Intel's Future Chips: News, Rumours & Reviews

[-Fran-]

I'm struggling on my crappy phone to copy paste the link but its on Bloomberg, Intel in talks to licence graphics IP from amd!!! Very interesting.

That sounds a bit odd... They cut ties with Imagination to develop their own IP IIRC, so now going to AMD for it sounds a bit weird.

Cheers!

 

[Cazalan]

Xeon FPGA details

http://www.theregister.co.uk/2016/03/14/intel_xeon_fpga/

 

[logainofhades]

I'm struggling on my crappy phone to copy paste the link but its on Bloomberg, Intel in talks to licence graphics IP from amd!!! Very interesting.

Found a link. http://techreport.com/news/29876/rumor-intel-may-be-in-talks-to-license-graphics-tech-from-amd

 

[Tommylincon]

a li'l recap:
haswell:
Analysis: Intel shows off Haswell, minus the important bits
http://semiaccurate.com/2011/09/21/analysis-intel-shows-off-haswell-minus-the-important-bits/
Intel admits Haswell uses transactional memory
http://semiaccurate.com/2012/02/07/intel-admits-haswell-uses-transactional-memory/
Haswell is a graphics monster
http://semiaccurate.com/2012/02/08/haswell-is-a-graphics-monster/
Haswell’s GPU prowess is due to Crystalwell
http://semiaccurate.com/2012/04/02/haswells-gpu-prowess-is-due-to-crystalwell/
Haswell-EP to use the same socket, just totally different
http://semiaccurate.com/2012/07/09/haswell-ep-to-use-the-same-socket-just-totally-different/
Haswell graphics for GT1 and GT2 revealed
http://semiaccurate.com/2012/08/31/haswell-graphics-for-gt1-and-gt2-revealed/
Haswell GT3 uses shaders to save power
http://semiaccurate.com/2012/09/07/haswell-gt3-uses-shaders-to-save-power/
Crystalwell is very wide memory for Haswell GT3
http://semiaccurate.com/2012/09/10/crystalwell-is-very-wide-memory-for-haswell-gt3/
Intel to split up ULV with Haswell
http://semiaccurate.com/2012/12/05/intel-to-split-up-ulv-with-haswell/

Intel's Haswell Architecture Analyzed: Building a New PC and a New Intel
http://www.anandtech.com/show/6355/intels-haswell-architecture
Making Sense of the Intel Haswell Transactional Synchronization eXtensions
http://www.anandtech.com/show/6290/making-sense-of-intel-haswell-transactional-synchronization-extensions
Intel Haswell Info: Single Chip for Ultrabooks, GT3 GPU for Mobile, LGA-1150 for Desktop
http://www.anandtech.com/show/5078/intel-haswell-info-single-chip-for-ultrabooks-gt3-gpu-for-mobile-lga1150-for-desktop
Haswell at IDF 2012: 10W is the New 17W
http://www.anandtech.com/show/6248/haswell-at-idf-2012-10w-is-the-new-17w

Specifications of Intel's Mobile Haswell CPUs Leaked
http://www.tomshardware.com/news/intel-haswell-mobile-cpu-processor,19848.html
Leaked Slide Shows Intel Haswell Set for March-June 2013
http://www.tomshardware.com/news/Intel-Haswell-LGA-1150-Ivy-Bridge-CPU,14673.html
Haswell CPUs to Top Out at 3.5 GHz (desktop)
http://www.tomshardware.com/news/intel-haswell-leaked-clock-speeds,19719.html

latest:
Intel to Offer Nine Types of Haswell Processors in Four Types of Packages.
http://www.xbitlabs.com/news/cpu/display/20121227233038_Intel_tffer_Nine_Types_of_Haswell_Processors_in_Four_Types_of_Packages.html
Intel’s Haswell to Feature Secret Weapon: Integrated Voltage Regulator.
http://www.xbitlabs.com/news/cpu/display/20121226225930_Intel_s_Haswell_to_Feature_Secrete_Weapon_Integrated_Voltage_Regulator.html

broadwell:
Intel’s Broadwell CPU; more details leak
http://semiaccurate.com/2012/11/29/intels-broadwell-cpu-more-details-leak/
Exclusive: Intel’s Broadwell CPU chipsets
http://semiaccurate.com/2012/10/26/exlcusive-intels-broadwell-cpu-chipsets/

beyond broadwell:
ye olde rumor on skylake/skymont (says broadwell is 16nm. go figure.)
After Intel’s Haswell comes Broadwell, Sk……
http://semiaccurate.com/2011/03/31/after-intels-haswell-comes-broadwell-sk/

oh thanks man i love amd thangs shit

 

[Cazalan]

Get your new Xeon Phi CPUs with 16GB HMC. (Knights Landing Edition, 14nm)

http://dap.xeonphi.com/

5K for the tower unit
20K for the Quad rack unit

 

[-Fran-]

Get your new Xeon Phi CPUs with 16GB HMC. (Knights Landing Edition, 14nm)

http://dap.xeonphi.com/

5K for the tower unit
20K for the Quad rack unit

I have to say. The Tower unit is mighty tempting. The specs make it to be an excellent "lab on wheels" for any project. And at that price, is not that out of reason.

I am actually surprised, pleasantly.

Cheers!

 

[Cazalan]

Get your new Xeon Phi CPUs with 16GB HMC. (Knights Landing Edition, 14nm)

http://dap.xeonphi.com/

5K for the tower unit
20K for the Quad rack unit

I have to say. The Tower unit is mighty tempting. The specs make it to be an excellent "lab on wheels" for any project. And at that price, is not that out of reason.

I am actually surprised, pleasantly.

Cheers!

Yeah the price is reasonable considering the announced prices Nvidia had for their Pascal box. Those were like 12K per unit.

 

[Anthcny]

That is correct!

 

[-Fran-]

Weird. No one is talking about this, so here it is:

http://www.anandtech.com/show/10288/intel-broxton-sofia-smartphone-socs-cancelled

Can we assume they acknowledged low power performance (phone/tablet) is ARM territory and they're moving away from that turf war?

Cheers!

 

[gamerk316]

At this point, the lack of software more or less dooms X86 on mobile. Throw in the lack of being able to get power draw low enough (which I predicted years ago) and the fact growth in the segment is slowing, not to mention billion dollar losses, it makes sense why Intel would finally throw in the towel.

 

[logainofhades]

Saw this last night.

http://www.kitguru.net/components/cpu/matthew-wilson/intel-core-i7-7700k-kaby-lake-details-supposedly-leak/

 

[-Fran-]

Saw this last night.

http://www.kitguru.net/components/cpu/matthew-wilson/intel-core-i7-7700k-kaby-lake-details-supposedly-leak/

"(via wccftech)"

So, they should start calling it "Tick, tac, toe"?

Cheers!

 

[TehPenguin]

for all of you that want to upgrade in the near future: there'S an ongoing cashback reward for selected i5/i7 bundles

https://toptechcashback.com/

ends July 15th

 

[turkey3_scratch]

Saw this last night.

http://www.kitguru.net/components/cpu/matthew-wilson/intel-core-i7-7700k-kaby-lake-details-supposedly-leak/

Personally I could care less. We're going to get our 5% performance boost and it'll be a day.

 

[TehPenguin]

You guys think Haswell-E will get any cheaper with Broadwell-E now available?

 

[jed]

You guys think Haswell-E will get any cheaper with Broadwell-E now available?

No, the prices on E model CPU normally don't drop.

 

[TehPenguin]

You guys think Haswell-E will get any cheaper with Broadwell-E now available?

No, the prices on E model CPU normally don't drop.

Thought so as they still perform splendidly. Thanks!

 

[okcnaline]

Drop any lower and Intel will probably lose $$$.

 

[jimmysmitty]

Drop any lower and Intel will probably lose $$$.

Doubtful. Intel has a pretty massive margin on their CPUs, especially since they do all their own fabrication which makes it cheaper for them compared to AMD or nVidia who use third party companies.

 

[Guest]

In my opinion Intel will hit the limit one day . Making transistors is very hard to make in smaller dimensions . So they'll use crazy designs and different materials for more performance .

 

[gamerk316]

In my opinion Intel will hit the limit one day . Making transistors is very hard to make in smaller dimensions . So they'll use crazy designs and different materials for more performance .

There's a reason why theres a TON of research in optical chips, and even using relays to replace transistors [the idea being, while relays take FOREVER to throw, you can trivially get processors in the hundred GHz range]. Traditional CPU design basically dies off after the 7nm node, as we simply can't improve performance by adding transistors.

We're literally 4-5 years from the end of performance gains.

 

[jimmysmitty]

In my opinion Intel will hit the limit one day . Making transistors is very hard to make in smaller dimensions . So they'll use crazy designs and different materials for more performance .

That is why they are looking at 3D stacked designs like with HMC and 3DXpoint or even their FinFETs they currently have for CPUs.

In my opinion Intel will hit the limit one day . Making transistors is very hard to make in smaller dimensions . So they'll use crazy designs and different materials for more performance .

There's a reason why theres a TON of research in optical chips, and even using relays to replace transistors [the idea being, while relays take FOREVER to throw, you can trivially get processors in the hundred GHz range]. Traditional CPU design basically dies off after the 7nm node, as we simply can't improve performance by adding transistors.

We're literally 4-5 years from the end of performance gains.

Intel already made a fiber optic silicon chip:

http://www.pcworld.com/article/114778/article.html

Isn't their Omni Path part of that?

 

[gamerk316]

The main issue with Optical is how you generate the light; it's very power hungry and causes all sorts of cooling problems. Sure, in a lab, you can do it, but a consumer grade chip? That's the problem.

I know DARPA has a 100 GHz CPU built around relays. The idea being to clock it insanely high, and use very complex on-chip logic to limit as much as possible how many relays you need to throw [due to the rather INSANE latency you introduce each time you throw a relay].

And of course, Quantum Computing.

My suspicion is you'll see 7nm be the last cost-effective node to produce on, and you'll see two generations where everyone goes wide, before they figure out the software won't cooperate. That's when you'll see the big players [Intel/IBM] throw money at alternative designs.

 

[juanrga]

The main issue with Optical is how you generate the light; it's very power hungry and causes all sorts of cooling problems. Sure, in a lab, you can do it, but a consumer grade chip? That's the problem.

I know DARPA has a 100 GHz CPU built around relays. The idea being to clock it insanely high, and use very complex on-chip logic to limit as much as possible how many relays you need to throw [due to the rather INSANE latency you introduce each time you throw a relay].

And of course, Quantum Computing.

My suspicion is you'll see 7nm be the last cost-effective node to produce on, and you'll see two generations where everyone goes wide, before they figure out the software won't cooperate. That's when you'll see the big players [Intel/IBM] throw money at alternative designs.

Both Intel and IBM labs have been throwing money on alternative designs during decades.

 

[Vogner16]

In my opinion Intel will hit the limit one day . Making transistors is very hard to make in smaller dimensions . So they'll use crazy designs and different materials for more performance .

I believe that as intel hits this performance wall (2.5% increase in IPC each gen now) we will see intel not trying to improve their core design to improve FP or integer performance per say but more on the lines of ASIC's for specialist tasks which remove threads from the cpu's work load. take for example amd APU's. they integrate a UVD to remove video decode from the cores to a specialist section of the silicon. makes video decode faster than what cpu could do and allows the cpu to be open for other compute tasks.

of course this needs software to use specialist asic's on a cpu die but intel has the market by the balls... just snap you fingers!

security decode asic's already here. video decode asic's already here. compression asic? this stuff is whats next. not improvement in per core performance! or adding more transistors per core! sure they will add a few, but you get my point. asic's are the future for computing