Could Intel's Braidwood Lessen SSD Appeal?

[Rob Williams]

From our front-page news:

Does there exist a computing enthusiast right now who wouldn't want a solid-state disk in their PC? Of course not, and as it stands, there are many who've opted to upgrade their storage and make their computer all the more faster, from quicker application launching to smoother operation. What's not to like? Of course, there's the costs and densities, but even those are getting better as time goes on.

With SSDs hogging so much of the computing limelight lately, what on earth could dampen its outlook? Well according to a report written by Objective Analysis, a Los Gatos, California firm, Intel's Braidwood (seriously, that's a code name?) has a good chance at it. In the simplest of explanations, Braidwood would add one of Intel's NAND flash memory chips directly to the motherboard, allowing for improved performance.

Ranging in densities from 4GB - 16GB, and price-ranges of $10 - $20, this addition would promise to increase to reliability and overall system access speeds. It could act as a cache and allow data to be written to the hard drive in such a way that it won't lag your machine, which is one of the major benefits of SSDs. Plus, as the chip would be based on SLC technology, it would be reliable, fast, and still much cheaper than typical DRAM.

The report clearly states that the technology has real potential to hurt SSD sales, since the overall performance boost would almost negate the overall speed increase that an SSD would offer. Intel disagrees and claims Braidwood has huge reliability features, which is no doubt true. In the end, I find it hard to believe that Braidwood could actually lesson the desire for people to own SSDs, but it will be interesting to see how things pan out when the technology is launched next year.

An Intel spokesman disagreed with Objective Analysis' premise that Braidwood could hurt SSD sales, saying "It's not just the performance, but also the added reliability...[SSDs] can help facilitate versus a hard drive. We see a long life ahead for SSDs, and won't stop inventing a variety of other technologies that make computers faster and/or more energy efficient."

Source: Computerworld​

 

[Kougar]

I beg to disagree with the analyst that wrote this report. There were some bad assumptions made in the report which were used to support the conclusions reached.

For instance he pretty much claims SLC NAND in Braidwood is going to bring SSD-level performance to all Braidwood equipped motherboards, and even hints that because it is SLC it will outperform some MLC SSD's... simply ludicrous. Intel's MLC based SSD's have easily outperformed competing SLC flash drives in real world programs, the controller and number of flash channels in use being the primary reasons.

A 16GB data buffer will help hide the slower performance of HDD's but it isn't going to fix it. Also just one or two SLC NAND chips is going to yield very low performance by SSD standards, something completely glossed over by the report... MLC and SLC SSD's reach their current performance levels partly because they concurrently use anywhere from 4 - 10 flash channels.

Regardless of if I disagree or not, one major issue overlooked here is longevity. The specifications for MLC require a minimum of 10,000 cycles per cell, SLC is an order of magnitude greater with 100,000 writes per cell before it is no longer guaranteed to function normally. One, two, or even four NAND SLC chips that equal 16GB in capacity are not going to last very long with continuous writes, erases , and more writes. What happens when these chips fail... can they be disabled or bypassed, or does the motherboard have to be thrown out?

Some previous Intel designs used a miniature SODIMM style slot, but even then the cost of replacing failed NAND cache every couple years seems silly. A real SSD will last significantly longer regardless of if it is MLC or SLC based due to spreading out the writes over the entire capacity of the drive, write combining, and gigabytes of spare reserve capacity.

 

[Rob Williams]

Good insight, and I agree.

One, two, or even four NAND SLC chips that equal 16GB in capacity are not going to last very long with continuous writes, erases , and more writes. What happens when these chips fail... can they be disabled or bypassed, or does the motherboard have to be thrown out?

I was wondering the same thing. It's one thing to have an SSD with 8 - 16 SLC chips, but it's another to have just one. I'd have to assume that the user would have the ability to disable the function, or that the BIOS would realize that the chip is bad and then just bypass it. Still, nothing would stop Intel from making this a replaceable option. You could just replace the chip if it dies. We do the same thing with RAM.

I agree it's a foolish route to take, but it this feature is actually worth its weight, then we may not mind the extra factor in our computing.