Intel SoFIA & Broxton Killed

[Shivansps]

So time to buy z8000 devices while we still can?

 

[Ajay]

Hmm, can modern medicine get Andy Grove back in fighting condition?

 

[Arachnotronic]

Hmm, can modern medicine get Andy Grove back in fighting condition?

No cure for death yet

 

[CHADBOGA]

Unbelievable how bad Intel has been executing in the last few years.

 

[VirtualLarry]

Anyway, X86 in phones is done. The ARM ISA has won out here.

I blame Microsoft here, to a significant degree, for not innovating enough in the X86 phone space, and creating a value solution for business users that would let them use Windows apps on their phone.

Or maybe they have innovated, with Universal Windows Apps, but the hardware just wasn't there? I don't know.

I thought that the only significant builds of the Windows Phone OS were ARM though, and wouldn't run Windows (Win32/64) apps.

 

[Exophase]

I blame Microsoft here, to a significant degree, for not innovating enough in the X86 phone space, and creating a value solution for business users that would let them use Windows apps on their phone.

I don't see this making a huge difference. Microsoft has failed to capture the phone market about as badly as Intel has. Being able to run legacy Windows apps on a phone would not have been a big differentiator. All this time there could have been more x86 phones running standard Windows without MS's oversight but we've seen at most a small handful of exotic offerings (AFAIK marketed to Japan)

Or maybe they have innovated, with Universal Windows Apps, but the hardware just wasn't there? I don't know.

I thought that the only significant builds of the Windows Phone OS were ARM though, and wouldn't run Windows (Win32/64) apps.

UWP has a fat binary format, so the "same" app can run on both Windows Phone and Windows 10 PCs (and XBox One). Windows Phone can't run legacy Windows programs, but at least some new UWP apps can run on all of these platforms.

Not a very groundbreaking technology, it's the same for Android for instance.

 

[Ajay]

No cure for death yet

Heh, I almost said resurrect. I didn't know he passed this past March.
Guy was a freaking amazing CEO.

 

[Arachnotronic]

I blame Microsoft here, to a significant degree, for not innovating enough in the X86 phone space, and creating a value solution for business users that would let them use Windows apps on their phone.

Or maybe they have innovated, with Universal Windows Apps, but the hardware just wasn't there? I don't know.

I thought that the only significant builds of the Windows Phone OS were ARM though, and wouldn't run Windows (Win32/64) apps.

I think the idea of a "full windows phone" is highly overrated. There's a reason smartphones run the types of apps that they do. The UIs are fundamentally touch based.

 

[Arachnotronic]

Heh, I almost said resurrect. I didn't know he passed this past March.
Guy was a freaking amazing CEO.

Yup.

 

[cbn]

I think the idea of a "full windows phone" is highly overrated. There's a reason smartphones run the types of apps that they do. The UIs are fundamentally touch based.

In the US such a phone wouldn't be a big deal. (re: laptops and desktops abound)

But in some other less developed country, I would think having the extra utility would be welcome.

 

[IntelUser2000]

Was there Broxton SoC intended for both smartphones and tablets?

Originally, Broxton was supposed to be the SoC that brought core convergence to Smartphones/Tablets.

It looks like they are giving up the mobile market completely, Smartphones and Tablets. Apollo Lake seems a super low end PC deal.

Heh, I almost said resurrect. I didn't know he passed this past March.
Guy was a freaking amazing CEO.

I am not sure if he would have made a difference. He was a right person at the right time.

This is his greatest achievement: "As a result, he chose to discontinue producing DRAMs and focus instead on manufacturing microprocessors."

In order to continue, Intel needs to do it once again. I would put my bet on switching the fundamental computing architecture to a new one. I think it starts with Opt...

 

[StrangerGuy]

The hardware side of good enough consumer computing has been commoditized to oblivion and it's staring Intel right in the face for a while now. Software is the last holdout, even then we are talking about apps with networked ecosystems like Facebook than standalone calculator app #342424234243.

 

[Exophase]

Originally, Broxton was supposed to be the SoC that brought core convergence to Smartphones/Tablets.

It looks like they are giving up the mobile market completely, Smartphones and Tablets. Apollo Lake seems a super low end PC deal.

But isn't Apollo Lake a better tablet SoC than Clover Trail? Why would vendors not want to refresh with it? Why would Surface 4 not be using it? I can't see anything that it lacks, eg it supports LPDDR3/4.

Intel's reference platform is even a convertible.

The only place it matter is with providing data connections, but using an SoC with integrated LTE means the tablets would need to have variants with a non-LTE version of the SoC (since most tablet sells are wifi-only), as opposed to just not populating the cellular parts. So that may not be that attractive anyway, unless the data model has huge volume which isn't really true for anything that isn't iPad.

 

[ksec]

I still have a hard time believing Intel has gained anything on Apple's modem business.

 

[IntelUser2000]

But isn't Apollo Lake a better tablet SoC than Clover Trail? Why would vendors not want to refresh with it? Why would Surface 4 not be using it? I can't see anything that it lacks, eg it supports LPDDR3/4.

Well, I can only guess they are continuing the tradition that started with Clover Trail/Pine Trail.

Wouldn't your logic be similar to "why not use Braswell"?

Though looking at the presentation again, I am not sure what the differences are. Why do they aim it at "Hybrids and 2-in-1 Tablets" rather than pure Tablets?

What about the recent GFXBench result with Broxton Pro? Which article is the right one? Did they cancel Broxton because based on GFXBench results its "too competitive" with Core?

 

[ksec]

Well, I made my reply already yesterday : http://forums.anandtech.com/showthread.php?t=2471470.

I think they might try again once they have their 5G modem up to snuff and Core has shrunk a few more times, so like in the 2020 timeframe on 7nm. At that time, the PC shouldn't be hurting their bottom line anymore so much (and give relatively stable base income), their IoT and NAND+3D XPoint projects should have grown to a bigger size and be taking off and together with the whole DCG business (at >10% per year with silicon photonics, Xeon Phi, networking, FPGA and omni-path all established), Intel *should* be healthy and growing at a decent clip.

But if you're already spending at a $12B R&D rate, and you know that you need ever more money for Moore's Law, you know that you can't invest in businesses that don't contribute meaningful to the bottom line IF you're not growing like in 2015, you know you must reshuffle those resources to the profitable segments, which is about what Intel has done now.

So if you're an investor, February was a good time to go long, I'd say, at least if they start delivering from now on. (How many years is BK now talking about the "transformation" of the company to set it up to get growing, and how long have those "great" products been on the horizon, always 12-18 months ahead until we see they're delayed even further.)

5G is also a very broad term for Intel, it includes everything from Software Defined Network using their Chip, to Cloud RAN off load 5G processing to Cloud using Intel's Chip rather then on site, Connection and Interconnetion using MXC Silicon Photonic technology. All of these will have synergy with Datacenter Group as well. And of course the 5G modem business.

It is also worth mentioning what 5G means here, because a a lot of the above are already happening as we speak in terms of 3GPP Rel 13 and 14. Which technically speaking is more like Pre 5G or 4.9 G era.

I dont mind Intel's modem not being as good / features rich as Qualcomm, but they should at least have it on their 14nm Fab instead of relying on TSMC.

 

[Vesku]

Will they be disbanding their wireless modem team? Seems odd to be touting IoT and not continue on the path to very low power one chip, or perhaps one interposer, solutions which would include wireless communication.

Instead fully focusing on the server aspect of the IoT + datacenters future vision?

 

[coercitiv]

It looks like they are giving up the mobile market completely, Smartphones and Tablets. Apollo Lake seems a super low end PC deal.

There is no giving up on tablets: both iOS and Android will use the tablet as an entrance vector into the productivity space. Unless Intel has a clear strategy to deliver better perf/watt in passively cooled devices with lower platform cost than today's Core M, they'll be giving up on more than tablets.

 

[Thala]

I blame Microsoft here, to a significant degree, for not innovating enough in the X86 phone space, and creating a value solution for business users that would let them use Windows apps on their phone.

This is just wrong. Why would you force Microsoft to use an non-competitive SoC in their Smartphones? It is in the first place Intels task to deliver.

 

[MarkizSchnitzel]

In the US such a phone wouldn't be a big deal. (re: laptops and desktops abound)

But in some other less developed country, I would think having the extra utility would be welcome.

As a huge windows fan, there is not a single Win32 app I would want on my WP. What is the point? There are better alternatives on android/ios for every use case.
Which is why I'm on my last WP, after 5 years, x86 or not.

Business apps is a different story, but even so I don't think it's such a big deal, until donekle gives a few cery solid examples.

 

[bhtooefr]

There is no giving up on tablets: both iOS and Android will use the tablet as an entrance vector into the productivity space. Unless Intel has a clear strategy to deliver better perf/watt in passively cooled devices with lower platform cost than today's Core M, they'll be giving up on more than tablets.

The OP is claiming that they're giving up on Broxton tablets. (The question then becomes, is this meant to be tablet applications of the chip formerly known as Morganfield (which is the smartphone Broxton, comparable to Moorefield) - note that some Silvermont tablets were using Moorefield instead of Bay Trail-T? Or, is it both Morganfield and Willow Trail (which was the tablet Broxton, comparable to Bay Trail-T or Cherry Trail)?)

So, does that mean that low-cost tablets get Apollo Lake instead? That changes the thermals significantly compared to Broxton, presumably. (Unless, of course, they can get the TDPs on Apollo Lake down to Broxton levels, at which point it may actually be a better part than Broxton for some tablet applications - SATA SSDs become possible, there's no "you must use either eMMC or NVMe".)

Does that mean that Intel just keeps producing Cherry Trail until the big core-Y SKUs are cheaper to produce?

Or, does that mean that Intel starts making a KBL-Y Celeron SKU with a core (so it's a hyperthreaded single core) and most of the GPU fused off? But, that'd likely be slower than Broxton or even Cherry Trail, and increase the cost. Then again, if they can get the PCH on-die...

At this point, though, it sounds like the floor for an x86 application processor's TDP is now going to be approximately 3.5 W, with a 600 MHz cTDP-down Core m.

 

[scannall]

Then Apple may have still ended up ditching them for in-house designs after a couple generations.

Apple didn't start building their CPU team until after Intel told them no. The ARM chips on the market weren't all that hot at the time either. It is very Apple-like to say 'screw it' and design their own if they can't buy the parts they want. And they have, to great success.

 

[jpiniero]

Apple didn't start building their CPU team until after Intel told them no.

You mean "bought PA Semi"

 

[scannall]

You mean "bought PA Semi"

You have to start somewhere. Don't forget Intrinsity as well. It's a fast way to get great engineers, who already know how to work together.

 

[StrangerGuy]

You have to start somewhere. Don't forget Intrinsity as well. It's a fast way to get great engineers, who already know how to work together.

Yeah, it's so easy to make great ARM SoCs by buying both companies up that somehow a pure chip vendor like Intel didn't do it when they had the chance, yet they have no qualms losing billions into the black hole of their mobile division. Or to think PA Semi/Intrinsity would have been remotely successful without Apple's war chest.

Quoting Emperor Palpatine: "You will pay the price for your lack of vision."