Is OCZ "Agility 2" fastest SSD in 60GB range?

[poohbear]

Hey all, i think im finally gonna take the plunge and buy an SSD for my WIN7 OS and basic small programs (games will still hafta go on my WD Caviar Black 1TB as my Steam games folder is 200gb ). I can only afford SSDs in the 60GB range, i.e. ~$170-$190 range.

Is the OCZ Agility 2 60GB the fastest SSD in this price range? (from the benchmarks it sure looks like it) Any other drives worth considering? cheers.

 

[tweakboy]

Yes it is. You will get 250mbps maybe or a bit less. gg and gb and tc

 

[FishAk]

The vertex 2 is the one with the faster random writes. Don't know if you'd really notice that much difference.

My understanding is that games load sequentially, for the most part, so having them on your 1TB Black should be fine. Especially if you make a 250GB partition to keep them confined to the outside edge of your disk. Some of them might have a cache for player stats or whatever, that might benefit from being on the SSD.

Oh!, and GG, Gb, PA, Hp, Tr, and RR. Tc too.

 

[poohbear]

Oh!, and GG, Gb, PA, Hp, Tr, and RR. Tc too.

thanks, what does this mean?

oh & i found it for $145+tax after $25 MIR shipped to my door, that's a good price yea?

 

[mutz]

SF-1200 is the same SF-1200 for all drives,
OWC uses them and so the LE, patriot Inferno, RunCore ProV, Adata S599, Corsair Force L100 and few more,
seems the speed differs between firmwares and minor changes might be as well due to different PCB design and memory chips,
earlier firmware release do happen to work faster with compression rather then older on yet compensate on stability, so in general it's what gives you the best buck for the bang..
OWC for example gives 5 years warranty, OCZ gives 3,
at normal use your SSD could last 80 years or so,
so again, each manufacturer (or composer) of the same SSD gives his cake with different flavor,
it's the one which catches your eye which you take ,

you might as well want to consider firmware support for these drive (which is open for OEM's to change),
you can read here for some more info.

E:
p.s -

oh & i found it for $145+tax after $25 MIR shipped to my door, that's a good price yea?

for 60GB that's a great price, be sure to understand, the drive is using only 7% over provisioning which takes away some of it's life,
of course with normal usage 20GB a day, that should be fine.

 

[FishAk]

It seems Benchmark Reviews and AnandTech don't agree that SF-1200 is the same SF-1200 for all drives.

According to AnandTech, earlier SF releases used firmware closer to SF 1500, which Corsair changed on the Force series, to fix a bug, by disabling low-level power states. SF came up with their own fix for the power issue, and distributed it with it's Mass Production SF-1200 FW. However, they gave OCZ a MP SF-1200 FW that still performs like an SF-1500, which OCZ uses in the Vertex 2.

By the way mutz, How do you get your links to go into the little blue "here" text? That's pretty cool.

thanks, what does this mean?

Oh!, and GG, Gb, PA, Hp, Tr, and RR. Tc too.

That would be: Good Grief, Good boy, Praise Alla, Hot potato, Take a rest, Rock & Roll, and Transmission closed. Keep up poobear, you gotta keep up.

TTFN, IBB

Edited to try mutz's cool hyperlink

 

[poohbear]

Thanks for explaining everything guys!!

yea i read the vertex 2 has an uncapped firmware due to an exclusive deal between OCZ & Sandforce, so im going that route since its only $10 more.

FishAK thanks for explaining all the acronyms, thought they were SSD codes or something that i was missing out on.

 

[mutz]

By the way mutz, How do you get your links to go into the little blue "here" text? That's pretty cool.

lol, that's at the bottom of the page, you got the BB code link, all the samples are there ,

have a look under URL Hyperlinking....

 

[Old Hippie]

The OCZ line of SSDs is confusing as hell to me.

I was drive shopping last weekend and purchased and cancelled two different orders because I bought the wrong drives.

I finally purchased two OCZSSD2-2VTX50Gs for a RAID0 setup.

 

[alaricljs]

Hmm, would have gone with these personally: http://www.mwave.com/mwave/skusearch...02.5IN%20SATA2

MWave has had the 60GB Vertex 2's at $140 after MIR for a couple months now. Note that it's the same exact drive as the 50GB but with a firmware that changes the provisioning.

 

[mutz]

The OCZ line of SSDs is confusing as hell to me.

with the agility 2 you get only 7 percent over provisioning i.e - another 4GB per 60GB drive,
with the LE for example, you get 28 percent, which is ~14GB per 50GB drive.

basically, they're all about the same, no much difference for the casual user,

OP, consider taking a 100GB LE, they're sold at 295 newegg (or rather you could find any cheaper 100GB sandforce drive),
50GB is not enough when you place a video and a game, add windows and linux VM or whatever,
be sure to consider demands are growing with time, and after leaving a HDD for an SSD you don't really want to go back.

 

[poohbear]

well i went the vertex 2 60gb route, just win7 & my common apps on there, for my games and storage i have a 1tb WD Caviar Black which is the fastest 7200rpm HDDs out there, and 1tb of WD Caviar Green for storage (completely silent compared to the Black). its my first SSD so i'll see how i feel.

What sucks is ALL my steam games have to be in 1 folder, Steam doesnt allow u to pick and choose which drive to install your games, would've loved to have my 2 favorite games on the SSD and the rest on my WD Black, but alas Steam doesn't allow that yet.

 

[Old Hippie]

with the agility 2 you get only 7 percent over provisioning i.e - another 4GB per 60GB drive,
with the LE for example, you get 28 percent, which is ~14GB per 50GB drive.

basically, they're all about the same, no much difference for the casual user,

OP, consider taking a 100GB LE, they're sold at 295 newegg (or rather you could find any cheaper 100GB sandforce drive),
50GB is not enough when you place a video and a game, add windows and linux VM or whatever,
be sure to consider demands are growing with time, and after leaving a HDD for an SSD you don't really want to go back.

I'm pretty sure the drives I bought had more over provisioning than the regular Agility2 drives (hence the 50GB) and I bought 2 for RAID0 which should keep the writes pretty high without TRIM.

These are test replacements for my two Intel 80GB RAID0 units that lost 50% write speed in 6 weeks.

 

[FishAk]

Remember that with SF, not all data is written to the drive- only what the controller needs to write to keep track of everything. For instance, when you load Word. and Excel, only about 50% of the total gets put on the disk, since so many files are the same. The OS still sees all the files taking up their normal room, and the difference, in effect, is spare aria. I think this is a big reason SF can get away with the smaller over provisioning.

 

[mutz]

These are test replacements for my two Intel 80GB RAID0 units that lost 50% write speed in 6 weeks.

you can partition the X-25's at ~80 percent, the controller will use the unpartitioned area as a spare pool.

Remember that with SF, not all data is written to the drive

that's the SF compression.

 

[VirtualLarry]

you can partition the X-25's at ~80 percent, the controller will use the unpartitioned area as a spare pool.


that's the SF compression.

Pretty sure that you have to do a "SET MAX LBA" followed by a "SECURE ERASE" on the X25-M's to reset their over-provisioning. They don't just read the partition table.

 

[mutz]

no, seems like only a secure erase would make it:
anand review: Intel X25-V in RAID-0

 

[Old Hippie]

you can partition the X-25's at ~80 percent, the controller will use the unpartitioned area as a spare pool.

That's how they were set-up.

I have a 29.3GB unallocated partition on my C drive from two 80GB G2s in RAID0.

 

[mutz]

That's how they were set-up.

I have a 29.3GB unallocated partition on my C drive from two 80GB G2s in RAID0.

that's odd...

p.s - larry, it probably isn't necessary to secure erase the drives before partitioning them,
the controller should see the empty partitioned part as a spare pool even though the other part is full.

E:
so maybe larry was right, maybe you should have set the MAX LBA before erasing the drive?
anand wasn't specifying this.

 

[Old Hippie]

that's odd...

Ya think?

AFAIK, I'm the only one that has reported any type of long term results when using a dedicated increased spare area in a RAID0 set-up.

Either the drive isn't using this space to it's advantage or it just doesn't matter and the drives' read speed is going to plummet anyway.

My AS SSD benchmarks are in this thread.

maybe you should have set the MAX LBA before erasing the drive?
anand wasn't specifying this.

There's no option for setting the LBA on HDDErase.

You think he forgot a whole step?

PSSS....I just gotta ask.....do you have an SSD? There seem to be a lotta experts that don't even have one.

 

[mutz]

AFAIK, I'm the only one that has reported any type of long term results when using a dedicated increased spare area in a RAID0 set-up.

anand's review showed that leaving ~15-20 percent spare area after partitioning the array result in much harmonic behavior.

i'm not sure exactly how these drives work from inside in details but maybe sequential writing the entire free area would make the drive perform better.

There's no option for setting the LBA on HDDErase.

never thought there is actually, and couldn't understand where is the part where you actually set it, before the format, with some raid manager or whatever, but if larry guy said that so maybe there is, i've never used X25-M's and neither raid.

PSSS....I just gotta ask.....do you have an SSD?

yes, and even an SS, a very special I7 920 and nice memory,
but cut the silly crap.

E:
p.s - some nice reading over that thread (that sub.mesa guy is crazy ).

 

[Old Hippie]

(that sub.mesa guy is crazy ).

He's here, he's there, he's everywhere!

I've seen some of those HD parameter commands when I was using HDAT2 but never used them.

but cut the silly crap.

No sense in getting offended. It was a legitimate question.

 

[mutz]

o.k bud,

maybe sequential writing the entire free area would make the drive perform better.

have you tried it?
slow write might be caused due to the controller trying to write into an already occupied area,
can't entirely figure why random write slow down the drive operation but apparently it happen,

try writing sequential to the entire free space available on the drive,
set IOMeter without any specified number of sectors and the test file should fill the entire free space, then you can run 1MB sequential write test with 100 percent access specification and 100 percent write and it should clear out all the mixed blocks,
run the test for 20-30 minutes.

your case is still a bit dazzling though.

E:
o.k, this explains it:
http://www.anandtech.com/show/2738/9

the spare space coming or being partitioned at the drive, fills up at some point,
so read modify write happens inevitably and so would slow down the drive,
this is only a temporary solution.

not sure how sequential write might change this scenario, maybe writing only 00 to it so the controller sees the NAND as empty (or maybe not, 00 still counts as data).

E2:
so what should/might happen with sequentially filling the entire free area is simply the controller would over-write it.

a used drive should have many blocks filled with random data, some of the pages are full while the other are empty.
when the controller would like to write a certain amount of data to the NAND, it has to read a portion, modify it and rewrite it, which takes time.
the OS sees deleted files as an unoccupied area, while the data is still actually there at the drive.

the controller have to write over this data, sending the file to different portions of the drive which takes time,
and lower the drive throughput.
now when the drive is filled with sequential pattern which is marked as delete, the controller can over-write this data at an instant and continuously, without searching for free pages or rather cut the file into to many small pieces and so - sending it all over the drive, a procedure which can take away some of the READ bandwidth as well.

so sequentially filling an already full drive should make the controller work faster.

TRIM ofcourse is supposed to handle these deletes and free up space for the controller to use,
while the G1 and RAID'ed drives are using none, they're prone to suffer from these architectural caveats.

what means here, Intel *!REALLY!* screwed the G1 buyers..
(and anand was even asking them through "The SSD-Anthology" to support TRIM with the G1's as a gesture (and an healthy marketing technique)..
hopefully it wasn't possible to allow TRIM for the G1's, or did it..?

..
anyhow,
anand seems to have kind of skipped a stage..

 

[Old Hippie]

what means here, Intel *!REALLY!* screwed the G1 buyers..

I also had the G1s and upgraded to the G2s in hope that the newer tech would do better.

Unfortunately that wasn't the case and the G2s lost the same amount of read speed as the G1s.

hopefully it wasn't possible to allow TRIM for the G1's, or did it..?

The G1s never got TRIM and AFAIK the "possibility" question was never answered.

 

[mutz]

Unfortunately that wasn't the case and the G2s lost the same amount of read speed as the G1s.

your only benefit would be TRIM from that aspect, the cache DRAM is changed and the controller revised between the two and the drive works a bit faster or slower at read&write speeds.

with RAID you lose TRIM and so any benefit, the architectural operation of the drive is still the same -

spare area get filled with usage,
controller has to read modify write data and so slows down,
new data is being written scattered all over the drive area and so read speed has some delay,
that's the architecture, that's how the drive work, you cannot skip it.

the only thing you can do, is either secure erase the drive, which actually deletes the entire data from the drive and so brings it back to a "new" state.

or,
fill the entire free area with sequential data, as explained above,
instead the controller would have to take a block, read it and modify it within the cache, and then rewrite it,
it should simply over-write it.
it gets the data from the LBA mapping that this data is there but is marked as delete and re flash it.
anand was suggesting it too (just without explaining exactly what happens (or rather it is somewhere inside one of the reviews).

GL ..