Intel Broadwell Thread

[dahorns]

So what's the conclusion on Broadwell Y Core M - is it a failed chip not living up to expectations? And it won't be fixed by better designed cooling solutions than seen so far?

If so, should we give up our hopes of fanless Ultrabooks with decent performance until Cannonlake?

I think it'll clearly perform better than the Yoga 3 Pro with a better design.

But, I also think the best you can hope for is that a production model hits 80% performance of Intel's reference designs. That's probably only if Microsoft or Apple use it for a product. The OEMs will probably only produce things that match Haswell Y in performance, but in slighter thinner packages.

 

[Fjodor2001]

Ok, I see.

Another thing: Broadwell Y Core M is already released in actual products, but we've not seen any review of the CPU itself from reputable sites like Anadtech and Tom's Hardware. What's up with that? I find that really strange, and can't recall any previous major CPU release where those sites did not review new major Intel CPUs at or prior to release. :hmm:

 

[III-V]

Ok, I see.

Another thing: Broadwell Y Core M is already released in actual products, but we've not seen any review of the CPU itself from reputable sites like Anadtech and Tom's Hardware. What's up with that? I find that really strange, and can't recall any previous major CPU release where those sites did not review new major Intel CPUs at or prior to release. :hmm:

Notebookcheck has. AnandTech seems to be suffering from staffing issues atm.

 

[rootheday3]

Pcper had a review up of the Lenovo Yoga 3 as well. Waiting for a review of Dell or some other designs that aren't hamstringing the chip by config tdp down to 3.5w.

 

[Fjodor2001]

Clarification: I'm not looking for reviews of Notebooks containing Broadwell Y Core M, but reviews of the actual chip itself. I.e. uArch, iGPU design, etc. Just like Anandtech and similar sites normally do for new major Intel CPU releases. So far I have seen none of that, which is very strange indeed considering there are actual products with this CPU already released! :\

 

[teejee]

Clarification: I'm not looking for reviews of Notebooks containing Broadwell Y Core M, but reviews of the actual chip itself. I.e. uArch, iGPU design, etc. Just like Anandtech and similar sites normally do for new major Intel CPU releases. So far I have seen none of that, which is very strange indeed considering there are actual products with this CPU already released! :\

The performance of the CPU/GPU will depend on the device due to throttling and configurable SDP/TDP.
But a comparison between different devices would be great too see.

 

[III-V]

Clarification: I'm not looking for reviews of Notebooks containing Broadwell Y Core M, but reviews of the actual chip itself. I.e. uArch, iGPU design, etc. Just like Anandtech and similar sites normally do for new major Intel CPU releases. So far I have seen none of that, which is very strange indeed considering there are actual products with this CPU already released! :\

Yeah, I agree with you on that. You'd think Intel would want to show off what Core M can do, and therefore send some prototypes out for review.

 

[AtenRa]

Sending a better performing and very expensive + overweight Intel Tablet for reviews, will only make every OEM products in the market seem awful and those OEMs not happy :whiste:

 

[III-V]

Sending a better performing and very expensive + overweight Intel Tablet for reviews, will only make every OEM products in the market seem awful and those OEMs not happy :whiste:

I don't see it that way at all. And anyway, I don't think it needs to be a "heavier" tablet.

 

[AtenRa]

I don't see it that way at all. And anyway, I don't think it needs to be a "heavier" tablet.

If you dont use a heavy thick aluminum back plate, Core-M will again perform like all the other products.

 

[III-V]

If you dont use a heavy thick aluminum back plate, Core-M will again perform like all the other products.

It's good to have a reference point. Also, thus far, Core M's biggest issue has been with Lenovo's decision to limit it to 3.5W.

 

[witeken]

So what's the conclusion on Broadwell Y Core M - is it a failed chip not living up to expectations? And it won't be fixed by better designed cooling solutions than seen so far?

I'd like to see Broadwell reviewed by AnandTech.

If so, should we give up our hopes of fanless Ultrabooks with decent performance until Cannonlake?

I don't think so. We also don't even know what Skylake-Y will bring. DirectX 12 might also give a decent boost in graphics performance.

 

[witeken]

Clarification: I'm not looking for reviews of Notebooks containing Broadwell Y Core M, but reviews of the actual chip itself. I.e. uArch, iGPU design, etc. Just like Anandtech and similar sites normally do for new major Intel CPU releases. So far I have seen none of that, which is very strange indeed considering there are actual products with this CPU already released! :\

http://www.anandtech.com/show/8355/intel-broadwell-architecture-preview is still a valid article, but it doesn't contain benchmarks and it lacks additional (in-depth) information that was provided at IDF (e.g. Gen8 architecture, which Ryan Smith told me they were going to review with Broadwell-K).

 

[dahorns]

If you dont use a heavy thick aluminum back plate, Core-M will again perform like all the other products.

the 12.5 inch reference design weighed 670 grams. Significantly less than the surface pro 3. And the iPad air 2 weighs 430 grams with a much smaller screen. Doesn't strike me that the reference design was comparably heavy.

 

[IntelUser2000]

For PC, an SoC integrates the CPU and both North & South bridges. Core-M has integrated the Memory controller and GPU + PCI-e lanes and that is all. Core-M cannot operate alone when installed on a board, it needs an extra chipset for storage, USB etc.

Baytrail can operate alone, simple as that. It is a system on a single chip design, it has everything it needs to operate alone. Thus its called SoC.

AtenRa, while the PCH isn't integrated onto the die, its on the same package. That makes it functionally identical to Atom SoC per chip.

 

[IntelUser2000]

So based on this, do you think Apple will abandon the plans of a fanless 12" MBA based on Broadwell Y Core M?

And is there any chance we'll see that happening using a Skylake chip instead, or will we have to wait until Canonlake at 10 nm for MBA to go fanless?

I don't know. I am not Apple. :\

I think there's doubt whether even at 6W the performance shown in Intel benchmarks can be sustained.

If they want to keep the performance, let alone make it better than the Haswell chip they should use the Broadwell U(honestly though the Core M's results aren't helping favorably towards ANY Broadwell chips but that's another topic). I guess there's a good chance they will do that since new MBA is rumored to be early next year or something right?

One thing will surely gurantee fanless MBAs though. A9/A10 Macbook Air.

 

[AtenRa]

AtenRa, while the PCH isn't integrated onto the die, its on the same package. That makes it functionally identical to Atom SoC per chip.

The PCH may be on the same package but that doesnt make it an SoC, since they are two different dies.

 

[ShintaiDK]

The PCH may be on the same package but that doesnt make it an SoC, since they are two different dies.

Having the PCH ondie/integrated doesnt make it a SoC either but of marketing name.

 

[witeken]

The road to 14nm (and 10)

I was just digging a bit in the Q3'14 conference call, and I saw some interesting comments:

Overall we’ve made some good progress during this quarter. Yet, we have an important work left to do. Within our factory network 14-nanometer yields improved meaningfully, but we’re behind where we expect it to be. These challenges highlight just how difficult it has become to ramp advanced process technology.

But you are seeing in the fourth quarter, you’re seeing the front edge of the startup cost associated with the 10-nanometer and that’s kind of right in line with the historical timing of what you’d expect.

It would be interesting to know what this implies for 10nm HVM. Let's dig deeper. I found this in the Q3'12 call (so that's 2 years ago, Intel should be at the same stage with 10nm if Tick-Tock holds):

Underlying all of this, our manufacturing advantage is extending over the rest of the industry as we start to ramp-up Haswell, our new micro-architecture on 22-nanometer and begin to build-out our 14-nanometer factory network

I am not seeing anything that would cause it to be off of historical patterns. So if you look at what’s happened in odd number of years when we started the new process technology, you’ll see a increase in startup costs that hits pretty significantly in Q1, goes up some more in Q2 and then it will start to come down from there. So I am not seeing anything that would cause that to be different from historical patterns.

For the fourth quarter, we expect gross margin to decrease by 1 point to 61% as we increase factory spending on 14 nanometer.

Going earlier than that doesn't yield good results

Christopher J. Muse - Barclays Capital, Research Division
That's helpful. And as a quick follow-up, you talked earlier about your manufacturing leadership and clearly, no doubt there. I'm curious if you could provide an update on your plans for 14-nanometer ramp in terms of equipment, installation and when you think you'll come to market there?

Stacy J. Smith - Chief Financial Officer, Principal Accounting Officer and Senior Vice President
No.

Paul S. Otellini - Chief Executive Officer, President, Director and Member of Executive Committee
We'll give you some -- better granularity to that in the Analyst Meeting in May. But generally, it's every other year. From a rule-of-thumb standpoint, that's as granular as I think I want to be right now.

Let's analyze this a bit. So, far, nothing too worrisome. Intel obviously hasn't talked about Skylake this call, but they referred to IM. Even if they haven't started HVM, that shouldn't be too bad either since the Broadwell/Skylake cycle has been very irregular. But we do see however the first signs of increased spending on 14/10nm, although this time it's the "front edge", while 2 years ago it was unspecified. Let's go a bit further to see what we can expect in the following quarters (extrapolate by 2 years, reminder):

Q4 '12 (January '13):

Last, but not least, we will begin our transition to 14 nanometers as we begin the world's first 14 nanometer products towards the end of this year. We see the industry in a period of transition and hyper innovation. We are well-positioned to take advantage of these trends across the spectrum of computing from the lowest-power portable devices to the most powerful data center servers and everywhere in between.

And lastly, we built on our manufacturing lead and are well into the ramp of our 22-nanometer factories and will start production on the 14-nanometer process this year.

And, as I look forward into the business in 2014 and 2015 as we finish up the 14-nanometer factories and begin deployment of the construction of and equipping of the 10-factories, we need those factories principally for our view of the computing market, and in that that would include tablets and certainly the data center. So as we look at it, it gets used.

Remember the leading edge capacity is the lowest cost for us on a per unit basis, the highest performance and the lowest power. So regardless of what you think the size of the market is, the world's leading edge fabs are the single greatest asset that we have.

Sure, Hi, C.J., this is Stacy. I will take that. Let me break out that capacity related CapEx in a couple of different ways. First if you look at it by process, it's very much driven by building for the peak of 14 nanometer and then it's the start of the investment and 10 nanometer. So just building on what Paul said, if you think about that, it's really building for units that we expect in 2014 and 2015 because that’s where you start to see the peak of 14 nanometer and the start of the ramp of the 10.

This is the sort of comments you should look after in the Q1'15 earnings call, if Intel will start HVM in 2015 (replace 14 by 10):

Building on this leadership, we are on track to start production on 14 nanometer process technology in the back half of this year.

As we said at the beginning of the quarter, [...] and we had [in Q1] a big increase in startup costs as you kind of expect, based on where we are in the 14 nanometer ramp

We’ve also got the technology transition to the 14 nanometers. [unintelligible] a first order, all of our spending is focused on 14 nanometers , which gives us a fairly significant ramp capability. If demand for older products exceeds what we could build on 14, we could still build 22 for quite some time. So I really think it depends on whatever demand scenario you see out there. In any event, the most important thing for us is to make that transition to 14 and continue to have the leading edge.

We have 14-nanometer in its final stages of development [May 17], ready for production at the end of this year and moving into next year.

That was Paul Otellini up to here, BTW.
Q2'13:

We will try and bring those updates to our Investor Meeting that happens in November. As far as our 14 nanometer Core launch in our – just our general product launch, I think what we’ve said so far is, first half of 2014 and we’re not going to – we’re not ready to give any specifics beyond that.

So that should give some references as you'll hear about 10nm. At this time, the lack of information of 10nm is probably because they're still busy with 14nm and maybe also because BK doesn't want to talk about these long term things. I see no reason why 10nm couldn't launch in (early) Q3 of 2016.

 

[Nothingness]

Having the PCH ondie/integrated doesnt make it a SoC either but of marketing name.

Yes, but this packaging results in:

1. higher power consumption
2. bigger latency
3. larger package.

than what the competition, wrongly or not, calls a SoC.

 

[monstercameron]

http://www.notebookcheck.com/Test-HP-Envy-x2-j001ng-Convertible.129799.0.html
another broadwell review looks decent!

not trying to troll but flames: So broadwell is like a sexier beema equivalent apu?

benchmark-------| beema | broadwell
3dmark physics--| 19842 | 20946
3dmark graphics-| 35540 | 34908
power max-------| 24.8W | 18.2

http://www.notebookcheck.net/Lenovo-B50-45-Notebook-Review-Update.126712.0.html
http://www.notebookcheck.com/Test-HP-Envy-x2-j001ng-Convertible.129799.0.html

 

[Kallogan]

How disappointing. Nothing to see here. Next.

 

[monstercameron]

How disappointing. Nothing to see here. Next.

How so?

 

[Arachnotronic]

http://www.notebookcheck.com/Test-HP-Envy-x2-j001ng-Convertible.129799.0.html
another broadwell review looks decent!

not trying to troll but flames: So broadwell is like a sexier beema equivalent apu?

benchmark-------| beema | broadwell
3dmark physics--| 19842 | 20946
3dmark graphics-| 35540 | 34908
power max-------| 24.8W | 18.2

http://www.notebookcheck.net/Lenovo-B50-45-Notebook-Review-Update.126712.0.html
http://www.notebookcheck.com/Test-HP-Envy-x2-j001ng-Convertible.129799.0.html

That's a pretty bad 3DMark Physics score...Moorefield in a smartphone does better.

 

[dahorns]

That's a pretty bad 3DMark Physics score...Moorefield in a smartphone does better.

Maybe. Rest of the numbers seem pretty nice for a truly fanless device. Cant complain too much about being able to run bioshock infinite.


Battery life could be higher. It seems to do relatively better in the more intense battery loads than the simpler loads. That is an interesting result.