Storage
Intel mad for power, but stacked-up dies keep MELTING!
Moore's Law is going to be good for at least another decade, according to chip-maker Intel. "There's always physical limits to everything," Steve Pawlowski, senior fellow and head honcho on exascale research, told The Register at the European Research and Innovation Conference in Ireland. "But you can always come up with clever …
This topic is closed for new posts.
Who is Eric Moore when he is at home?
... or is this a new Moore's law we have not heard about?
Memory stacked on top of the processor?
If, down the way, somebody decides to patent this idea, I might suggest they look at the RaspberryPi.
Packages
Intel are trying to stack die-on-die, not package-on-package which has been done for yonks -- My old Atari ST 520 got a RAM upgrade that way with sixteen DIL chips stacked on the existing RAM chips and some legs-in-the-air (Oooh, Missus!) wiring to cope with the extra address line.
The RasperryPi chips are low-power and low-speed -- Intel want to stack DRAM dies with a couple of microns separation on top of a CPU putting out 40 watts of heat and more. I think they'll have to come up with some kind of nanoengineered active cooling system to get the heat away from the silicon at a great enough rate to prevent meltdowns or thermal noise problems at high data speeds.
Just for completeness
The chip the Raspberry Pi uses is s BRCM2835, not one designed by Raspberry Pi themselves. There is a 256MB ram package soldered on top of the 2835. Works pretty well - cooling not a problem because the SoC is so low power anyway, even with no heatsink.
Samsung have been stacking RAM, Flash and CPU inside a < 3mm high package for maybe 5 years.
of course it's ARM not x86
Intel may be a leader in x86, everything else not so much.
As The Lab Boys...
...are finding out, the real challenge to vertical assembly is thermal management.
It's much easier with beer, just drink it before it gets too hot.
Eric...
ERIC: European Research and Innovation Conference
http://www.intel.com/corporate/education/emea/event/irc/
Deja Vu?
I seem to recall that the Pentium IV was going to scale to 10GHz too.
"I’m kind of interested in it for a number of reasons, but is it going to take over everything and be the new technology that’s going to drive us to exascale? I don’t believe it,"
What I reckon he is really saying is that Intel have no solutions beyond silicon and so have been reduced to squeezing the last dregs from the current approach, hence multi-core and now stacking.
I have an old 3GHz P4 laptop
and if they ever pushed P4 to 10GHz I would suggest having the fire brigade on standby, and an ambulance to take me to the burns unit.
Why not just get the increased speed by writing the software properly.
yawn
Stacked dies are hardly new, and many novel 3D schemes have been proposed in the past. Maybe the Intel wanker is a newbie (and he shouldn't waste his 15 minutes of fame demonstrating his lack of perspicacity).
OTOH, he is right that Moore's law impending death is greatly exaggerated. But that is hardly news except perhaps to the yellower segments of the mewling media.
On the other hand
That's Intel w*****r is paid a lot more than you, and has a lot more knowledge on the subject.
On the third hand,
perhaps AC OP is a senior chip engineer, and knows a lot more about a lot more than you think, AC2 (and is possibly paid more than both of us combined.)
Or perhaps before presuming to know who an AC is, you'd like to present your own bona fides?
No bona fides I'm afraid.
I posted AC for a reason. See if you can guess what it is?
Pretty sure
That a senior chip engineer and one who earns considerable more that the OP's above would not have resorted to using the word 'wanker' . That smacks of illiterate trolling, not brainpower.
I could be wrong. Maybe the world of top end chip design is full of people who have to resort to bad language to make a point. I doubt it though, nobody here needs to.
@James Hughes 1
Illiterate trolling and brainpower are, sadly, very much not mutually exclusive.
Furthermore, there is a difference between choosing to use "bad*" language and needing to. In my experience, experts often choose to use such language when talking about a topic that is close to their field of expertise, because like many humans, they are somewhat emotionally involved. I learned long ago that judging someone's intelligence from their outward appearance or choice of language is a sure way to alienate oneself from some very intelligent people.
My primary point was that AC2 assumed a lot of things from very little information. My secondary point was that posting AC in response to an AC in a conversation that very much hinges on identity-related information (expertise and salary) is somewhat ironic.
* and really, "wanker" is your limit? Really? You must have a hard time reading the comments here. Sure, nobody here needs to use such language, but they do -- very frequently.
Hmm. 100 times faster in 6 years.
Somehow moors law doesn't seem at all adequate for that.
My expectation is that you won't see a factor of more than 4 or 6 by 2018.
And...
What good is additional speed on the procs if everything they store goes to a media thats a good 3 orders of magnitude slower.
prior art exists
the Cray II, I believe it was, had chips three-deep in the processor. that's why they had to run it submerged in flowing cooled freon. try that on your laptop.
It wasnt Freon, it was Fluorinert. Different Chemical, same idea though. Its use with the Cray-2 was single-stage (liquid), Though you can also use Fluorinert in a dual stage configuration where it turns gaseous and then condenses back into a liquid in an exchanger. Apparently you can drink it, but not breathe it.
This topic is closed for new posts.
