Bay Trail benchmark appears online, crushes fastest Snapdragon ARM SoC

[nonworkingrich]

Again, don't underestimate the business side of Intel - they have some of the best and brightest minds available. They want domination of this market, in fact they're hellbent on it.

Maybe they should go back to the drawing board and design something that can actually compete. Like they had better done when they were hellbent on taking over the graphics market.
Have a look here: http://wenku.baidu.com/view/678a3a36dd36a32d73758148.html
Lots of Silvermont Pentiums and Celerons... heavy binning anyone? And that with the world's best fabs with the worlds bestest process. Weird, isn't ist? What would an ARM SoC cost if they had to scrap two thirds because they dont reach the perf/Watt targets?
But lucky Intel can always make money from scrap, because it has its devoted army of OEMs to to shove that garbage down the throat of unsespecting PC buyers. Say goodbye to Ivy Bridge Celerons, be happy to pay the same for slower chips in the future. Thats progress the Intel way... but I guess someone has to make sacrifices when Intel is hellbent on doing cool things. I know it won't be me.
Anybody who thinks Intel could sustain a mobile CPU business with Atom at the usual margins in that market, keep on dreaming. They have to heavily subsidize the mobile business line - Atom for mobile devices can not support itself, in contrary to the ARM SoC business. Or in other words: Atom still cannot compete, but Intel is hellbent on not letting the world know this time

 

[blackened23]

Welcome to the forums on your first post. As far as I can tell, Intel already has ARM SOCs beat in terms of efficiency - the Haswell ULV provides 20x more performance while still being more efficient than ARM SOCs. You don't think the Bay Trail will maintain the efficiency lead? It is on the same process as ULV Haswell with all of the same power and efficiency optimizations, I don't think ARM SOCs have a chance to beat intel (my opinion). I mean, it's debatable, but the product will be released next month fortunately - so it will be interesting to see how things turn out.

 

[Zucker2k]

So Intel is going to go to all these lengths to try to penetrate a market in which it is a minor player, if even that, only to price itself out? Surely, you don't mean it? Intel does what it does in the x86/desktop universe because it can. Things are different in the mobile universe.

 

[nonworkingrich]

Welcome to the forums on your first post.

thx
You surely are familiar with CPU binning?
I dont have the slightest doubt that Intel can output dies that can compete with the latest and greatest ARM SoC in production (produced on a much cheaper process). But I have massive doubts the ratio of competitive dies / incompetitive dies is good. So the question araises: What will happen to the incompetitive ones? Is this business model for the good of someone that is not Intel? Therefore IMO Atom is still no viable alternative to ARM in mobile space.

 

[sushiwarrior]

the Haswell ULV provides 20x more performance while still being more efficient than ARM SOCs. You don't think the Bay Trail will maintain the efficiency lead? It is on the same process as ULV Haswell with all of the same power and efficiency optimizations

Can you explain how Haswell has more efficiency than ARM? Just a few benchmarks and power consumption numbers. Because I can't say I fully agree there.

Bay Trail is a new core design and has none of the "power and efficiency optimizations" from Haswell, it's completely different.

 

[monstercameron]

Can you explain how Haswell has more efficiency than ARM? Just a few benchmarks and power consumption numbers. Because I can't say I fully agree there.

Bay Trail is a new core design and has none of the "power and efficiency optimizations" from Haswell, it's completely different.

how would we[by that I mean reviewers] measure perf/watt cross uarch?

 

[blackened23]

Can you explain how Haswell has more efficiency than ARM? Just a few benchmarks and power consumption numbers. Because I can't say I fully agree there.

Bay Trail is a new core design and has none of the "power and efficiency optimizations" from Haswell, it's completely different.

Well, it's hard to do a direct comparison here due to battery sizes, but you can do rough extrapolations with current products. The current generation Surface Pro with an i5 gets, I believe, around 5 hours of battery life with a similar battery size as the ipad, while the ipad gets near 10 hours of battery life.
A Haswell ULV configured similarly to the current Ivy Bridge Surface Pro *should* get 11-2 hours of battery life - using the macbook air as a guide, the Haswell ULV seems to be easily getting double the battery life of the prior generation IVB counterparts. As well, as far as I can tell, the surface pro is pretty close to ipad size but i'm not sure on battery sizes.

It's hard to do a direct comparison because they are different form factors and sizes, yet considering the Bay Trail will use all of the same power optimizations as Haswell ULV does I really think the Bay Trail will have a commanding lead in terms of efficiency. Even the core Haswell is now competetive with ARM SOCs in terms of battery life - many ARM SOC products are getting 7-9 hours. Haswell can easily meet that - based on that I can see Bay Trail having a commanding lead over ARM SOCs, as mentioned.

** this is all speculation, but we'll see next month I suppose. I concede that I could be entirely wrong, just speculation on my part.

 

[SiliconWars]

Intel has rules for ensuring the most power efficient components are used in the ultrabooks that you read about on AT reviews, this is one reason why there is so much variation between ultrabooks.

They aren't going to be demanding the same thing on their low end. Of course what you'll see reviewed on AT is the best possible outcome because AT only gets sent the best possible systems for review.

 

[Imouto]

** this is all speculation, but we'll see next month I suppose. I concede that I could be entirely wrong, just speculation on my part.

Why you keep repeating this then? It's been a dozen of times just in this thread, several hundreds in this forum since Haswell launch.

 

[sushiwarrior]

Well, it's hard to do a direct comparison here due to battery sizes, but you can do rough extrapolations with current products. The current generation Surface Pro with an i5 gets, I believe, around 5 hours of battery life with a similar battery size as the ipad, while the ipad gets near 10 hours of battery life.
A Haswell ULV configured similarly to the current Ivy Bridge Surface Pro *should* get 11-2 hours of battery life - using the macbook air as a guide, the Haswell ULV seems to be easily getting double the battery life of the prior generation IVB counterparts. As well, as far as I can tell, the surface pro is pretty close to ipad size but i'm not sure on battery sizes.

It's hard to do a direct comparison because they are different form factors and sizes, yet considering the Bay Trail will use all of the same power optimizations as Haswell ULV does I really think the Bay Trail will have a commanding lead in terms of efficiency. Even the core Haswell is now competetive with ARM SOCs in terms of battery life - many ARM SOC products are getting 7-9 hours. Haswell can easily meet that - based on that I can see Bay Trail having a commanding lead over ARM SOCs, as mentioned.

** this is all speculation, but we'll see next month I suppose. I concede that I could be entirely wrong, just speculation on my part.

Okay, I see what you're extrapolating based off of, and I've gotta say your reasoning is completely flawed at best.

1. Macbook Air gained battery life by cutting the clock speed in half
2. Good idle/web browsing battery life =/= good perf/w
3. OSX Battery life =/= windows/android battery life
4. Proper comparisons between ARM and x86 are almost impossible because of platform differences (mainly, OS differences).

If you ran an MBA with an ARM chip, I'm sure it could get 20 hours of battery life at idle. The difference is battery sizes, OS, and use (ie. Haswell could get 12 hours idle, but run it full tilt and it lasts an hour or two).

 

[FwFred]

M
Lots of Silvermont Pentiums and Celerons... heavy binning anyone? And that with the world's best fabs with the worlds bestest process. Weird, isn't ist? What would an ARM SoC cost if they had to scrap two thirds because they dont reach the perf/Watt targets?
But lucky Intel can always make money from scrap, because it has its devoted army of OEMs to to shove that garbage down the throat of unsespecting PC buyers. Say goodbye to Ivy Bridge Celerons, be happy to pay the same for slower chips in the future. Thats progress the Intel way... but I guess someone has to make sacrifices when Intel is hellbent on doing cool things. I know it won't be me.
Anybody who thinks Intel could sustain a mobile CPU business with Atom at the usual margins in that market, keep on dreaming. They have to heavily subsidize the mobile business line - Atom for mobile devices can not support itself, in contrary to the ARM SoC business. Or in other words: Atom still cannot compete, but Intel is hellbent on not letting the world know this time

Have you ever considered Silvermont may be good enough to replace the mobile Celeron/Pentium lineup? If so, why would Intel use expensive Ivy/Haswell dies with a bunch of things fused off? Wouldn't it want to save cost (i.e. higher margin) by selling Bay Trail silicon at a higher TDP?

 

[FwFred]

http://wenku.baidu.com/view/678a3a36dd36a32d73758148.html
Lots of Silvermont Pentiums and Celerons... :

Actually, this would make a pretty sweet fanless netbook or chromebook

Pentium N3510 No QC 2.00 GHz Up to 750MHz 4.5W/7.5W BayTrail-M

 

[nonworkingrich]

Have you ever considered Silvermont may be good enough to replace the mobile Celeron/Pentium lineup? If so, why would Intel use expensive Ivy/Haswell dies with a bunch of things fused off? Wouldn't it want to save cost (i.e. higher margin) by selling Bay Trail silicon at a higher TDP?

I'd rather ask the question if it is as good as the current products - and if not, if price will go down.
Of course Intel wants to command higher margins with the desktop Atoms, that we shall no longer call Atoms. Higher margins (mostly) on the expense of uninformed buyers. Why dont call these things what they are, Atoms?
And if you look at the document I linked to, for Bay Trail Pentiums and Celerons a higher TDP goes together with lower clocks which at least to me hints at some heavy binning going on.

 

[Arachnotronic]

Actually, this would make a pretty sweet fanless netbook or chromebook

Pentium N3510 No QC 2.00 GHz Up to 750MHz 4.5W/7.5W BayTrail-M

That's the point

Full Windows PCs with no fans and great battery life for everybody.

 

[FwFred]

Full Windows PCs with no fans and great battery life for everybody.

I always liked the netbook concept, just not the implementation. I'd be pretty happy with a 10-11" clamshell with at least a 768p screen, 9+ hours battery, and fanless.

Windows XP, Atom 270 (+ancient chipset/graphics), and 1024x600 screens killed what should have been a solid PC, even after the iPad.

 

[Arachnotronic]

I always liked the netbook concept, just not the implementation. I'd be pretty happy with a 10-11" clamshell with at least a 768p screen, 9+ hours battery, and fanless.

Windows XP, Atom 270 (+ancient chipset/graphics), and 1024x600 screens killed what should have been a solid PC, even after the iPad.

Oh, I agree...Intel was far, far too afraid to cannibalize its Core sales with a solid Atom, and the "cheap" parts like the Celeron/Pentium were power guzzling junk that led to bulky POS laptops.

Intel is forced to make the cheap Windows notebook/PC something that is actually desirable because it really has no choice; every moment that Intel dithers on this front, a competitor's chip sells into a device that may be more desirable/within the price point of a competitor, leaving Intel selling nothing.

Intel has ZERO choice at this point - either it bites the bullet and goes all-in on Atom while at the same time trying to make Core products even more desirable, or it will end badly.

From Intel's aggressive roadmap beginning with Silvermont, it seems that the company made the right decision. Intel also realizes that graphics is very important which is why you're seeing them finally go all-out on graphics. Haswell ULT is more GPU than CPU. Something to think about...

 

[Sweepr]

That's the point

Full Windows PCs with no fans and great battery life for everybody.

A 7.5W quad-core Bay Trail-T would be great for a full Windows 8.1 iPad-like tablet. The reference design they presented at Computex kinda shows whats possible: www.youtube.com/watch?v=Xe_nQaIaqmw

 

[Arachnotronic]

A 7.5W quad-core Bay Trail-T would be great for a full Windows 8.1 iPad-like tablet. The reference design they presented at Computex kinda shows whats possible: www.youtube.com/watch?v=Xe_nQaIaqmw

7.5W? No, no...Bay Trail-T will come at a significantly lower TDP since it doesn't need to integrate all of the PC I/O junk that the BYT-M and -D need to,

3W.

 

[Sweepr]

7.5W? No, no...Bay Trail-T will come at a significantly lower TDP since it doesn't need to integrate all of the PC I/O junk that the BYT-M and -D need to,

3W.

4 cores running @ >2GHz and 3W TDP? Impressive.

 

[Arachnotronic]

4 cores running @ >2GHz and 3W TDP? Impressive.

2W SDP, probably 3W TDP (with 4W max power draw as we saw with Clover Trail).

 

[Khato]

You surely are familiar with CPU binning?
I dont have the slightest doubt that Intel can output dies that can compete with the latest and greatest ARM SoC in production (produced on a much cheaper process). But I have massive doubts the ratio of competitive dies / incompetitive dies is good.

Why precisely do you have doubts regarding Intel's yields? There's no question that some amount of binning takes place, but the amount of variation is relatively small. If your basis for doubting Intel's ability to produce low power Baytrail silicon is that chart of Industrial, Mobile, and Desktop SKUs showing TDPs well in excess of what we're told to expect for the Tablet SKUs... why precisely wouldn't that be expected? In non-tablet applications there's no need to limit the design from frequently running at peak turbo frequency.

Can you explain how Haswell has more efficiency than ARM? Just a few benchmarks and power consumption numbers. Because I can't say I fully agree there.

Sadly we'll have to wait for actual useful benchmarks and power consumption numbers. But I can easily tell you how it can happen courtesy of Anand's recent investigation into the i5 vs i7 Haswell MBA, specifically the observed voltages on page 3. (Well, that combined with the little tidbit in the most recent Anandtech Podcast 21 when they were talking about Snapdragon 800 and how these high performance ARM SoCs are running at anywhere from 1.25V to 1.3V under load.) Because an ARM SoC running its cores at 1.3V in order to hit say 2.3 GHz versus Haswell running at 0.842V to hit the same speed... We already know that Haswell idle power consumption is roughly equivalent to current ARM SoCs. Then the basic equation for dynamic power consumption is (activity factor)*(capacitance)*(frequency)*(voltage squared) - the difference in voltage (1.3^2/0.842^2) basically means that Haswell (activity factor)*(capacitance) would have to be 2.38x that of the ARM SoC or greater in order for it to be less energy efficient. Do you really think ARM has some magical properties to its architecture that allow it to be that much better than what Intel can come up with? (And yes, the differences in operating voltage we're seeing here are quite simply obscene - I'll freely admit that I was quite surprised to see operating at voltages that low for Haswell.)

Intel has rules for ensuring the most power efficient components are used in the ultrabooks that you read about on AT reviews, this is one reason why there is so much variation between ultrabooks.

Somewhat true - the 2013 Haswell Ultrabook requirements do include >= 9 hours of windows idle battery life and >= 6 hours of video playback battery life. Now that can be achieved by a larger battery instead of more efficient components... but considering the design aid that Intel is offering for high efficiency I wouldn't be surprised if most Haswell ultrabooks do pretty well in that regard. Battery life is one of the easily marketed characteristics of a system after all.

If you ran an MBA with an ARM chip, I'm sure it could get 20 hours of battery life at idle. The difference is battery sizes, OS, and use (ie. Haswell could get 12 hours idle, but run it full tilt and it lasts an hour or two).

No it wouldn't. According to Anand in the latest Anandtech Podcast 21, when running the same tablet benchmark suite the Haswell based MBA was showing better minutes per Wh of battery capacity than the iPad 4. aka, put an ARM chip in the MBA and battery life would actually decrease by a very small amount. Now sure, under load battery life would increase, but nowhere near so much as the performance would decrease.

4 cores running @ >2GHz and 3W TDP? Impressive.

Indeed it is. After seeing the voltage numbers for Haswell in the i5 vs i7 MBA article I'm relatively convinced that Silvermont was designed to run up to at least 2.4 GHz on only 0.75V. (I've suspected such might be the case ever since Intel's 22nm SoC process briefing included transistor characteristics at 0.75V instead of the usual 1.0V.) It would certainly explain where a lot of the power efficiency gains are coming from.

 

[Piroko]

A Haswell ULV configured similarly to the current Ivy Bridge Surface Pro *should* get 11-2 hours of battery life - using the macbook air as a guide, the Haswell ULV seems to be easily getting double the battery life of the prior generation IVB counterparts.

edit, derped the numbers -
You do have to realize that the MBA only gained 47% idle battery life with all improvements (including different SSD and Wlan modules). Haswell did a healthy jump in efficiency, but doubling the battery life is a massive overstretch.

 

[Vesku]

Why precisely do you have doubts regarding Intel's yields? There's no question that some amount of binning takes place, but the amount of variation is relatively small. If your basis for doubting Intel's ability to produce low power Baytrail silicon is that chart of Industrial, Mobile, and Desktop SKUs showing TDPs well in excess of what we're told to expect for the Tablet SKUs... why precisely wouldn't that be expected? In non-tablet applications there's no need to limit the design from frequently running at peak turbo frequency.

He said why, the prevalence of lower clocked yet with a hefty TDP jump SKUs. This suggests that the truly low power dies are a small subset or that they are making Bay Trail on more than just their ultra-low power 22nm process they've been talking about in their push for mobile.

 

[Dresdenboy]

BTW, Francois Piednoel agreed with David Kanter on Twitter, that the AnTuTu benchmark is not so well suited and can be cheated. A mobile SPEC suite would be better.

 

[inf64]

BTW, Francois Piednoel agreed with David Kanter on Twitter, that the AnTuTu benchmark is not so well suited and can be cheated. A mobile SPEC suite would be better.

I can't wait for the results and comparing them to antutu relative scores for all CPUs involved