REGISTER | LOGIN
Breaking News
News & Analysis

Intel's Xeon Tackles Real Time

E7v4 minimizes analytics latency
6/6/2016 03:01 AM EDT
12 comments
NO RATINGS
More Related Links
View Comments: Newest First | Oldest First | Threaded View
Page 1 / 2   >   >>
R_Colin_Johnson
User Rank
Author
Re: Scale Up and Real Time?
R_Colin_Johnson   6/7/2016 1:07:31 PM
NO RATINGS
Thanks for the insights, Mike. You always have an interesting take on the marketing/supply chain situation. Didn't now there were so many E7v2s still on the distributor shelves. The customers that do buy them at bargain prices will have a clear upgrade path since v3s and v4s are already available and pin-compatible!

perl_geek
User Rank
Author
Really real-time?
perl_geek   6/7/2016 12:21:55 PM
NO RATINGS
In one sense, every useful control system has to be "real-time", i.e. reliably produce a response faster than the occurrence of the phenomenon that it's controlling.

The question is, what's a meaningful time interval for the event? In the case of an accounting transaction, it could be measured in weeks, though now we're used to sub-second responses. For a big boiler, with lots of thermal inertia, a few minutes to turn the heat on or off would probably be adequate, while for an unstable aircraft's elevator, milliseconds could be critical. High-frequency traders might be concerned about nanoseconds.

(If anyone in those fields wants to correct me by an order of magnitude or so, feel free; it doesn't really change the principle.)

The operational definition has always seemed to be "a system response faster than a human could  achieve, and guaranteed fast enough to be useful". What was real-time a generation ago would be rated as glacial today.

 

elizabethsimon
User Rank
Author
Re: real-time not
elizabethsimon   6/7/2016 12:04:46 PM
NO RATINGS
Exactly right. He fell into the trap of thinking that real-time == fast while in many cases predictability is much more important.

I design equipment used in the power industry. Sampling and analyzing a 60Hz waveform doesn't necessarily require speed but we absolutely have to finsh processing before the next sample comes in or really bad things can happen. And of course it's the times when you have a lot of inputs changing (requiring extra processing time) when you REALLY can't afford to miss a sample.

R_Colin_Johnson
User Rank
Author
Re: real-time not
R_Colin_Johnson   6/7/2016 8:09:05 AM
NO RATINGS
Yes, that's definitely the best definition of realtime I've heard, which of course will be different depending on the application. None of Intel's applications meet this realtime definition, so I guess we will have to call Intel's unconventional at the least. Thanks for the clarification.

sw guy
User Rank
Author
Re: real-time not
sw guy   6/7/2016 8:03:25 AM
NO RATINGS
Mainly, what he missed (most probably on purpose) is predictability:
Does not matter that the mean response time is well below expectation if you cannot guaranty that in no case a computation will terminate too late to drive whatever (vital) external interface

R_Colin_Johnson
User Rank
Author
Re: real-time not
R_Colin_Johnson   6/7/2016 7:58:43 AM
NO RATINGS
Yes, I was a a little uncomfortable with Intel's use of realtime, but they defined it as "less than a second" and gave only examples of tasks that RTOS programmers would never tackle. I finally accepted it at face value, but you are right--he definitely redefined what engineers ordinarily consider as realtime (which of course depends on the task, but would ordinarily be measured in microseconds).

sw guy
User Rank
Author
real-time not
sw guy   6/7/2016 6:32:37 AM
NO RATINGS
Someone to tell Patrick Buddenbaum he is redefining "real time" in a way that arrange him but does not fit needs of people used to tackle critical real-time task inside, say, a plane or a computer driver drilling machine

trtewell
User Rank
Rookie
Re: A comment on semiconductor Distributors
trtewell   6/6/2016 7:58:46 PM
NO RATINGS
Makes me kind of sad... :( As Kettering said: "~ People are very open-minded about new things, as long as they're exactly like the old ones."

Susan Rambo
User Rank
Author
Re: A comment on semiconductor Distributors
Susan Rambo   6/6/2016 5:07:50 PM
NO RATINGS
Thanks. I fixed it.

realjjj
User Rank
CEO
Re: A comment on semiconductor Distributors
realjjj   6/6/2016 4:57:24 PM
NO RATINGS
"The deal is bending regulatory approval"


Lovely typo.


Page 1 / 2   >   >>
Like Us on Facebook
EE Times on Twitter
EE Times Twitter Feed