OCZ to come out with new VERTEX SSD,S equiped with 64MB OF CACHE starting at $129

[Freonor]

30GB for OS/WoW/A few apps. Possibly 2x30GB raid 0 if I see the need.

[Eagleclaw]

And from Cdfreaks.com

http://www.dailytech.com/article.aspx?newsid=14257

reviews coming soon from anandtech

[Tony]

Hmm... wonder why they didn't show the smaller read/writes-

The performance of those drives has massively changed with each firmware revision. That bench was right at the start of development where we just wanted to show what we predict max speed would be for a vertex 2 if we internally raided it.

Call it a working concept.

Ohh I saw over 2500IOP's recently on Anands useless IOmeter test 10GB test file if that matter to you with a prototype vertex

[Lu(ky]

Tony is there a big difference in cache size in performance? 32 vs 64 in a raid -0-?

[Tony]

Not been able to test that...sorry.

Plus i can't post performance numbers (exactly) till reviewers do.

[One_Hertz]

Ohh I saw over 2500IOP's recently on Anands useless IOmeter test 10GB test file if that matter to you with a prototype vertex

What were the settings? I am considering two vertexes along with my two Intels for more fast storage.

[Tony]

IOmeter settings?

sectors to =10GB
starting sector 1024
4k-RW ran for 2 mins

Also going from SATA controller to SATA controller makes a difference on outcome...keep that in mind

last thing regards IOmeter...you all comparing IO's from drive to drive need to understand you do NOT need 2500 or 3000 or 100000 IO's to have a fast stable stutter free system, infact less that 100 on the 4k test will be fine, in some cases less than 50.

So this IOmeter test is just turning into a bragging bench IMO...it means nothing to an average user.

[One_Hertz]

That's like saying that you don't need a 50GHZ CPU because a 10GHZ will be just fine. Faster is faster.

Although, generally, I agree. I basically care only about how quick apps/games load up and for most things, CPU limit comes first anyways. I just hate waiting for things; makes my blood boil.

[m^2]

I think that better comparison would be core count - for home use there's just no difference whether you have 4 or 400.

The thing that matters is access time when you do your regular stuff=much under 1000 IOPS.

Measuring IOPS with 1-2 queue length would be quite similar, but with one exception: it would give Vertex less time for buffer flushing.

Tony, am I right?

[One_Hertz]

I think that better comparison would be core count - for home use there's just no difference whether you have 4 or 400.

Not really. More IOPS does make things faster, while more cores will do zero. The thing is, more IOPS does reduce the time for some operations to complete from like 0.1s to 0.01s so the difference probably won't be felt, but it is still there. 300fps versus 100fps is another good comparison.

[m^2]

Not really. More IOPS does make things faster, while more cores will do zero. The thing is, more IOPS does reduce the time for some operations to complete from like 0.1s to 0.01s so the difference probably won't be felt, but it is still there. 300fps versus 100fps is another good comparison.

A 1000 drive RAID 10 will make huge IOPS and be slower than a single good SSD, for a home user of course. Above some point it doesn't reduce anything. Exactly like core count.

[One_Hertz]

A 1000 drive RAID 10 will make huge IOPS and be slower than a single good SSD, for a home user of course. Above some point it doesn't reduce anything. Exactly like core count.

Sure, because that HDD array will take the same 12ms (or however fast the drives are) to do 1 IO and several hundred IOs. If the task requires less than that several hundred IOs per 12ms then the number of max IOs doesn't matter anymore (it might as well be millions per 12ms) as the access time becomes the constraint.

You have to realize that with SSDs those IOPs limits are ~100 times higher than that of the HDDs due to the access times. I highly doubt things are truly maxed out in terms of the IOPs at the current performance numbers.

Is there a program that records the number of IOs going to the array that we can run in the background to measure how much apps really use when they load?

[FoxRacR17]

I'd be happy with a 60gb drive for OS/games if they don't stutter and if they ever come out anytime soon.

Same here. I'm planning on starting with a single 60, then down the line adding a second for raid

[Tony]

My point is you don't need 3k IOPs random write to have a fast system...many of our Core V2 and Apex users have proven this.

I said it before and I will say it again, some reviewer (as no one believes engineers) needs to sit down and find the minimum IO's needed...forget the max...tell us what the minimum is for XP...Vista and W7.

[m^2]

Is there a program that records the number of IOs going to the array that we can run in the background to measure how much apps really use when they load?

There's professional stuff to tune your arrays, but I know very little about it.

My point is you don't need 3k IOPs random write to have a fast system...many of our Core V2 and Apex users have proven this.

I said it before and I will say it again, some reviewer (as no one believes engineers) needs to sit down and find the minimum IO's needed...forget the max...tell us what the minimum is for XP...Vista and W7.

OK, so what you meant is "drive can run smooth with little IOPS".
I guess that most people here want smooth and fast. And finding minimum is pointless for this purpose, we'd rather search for maximum that helps.

And seriously, how do you expect reviewers to find out?? What experiment could a person that can't change anything with hardware / firmware do? It's OCZ and other SSD vendors who should investigate it....if the problem is well defined at all.

[FoxRacR17]

My point is you don't need 3k IOPs random write to have a fast system...many of our Core V2 and Apex users have proven this.

I said it before and I will say it again, some reviewer (as no one believes engineers) needs to sit down and find the minimum IO's needed...forget the max...tell us what the minimum is for XP...Vista and W7.

I think its kinda like this: If money was no option, would you get the Ferrari that can do 200+MPH, or the Toyota Camery. Sure the Camery is good enough to get where you want to, and you will never get to use all the potential of the Ferrari because domestic of speed limits and such, but I know I would rather have the Ferrari

[Halk]

I'm not so sure. Tony seems to be saying there's no impact, that it has no effect rather than it's got an effect that we find difficult to measure.

[One_Hertz]

My point is you don't need 3k IOPs random write to have a fast system...many of our Core V2 and Apex users have proven this.

I said it before and I will say it again, some reviewer (as no one believes engineers) needs to sit down and find the minimum IO's needed...forget the max...tell us what the minimum is for XP...Vista and W7.

Some people want to know the minimum, while I am personally more interested at the absolute maximum that I can use i.e. at which magic IOPS number games/apps/windows will no longer load ANY faster. Although there would be too many other variables...

[alfaunits]

That's a marketing way of putting it, Tony... if you needed just 100 write IOPS, stick to VelociRaptor... that point of yours makes SSDs useless

[Tony]

Its nothing to do with any of what you guys are focused on...its to do with cost and selling drives. Nothing expensive is selling in large quantity at this time, fast IO processors cost money hence drives are expensive. I need to sell drives to 100k people...NOT 100 who want to break world records etc.
So my point is totally valid...i want to build an SSD that is dirt cheap, meets specs for IOPs so windows works just fine and sell them by the bucket load all over the world.

Now we have a fast drive, from what I have just seen Vertex doing im pretty damn impressed and you will be also...but i still want a cheap drive...a $50 special if you will that is perfect for Joe Blow who is not a speed freak.

[NeedMoMegaHurtZ]

To help assess studdering, "maximum write response time" in iometer is what you want to look at, under different scenarios, %reads/writes, %random, different queue depths

I don't think you can say at "x number of iops you will/will not get studdering" - it will vary by scenario, queue depth, size, %writes, % random etc...

e.g. Mechanical hd score tiny # of iops, but their "maximum write response time" isn't 900-1000ms like some of these cheaper ssd's - they don't studder and their max write latency 5-10x less than the cheap jmicorn mlc ssd's. Same thing can be said about Mtron SLC mobi -e.g. they get only get ~20-30 iops @ 4k 100% write , but the max write latency is 50-150 under most workloads - so you don't get any perceptible studdering

But Tony is right - most desktop users shouldn't even care about iometer - usage patterns are much different

[One_Hertz]

Its nothing to do with any of what you guys are focused on...its to do with cost and selling drives. Nothing expensive is selling in large quantity at this time, fast IO processors cost money hence drives are expensive. I need to sell drives to 100k people...NOT 100 who want to break world records etc.
So my point is totally valid...i want to build an SSD that is dirt cheap, meets specs for IOPs so windows works just fine and sell them by the bucket load all over the world.

Now we have a fast drive, from what I have just seen Vertex doing im pretty damn impressed and you will be also...but i still want a cheap drive...a $50 special if you will that is perfect for Joe Blow who is not a speed freak.

This is XS? no?

[Tony]

Yes but you comment on cheap value drives as if they are high performance drives...just look at the bashing Core got on here and you will understand my point.

When OCZ release a value part you guys still think it should be good for breakin records, fact is to make money these days you have to focus away from the top end enthusiast as the market is to small until the price of the high end components comes down.

Sorry fact but true.

Its like selling top line enthusiast motherboards, 500 to 1000 pieces made total, mid range boards...40 to 50k Pieces made...
I have a feeling you guys are going to see less high end parts coming thru due to the recession.

[One_Hertz]

Yes but you comment on cheap value drives as if they are high performance drives...just look at the bashing Core got on here and you will understand my point.

When OCZ release a value part you guys still think it should be good for breakin records, fact is to make money these days you have to focus away from the top end enthusiast as the market is to small until the price of the high end components comes down.

Sorry fact but true.

Its like selling top line enthusiast motherboards, 500 to 1000 pieces made total, mid range boards...40 to 50k Pieces made...
I have a feeling you guys are going to see less high end parts coming thru due to the recession.

Why are you blaming us for thinking of them as high end drives when one of the stickies on your own forums says something along the lines of core V2 drives breaking some world records?

I agree entirely with what you are saying about making money, ESPECIALLY in the recession, but when there are admins on your own forums saying that a few core drives will beat the Intel drives (I know you stopped that now, but things like this were said in the past, and is why I bought my two core V2s before buying Intels) how do you expect us not to compare them and think of them as same segment products?

[XS Janus]

I wanna see how many will cap ich10R... that's how much I'll be getting!