JMicron SSD Testing With A19, L85A & L85C Memory

When we go back a few years, JMicron was one of the first in the SSD business to get the ball rolling and provide an SSD controller for SSD manufacturer use.  This controller was released to manufacturers in 2008 and was the JMF602 controller.  Companies such as  AData, G-Skill, Kingston, OCZ, Patriot, PQI, Runcore, Solidata, Supertalent and Transcend, were all quick to grab this controller and it hit the streets faster than it ever should have.  Even JMicron couldn’t have guessed that things would take off as they did.

There was a problem with the 602 controller though.  It had severe latency issues, these being first identified by Anandtech, although too late to stop the flood of consumer SSD sales.  To the average consumer unfortunate enough to have purchased an SSD at that time, these issues manifested themselves in the form of stuttering and extended delays getting information written to the drive.  This controller was labeled the ‘602 stuttering’ controller and had the potential to corrupt a SSD industry that was still in its infancy. Similarly, very early Samsung and SanDisk controllers also suffered this same latency issue which left Intel as the light at the end of the tunnel in the SSD industry at that time.

This issue was corrected by Samsung, and also by JMicron through the addition of a 16Kb cache buffer, JMicron’s newly released controller designated as the JMF602B.

Those were the early days of the SSD industry, only to be followed by the next step where third-party companies would buy complete SSD packages from Samsung and Intel, only to apply a branding sticker and call them their own.

This is the one of the first Crystal DiskMark benchmarks we ever conducted, and it was of the original release of the Samsung 64GB MLC FlashSSD.

Fast forward a bit and today we have in front of us three prototype SSDs, all containing the JMF667H controller, and each with different NAND flash memory on board.  We are going to take a look at this controller in a new light, however, it is by no means new (other than its firmware) and has been on the market for just over a year now.  Our previous reviews have it benched in the SuperTalent USB 3.0 Express RC4 Flash Drive, Kingspec Multi-Core 1TB PCIe SSD, and most recently, even in the SuperTalent PCIe DX1 M.2 SSD.  Having shown the JMF667H in an exclusive SSD report at 2.5GB/s, and also as the heart and soul of the latest M.2 form factor, we can support that the JMF667H controller was a solid controller even before today’s testing.

JMICRON JMF667H REFERENCE DESIGN

In our package from JMicron, we received three SSDs, representing the JMicron JMF667H controller with Toshiba A19, and IMFT L85A and L85C NAND flash memory.

From a readers standpoint, very least this will be an excellent representation of how all components within a SSD affect the overall performance of that drive.  This report will definitely demonstrate how a difference in memory alone can affect SSD performance.

TOSHIBA A19 SAMPLE

The Toshiba A19 sample contains 16 modules of Toshiba A19 19nm Toggle Mode NAND flash memory, the JMF667H 6Gbps 4 channel SSD controller, and one module of ESMT (Elite Semiconductor Memory Technology) DDR3 SDRAM cache.

This SSD is of a 128GB capacity, with each of the memory modules having a RAW capacity of 8GB each; we have never seen this memory used in a SSD prior.

IMFT L85A SAMPLE

The IMFT L85A sample was the only containing memory that we have seen in other SSDs prior, specifically the Intel 730 and DC3500.  It contained the JMicron JMF667H controller, eight modules of NAND flash memory, as well as a module of NANYA DDR3 DRAM cache.  No memory was contained on the reverse of this PCB.

This memory is 20nm with each module having a RAW capacity of 32GB in size, for a total formatted volume of 238GB.

IMFT L85C SAMPLE

Once again, the IMFT L85C sample contains the JMicron JMF667H controller, ESMT SDRAM cache, along with 8 pieces of IMFT’s 20nm NAND flash memory.

This SSD is only 128GB in capacity, 119GB after formatting, which means that every flash chip has a RAW capacity of 16GB.  We have never seen this memory in use in any other SSDs, as well,