New JMicron JMF602/JMF601 Revision B supposed to fix stuttering writes on lower priced SSDs.

[fuzzybabybunny]

http://www.dailytech.com/article.aspx?newsid=14004

These JMF602B chips have been suspected to be in the Runcore 1.8" PATA ZIF drives that were just released less than a month ago as well as in the new G.Skill Titan. The stuttering does seem to be fixed and the performance numbers have improved quite a bit on the 4K random write tests to be even faster than Samsung SLCs... price/performance/storage ratios could now be very competitive to Intel.

http://forums.anandtech.com/me...=2269761&enterthread=y

JMicron states that the new chips should be in all new SSDs:

Currently, all JMicron customers are using latest version, including ASUS NB/EeePC, OCZ, Super Talent, Transcend, etc. The B version improves the write latency a lot.

Woot!

 

[taltamir]

excellent... I wonder if we will soon see a new samsung SLC controller.

 

[meson2000]

That is great news. I hope Anand does another round up with the latest drives like the Titan and the new Vertex drives from OCZ...

 

[coolVariable]

I hope that is NOT the "new" controller in the vertex.
G.skill has it and while that SSD is better than other JMicron SSDs, it still does not reach intel's performance.

Also ssems like not everything is well in g.skill-ville:

http://ht4u.net/reviews/2009/gskill_titan_256gb_ssd/index7.php

The last three benches are writing access times.

Avg. write access time is 100x SLOWER than the intel
Min. write access time is 25x times slower than the intel
Max. write access time is 1.5x times slower than the intel (and 20x times that of a Western Digital HDD)


And these ATTO tests also do not bode well:
http://ht4u.net/reviews/2009/gskill_titan_256gb_ssd/index14.php

 

[imported_TimBob]

Originally posted by: coolVariable
I hope that is NOT the "new" controller in the vertex.
G.skill has it and while that SSD is better than other JMicron SSDs, it still does not reach intel's performance.

Also ssems like not everything is well in g.skill-ville:

http://ht4u.net/reviews/2009/gskill_titan_256gb_ssd/index7.php

The last three benches are writing access times.

Avg. write access time is 100x SLOWER than the intel
Min. write access time is 25x times slower than the intel
Max. write access time is 1.5x times slower than the intel (and 20x times that of a Western Digital HDD)


And these ATTO tests also do not bode well:
http://ht4u.net/reviews/2009/gskill_titan_256gb_ssd/index14.php

From what I've read, the Vertex does not use a jmicron, but the Apex does.

 

[fuzzybabybunny]

Originally posted by: coolVariable
I hope that is NOT the "new" controller in the vertex.
G.skill has it and while that SSD is better than other JMicron SSDs, it still does not reach intel's performance.

Also ssems like not everything is well in g.skill-ville:

http://ht4u.net/reviews/2009/gskill_titan_256gb_ssd/index7.php

The last three benches are writing access times.

Avg. write access time is 100x SLOWER than the intel
Min. write access time is 25x times slower than the intel
Max. write access time is 1.5x times slower than the intel (and 20x times that of a Western Digital HDD)


And these ATTO tests also do not bode well:
http://ht4u.net/reviews/2009/gskill_titan_256gb_ssd/index14.php

I wish I could read German...

I'm having a hard time understanding the tests. On the page you linked there are three graphs with the exact same title: Zugriffszeiten H2BenchW (Schreiben), which translates into Access Times for Write, I believe, and then average, min, and max. What were the parameters of the three different graphs? 4K random writes? 512K non-random writes? Anything about I/O depth?

The average write access times being 100X slower than Intel is problematic. The minimum I don't think so much because that might be splitting hairs there (0.1ms vs. 1.4ms - but here's another thing I don't understand - the Intel is 0,1ms and the GSkill is 1,4ms but it's 25X longer? Huh?)

In the Runcore thread the Runcore drive got only 2.15ms average latency for 4K 100% random 100% writes using IOmeter, quite different from the 190ms figure in the tests you linked. And we think the Runcore drives run on the JMF602B chipset as well.

 

[coolVariable]

Originally posted by: TimBob


From what I've read, the Vertex does not use a jmicron, but the Apex does.

From what I've read it uses a new controller ... and I don't think OCZ makes its own controllers.

 

[frostedflakes]

Yeah this isn't exactly new, IIRC the Patriot Warp V2 and probably most SSDs released in the last few months have been using the JMF602B controller. I used to own a Warp V2 and it worked pretty well, but there was still some stuttering. I have a Samsung MLC drive now and it doesn't exhibit any stuttering.

Nice to hear that JMicron is addressing the issue, though, they have received a lot of bad press on this. I'm anxious to see how the controller they plan to release Q309 performs, sounds like it will be a complete redesign, as opposed to the tweaked JMF602 rev B.

 

[emo65170]

I purchased an older model G.Skill 128 SSD that has the much maligned JMF602B controller on Craigslist and experienced the stuttering issue as everyone else. This drive is not the Falcon or Titan. I was so frustrated with the performance that I was about to toss it.
After much searching online, I discovered that the firmware update tool, ISP.exe, used to update the JMicron 602B controller firmware on OCZ's core drives also works on my G.Skill 128. I connected the G.Skill 128 drive to USB and ran the ISP.exe tool and flashed the drive to the latest firmware. Here's the logfile, mo.txt, that is created after the ISP.exe tool completes.


System mode !
Auto detect...
Flash model=K9HCG08U1M
(Ch,Bank,CE,DrvCur)=(8,4,2,2)
Part 1 : SATA
Part 2 : USB
WARNING: drive current not match, det=02 old=03

Create original setting file : old.txt
FW updated count : 0x01
Current firmware version= 081016
New updated firmware version : 090508S
Preformat...pass
Download code...pass
Upgrade firmware : completed
Please restart the system.

ISP End.

Note that you will lose everything on the drive and you'll have to reinstall/reimage but after you do, you'll find no more stuttering and a much smoother experience overall.

I downloaded the Iometer software to try to benchmark it but I couldn't figure out how to configure it. If anyone wants to give me some tips on how to use Iometer to measure the Random Write performance, I'd be happy to run those tests.

 

[coolVariable]

Way to go to revive an old thread.
Considering SSD price, I don't know why anyone would touch one of the early SSDs with a 10 foot pole at this point.

 

[emo65170]

Way to go to revive an old thread.
Considering SSD price, I don't know why anyone would touch one of the early SSDs with a 10 foot pole at this point.

With all the bad press the drive got when it came out, you might be able to acquire one for cheap, and with the "new" firmware, it's worlds faster than the old 120GB 5400rpm drive that it replaced.

 

[F1shF4t]

If you can get these older SSDs for cheap they would still work quite nicely for external data drive or lappy drive.

I'll not comment on how it would work as a desktop boot drive cause I never used one. But since chucking my G1 x25-m ssd into a usb enclosure for data cause it was more sluggish than a x25-v, I wouldn't have much hope.

 

[coolVariable]

If you can get these older SSDs for cheap

That would need to be "extremely cheap" considering current prices - still not worth it.