Intel Broadwell Thread

[III-V]

IGZO does reduce backlight power. The parts of the display that block backlight are significantly reduced(wires and transistors). IGZO is able to use 50% less power with less flicker.

Low temperature poly silicon is actually one step further than this but is more expensive.

What interests me more is the fact that Intel has been experimenting with similar materials for 10nm.

Cool, didn't know that. Thanks.

 

[AtenRa]

New roadmap,

Edit: 14nm ATOM is one year late. Will Intel stop Contra Revenue in 2014 or will they continue for the first half of 2015 ??

http://www.zdnet.com/what-we-learned-about-intel-this-week-7000036068/

 

[Keysplayr]

Can the OP be updated with potential launch dates for desktop Broadwell? I has probably been said but I don't think anyone wants to sift though 1100+ posts to find that.

 

[Sweepr]

New roadmap,

Edit: 14nm ATOM is one year late. Will Intel stop Contra Revenue in 2014 or will they continue for the first half of 2015 ??

A few days ago they said Cherry Trail should be out by the end of the year so here's hoping we will see some products sometime in Q1/2015. ARM competition has never been stronger (20nm/14nm A57/A53, custom cores, etc), but there's no x86 Mullins successor planned for 2015. They should be able to increase increase their competitiveness a lot on the gfx side and Airmont cores should be extremelly power efficient @ 14nm.

 

[dahorns]

A few days ago they said Cherry Trail should be out by the end of the year so here's hoping we will see some products sometime in Q1/2015. ARM competition has never been stronger (20nm/14nm A57/A53, custom cores, etc), but there's no x86 Mullins successor planned for 2015. They should be able to increase increase their competitiveness a lot on the gfx side and Airmont cores should be extremelly power efficient @ 14nm.

Yeah, that slide was from the investor's meeting. I don't think (but am not sure) it is graphically showing relative availability in 2015. I believe they were just showing the products that would be available in 2015, but silent as to when in 2015. Earlier in the investor meeting they indicated cherrytrail would be out by "end of year." At the time, it seemed they were referring to 2014. Again, it wasn't 100% clear.

If that were a traditional roadmap, it would make no sense. Moorefield is out now. It isn't being released in the second half of 2015.

 

[Arachnotronic]

Yeah, that slide was from the investor's meeting. I don't think (but am not sure) it is graphically showing relative availability in 2015. I believe they were just showing the products that would be available in 2015, but silent as to when in 2015. Earlier in the investor meeting they indicated cherrytrail would be out by "end of year." At the time, it seemed they were referring to 2014. Again, it wasn't 100% clear.

If that were a traditional roadmap, it would make no sense. Moorefield is out now. It isn't being released in the second half of 2015.

Cherry Trail will qualify for production this quarter, per Intel. I'd imagine that it'll be a couple of quarters before we see Cherry Trail in big volumes.

 

[mikk]

Q2 2015 for Cherry Trail then as I said.

 

[witeken]

New roadmap,

Edit: 14nm ATOM is one year late. Will Intel stop Contra Revenue in 2014 or will they continue for the first half of 2015 ??

http://www.zdnet.com/what-we-learned-about-intel-this-week-7000036068/

Intel will stop contra-revenue when they stop selling Bay Trails.

 

[witeken]

Can the OP be updated with potential launch dates for desktop Broadwell? I has probably been said but I don't think anyone wants to sift though 1100+ posts to find that.

(Very) Early spring.

 

[mikk]

(Very) Early spring.

Q3 isn't early spring.

 

[witeken]

Q3 isn't early spring.

Who says it's Q3?

 

[III-V]

Who says it's Q3?

Leaked roadmaps? It's been known to be summer '15 for ages.

 

[mikk]

Who says it's Q3?

CPU-world, Vr-zone, digitimes and maybe some more.

 

[chucky2]

How is it possible Broadwell Desktop is going to be release 3Q15 and Skylake will hit, for the sake of argument, the last day of 2015? Or is it Skylake for Mobile in 2H15 and we can expect Skylake Desktop sometime in 2016?

I wish Intel would just drop Broadwell Desktop and put out Skylake Desktop sometime reasonable in 2H15...

 

[witeken]

CPU-world, Vr-zone, digitimes and maybe some more.

Okay, then Kirk is obviously lying in front of all those investors...

http://intelstudios.edgesuite.net/im/2014/archive/ks/archive.html 28 and some seconds.

Transcript:

When we launch 5th generation Core in early next year, we're going to have a 100X increase in the graphics performance.

Edit, and at the 30 minute mark:

Second is we're launching 5th generation Core, that's the Broadwell products. I'm expecting one of, if not, the fastest ramp of a new process technology in our history. Core i7 down to Celeron, launching early spring.

Edit: he could also be talking about the laptop SKUs, but talking about Broadwell with a 100X graphics improvement (that likely won't happen with Broadwell-U), in a part of the presentation where he's talking about the performance segment with Devil's Canyon and Haswell-E, is unambiguous. Gamers are not going to buy Apple Macbook ultrabooks.

 

[Arachnotronic]

Okay, then Kirk is obviously lying in front of all those investors...

http://intelstudios.edgesuite.net/im/2014/archive/ks/archive.html 28 and some seconds.

Transcript:



Edit, and at the 30 minute mark:

witeken

He's talking about Broadwell-U.

 

[Arachnotronic]

Intel will stop contra-revenue when they stop selling Bay Trails.

...and Clover Trail+.

They mentioned Clover Trail+ having a $15 BoM deficiency, too. Crazy to think that this, perhaps more than anything else, is what has hampered Intel's mobile efforts thus far.

 

[witeken]

witeken

He's talking about Broadwell-U.

Read my edit.

...and Clover Trail+.

They mentioned Clover Trail+ having a $15 BoM deficiency, too. Crazy to think that this, perhaps more than anything else, is what has hampered Intel's mobile efforts thus far.

Does CT+ have contra-revenue?

 

[Enigmoid]

So it looks like 33% gain in graphics.

75% -> 100%

Meh

Edit: The 75% is Iris Pro.

Too early to tell. That would be an amazing gain for GT2 but terrible for GT4.

 

[witeken]

So it looks like 33% gain in graphics.

75% -> 100%

Meh

Edit: The 75% is Iris Pro.

Too early to tell. That would be an amazing gain for GT2 but terrible for GT4.

Broadwell GT4 does not exist. Note that the actual gain in FLOPS for Broadwell is 20% from the additional EUs. If 3DMark06 (the benchmark that gave them this 100x) didn't suffer from (non-EU) bottlenecks (that caused it too perform less than its FLOPS suggested), this is the performance you'd expect.

---
Edit: spring could still mean June at Computex, though. Exactly 1 year after launching DC.

 

[dahorns]

Some more benchmarks out, this time for the Helix 2 (also some others from Dell and HP).

It seems Lenovo has the Helix 2 capped at the same TDP as the Yoga 3. But the scores are more consistent (possible because the Yoga had a different performance setup for each mode).

Anyway,

3 Helix 2 with 5y70 Geekbench scores now available, all are around

2300 single, 3950 multi.

Example: (http://browser.primatelabs.com/geekbench3/1342140)

The interesting thing is the 5y10 model actual performs better in multi:

2069 single, 4271 multi (http://browser.primatelabs.com/geekbench3/1342330)


More information on Helix 2 here: http://forum.tabletpcreview.com/threads/thinkpad-helix-2.63682/page-20

Also available is another Dell Venue 11 pro and an HP Envy x2 (15) with similar scores for the 5y10.

http://browser.primatelabs.com/geekbench3/compare/1331430?baseline=1347151

I really want to see someone push the TDP on the 5y70 to 6.5. May have to wait for Microsoft or Apple to see it.

 

[witeken]

I was reading the AT Chromebook article, and decided to do a quick Google Ark search for the mentioned Haswell CPUs.

Core Celeron 2955U TRAY: $107.00 is in the $180-200 Acer C720.

Core i3 4005U TRAY: $275.00 is in the $350 version.

The i3 costs more than 75% of the whole Chromebook. Ark doesn't say what Core M's price is (anymore?), but it's $281.

Core M pricing really doesn't seem that outrageous, compared to those 2 SKUs (the price of Core M 2-in-1s is more than 2x the price (>$599) than the SoC alone, compared to these's <2x ratio). Core M probably also has a lower BOM.

 

[CP5670]

Can the OP be updated with potential launch dates for desktop Broadwell? I has probably been said but I don't think anyone wants to sift though 1100+ posts to find that.

I would like to know this too. I'm trying to decide whether to jump on a Haswell/Z97 deal right now or wait for this. If we're talking Q1 next year, I would rather wait.

 

[witeken]

I would like to know this too. I'm trying to decide whether to jump on a Haswell/Z97 deal right now or wait for this. If we're talking Q1 next year, I would rather wait.

Q1 is not going to happen. Broadwell is being ramped in multiple (I guess 3) fabs in Q4 (up from 1), and increasingly more significant volume will come out of those fabs for a Broadwell-U release in very early Spring. Broadwell-K will probably launch at Computex, although Kirk suggested early 2015, while others are saying it's Q3.

The main improvements of Broadwell unlocked will be its lower power consumption and much improved Iris Pro graphics. Overclocking potential is unknown.

 

[jpiniero]

I would like to know this too. I'm trying to decide whether to jump on a Haswell/Z97 deal right now or wait for this. If we're talking Q1 next year, I would rather wait.

Still believe it's June.

Core M pricing really doesn't seem that outrageous, compared to those 2 SKUs (the price of Core M 2-in-1s is more than 2x the price (>$599) than the SoC alone, compared to these's <2x ratio). Core M probably also has a lower BOM.

14 nm at this point has to be very expensive, not to mention all the money spent on the specific process for Core M. The bigger problem is that there is no market for >400 non-Apple tablets, and even then the tablet market is being eclipsed by phablets.