HP Destroys a Dream Computer to Save It

http://bits.blogs.nytimes.com/2015/06/03/hp-destroys-a-dream-computer-to-save-it/

by luu • 11 years ago

56 45 11 years ago

bits.blogs.nytimes.com

ghshephard• 11 years ago

320 terabytes of memory - impressive.

bra-ket• 11 years ago

I hope they don't put it on a single IO-constrained box

dasmoth• 11 years ago

Why not? If you've got reasonably static 100TB datasets you want to be able to fire queries at, then -- if the price is right -- a single huge-memory box could be a great solution.

Need to run more queries than the box can handle? Buy another one.

alphakappa• 11 years ago

some people are working on the Machine with an eye to putting the technology into printers. That way, even a simple business printer could contain and analyze all of a corporation’s documents.

Are businesses really asking for a printer that can analyze their documents, or is this an application that's being shoehorned into a product that HP just happens to have?

Tloewald2• 11 years ago

It does seem like a rather odd line. A scanner that could do this might be useful -- a printer doing it would be kind of nutty.

strictnein• 11 years ago

I'm really guessing they mean one of those big all-in-one office machines, and not a deskjet, but who knows really.

vidarh• 11 years ago

For larger business, probably. For smaller businesses, the printer and scanner is very often an all in one unit these days, and a NAS is one more device we have to maintain, and one of the most frequent activities is to scan documents and save them in a document store, or pick documents and print them. If you can stuff enough capacity in the printer/scanner to make it the NAS too, cheaply enough, and add capabilities to e.g. have it print files directly, then that could be quite useful. A lot of printers already can print directly from memory cards etc., so why not straight from the file server?

rrss11222• 11 years ago

That happens when HP wants to salvage a technology. I remember a few years ago they talked about putting webOS on printers, for example.

frik2• 11 years ago

They sold webOS to LG, nowadays LG SmartTVs are powered by webOS. Today we have also ChromeOS and FirefoxOS which have a similar role.

smacktoward• 11 years ago

I bet the NSA is asking for it!

woodman• 11 years ago

It is kind of an interesting idea, from the prospective of information fusion. Unfortunately I spend the majority of my development time doing software integration, to achieve the whole synergy thing - for thinking outside of the box with icons in cornflower blue. These systems are generally rdbms backed, but politics and interesting interpretation of security standards often gets in the way. So most of the work is happening further up the stack, where automation is more difficult and common data formats are more rare. Except... PostScript [0] and PCL [1] (guess who owns that). This shortcuts the whole authorization problem, if you can print the data then you are probably authorized to use it. It shortcuts departmental kingdom building, and everybody already ignores security when it comes to printers... Lots of interesting possibilities. Of course then you'd have a Rube Goldberg machine of data format conversion, so I guess my job is safe for now - but I'm curious to see how HP would address that.

[0] http://en.wikipedia.org/wiki/PostScript

[1] http://en.wikipedia.org/wiki/Printer_Command_Language

Animats• 11 years ago

"some people are working on the Machine with an eye to putting the technology into printers. That way, even a simple business printer could contain and analyze all of a corporation’s documents."

Analyze them to send information to whom? NSA? FinCen? Hedge fund managers? I want to see the EULA on that thing.

omarforgotpwd• 11 years ago

that quote sounds like a joke

concernedctzn• 11 years ago

I remember the first time I read a hype article about the memristor. They really made it sound like it was coming out in a year or two. Guess I should've known better.

errtnsd• 11 years ago

Please refer to this handy chart: https://xkcd.com/678/

badsock• 11 years ago

You can buy devices with memristors today:

http://www.mouser.com/ProductDetail/Panasonic/MN101LR05DXW/?...

They've been delivered as promised. I read this as HP not getting the performance they need to surpass SRAM at its own game and also be nonvolatile. Which, to me, is kind of duh. You have to bring new tech in from the edges - niche, but important, products that are very ill-served by existing tech. Once you've got your production ramped up and the kinks worked out, then you take on at the really central stuff like CPUs.

Of course, it's very easy to say all this from my armchair :)

rdlecler1• 11 years ago

I'm not sure an organization would want the ability of a smartphone to pull all of the corporate data... Maybe it's a little safer operating behind cloud services.

junichiro• 11 years ago

Lol right! Can you imagine...

hackuser• 11 years ago

Devices that now rely on cloud computing for much of their functionality, like smartphones, could become self-contained objects, capable of memory-intensive things like voice recognition and language translation without calling on external computers

That could have a lot of impact on privacy and end-user control, and on businesses that sell cloud services.

higherpurpose• 11 years ago

I expect Google to be a laggard (at best) in that area because it wants you to have to connect to its servers, so hopefully others will pick up the slack.

Sanddancer• 11 years ago

Google wants you to be able to search. Being able to have lower latency for things like text to speech means that you're more apt to do voice searches and the like. You've got to keep in mind that the high end google phones already tend to have more RAM than their apple counterparts.

Tloewald2• 11 years ago

So, it's all going to happen just like they said it would if for "breakthrough new computer technology" you read "existing computer technology" and for "new operating system" you read "Linux port".

insulanian• 11 years ago

And I hoped for a breakthrough in OS research... The Machine is dehyped.

fra• 11 years ago

this is an unmitigated disaster for HP. What this signals to me is that they have given up on commercializing memristors any time soon. That's a $500M write-down (not as bad as autonomy I guess) and the end of their most promising R&D project.

rrss11222• 11 years ago

It is a shame that the hype could not match reality. They actually managed to get me excited for the memristor future.

uxcn2• 11 years ago

Interesting research[1] is still happening.

[1] http://spectrum.ieee.org/nanoclast/semiconductors/materials/...

aaggarwal2• 11 years ago

Actually, HP had been doing that since quite some time. In 2008, they announced that they have discovered the memristor and would bring it in the market by 2012, which then shifted to 2014, then 2015, and now they are still not sure whether they will be able to utilize this by 2020. No surprise here.

norea-armozel• 11 years ago

I think they underestimated the time table to bring it to market. It would've been better to focus on introducing memristors into specialized processing chips where they could be leveraged immediately.

On the other hand, it's good they've tried for a Moon shot than nothing. Maybe someone could buy the project from them since I think non-von-Neumann architectures are the future. It's just they don't have the budget or the share holders with the patience for it.

aaggarwal2• 11 years ago

I agree, plus they haven't been able to bring memristors to market as they promised again and again in 2008, 2012, 2014 and even now.

devonkim• 11 years ago

HP has had a strong recent history of strongly preferring to launch products made by other companies originally resulting in lagging pretty far behind and failing to become a market leader in any higher margin business with actually realized growth. Internally developed products from R&D got far, far less backing compared to a salesforce of hundreds to thousands pushing acquisition products like Fortify and Autonomy or 3PAR and webOS based tablets on the hardware side.

I'm not sure if you meant to make a pun, but HP literally had a project called Project Moonshot for a while that was a bet on the rise of microservers. It was also highly coupled with HP's bet on OpenStack taking mass control over internal datacenters of the Fortune 500. I, myself, kind of bet fairly hard on Openstack but after seeing the rather slow progress of adoption for it after the better part of a decade of its existence in some form, I'm thinking that people will start pulling out just from lack of progress.

serve_yay• 11 years ago

People bitch about Apple being driven by marketing but I really just read an entire article where the author credulously used the phrase "the Machine" throughout.

OedipusRex• 11 years ago

They realize that this product is far from being a reality and calling it anything other than "the Machine" would imply they are farther along in R&D or even production than they are.

swalsh• 11 years ago

But that's actually the name of the product! http://www.hpl.hp.com/research/systems-research/themachine/

PhantomGremlin• 11 years ago

When I read the headline I immediately thought of When HARLIE Was One[1] and of the Great Machine[2] from Babylon 5.

Apparently some people at HP are also Sci-Fi fans.

Edit: BTW I think it was Henry Spencer who, long ago, quipped something like "it's not a supercomputer unless it costs a billion dollars". Looks like they're actually budgeting for half of that so with inevitable cost overruns they might get there.

[1] https://en.wikipedia.org/wiki/When_Harlie_Was_One [2] https://en.wikipedia.org/wiki/Great_Machine

uxcn2• 11 years ago

Is it actually named after the machine on Epsilon 3, or is it just a coincidence? There's a B-Film called The Machine as well.

codewithcheese• 11 years ago

I believe the goal of the new OS was to try and take full advantage of RAM only computing and its implications that memristors required. Seems like they have realized that memristors are further away and more costly to R&D than they can stomach right now so they are hoping to usher in a new RAM only paradigm via a linux distro to further justify their continued R&D costs.

The whole printer thing is pure politics. To try and tack some justification for this project onto their profit center.

Dylan16807• 11 years ago

Devices that now rely on cloud computing for much of their functionality, like smartphones, could become self-contained objects, capable of memory-intensive things like voice recognition and language translation without calling on external computers.

As far as I understand it those things do not at all need large amounts of RAM.

Not to mention how shoving a million DRAM chips into a server has absolutely no impact on how you would cram more memory into a phone.

frik2• 11 years ago

A bit a disappointment, I thought memristor would be big in near future.

Oh, and I somehow missed that HP will split itself in two companies: Hewlett-Packard Enterprise (computer servers and data storage equipment), HP Inc (personal computers and printers). So they split out the former Compaq devision and their printer devision, or the other way around.

CoffeeDregs• 11 years ago

Totally agreed. I know that I shouldn't believe announcements about future technology... But the early announcements from HP were so confident and "no, it's really going to be here in 18 months" that I bought in... I was rooting for a big disruptive result... Still am, but with less enthusiasm.

shkkmo• 11 years ago

Just as important, the company seems to have realized that even great new computing technologies have to attract lots of software developers. Since these people already have lots of work, they tend to dislike learning new systems, even if they are more efficient.

Um... I don't think that description is accurate at all.

HP Destroys a Dream Computer to Save It

Related Stories