary by up to 25% depending on controller, even discarding outliers due to controller incompatibilities. What I do care about is the accuracy of your company's advertisements and the lack of mentioning steadystate, which I also do not consider a viable solution. A viable solution would have been adding cache to the V1 drive's controller, like all plattered hard disks have, and releasing that as V2 but instead OCZ pushes steadystate at end users. My main rig is also pulling double duty as my home workstation and I need much more than 4gb of RAM for my 3D work in addition to a disk system that can handle both small and large writes.

[Spoiler]

I don't care about benchmarks; disk benchmarks can vary by up to 25% depending on controller, even discarding outliers due to controller incompatibilities. What I do care about is the accuracy of your company's advertisements and the lack of mentioning steadystate, which I also do not consider a viable solution.

The suggestion of using steadystate is a way of improving a particular aspect of the drive. The drive will still work without steadystate. It might not do what you want it to out of the box, but that is just the way it is with the MLC SSD's of its time (Intel's MLC SSDs came out a bit later). I doubt too many companies will advertise their new tech products with a bunch of "disclaimers". If they did, the general public would just buy the other brand that failed to include them.

A viable solution would have been adding cache to the V1 drive's controller, like all plattered hard disks have, and releasing that as V2 but instead OCZ pushes steadystate at end users.

My guess is that after the V1's, they started planning for the new controller w/ more cache. But since that takes time, it was probably easier to make small enhancements via Core V2 until the "Vertex" would be ready.

My main rig is also pulling double duty as my home workstation and I need much more than 4gb of RAM for my 3D work in addition to a disk system that can handle both small and large writes.

It sounds like the Core series are not for you then. Unless of course you want to use them with your raid controller.


Tony-

Can't wait to see how the Apex and Vertex holds up. Thanks for the info too!

[SkOrPn]

@Tony,

I would like to know if there is a limit to how many Vertex's can be raided? I have a 8 port raid card and would like to do 7 Vertex's in RAID 0 for OS/apps/games and the 8th port for a SAS or SATA backup drive. Not sure if this setup works though, especially considering I will be mixing conventional drives with ssd's on the same controller.

Also, can these SSD drives lifespan be enhanced by using in raid 0 mode? My theory is since the controller is splitting data near equally per drive does this help the drives to last longer possibly? just curious

Thanks

[chunkylover77]

Tony once again thanks for taking the time to show us this sneak peek. I think I'm going to finally make the SSD jump.

[Gogeta]

Thank you, Tony. Its great to see OCZ leading the mainstream SSD marketing push with the sort of material you've made available.

[NickK]

Hmm all this talk about internal RAID0. I suppose it's good to convey the design idea but I suspect it's actually flash bank interleaving (same as your normal RAM).

When you read/write the data is broken up at a flash read/write level. The optimum blocks size is dependent on the flash chip and the timings (as well as the operations such as erase etc) so you'll not have the ability to alter the block size (nor would you really want to).
So the best treat it as a single drive and forget about the 'Raid 0' marketing.

There will probably still be issues once you max out the drive itself but you'll probably not do that with the SATA 3GB/sec interface.

If you really need to I/O quick, it still makes more sense to plumb for 16-32GB of ram instead. However for laptops - I'm certainly interested in 120-160GB to replace my MBP internal drive..

[skycrane]

Tony,

thanks for the great info, my next build will definatly be a few SSD in raid 0

will you guys be making any smaller SSD maybe 15gb or so?
after 7 yrs or so of running 24/7, some of my 6gb HD are starting to show thier age in my crunchers in the basement. so ill need to start replacing them soon

[Stanley Pain]

The real question is if it still has that micro stutter bull happening. All the read/write speed in the world means nothing when your teeny, tiny file i/o pauses for a second.

[RyderOCZ]

The real question is if it still has that micro stutter bull happening. All the read/write speed in the world means nothing when your teeny, tiny file i/o pauses for a second.

Read the OCZ forums, we have users that are using the drives.