REGISTER | LOGIN
Breaking News
News & Analysis

Rambus, Microsoft Heat Up With Cold DRAM

4/17/2017 04:30 PM EDT
4 comments
NO RATINGS
Page 1 / 4 Next >
More Related Links
View Comments: Newest First | Oldest First | Threaded View
DMcCunney
User Rank
Author
Re: Achieving cryogenic temperature
DMcCunney   4/26/2017 8:30:35 PM
NO RATINGS
@Rambus: Good question – how one cools devices to <93K is indeed a challenge, but one that can be dealt with.

I thought it was a good question, but your answer just assures me it can be dealt with.  I'd like to know how.

Efficient cryogenic cooling isn't practical for small scale devices such as mobile phones or tablets.  But when one gets to the scale of a datacenter then refrigeration technology gets reasonably efficient and reliable.

I never thought it would be practical for the sorts of things we carry around.  But I suspect full application in a datacenter will require fundamental changes in the way datacenters are built and provisioned. It doesn't sound like something that can be economically retrofitted.

Regarding the question of what happens should the refrigerator fail, the answer is that it's relatively easy to engineer a "soft landing".

That, too, was about what I expected.  I didn't see a catastrophic failure - just a condition where you wouldn't get the performance benefits you expected from a working system.


>Dennis

junko.yoshida
User Rank
Author
Re: Achieving cryogenic temperature
junko.yoshida   4/26/2017 6:13:52 PM
NO RATINGS
Thanks for your further explanation, @rambus. Much appreciate it.

@DMcCunney , here's the answer from the horse's mouth.

Rambus
User Rank
Rookie
Re: Achieving cryogenic temperature
Rambus   4/26/2017 10:33:08 AM
NO RATINGS
Good question – how one cools devices to <93K is indeed a challenge, but one that can be dealt with.  Efficient cryogenic cooling isn't practical for small scale devices such as mobile phones or tablets.  But when one gets to the scale of a datacenter then refrigeration technology gets reasonably efficient and reliable.  Thermodynamics determines the theoretical maximum efficiency for a refrigerator (the Carnot efficiency) and then the mechanical efficiency of the refrigerator defines how close one gets to that theoretical limit.  The mechanical efficiency of a cryogenic refrigerator can be 30% or higher which makes the energy gains from cooling memory more than offset the energy cost of cooling.

Regarding the question of what happens should the refrigerator fail, the answer is that it's relatively easy to engineer a "soft landing".  The memory devices don't fail catastrophically when they warm to room temperature .  They simply don't meet all of their specifications.  One can monitor this and then take the memory offline for repair at a convenient time.  It's also worth pointing out that cryogenic refrigeration is widely used in the medical industry  for MRI imaging and the reliability of these systems has been proven in the field for many, many years.

DMcCunney
User Rank
Author
Achieving cryogenic temperature
DMcCunney   4/18/2017 8:28:26 PM
NO RATINGS
DRAM may indeed be adaptable for the things suggested in the article.

The missing piece here is precisely how folks using it intend to get operating temperatures that low.  How do you cool DRAM and other components to minus−180 °C or minus−292.00 °F or 93.15 Kelvin?

(And for that matter, what happens if you can't maintain those temperatures?   You may not get a system failure, depending on how fast things warm up, but you won't get the desired performance.)

I'd love to hear more about the proposed cooling methods.

>Dennis

 

 

 

Like Us on Facebook
EE Times on Twitter
EE Times Twitter Feed