Well, I've gone crazy

[Tony]

with all this testing I hope you guys aligned the partitions or you wasted your time.

vertex is an IO monster BTW but remember this is mainly cache induced, linear reads writes are not much better than Apex.

SSD testing needs a total revamp, none of the common HDD tests tell the truth OR give an end user any idea about whether a cheap drive will work ok on their system or they need a more expensive drive.


Here is a question for ya'll

What is the minimum IOP's you need 4K random write 10GB test file for 2 mins for a drive to work without causing the sata controller to pause and cause a stutter?

Not 1 review on a forum or a website yet has proven what is needed and what is not...lets see if you guys can actually come up with the answer

[One_Hertz]

with all this testing I hope you guys aligned the partitions or you wasted your time.

vertex is an IO monster BTW but remember this is mainly cache induced, linear reads writes are not much better than Apex.

SSD testing needs a total revamp, none of the common HDD tests tell the truth OR give an end user any idea about whether a cheap drive will work ok on their system or they need a more expensive drive.


Here is a question for ya'll

What is the minimum IOP's you need 4K random write 10GB test file for 2 mins for a drive to work without causing the sata controller to pause and cause a stutter?

Not 1 review on a forum or a website yet has proven what is needed and what is not...lets see if you guys can actually come up with the answer

I did a sector to sector copy when moving between drive setups so everything was EXACTLY the same.

That's a good question on the minimum IOPS . It can be tested for but somebody is going to have to sit down a spend weeks doing that. If only I didn't have full time work and part time university . There is definitely a number of IOPS past which it just doesn't help any more. Same with STR. I know my X25-Es are well beyond the helpful IOPS number.

[Brahmzy]

with all this testing I hope you guys aligned the partitions or you wasted your time.

vertex is an IO monster BTW but remember this is mainly cache induced, linear reads writes are not much better than Apex.

SSD testing needs a total revamp, none of the common HDD tests tell the truth OR give an end user any idea about whether a cheap drive will work ok on their system or they need a more expensive drive.


Here is a question for ya'll

What is the minimum IOP's you need 4K random write 10GB test file for 2 mins for a drive to work without causing the sata controller to pause and cause a stutter?

Not 1 review on a forum or a website yet has proven what is needed and what is not...lets see if you guys can actually come up with the answer

Exactly. This will get most SSDs... regardless of cache. I have a 13GB backup folder which I copy between multiple machines weekly at home.
It has tons of small .jpeg's, tons of large .jpeg's, huge 700mb single files, and a collection of everything in between.

My Velociraptors would chew this thing up without thinking about it. Amazing write speed - as fast as my other WD6400AAKS could send it. When I went to copy this huge folder onto my new Titan SSD RAID0 array, it actually handled it amazingly well, but I did experience quite a few half-second pauses while writing. I use robocopy which shows me EXACTLY where the SSDs slow down during the write process. It's always when it gets to a folder with a ton of tiny little .jpeg's or similar.

The folder typically took just about 3 minutes to write to my 3 Velociraptors and it took about 4 1/2 minutes to write to the 2 Titans. Not bad, but clearly slower. There comes a point where, depending on what types of files are being written and how much of them, where cache or no cache doesn't matter. Cache gets saturated when the write requests keep coming after many seconds and certainly minutes. It fills up and then still has to wait on the SSD(s) to write it all. Now day to day multitasking? Yeah, cache is the answer.

Now I rarely ever write that folder to the SSDs - only during a system rebuild. It is always READ from the SSDs which is ridiculously fast (faster than my VR's could ever dream) for LAN backups.

[NapalmV5]

with all this testing I hope you guys aligned the partitions or you wasted your time.

vertex is an IO monster BTW but remember this is mainly cache induced, linear reads writes are not much better than Apex.

SSD testing needs a total revamp, none of the common HDD tests tell the truth OR give an end user any idea about whether a cheap drive will work ok on their system or they need a more expensive drive.


Here is a question for ya'll

What is the minimum IOP's you need 4K random write 10GB test file for 2 mins for a drive to work without causing the sata controller to pause and cause a stutter?

Not 1 review on a forum or a website yet has proven what is needed and what is not...lets see if you guys can actually come up with the answer

thank you for bringing more common sense to the subforum.. it has gone randomly/ioply mad

we need ssd friendly controllers/benchmarks/ultimately os/apps

[NeedMoMegaHurtZ]

Here is a question for ya'll

What is the minimum IOP's you need 4K random write 10GB test file for 2 mins for a drive to work without causing the sata controller to pause and cause a stutter?

Not 1 review on a forum or a website yet has proven what is needed and what is not...lets see if you guys can actually come up with the answer

Can you define a "stutter". How do you measure it? Is it a subjective experience for the user or can you objectively measure it? Would that "threshold value" be different for different users?

I think there is a high correlation from what many would describe as a "stutter" and IOmeter's reported "maximum write response time". But there I don't think there is necessarily a linear relationship between IOPS and "stutter". I don't think you can say "at this IOPS, you get stutter" . For example you can use a regular HD, and have very low 4K random write IOPS, but it's not the same perceptible experience as "stutter" on a SSD.

I think best tests are the ones that test real usage patterns/conditions

[flopper]

Not 1 review on a forum or a website yet has proven what is needed and what is not...lets see if you guys can actually come up with the answer

Your actually thinking that highly of review sites and reviewers?
I use salt, a pinch of finest highland malt as there is no way they know what they do.

Even a prominent site as benchmark review havent got to a good benchmark yet. And those guys should know the in and outs of harddrives.

just smack 1gb cache on your ssd´s and they run just fine.
1gb ram nowadays isnt that much in cost.
Much easier solution instead of using limited controllers.
Who ever made that decision to use jmicron should be shot.
It was a bad business decison.
IMO.

Since ssd are getting raided in all sorts of ways, we will see as your own tech simulation shows, a speed increase where its a must to get raided cards as sata 2 and then sata 3 is not enough for millions of users.
The train is about to leave the station and what do we get? sata 3 which is to slow for us.
If someone would be smart out there, they be doing such solution, providing ssd with pci-e out of the box.
ah yes, called fusion io drive.

The standards is to slow to adopt to the market.
SSD will simply explode this year and next, and what do people say when simply, this drive isnt any good to buy (vertex 2) since its limited by my motherboard?

Ah well, I get a raid card then

[m^2]

Since ssd are getting raided in all sorts of ways, we will see as your own tech simulation shows, a speed increase where its a must to get raided cards as sata 2 and then sata 3 is not enough for millions of users.
The train is about to leave the station and what do we get? sata 3 which is to slow for us.
If someone would be smart out there, they be doing such solution, providing ssd with pci-e out of the box.
ah yes, called fusion io drive.

Micron and RunCore are on the way.

[BenchZowner]

Your actually thinking that highly of review sites and reviewers?
I use salt, a pinch of finest highland malt as there is no way they know what they do.

Even a prominent site as benchmark review havent got to a good benchmark yet. And those guys should know the in and outs of harddrives.

If I were you I'd take that back.
There are several reviewers ( not necessarily working on the well-known sites ) that know lots of stuff, and some of them are even working in the industry and/or engineering labs and are college graduates.
I can name at least 5 of them, but for various reasons I won't.

Friendly,
Bill

[Tony]

Your actually thinking that highly of review sites and reviewers?
I use salt, a pinch of finest highland malt as there is no way they know what they do.

Even a prominent site as benchmark review havent got to a good benchmark yet. And those guys should know the in and outs of harddrives.

just smack 1gb cache on your ssd´s and they run just fine.
1gb ram nowadays isnt that much in cost.
Much easier solution instead of using limited controllers.
Who ever made that decision to use jmicron should be shot.
It was a bad business decison.
IMO.

Since ssd are getting raided in all sorts of ways, we will see as your own tech simulation shows, a speed increase where its a must to get raided cards as sata 2 and then sata 3 is not enough for millions of users.
The train is about to leave the station and what do we get? sata 3 which is to slow for us.
If someone would be smart out there, they be doing such solution, providing ssd with pci-e out of the box.
ah yes, called fusion io drive.

The standards is to slow to adopt to the market.
SSD will simply explode this year and next, and what do people say when simply, this drive isnt any good to buy (vertex 2) since its limited by my motherboard?

Ah well, I get a raid card then

1GB cache...performance Vs cost would mean no one would buy the drives...plus its the controller that dictates the cache and i don't know of one that will do that high yet.

I was seeing 1XXX+ IOP's on vertex on the infamous AT 4k bench, thats well enough for day to fay use

[Chosen.]

Here is a question for ya'll

What is the minimum IOP's you need 4K random write 10GB test file for 2 mins for a drive to work without causing the sata controller to pause and cause a stutter?

Not 1 review on a forum or a website yet has proven what is needed and what is not...lets see if you guys can actually come up with the answer

Is this you or the OCZ R&D team asking? Because after that triple failure (Core, V2, Solid Series) I wouldn't be too surprised if OCZ was asking forum users for answers they couldn't find on their own