Bay Trail benchmark appears online, crushes fastest Snapdragon ARM SoC

[Imouto]

@blackened23

Intel better ensure to buy a huge warehouse to carefully stockpile all those amazing chips.

 

[dealcorn]

In due course, independent testing will clarify who is in denial. However, I want to to overstate one common blind spot which prevents most folks from recognizing Intel's threat to ARM. If Intel has to build a new fab, it's most current process node is cheapest. At 22nm, Intel is on record capping the bottom of BT's tablet ambitions at $200. ARM has lots of cheaper stuff.

Intel's cheapest transistors come from fully depreciated n-1 fabs. Assuming Intel's 32nm fabs become fully depreciated about the time BT goes live, that means Intel's cost to produce a CT+ SoC drops a lot, say 50% for illustrative purposes. While BT is tasked with removing ARM from mainstream (non gaming) tablets that cost more than $200, CT+ becomes available to trash the profitability of many ARM middle market SoC's that go into tablets costing less than $200.

ARM should retain tablet market share at SoC price points like $10-$12.50 which are produced at n-1 fabs. However, that market is not sufficient to support the substantive investment in better process technology and more capable (not piggy) designs necessary to be a credible competitor.

The rear view mirror is not the best way to look forward.

 

[SOFTengCOMPelec]

There is no room for anyone but 1 in the endgame. The IC design might actually be the most expensive and not the fab.

I tend to agree, a lot with that.

Many, many years ago, we sort of already had it happen in the computer industry with operating systems.

Ok, there are alternatives to Microsoft Windows, such as Apples (mac)OS-X, Linux, Googles android (sort of, as more geared towards hand held devices), etc etc.

But, to many home consumers, a desktop Windows(R) is what a home PC or business PC is all about.

But, I'm not convinced that 'Microsoft's' (what I think is) a MONOPOLY, is a good thing for consumers and/or businesses, that need/want to use Microsofts products.
E.g. Pricing seems to be on the high side.
Microsoft deciding annoying rules, e.g. Any Xbox-One used games belongs to "Microsoft" and/or CAN'T sell it, tough. MUST always connect to internet before/during playing games. etc. (policy later retracted, I know).

 

[ShintaiDK]

I tend to agree, a lot with that.

Many, many years ago, we sort of already had it happen in the computer industry with operating systems.

Ok, there are alternatives to Microsoft Windows, such as Apples (mac)OS-X, Linux, Googles android (sort of, as more geared towards hand held devices), etc etc.

But, to many home consumers, a desktop Windows(R) is what a home PC or business PC is all about.

But, I'm not convinced that 'Microsoft's' (what I think is) a MONOPOLY, is a good thing for consumers and/or businesses, that need/want to use Microsofts products.

Not sure why you compare it with software. Software uses an entirely different business model and got no entry barrier.

 

[SOFTengCOMPelec]

Not sure why you compare it with software. Software uses an entirely different business model and got no entry barrier.

It has similarities, because it shows what can happen, when the range of suppliers, ends up being ONE.
Also, it is similar, because developing new operating systems (big scale, like a new version of Windows), needs huge resources, making it very difficult for competitors to join the market. Just like developing a new cpu architecture.

But, I agree, they are significantly different, in a number of ways.
I was trying to show an example/analogy/similarity between them.

 

[Arachnotronic]

@blackened23

Intel better ensure to buy a huge warehouse to carefully stockpile all those amazing chips.

Why? They'll be in tablets delighting both Android and Windows fans this holiday season. I know that Bay Trail is the tablet platform I've been waiting for...7" Windows 8 tablet with performance that blows the doors off of what's out there currently?

Yes, please

 

[Arkaign]

Intel is hell bent on efficiency improvements. They also have more R&D and fab investment/expertise than basically everyone else in the world put together.

It's not a question of 'if' they overtake the competition in every metric, but 'when'. I'm not even a fan of Intel in any way. I couldn't possibly care less whose name is on 'that' company's buildings and products. But facts are facts.

There are multiple reasons that Intel as yet hasn't blown away the phone/tablet world, chiefly their components were too expensive, too power hungry/hot, and too large to be competitive in those spaces. Every single one of those metrics is about to go out the window, and within 24 months we're going to see ramping up to 10nm. What will this mean? This will mean products that absolutely sip power, while enabling more performance from smaller/cooler implementations. Think something in the power range of a 2Ghz i7 in a form factor like an iPhone or Galaxy phone, with battery life that exceeds anything currently on the market. Followed up with 7nm, 5nm and beyond.

The only logical alternative is someone else matching their immense investments in R&D and fabrication efforts. That is extremely unlikely. Intel's expansion in Israel alone is shown to be probably 10 billion+.

Not 'if'.

When.

 

[Imouto]

Why? They'll be in tablets delighting both Android and Windows fans this holiday season. I know that Bay Trail is the tablet platform I've been waiting for...7" Windows 8 tablet with performance that blows the doors off of what's out there currently?

Yes, please

Like all those amazing Tegra 4 chips.

 

[Arachnotronic]

Like all those amazing Tegra 4 chips.

Bay Trail is faster than T4, supports full Windows 8 (hello Windows 8 chromebook like cheap notebooks), and will also be headed to Android.

 

[Imouto]

T4 is on par with the top tier from Qualcomm and Samsung yet it didn't score any major design win. Having a good chip doesn't guarantee you a thing. Chip makers could play the same game as Intel back when it bribed all the major OEMs... wait, all the major OEMs are also chip makers now.

They're all allied to leave Intel out. The HSA foundation is just a visible front.

And Windows 8 is a failure, get it already.

 

[Arachnotronic]

T4 is on par with the top tier from Qualcomm and Samsung yet it didn't score any major design win. Having a good chip doesn't guarantee you a thing. Chip makers could play the same game as Intel back when it bribed all the major OEMs... wait, all the major OEMs are also chip makers now.

They're all allied to leave Intel out. The HSA foundation is just a visible front.

And Windows 8 is a failure, get it already.

Is Qualcomm an OEM? No? It sells all of its products to device vendors. So your argument is pure malarkey.

 

[blackened23]

Like all those amazing Tegra 4 chips.

T4 lost all of it's design wins because the Snapdragon 600/800 are better by every metric. And the T4 is late, with nvidia scrambling to modify the chip for LTE support - basically nvidia screwed the T4 up in every way fathomable. That's why Qualcomm is getting design wins over the T4. Do you ever wonder why nvidia suddenly decided to license Kepler GPU technology? They need a revenue stream (on top of their dGPU revenue) since the T4 isn't going to sell as they had hoped.

I find it pretty funny that the Bay Trail is going to crush every ARM SOC by every metric, but for some reason, you think they won't sell. Whatever, if you say so. The prior atom didn't sell because it was a terrible chip. Intel went to the drawing board while being hellbent on beating ARM at their game, only this time, they're going to crush ARM at their own efficiency game. "Intel will never beat us in efficiency". Seeing ARMH's hubris catch up to them will be quite funny indeed.

 

[SiliconWars]

Intel is hell bent on efficiency improvements. They also have more R&D and fab investment/expertise than basically everyone else in the world put together.

It's not a question of 'if' they overtake the competition in every metric, but 'when'. I'm not even a fan of Intel in any way. I couldn't possibly care less whose name is on 'that' company's buildings and products. But facts are facts.

There are multiple reasons that Intel as yet hasn't blown away the phone/tablet world, chiefly their components were too expensive, too power hungry/hot, and too large to be competitive in those spaces. Every single one of those metrics is about to go out the window, and within 24 months we're going to see ramping up to 10nm. What will this mean? This will mean products that absolutely sip power, while enabling more performance from smaller/cooler implementations. Think something in the power range of a 2Ghz i7 in a form factor like an iPhone or Galaxy phone, with battery life that exceeds anything currently on the market. Followed up with 7nm, 5nm and beyond.

The only logical alternative is someone else matching their immense investments in R&D and fabrication efforts. That is extremely unlikely. Intel's expansion in Israel alone is shown to be probably 10 billion+.

Not 'if'.

When.

TSMC is already catching up on Intel's capex spending, and all they do is build fabs. They may already spend more on fabs than Intel does, certainly with globalfoundries both easily overtake Intel's fab spending. Samsung is about equal to Intel as well (not on fabs obviously).

Your problem is you read too much on Anandtech forums and not enough elsewhere.

 

[Sweepr]

Bay Trail-T SoCs for tablets will be released from August 28 to September 13. The slide does not show their processor numbers, but it does list Bay Trail-M SKUs, that will be available "after PRQ".

Wow, never though Bay Trail was coming that soon. Heres hoping we'll get some nice x86 7-8.9" Windows 8.1 tablets for a reasonable price.

 

[MisterMac]

Is Baytrail supposed to include Intel GEN gpu XTORS?


...Will be fun to see final silicon @ 2,4ghz and some power numbers.

That francois piednoel guy must be having a blast from his little secret twitter marketing pr chamber!

 

[sontin]

T4 is on par with the top tier from Qualcomm and Samsung yet it didn't score any major design win. Having a good chip doesn't guarantee you a thing. Chip makers could play the same game as Intel back when it bribed all the major OEMs... wait, all the major OEMs are also chip makers now.

They're all allied to leave Intel out. The HSA foundation is just a visible front.

And Windows 8 is a failure, get it already.

Yeah, what?
Toshiba, HP and Asus announced Tablets with Tegra 4. Toshiba is shipping them in Japan since 2 weeks.

 

[SiliconWars]

Yeah, what?
Toshiba, HP and Asus announced Tablets with Tegra 4. Toshiba is shipping them in Japan since 2 weeks.

And yet Tegra continues to leak cash for Nvidia. Tegra is a disaster that is sold at firesale prices and that's why it gets any design wins at all, get over it already Sontin.

 

[Imouto]

Is Qualcomm an OEM? No? It sells all of its products to device vendors. So your argument is pure malarkey.

Samsung and Apple account for 50% of market share, Qualcomm is likely to sell way cheaper chips than Intel targetting cheaper OEMs. I don't see Amazon trying to make its Kindle the very best out there.

What's left for Intel?

That leaves OEMs willing to draw a wild card like LG, ASUS, ZTE and Huawei or failures as Nokia and BlackBerry.

T4 lost all of it's design wins because the Snapdragon 600/800 are better by every metric. And the T4 is late, with nvidia scrambling to modify the chip for LTE support - basically nvidia screwed the T4 up in every way fathomable. That's why Qualcomm is getting design wins over the T4. Do you ever wonder why nvidia suddenly decided to license Kepler GPU technology? They need a revenue stream (on top of their dGPU revenue) since the T4 isn't going to sell as they had hoped.

It's still a "good enough" chip. An inbreed console and a lol HP lol device? That chip deserved better.

I find it pretty funny that the Bay Trail is going to crush every ARM SOC by every metric, but for some reason, you think they won't sell. Whatever, if you say so. The prior atom didn't sell because it was a terrible chip. Intel went to the drawing board while being hellbent on beating ARM at their game, only this time, they're going to crush ARM at their own efficiency game. "Intel will never beat us in efficiency". Seeing ARMH's hubris catch up to them will be quite funny indeed.

And you deduced all that by a skewed bench score, oh my.

Intel isn't selling in the mobile market because everyone is actively avoiding it. It could make its own branding or pay some OEMs to make a device with the "Intel Inside" logo. Nothing, nada. Intel is out of the mobile space.

None of the top OEMs will go back to chains with Intel and no one will change the perception of lesser OEMs as being cheap chinese junk.

Yeah, what?
Toshiba, HP and Asus announced Tablets with Tegra 4. Toshiba is shipping them in Japan since 2 weeks.

Yeah, don't forget to add those Tesla cars with 50k units sold. JFYI the first Toshiba tablet to show up at market share charts is at 0.5% and HP lagging behind at 0.1%. Hell yeah.

 

[inf64]

I hope Exophase won't mind me quoting him from seekingalpha

ntel Breaks ARM, Sends Shares Down 20% [View article]

To those who are following the latest blitz of performance numbers of Saltwell vs A15, pushed by Intel, obvious proxy analysts behind paywells, and Ashraf's posts like this one: take heed, because AnTuTu doesn't really tell the full story. And if it wasn't already obvious this study is just running AnTuTu.

First, just to get some context, look at any other CPU bench for Saltwell vs A15 or similar processors - Geekbench, Phoronix, 3DMark physics, even the usual Javascript nonsense is favoring A15 in peak perf across the board. Yet when it comes to AnTuTu we find that it takes twice as many A15 cores to match Saltwell, and this is with a benchmark that scales unreasonably well with core count by simply duplicating run instances. How can this be the case? By virtue of the closed nature of this benchmark it's hard to tell entirely, but you can tell this much from poking around the binary:

a) The x86 target is compiled with ICC
b) The ARM target only supports untweaked NDK GCC - more importantly, just VFP w/16 registers and no NEON path is supported

Normally Intel would have some merit in making a better compiler available, but ICC is not a real option for most Android developers because it's not part of the NDK. It's strange that it's there in the first place - almost certainly a result of Intel paying AnTuTu to use an exotic custom solution, possibly one that Intel helped implement. And there's no good reason for AnTuTu not to support a NEON-capable path (NDK binaries will happily include both)

AnTuTu's CPU tests appear to be based on nbench, an ancient, highly synthetic benchmark that Intel would have had years to crack with special compiler paths. It's not even remotely representative of any real performance characteristics. The difference that Intel's vectorization vs no vectorization can make for these toy functions can be very significant.

If you want to dismiss this and acknowledge the scores anyway go ahead, lots of other popular benchmarks are very bad too. But you should at least not draw sweeping conclusions from a single benchmark, especially when used in marketing pitches.

There you go, the AnTuTu benchmark is BS essentially.

 

[sontin]

Yeah, what a luck for the ARM world that Intel gets only the same design wins:
Asus, Acer, HP, Toshiba...

 

[blackened23]

Did asus announce a T4 tablet? I was under the impression that their zenbook infinity will also be using the Qualcomm SD 800.

Also, nearly 70% of all Tegra 3 sales were the Google Nexus. Now that Google swapped to the Snapdragon, that's going to be a huge and painful drop in revenue. If Toshiba and HP are the extent of current T4 design wins, I wish nvidia luck. They'll need it.

 

[Roland00Address]

Did asus announce a T4 tablet? I was under the impression that their zenbook infinity will also be using the Qualcomm SD 800.

http://www.anandtech.com/show/7014/asus-new-transformer-pad-infinity-tegra-4-2560-x-1600-ips-display

 

[Sweepr]

Did asus announce a T4 tablet? I was under the impression that their zenbook infinity will also be using the Qualcomm SD 800.

Also, nearly 70% of all Tegra 3 sales were the Google Nexus. Now that Asus/Google swapped to the Snapdragon, that's going to be a huge and painful drop in revenue. If Toshiba and HP are the extent of current T4 design wins, I wish nvidia luck. They'll need it.

Zenbook Infinity uses i7 4558U, an ULT Haswell 2C/4T @ 2.8-3.3GHz with Iris graphics (~2x ULV HD4000 performance). A design win like a Nexus tablet would be huge for Intel. Bay Trail is certainly coming to x86 Windows 8.1 tablets, but they also need to at least try to get some Android marketshare.

 

[AtenRa]

The only problem i see is the x86 license, everyone can design/manufacture ARM, only Intel can manufacture ATOM.

 

[aigomorla]

The only problem i see is the x86 license, everyone can design/manufacture ARM, only Intel can manufacture ATOM.

well AMD also has the x86 license.


And it seems like intel isnt giving AMD any breathing room dude.
Neither is it giving TSMC any nudging room either.

They straight up bulldozed though the walls and said lookie here, you guys are in for a beating.

It looks like a bigggggggggg chip