Intel Broadwell Thread

[jdubs03]

Except it will be the same performance +5%. Yawn. 2500k till 2020 at this rate.

i suspect the performance gain to be considerably higher, 10-15% is probably more realistic.

 

[Sable]

i suspect the performance gain to be considerably higher, 10-15% is probably more realistic.

And what do you base that on? Intel's record of adding 5% performance for Ivy, Haswell and now Broadwell?

 

[meloz]

Sable are you actually being limited by single-threaded performance? Most people aren't. So as long as there is no regression in single-threaded performance and we get higher performance/watt, I am happy.

 

[JDG1980]

But even that’s still not enough, and for Core M Intel went so far as to give Broadwell-Y its own die and design a low-power optimized version of their 14nm process just for it. This variant is designed to further reduce power consumption by optimizing the resulting transistors for lower power, lower voltage, lower clockspeed operation. By doing this Intel was able to further reduce power consumption in all of the major areas over what would be a traditional 14nm Intel process.

That actually could be good news for enthusiasts. If Intel is creating a special version of the 14nm process for Core-M, then this means that their "mainstream" 14nm process can be better optimized for higher clock speeds since it won't have to be bogged down by the need to optimize power consumption at all costs.

 

[witeken]

Won't that be a step back from HW GT1? Or at most a minor increase?

Bay Trail has 4 EUs. Cherry Trail will double and quadruple this to 8 and 16 EUs.

There is no information about Broadwell GT1.

 

[witeken]

Considering what the relatively underpowered GPU in baytrail manages in windows gaming, even getting within sight of a 4200 would be amazing for the res I'm running most of the games at. I wish AMD's idea for a laptop base for your tablet that included a fan allowing double TDP when docked had taken off. Talk about the best of both worlds.

I don't know if it was a rumor but I read somewhere that if you switch to laptop mode, the TDP increases by 4W.

Now the hard decision is breaking the bank for a core M or going for CT. When is the successor to Cherry trail expected?

Broxton should be mid-2015 with Gen9 graphics and Goldmont architecture.

 

[TrulyUncouth]

I don't know if it was a rumor but I read somewhere that if you switch to laptop mode, the TDP increases by 4W.


Broxton should be mid-2015 with Gen9 graphics and Goldmont architecture.

Is mid-'15 for actual availability? I thought Cherry trail wasn't coming til the beginning of the year- will they really be releasing another update so soon?

 

[witeken]

Mid-2015 is for actual availability, although I do not know if Broxton's launch is impacted by the delay of 14nm/CT. Since Broxton is for both tablet and smartphones (and Moorefield won't get a separate die shrink to 14nm), I think Broxton won't be delayed (much).

 

[Sweepr]

We were allowed to handle a working prototype in the form of a sexy 7mm-thin tablet, but were not allowed to run any software or examine specifications via the control panel. Until we have a working model to test, we'll have to take Intel's word that Broadwell-Y provides a "greater than 2x reduction in TDP with better performance than Haswell-Y".

We do have some specifics when it comes to the die and package sizes though. The Broadwell-Y chip is 82mm2, scaled down about 63% compared to Haswell-Y's 130mm2 die size. As for the board package, Broadwell-Y has a 50% smaller surface area and 30% thinner package compared to Haswell-Y.

www.tomshardware.com/reviews/intel-14nm-broadwell-y-core-m,3904-2.html

Really small die. Quick comparison:
- Bay Trail-T: 102mm²
- Kabini: ~110mm²
- Apple A7: 102mm²
- Snapdragon 800: 118.3mm²

 

[liahos1]

www.tomshardware.com/reviews/intel-14nm-broadwell-y-core-m,3904-2.html

Really small die. Quick comparison:
- Bay Trail-T: 102mm²
- Kabini: ~110mm²
- Apple A7: 102mm²
- Snapdragon 800: 118.3mm²

wow that's really small. i wonder if intel should have kept die size ~100mm and used the extra real estate for more performance. piednoel cryptically laughed off denver k1's performance in a reply to a tweet yesterday (again). interested in seeing benchmarks for this.

has anyone run the calculus of how this thing should bench against an A7 or Denver K1? I recall seeing one or two comments estimating core m performance similar to the surface pro 1. if true would put it at multiples faster than an a7 and k1

if intel's real competition for the next 6 months is 28nm soc's with a few 20nm soc's throw in for measure i dont see how they dont destroy the perf/watt comparisons.

 

[witeken]

www.tomshardware.com/reviews/intel-14nm-broadwell-y-core-m,3904-2.html

Really small die. Quick comparison:
- Bay Trail-T: 102mm²
- Kabini: ~110mm²
- Apple A7: 102mm²
- Snapdragon 800: 118.3mm²

We already know the die size since IDF 2013.

Also, this 0.63x die size reduction includes 20% more transistors in Broadwell-Y compared to Haswell.

 

[witeken]

wow that's really small. i wonder if intel should have kept die size ~100mm and used the extra real estate for more performance. piednoel cryptically laughed off denver k1's performance in a reply to a tweet yesterday (again). interested in seeing benchmarks for this.

Note that this 80mm² doesn't include the new 32nm PCH, so total die area will be more. Broadwell-Y obviously also doesn't have an integrated modem.

has anyone run the calculus of how this thing should bench against an A7 or Denver K1? I recall seeing one or two comments estimating core m performance similar to the surface pro 1. if true would put it at multiples faster than an a7 and k1

What Intel was basically trying to tell us yesterday is that Broadwell-Y didn't regress in performance, despite a TDP reduction of >2X from 11.5W to 4.5W.

 

[liahos1]

Note that this 80mm² doesn't include the new 32nm PCH, so total die area will be more. Broadwell-Y obviously also doesn't have an integrated modem.


What Intel was basically trying to tell us yesterday is that Broadwell-Y didn't regress in performance, despite a TDP reduction of >2X from 11.5W to 4.5W.

so its comparable in performance to the i3 found in the surface pro 3?

edit..

actually nevermind the i3 in the sp3 doesnt turbo and has a base clock of 1.5ghz. so this should be faster given scaling to 2ghz+ and a 5% improvement in ipc? ive got hungry eyes for a product like this.

 

[witeken]

Exactly. On the other hand, the base clock of Core M is only 800-1100MHz, but that might be because of the nature of tablets; they have smaller batteries and lower power requirements.

 

[liahos1]

if you take the denver results from the ppt

Geekbench 3 Single-Core
Baytrail (Celeron N2910): 0.65x
S800 (Krait 400 8974AA): 0.80x
Tegra K1 (R3 Cortex A15): 1.00x
A7 (Cyclone): 1.20x
Haswell (Celeron 2955U): 1.20x
Tegra K1 (Denver): 1.65x

gets you a denver single core k1 score of 2050 = 1.65/1.2 x 1491

if you look at the i5 version of haswell at 2981 there still seems to be a pretty big performance gap even assuming slightly lower performance relative to i5 version of surface pro 3.

is there a comparable specfp and specint results archive?

am i thinking about this wrong. i know that geekbench itself is cited as a worthless/stupid benchmark by linus on realworldtech but it would be interesting to see how these compare.

 

[witeken]

I think that it's Intel's intention to get Broadwell-Y in phones eventually. I suspect Skylake will aim for smaller tablets and Cannonlake will go into (i?)phones.

 

[liahos1]

I think that it's Intel's intention to get Broadwell-Y in phones eventually. I suspect Skylake will aim for smaller tablets and Cannonlake will go into (i?)phones.

is k1 suppose to be going into any phones? If that is i dont see how a broadwell-y couldnt be jammed into a phablet haha.

 

[Sweepr]

Note that Celeron N2910 is a 1.6GHz chip, one of the slowest Bay Trail-M around. Pentium N3540 operates at 2.66GHz (Turbo). NVIDIA is comparing Denver to a capped 1.4GHz Haswell (2C/2T 2MB L3) while Broadwell-Y (2C/4T 4MB L3) not only improves IPC (>5% according to Intel) but also Turbo @ up to 2.6GHz according to leaked specs. If Broadwell-Y can hit >2GHz in ST/MT workloads @ 4.5W TDP then I think Intel has a winner in their hands. I hope pricing is also competitive, a Surface Pro-like Windows device running Core M for $600 would be great.

 

[witeken]

is k1 suppose to be going into any phones? If that is i dont see how a broadwell-y couldnt be jammed into a phablet haha.

Not likely, but I'm not sure that's because of its power consumption. With all the changes Intel made to BDW and the 14nm process, it should perform at least decently even in 5" phones, but BDW-Y is simply not designed for those. I suspect that Intel will use the next 2 years to do that. Somewhere in 2016, we might see Intel do another presentation like yesterday's and announce that they used the 10nm process with a >2X efficiency improvement and their integrated modem to go into phones with Core.

 

[exar333]

so its comparable in performance to the i3 found in the surface pro 3?

edit..

actually nevermind the i3 in the sp3 doesnt turbo and has a base clock of 1.5ghz. so this should be faster given scaling to 2ghz+ and a 5% improvement in ipc? ive got hungry eyes for a product like this.

Same here.

I was really tempted by the SP3, but Broadwell really looked to be even more of a game-changer. Can't wait to get my hands on one of those...

 

[ShintaiDK]

I think that it's Intel's intention to get Broadwell-Y in phones eventually. I suspect Skylake will aim for smaller tablets and Cannonlake will go into (i?)phones.

Core products stops at the highend tablets. Anything else is Atom and successors.

 

[mikk]

www.tomshardware.com/reviews/intel-14nm-broadwell-y-core-m,3904-2.html

Really small die. Quick comparison:
- Bay Trail-T: 102mm²
- Kabini: ~110mm²
- Apple A7: 102mm²
- Snapdragon 800: 118.3mm²

130mm is a bad estimate for Haswell-Y. Haswell-Y is 126mm² big according to my calculation, means BDW-Y must be 79mm² big by using a 0.63 shrink factor. And as a second proof by using Intels 30x16.5mm for the package size I get 79mm² too. So I doubt BDW-Y is over 80mm².

 

[mikk]

Broxton should be mid-2015 with Gen9 graphics and Goldmont architecture.

Surely not. CHT after all delays is coming Q1-Q2 2015. You should better not expect Broxton for mid-2015. In the last Roadmap Intel already corrected it to end 2015, after the latest delays I don't think Broxton is coming 2015.

 

[liahos1]

Surely not. CHT after all delays is coming Q1-Q2 2015. You should better not expect Broxton for mid-2015. In the last Roadmap Intel already corrected it to end 2015, after the latest delays I don't think Broxton is coming 2015.

broxton is the converged tablet smartphone/tablet part. if they dont introduce it next year they wont have a hero soc for the smartphone market?

also the indiciative pricing of this seems interesting

http://www.engadget.com/2014/06/05/intel-core-m-devices-will-cost-as-little-as-599/

599 for a fanless 2-1 that is somehwere between i3/i5 haswell performance with 9+ hours of battery life. could be really good...maybe cheaper still on android.

 

[Sweepr]

Cherry Trail-T based devices should be available in February and Broxton comes in H2-2015 according to the latest DigiTimes rumour. Broxton should not be affected by the introduction of a new fab process like Cherry Trail-T was and I don't think Intel would slow down their mobile plans +6 months if they really want to be competitive. Also 2016 is way too late for a Moorefield replacement.