Samsung's 32nm NAND suffers major performance setback

Status
Not open for further replies.

Matthew DeCarlo

Posts: 5,271   +104

Folks that have been waiting for 32nm NAND flash production to pick up might be in for a disappointment. DailyTech is reporting that Samsung's 32nm NAND chips are problematic and too slow for use in SSDs. The site has talked with half a dozen players in the SSD market who use Samsung's flash chips, and all of them noted significant write speed issues with Sammy's parts.

Samsung's chips have performed so poorly in fact, that SSD manufacturers are unable to use them and are relying on other flash makers like IM Flash Tech and Toshiba for supply. "Of course every transition to a new process has its problems," said an anonymous source. "We just weren't expecting it to be this slow." Toshiba's 32nm NAND faced similar setbacks, but most of the kinks have been worked out, and those chips will be used in drives like OCZ's Agility Series.

Samsung is working to resolve the issues, and is still trying to move its troubled chips in the embedded scene. The company needs to get its act together if it plans to capitalize on the hot market. With Intel's new CPU lineup, cheap DDR3, DirectX 11 GPUs, and Windows 7 all available this holiday season, people will surely want to pick up an SSD to complete the package.

Permalink to story.

 
Why buy a SSD when I have a 1TB Spinpoint F3 with 2(!) plates? ;)
 
This problem will only continue to get worse, and despite the hype from SSD makers, and some band-aids for performance at 34nm, there's no solution.

Sun's Chief Technologist for Flash recently stated (at the Flash Memory Summit conference in August) that "in two process generations NAND will have higher latency than HDD". That means before the end of 2011.

The real problem is that NAND Flash -- by it's very fundamental nature -- gets dramatically slower as the bit-density (chip capacity) gets higher.

Toshiba's Ken Takeuchi, inventor of MLC flash and probably the world's leading expert on the problem described the phenomenon in the IEEE Journal of Solid State Circuits in April.

What this means is that the cost/performance of Flash is as good as it's ever going to get -- there are no more Flash cost-decreases coming that don't ALSO come with dramatically reduced performance. The low-end consumer markets overwhelmingly drive volume and therefore dictate the Flash technology roadmap, and these markets have voracious demands for more capacity and lower cost -- but don't care about performance. As a result, these performance declines are inevitable, and so the concept of NAND-Flash based SSDs in enterprise storage contexts is, surprising to almost everyone -- a dead end. The once-huge performance advantage of Flash over HDD will be gone by 2012.

So NAND based SSD is a "Flash-in-the-pan" indeed. IMO, this is why the founders of STEC sold off the majority of their stock holdings in August.

There may someday be a solid-state technology to replace disk drives, but it's not Flash.
 
Guest (above): good info. I'd like to read more. Have any sources?
 
there are numerous technologies ready to supplant NAND flash -- Samsung's phase-change memory (I believe the story is on TechSpot), and memristor based storage. Whatever it is, the industry will move to the highest density and fastest (for the lowest cost) technology in SSDs. Don't fret, it's not the end of the world :p
 
Status
Not open for further replies.
Back