Intel 320 SSD aka G3

[Computurd]

jesus still very expensive!

[chuchnit]

Anvil have these shown up in stock anywhere in your country? Does anyone know the "official" launch date on these?

[Anvil]

No, and the listings have been removed from the websites.
(available documents list March 2011 as First Availablity)

There should be some more info within March I guess, lets see what materializes first , Crucial M4, Intel 320 or the Vertex 3.

[johnw]

I don't think there is an official launch date from Intel yet. I saw a rumor a couple weeks ago that it would be April. Before that, the rumor was March, and before that February, and before that, 2010 Q4.

[czesiu]

Apparently, things are going so well over at Intel that they are going to launch their 25nm MLC based 320 Series (Postville Refresh), ahead of schedule, on the 28th of March. Intel 320 series SSD is set to replace the current X25-V and X25-M series at the same price point targeting mainstream consumers, SMBs, data centres and embedded markets.
(...)
Available in 5 capacities; 40GB, 80GB, 120GB, 160GB, 300GB and 600GB, they are estimated to be priced at $109, $189, $239, $329, $569 and $1119 respectively.

http://vr-zone.com/articles/intel-fi...led/11655.html

[Anvil]

Thanks,

So, another week to wait.

Not overly impressed by these figures but lets see how they perform, they are supposed to use an onboard cache.

-Up to 39,500 IOPS random 4 KB reads; up to 23,000 IOPS random 4 KB writes
-Full Disk Encryption via AES 128 bit Encryption Technology
-Enhanced power loss management
-1.2M hours MTBF
-Intel® 25nm compute NAND

[johnw]

available in 5 capacities; 40gb, 80gb, 120gb, 160gb, 300gb and 600gb,

5? Heh.

[czesiu]

5? Heh.

haha maybe they have more typos there

[saint-francis]

Those prices are markedly unimpressive.

[fatherTime27]

I want one

[Ao1]

Those prices are markedly unimpressive.

Indeed. It seems 4k performance has dropped with 25nm and sequential speeds have gone up. If that was offset with a 50% price reduction then hey, but over $1,000 for 600GB drive is just lunacy.

[czesiu]

It seems 4k performance has dropped with 25nm

4K writes are much higher compared to G2 http://download.intel.com/design/fla...eam/322296.pdf

[alfaunits]

Yeah, since Intel intentionally limited the write speeds on G1/G2
G3 is a minor evolutionary step - we need a revolutionary step!

[Ao1]

The new drives have much larger DRAM and a power cap. 8k pages and 25nm seem to equal slower 4K read performance. Write performance is improved, but that is the result of cache. 4K write and faster sequential speeds don’t really do it for me if they come at the expense of higher 4K read latency. l might just but another G2 before they disappear.

[Anvil]

I'm not going to order more of the G2s, I'm confident that the G3s are better drives all things considered and 4K low QD performance wasn't G1/G2s strongest points.

With this new info on 4K 25nm performance it looks like the SF controller is very strong but I'll definately be ordering some more of the G3s.

We might also see a need for more of the 8KB benchmarks, the iometer workstation and database patterns are already 100% 8KB benchmarks.

[Anvil]

Some more info on the G320

g3_info.JPG

No info on the 600GB yet.

[One_Hertz]

Cluster size needs to move to 8kb on these new SSDs... 4kb does not make sense if pages are 8kb. If the cluster is 8kb then we need to test 8kb performance.

[Anvil]

A lot of things might need adjustments

Here's an old HDTune I found last year (2010/10), notice the 4KB QD1 result, that's what's interesting, I know HDTune isn't a great tool for SSDs but it shows QD1 alright.

This is an older revision 2BW and the new one is 2CW and so things might have changed.

(I can't vouch for the validity of this, but it looks promising)
g3_hdtune.jpg

[Anvil]

I found a test of the 80GB 320 Series and it's not what I hoped for. (tested on an AMD MB, Asus M4A88TD-M, 880G/SB850)

4KB low QD results looks like any other 25nm I've seen, not impressed iow.

Hopefully, the >= 160GB will scale better on randoms.

No sign of the 256MB cache, just a small 32MB cache according to screenshots.

[Ender17]

I found a test of the 80GB 320 Series and it's not what I hoped for. (tested on an AMD MB, Asus M4A88TD-M, 880G/SB850)

4KB low QD results looks like any other 25nm I've seen, not impressed iow.

Hopefully, the >= 160GB will scale better on randoms.

No sign of the 256MB cache, just a small 32MB cache according to screenshots.

here's the review if anyone else is interested
http://www.expreview.com/14414-1.html
controller is the same chip as G2!?!?

[Brahmzy]

Hmmmm. Yeah. I get the feeling our current benchmark utilities are not showing us the real story. Need more options that will better reflect real-world.

[Ender17]

Hmmmm. Yeah. I get the feeling our current benchmark utilities are not showing us the real story. Need more options that will better reflect real-world.

why do you say that?
same controller + same size cache + 25nm NAND
no reason it should be faster

[Anvil]

Some examples of the 320 160GB

Link to forum (google translate)

The raid-0 tests are w/o WBC and so the results could have been much higher.

Grabbed from that same forum

Intel_SSD_320_G3_160GB-031.png Intel_SSD_320_G3_160GB-032.png

The score is about 80 points down from a clean G2 of the same size.

[felix_w]

The score is about 80 points down from a clean G2 of the same size.

I don't own a G2 160GB, but the AS-SSD results i could find (1,2) do not show this, do you happen to have any benches handy ?

EDIT : Maybe you meant that the G3 is above the G2 by 80 points in total score ? (Just clearing up, no purpose to offense)

[Anvil]

Sure, I posted performance on the old/new G2 firmware some time ago here at XS

Link to thread



This comparison makes me wonder if the score is weighted too much on the 64K Thread part of the test.
It just doesn't look right.