IBM's Racetrack Memory Speeds Past Moore's Law


IBM sure has some seriously crazy semiconductor researchers locked in its basement. These guys question everything when it comes to advancing chip technology. Today IBM announced a new type of memory that will combine some of the best properties of Flash with the best properties of hard drives. Like Flash, it won’t have any moving parts, but like hard drives, it will store more information and be able to quickly write data.

For example, this technology could enable a handheld device such as an mp3 player to store around 500,000 songs or around 3,500 movies — 100 times more than is possible today — with far lower cost and power consumption.

IBM calls the technology “racetrack memory,” because it takes advantage of electrons traveling around a “racetrack” of wire. Somehow IBM has figured out how to store information in the boundaries between magnetic regions in magnetic materials. If I were a physicist I could tell you how it works, but I’m not, so if you’re interested, check out IBM’s site.

One of the cooler takeaways is that these chips won’t be flat, the way most integrated circuits are today, but will have the ability to be built up. IBM says the 3D nature of the chips could be another way around Moore’s Law. Of course, they’ve made that claim in the past with some of their DNA chip research. Maybe they just want a law named after someone who works at Big Blue?

Om adds on Friday: It is painful to see the story being dumbed down and then being dumbed down further. Don’t believe me, then read these headlines: New chip will let an iPod store 500,000 songs and IBM breakthrough means 500,000 song iPod


Seth Weintraub

What’s dumber? Making Racetrack memory exciting and relevant to a readerbase – like iPods to mac fans…or dissing (and linking) to a site that will never link to anything at Gigaom again.



@Om’s comment re: headlines

I don’t think those headlines are sad or unusual at all. In j-school, they teach that, like it or not, the average reader of a newspaper is reading at a 4th grade level and that the great challenge of the competent journalist is to take complicated stories and tell them in a way that makes sense to those readers. Readers know what ipods are, they may not be as intimately familiar with megabytes and burn rates. Sure, this may mean some “dumbing down”, but that’s the way (good) general media writing goes.

Chris Edwards

Om, Stacey,

Sorry for the name mixup: I read the piece…but in NetNewsWire. I just didn’t check the byline: in the feedreader it’s not quite as obvious as it is on the site here. However, I should have checked that.

The reason for picking on this particular item to kick off my post was simply because it helped make the point that IBM has tried to portray domain-wall memory as if it’s all IBM on its own (Big Blue has form for this, especially when it comes to memory technology). The line “these guys question everything” was too handy to pass up.

Stacey Higginbotham

@Lou, I should have emphasized that this is research rather than anything even close to production, but I don’t think the story is too sensational. Pure science IS sensational until the realities of commercialization and production intrude.


amazing that someone thinks 500,000 songs is a hook… the technology is a lot more interesting than the fact that i can own more bad music than i can listen to in the next several incarnations

Om Malik

@ Lou

Thanks for the link to Chris Edwards post. I take issue with how he characterizes our report, which is just a simple news report pointing out a breakthrough by IBM Research. We like to do that. Secondly, Edwards incorrectly says I wrote the piece, while Stacey is the writer. Sort of tells me, he didn’t read our little report. Otherwise it is an awesome piece.

Having said that, we don’t make any assertion that this is showing up tomorrow. In fact we left the sensationalization of the press release to The Times of London, the bastion of tech reporting these days.


IBM has been on a big campaign since 2006 (and maybe earlier) coming up with different implementations of Storage Class Memory, which blurs the distinction between memory (fast, expensive, volatile) and storage (slow, cheap, non-volatile). Some of the candidates include or have included:
-Improved Flash
-FeRAM (ferroelectric RAM)
-MRAM (magnetic RAM…also known as Racetrack memory outlined above)
-RRAM (resistive & polymer memory)
-Solid Electrolyte
-PC-RAM (Phase Change RAM)
These were outlined in detail by IBM researchers as the FAST08 conference in San Jose this past February.

The good news is that it appears like we’ll have plenty of options going forward. But knowing that many new memory types are scheduled to arrive in the coming years, we need to turn our attention to how to make the most effective use of these various memory types in data center systems.

Comments are closed.