Broadwell Won’t be Available to Consumers Until 2H 2014

Status
Not open for further replies.
not a surprise. Intel has been running into the physics issues involved in sub 22nm engineering for a while. it's what's holding up AMD and Nvidia from going to 20nm... the engineering problems at that size are quite large. the physics stops working right as you approach atomic sizes, bringing up whole new issues.
 

InvalidError

Titan
Moderator

Haswell-E would be on the same 22nm as all the Ivy Bridge and other Haswell designs so there is no reason for those to be affected by the 14nm hiccups.
 

Draven35

Distinguished
Nov 7, 2008
806
0
19,010


I hope we all appreciate the irony of big problems working at tiny sizes :D
 

InvalidError

Titan
Moderator

If you like huge machines working with tiny stuff, you must love the LHC!

The irony of "big.LITTLE" does not get much better than that... at least not within the scope of what science can be carried out directly on this planet.
 

Draven35

Distinguished
Nov 7, 2008
806
0
19,010


Sure, I got an interesting story about the use of GPU processing at the LHC from a CERN guy I met at GTC 2013.

 

Wamphryi

Distinguished


Or right up until your Motherboard dies and you can't find a replacement which is normally the case. Right now there is a large number of Lynnfield and Core 2 CPU's gathering dust with no Motherboards to call home for example.
 

MANOFKRYPTONAK

Distinguished
Feb 1, 2012
952
0
19,060
Things are always impossible. We could not figure out how Bees fly. Flying was impossible. I could list plenty more, the point is these barriers are puzzles that get figured out and we work around or with them. I believe that the same will hold true for computer parts.
 

gmax9000

Distinguished
Dec 30, 2011
10
0
18,510
You can do it Intel!! My good old 2600k still runs at a spritely 4.8 Ghz, broadwell will get ironed out by summer. I'm looking forward to DDR4 and PCI-E 4 hopefully all these updates come out around the same time, I'm gonna build me a sweetheart of a machine!
 

InvalidError

Titan
Moderator

Flying does not require pushing the fundamental laws of physics as we currently understand them to the breaking point. Chipmaking on the other hand is pushing the limits of how close to theoretical limits economically viable manufacturing can go: you cannot make traces inside chips thinner than the minimum required to ensure continuity including a margin to accommodate the manufacturing processes' variances and getting closer to that theoretical minimum requires tightening tolerances that much further across the whole process.
 

Grandmastersexsay

Honorable
May 16, 2013
332
0
10,780
Don't die shrinks only increase profitability of chips and not performance? There are other ways of accomplishing that goal that don't involve disrupting the space time continuum. How about cheaper manufacturing through better automation? I'm also sure that a company as large as intel could trim some fat if AMD made them. I'm sure there are ecconomy of scale solutions as well. Does intel really need to offer so many variants and do they really need to release a new chip every year? I'm sure the cost of R&D rivals that of the actual silicone.
 

InvalidError

Titan
Moderator

Smaller CMOS transistors have lower gate capacitance which allows faster switching using less power and shorter distances reduces propagation delays between gates. Both of those factors should allow faster chips.

A fair chunk of those margins is lost to reduced core voltages which reduce CMOS gate drive voltage and the associated output currents leading to smaller switching time improvements than otherwise possible and most of the rest is spent on cramming extra logic between flip-flops to improve IPC - this would be the other reason (aside from TIM under the heat-spreader) why Intel's newer CPUs have generally lower overclock margins: critical paths got longer/deeper logic-wise.
 

kirilmatthew

Honorable
Jul 24, 2013
1,377
1
11,660


Right. Process node shrinks do not directly increase performance, but decrease energy consumption and size, leading to chips consuming less energy, or that can be exchanged for greater performance. Intel has mostly been focusing on only the lower power usage and better GPU. AMD are gearing to make their chips faster on the flip side.
 

Wamphryi

Distinguished
The technology is going to reach a limit but is this really an issue? Real world computing barely nudges my stock Ivy Bridge i7 over 20%. The odd game might push it to 50% and rendering 100%. If there comes a time when applications really start pushing CPU resources then dual socket motherboards will appear in the mainstream.
 

InvalidError

Titan
Moderator

The thing with rendering and most other stuff that "really starts pushing CPUs" is that most of it (audio/image/video/AI/etc.) would be much better suited for GPGPU than CPUs so the most logical thing to do there is hybrid computing... and with hUMA/HSA, it looks like that is exactly where AMD is trying to go next.

Hopefully this won't take 12+ years to catch on like multi-threaded programming in mainstream software that could benefit from it.
 
Status
Not open for further replies.