Back in the day it didn’t take much for me to get excited about a new technology. The rapid progressions we saw from the late 90s through to the early 2010s had us all fervently awaiting the next big thing as it seemed nearly anything was within our grasp. The combination of getting older and being disappointed a certain number of times hardened me against this optimism and now I routinely attempt to avoid the hype for anything I don’t feel is a sure bet. Indeed I said much the same about HP’s The Machine last year and it seems my skepticism has paid dividends although I can’t say I feel that great about it.
For the uninitiated HP’s The Machine was going to be the next revolutionary step in computing. Whilst the mockups would be familiar to anyone who’s seen the inside of a standard server those components were going to be anything but, incorporating such wild technologies as memristors and optical interconnects. What put this above many other pie in the sky concepts (of which I include things like D-Wave’s quantum computers as the jury is still out on whether or not they’re providing a quantum speedup) is that it was based on real progress that HP had made in many of those spaces in recent years. Even that wasn’t enough to break through my cynicism however.
And today I found out I was right, god damnit.
The reasons cited were ones I was pretty sure would come to fruition, namely the fact that no one has been able to commercialize memristors at scale in any meaningful way. Since The Machine was supposed to be almost solely based off of that technology it should be no surprise that it’s been canned on the back of that. Now instead of being the moonshot style project that HP announced last year it’s instead going to be some form of technology demonstrator platform, ostensibly to draw software developers across to this new architecture in order to get them to build on it.
Unfortunately this will likely end up being not much more than a giant server with a silly amount of RAM stuffed into it, 320TB to be precise. Whilst this may attract some people to the platform out of curiosity I can’t imagine that anyone would be willing to shell out the requisite cash on the hopes that they’d be able to use a production version of The Machine sometime down the line. It would be like the Sony Cell processor all over again instead of costing you maybe a couple thousand to experiment with it you’d be in the tens of thousands, maybe hundreds, just to get your hands on some experimental architecture. HP might attempt to subsidise that but considering the already downgraded vision I can’t fathom them throwing even more money at it.
HP could very well turn around in 5 or 10 years with a working prototype to make me look stupid and, honestly, if they did I would very much welcome it. Whilst predictions about Moore’s Law ending happen at an inverse rate to them coming true (read: not at all) it doesn’t mean there isn’t a few ceilings we’ve seen on the horizon that will need to be addressed if we want to continue this rapid pace of innovation. HP’s The Machine was one of the few ideas that could’ve pushed us ahead of the curve significantly and its demise is, whilst completely expected, still a heart wrenching outcome.
Whilst computing has evolved exponentially in terms of capabilities and raw computing performance the underlying architecture that drives it has remained largely the same for the past 30 years. The vast majority of platforms are either x86 or some other CISC variant running on a silicon wafer that’s been lithographed to have the millions (and sometimes billions) of transistors etched into it. This is then all connected up to various other components and storage through the various bus definitions, most of which have changed dramatically in the face of new requirements. There’s nothing particularly wrong with this model, it’s served us well and has fallen within the bounds of Moore’s Law for quite some time, however there’s always the nagging question of whether or not there’s another way to do things, perhaps one that will be much better than anything we’ve done before.
According HP their new concept, The Machine, is the answer to that question.
For those who haven’t yet read about it (or watched the introductory video on the technology) HP’s The Machine is set to be the next step in computing, taking the most recent advances in computer technology and using them to completely rethink what constitutes a computer. In short there are 3 main components that make it up, 2 of which are based on technology that have yet to see a commercial application. The first appears to be a Sony Cell like approach to computing cores, essentially combining numerous smaller cores into one big computing pool which can then be activated at will, technology which currently powers their Moonshot range of servers. The second piece is optical interconnects, something which has long been discussed as the next stage in computing but as of yet hasn’t really made in roads at the level HP is talking about. Finally the idea of “universal memory” which is essentially memristor storage which HP Labs has been teasing for some time but has failed to bring any product to light.
As an idea The Machine is pretty incredible, taking the best of breed technology for every subsystem of the traditional computer and putting it all together in the one place. HP is taking the right approach with it too as whilst The Machine might share some common ancestry with regular computers (I’m sure the “special purpose cores” are likely to be x86) current operating systems make a whole bunch of assumptions that won’t be compatible with its architecture. Thankfully they’ll be open sourcing Machine OS which means that it won’t be long before other vendors will be able to support it. It would be all too easy for them to create another HP-UX, a great piece of software in its own right that no one wants to touch because it’s just too damn niche to bother with. That being said however the journey between this concept and reality is a long one, fraught with the very real possibility of it never happening.
You see whilst all of these technologies that make up The Machine might be real in one sense or another 2 of them have yet to see a commercial release. The memristor based storage was “a couple years away” after the original announcement by HP however here we are, some 6 years later, and not even a prototype device has managed to rear its head. Indeed HP said last year that we might see memristor drives in 2018 if we’re lucky and the roadmap shown in the concept video shows the first DIMMs appearing sometime in 2016. Similar things can be said for optical interconnects as whilst they’ve existed at the large scale for some time (fibre interconnects for storage are fairly common) they have yet to be created for the low level type of interconnects that The Machine would require. HP’s roadmap to getting this technology to market is much less clear, something which HP will need to get right if they don’t want the whole concept to fall apart at the seams.
Honestly my scepticism comes from a history of being disappointed by concepts like this with many things promising the world in terms of computing and almost always failing to deliver on them. Even some of the technology contained within The Machine has already managed to disappoint me with memristor storage remaining vaporware despite numerous publications saying it was mere years away from commercial release. This is one of those times that I’d love to be proven wrong though as nothing would make me happier than to see a true revolution in the way we do computing, one that would hopefully enable us to do so much more. Until I see real pieces of hardware from HP however I’ll remain sceptical, lest I get my feelings hurt once again.
With the amount of NVRAM that’s used these days the amount of innovation in the sector has been comparatively little. For the most part the advances have come from the traditional avenues, die shrinks and new gate technologies, with the biggest advance in 3D construction only happening last week. There’s been musings about other kinds of technology for a long time like memristors which had their first patent granted back in 2007 and were supposed to making their way into our hands late last year, but that never eventuated. However news comes today of a new memory startup that’s promising a lot of things and whilst they don’t say it directly it looks like they might be one of the first to market with memristor based products.
Crossbar is a new company that’s been working in stealth for some time on a new type of memory product which, surprisingly, isn’t anything particularly revolutionary. It’s called Resistive RAM (RRAM) and a little research shows that there’s been companies working on this idea as far back as 2009. It’s based around a fairly interesting phenomena whereby a dielectric, an electric insulator, can be made to conduct through the application of high voltage. This forms a filament of low resistance which can then be reset, breaking the connection, and then set again using another high voltage jolt. This idea lends itself well to applications in memory as the two states translate perfectly to binary and if the specifications are anything to go by the performance that will come out of them should be quite spectacular.
If this is sounding familiar then you’re probably already familiar with the idea of memristors. These are the 4th fundamental component of electronic circuits that were postulated back in 1971 by Leon Chua and were made real by HP in 2007. In a basic sense their resistance is a function of the current following through them and when the current is removed that resistance is remembered, hence their name. As you can see this describes the function of RRAM pretty well and there is a solid argument to be made that all RRAM technologies are in fact memristors. Thus whilst it’s pretty spectacular that a start up has managed to perfect this technology to the point of producing it on a production fab it’s actually technology that’s been brewing for quite some time and one that everyone in the tech world is excited about.
Crossbar’s secret sauce could likely come from their fabrication process as they claim that the way they create their substrate means that they should be able to stack them, much in the same way that Samsung can now do with their VNAND. Now this is exciting because previously HP alluded to the fact that memristor based storage could be made much more dense than NAND, several orders of magnitude more dense to be precise, and considering the density gains Samsung got with their 3D chips a layered memristor device’s storage capacity could be astronomical. Indeed Crossbar claims this much with up to 1TB for a standard chip that could be stacked multiple times, enabling terabytes on a single chip. That puts good old fashioned spinning rust disks on notice as they just couldn’t compete, even when it comes to archival storage. Of course the end price will be a big factor in this but that kind of storage potential could drive the cost per GB through the floor.
So the next couple months are going to be quite interesting as we have Samsung, the undisputed king of NAND, already in the throws of producing some of the most dense storage available with Crossbar (and multiple other companies) readying memristor technology for the masses. In the short term I give the advantage to Samsung as they’ve got the capital and global reach to get their products out to anyone that wants them. However if memristor based products can do even half of what they’re claimed to be capable of they could quickly start eating Samsung’s lunch and I can’t imagine it’d be too long before they either bought the biggest players in the field or developed the technology themselves. Regardless of how this all plays out the storage market is heading for a shake up, one that can’t come quick enough in my opinion.
The computer (or whatever Internet capable device you happen to be viewing this on) is made up of various electronic components. For the most part these are semiconductors, devices which allow the flow of electricity but don’t do it readily, but there’s also a lot of supporting electronics that are what we call fundamental components of electronics. As almost any electrical enthusiast will tell you there are 3 such components: the resistor, capacitor and inductor each of them with their own set of properties that makes them useful in electronic circuits. There’s been speculation of a 4th fundamental component for about 40 years but before I talk about that I’ll need to give you a quick run down on what the current fundamentals properties are.
The resistor is the simplest of the lot, all it does is impede the flow of electricity. They’re quite simple devices, usually a small brown package banded by 4 or more colours which denotes just how resistive it actually is. Resistors are often used as current limiters as the amount of current that can pass through them is directly related to the voltage and level of resistance of said resistor. In essence you can think of them as narrow pathways in which electric current has to squeeze through.
Capacitors are intriguing little devices and can be best thought of as batteries. You’ve seen them if you’ve taken apart any modern device as they’re those little canister looking things attached to the main board of said device. They work by storing charge in an electrostatic field between two metal plates that’s separated by an insulating material called a dielectric. Modern day capacitors are essentially two metal plates and the dielectric rolled up into a cylinder, something which you could see if you cut one open. I’d only recommend doing this with a “solid” capacitor as the dielectrics used in other capacitors are liquids and tend to be rather toxic and/or corrosive.
Inductors are very similar to capacitors in the respect that they also store charge but instead of an electrostatic field they store it in a magnetic field. Again you’ve probably seen them if you’ve cracked open any modern device (or say looked inside your computer) as they look like little circles of metal with wire coiled around them. They’re often referred to as “chokes” as they tend to oppose the current that induces the magnetic field within them and at high frequencies they’ll appear as a break in the circuit, useful if you’re trying to keep alternating current out of your circuit.
For quite a long time these 3 components formed the basis of all electrical theory and nearly any component could be expressed in terms of them. However back in 1971 Leon Chua explored the symmetry between these fundamental components and inferred that there should be a 4th fundamental component, the Memristor. The name is a combination of memory and resistor and Chua stated that this component would not only have the ability to remember its resistance, but also have it changed by passing current through it. Passing current in one direction would increase the resistance and reversing it would decrease it. The implications of such a component would be huge but it wasn’t until 37 years later that the first memristor was created by researchers in HP’s lab division.
What’s really exciting about the memristor is its potential to replace other solid state storage technologies like Flash and DRAM. Due to memristor’s simplicity they are innately fast and, best of all, they can be integrated directly onto the chip of processors. If you look at the breakdown of a current generation processor you’ll notice that a good portion of the silicone used is dedicated to cache, or onboard memory. Memristors have the potential to boost the amount of onboard memory to extraordinary levels, and HP believes they’ll be doing that in just 18 months:
Williams compared HP’s resistive RAM technology against flash and claimed to meet or exceed the performance of flash memory in all categories. Read times are less than 10 nanoseconds and write/erase times are about 0.1-ns. HP is still accumulating endurance cycle data at 10^12 cycles and the retention times are measured in years, he said.
This creates the prospect of adding dense non-volatile memory as an extra layer on top of logic circuitry. “We could offer 2-Gbytes of memory per core on the processor chip. Putting non-volatile memory on top of the logic chip will buy us twenty years of Moore’s Law, said Williams.
To put this in perspective Intel’s current flagship CPU ships with a total of 8MB of cache on the CPU and that’s shared between 4 cores. A similar memristor based CPU would have a whopping 8GB of on board cache, effectively negating the need for external DRAM. Couple this with a memristor based external drive for storage and you’d have a computer that’s literally decades ahead of the curve in terms of what we thought was possible, and Moore’s Law can rest easy for a while.
This kind of technology isn’t you’re usual pie in the sky “it’ll be available in the next 10 years” malarkey, this is the real deal. HP isn’t the only one looking into this either, Samsung (one of the world’s largest flash manufacturers) has also been aggressively pursuing this technology and will likely début products around the same time. For someone like me it’s immensely exciting as it shows that there are still many great technological advances ahead of us, just waiting to be uncovered and put into practice. I can’t wait to see how the first memristor devices perform as it will truly be a generational leap ahead in technology.