Mobile Haswell achieves all day battery life - beginning of the end for ARM?

Page 3 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

CHADBOGA

Platinum Member
Mar 31, 2009
2,135
832
136
At least the rather dubious person you replied to was banned as well.
Not me who replied, that was NTMBK.

However that was the dubious person's multiple recent post of that nature and NTMBK's first in God knows how long.

If NTMBK hadn't replied, the dubious guy would have been allowed to once again get away with it, because he follows the forum orthodoxy.
 

Blandge

Member
Jul 10, 2012
172
0
0
IT'S like you didn't read the thread at all. Point out one post in which someone said tablets are a fad. Go on, do it. I think someone else is delusional.

It's great when someone jumps in for a post and doesn't even read the thread for context.

I did indeed read every post before I wrote that. After considering the wording carefully for a few minutes I went back and revised out the initial statement. It was not my intention to come off so hostile.
 

Blandge

Member
Jul 10, 2012
172
0
0
While I agree with the direction, I don't think it will happen "by 2014." That's less than a year, and there's a huge discrepancy in many aspects between ultrabooks and tablets.

I didn't mean the beginning of 2014. Broadwell in 2014 will likely introduce fanless Core based ultrabook/tablets. This is the the key.
 

Puppies04

Diamond Member
Apr 25, 2011
5,909
17
76
Not me who replied, that was NTMBK.

However that was the dubious person's multiple recent post of that nature and NTMBK's first in God knows how long.

If NTMBK hadn't replied, the dubious guy would have been allowed to once again get away with it, because he follows the forum orthodoxy.

I actually just signed up over there (get all my hardware forum joys here usually). I might give it a miss though.

Anyway back to topic.

Is there a NDA relating to mobile haswell chips that we know about yet or is it all still under wraps. I asked at my supplier earlier today and they haven't heard a thing about any release dates for haswell.
 

NTMBK

Lifer
Nov 14, 2011
10,248
5,045
136
Hey, there are people on AlienBabelTech who call this place AMD biased too :p S|A does have its number of dissenters, like the guy who insists that the new consoles can't possibly be AMD x86 and must be POWER based, etc. They tend to get a lot of AMD fans due to the type of article that Charlie writes, but I don't notice the mods being that biased as such. In all fairness a mod in that thread had just warned us not to get personal and I totally ignored that, so I honestly had it coming. I could have rebutted his comments without personal remarks.
 

scannall

Golden Member
Jan 1, 2012
1,946
1,638
136
I remember when the desktop pc was called a toy computer.

I do too. When I got started in the computer industry in the mid 70's everyone thought I was wasting my time. Computers in the home? Preposterous! Computers are just for banks and governments. I didn't care though. I thought they were wrong, and my passion and fascination kept me on course. It does make it nicer that I was right though. ;)
 

zephyrprime

Diamond Member
Feb 18, 2001
7,512
2
81
I will also believe it when i see it. Haswell tablets are still a ways off.

Plus, the benchmarks above are very misleading. The benchmark in the graph is heavily dependent on the browser/OS used. The Nexus 7 and Surface RT above use practically the same Tegra 3 chipset (T30L versus T30) but the Surface RT takes nearly half the time to complete the benchmark.

True enough and I believe IE 10 is the only browser on the surface RT but the chart also includes the a Surface Pro run using Chrome which is the default browser on the Nexus 7 so in that case, it's pretty much an apples to apples comparison.
 

Roland00Address

Platinum Member
Dec 17, 2008
2,196
260
126
Unless intel can design and produce haswell and successor chips at $10 a piece, it won't compete with arm where it matters. Sure, it'll compete with arm in performance and eventually battery life, but not in price.

But atom in the other hand...

Maybe that is why I am looking more forward to silvermount atom tablets instead of haswell tablets. Currently the atom chips are about 20% the performance than a ulv i5. If silvermount doubles the cpu speed due to having 4, moving to an out of order design, and higher clock speeds then that means haswell will only be 2.5 to 3 times faster.
 

pablo87

Senior member
Nov 5, 2012
374
0
0
Unless intel can design and produce haswell and successor chips at $10 a piece, it won't compete with arm where it matters. Sure, it'll compete with arm in performance and eventually battery life, but not in price.

But atom in the other hand...

You're probably right but I think Intel is going at this market with both barrels firing.

Based on capex and current volumes, I can't see Intel making Haswell's for any less than $50. However, Intel has been cross-subsidizing Celerons and Pentiums for a long time (now moved to 22nm so definitely not built on fully depreciated fabs) and while it was very strategic at one point (Andy Grove compared it to Rebar), its much less so today.

However, Tablets - Chromebooks - Smartphones - are extremely strategic. You can bet your boots Intel will get aggressive on Haswell for those markets at some point - they might not initially but when they sense that only price is keeping them from significant market share gains, they will.

This to me explains the very heavy capex (yes, its a feedback loop). They need to get to 14nm and maybe even 10nm? in order to get the die size down and the power consumption down even further so that they can ~ the all day battery life vector set by ARM based product.
 
Last edited:

blastingcap

Diamond Member
Sep 16, 2010
6,654
5
76
Let's face it. The only reason ARM is relevant is because despite how hilariously low performance the current parts are, all day battery life is nice.

Wrong. ARM chips are cheap. Intel? Not so much. Mobile is a cutthroat business where even a few cents can mean the difference between winning a contract and not winning a contract.
 

cbn

Lifer
Mar 27, 2009
12,968
221
106
They need to get to 14nm and maybe even 10nm? in order to get the die size down and the power consumption down even further so that they can get hit the all day vector with ~ battery as ARM.

Yep, and it is amazing to me Intel is doing just that with high power/high leakage process tech for "Core" (Basically a node ahead of ARM, but with greater static power losses than what they could achieve with the SOC/smartphone process.)

I just wonder how much power saving improvement the ULT/ULX chips would see on the low power process (ie, 14nm LP vs. 14nm HP)? <---No doubt, the max turbo clock would take a hit, but the lower clock speed range stands to be more efficient. But maybe Intel doesn't care about improving the low (i3 and below) parts if the more expensive i5 and i7s ULVs ended up needing to carry a less impressive turbo boost spec?

I guess it depends on Intel's volumes? I wonder how many i5s and i7s ULVs do they sell compared to i3, Pentium and Celeron ULV?
 

cbn

Lifer
Mar 27, 2009
12,968
221
106
Speaking of Intel, ULV and turbo boost scaling (on high leakage process tech)....I am so looking forward to the day they fix the following problem:

http://www.anandtech.com/show/6194/asus-ux31a-putting-the-ultra-in-ultrabooks/8

In fact, you can set the UX31A to 25W TDP, but it appears the cooling solution isn’t actually able to deal with the higher TDP for longer periods of time and thus the CPU ends up dropping back to 17W after a few minutes of heavy lifting. That’s hardly surprising, considering how thin the UX31A is—there’s just not much space for air to flow through.

If the ULV silicon is built on high leakage silcon (like desktop).....then letting us make better use of the potentially extended capabilities of the chip would be high yield.
 

Lepton87

Platinum Member
Jul 28, 2009
2,544
9
81
NTMBK went against the prevailing orthodoxy of that forum, namely AMD = Good & Pure, Intel = Bad & Evil.

Thus NTMBK had to be punished. :D

And here I thought it was because he called an other member a moron and suggested that the same member should defecate before posting :D "go blow it out your arse"
 

pablo87

Senior member
Nov 5, 2012
374
0
0
Yep, and it is amazing to me Intel is doing just that with high power/high leakage process tech for "Core" (Basically a node ahead of ARM, but with greater static power losses than what they could achieve with the SOC/smartphone process.)

I just wonder how much power saving improvement the ULT/ULX chips would see on the low power process (ie, 14nm LP vs. 14nm HP)? <---No doubt, the max turbo clock would take a hit, but the lower clock speed range stands to be more efficient. But maybe Intel doesn't care about improving the low (i3 and below) parts if the more expensive i5 and i7s ULVs ended up needing to carry a less impressive turbo boost spec?

I guess it depends on Intel's volumes? I wonder how many i5s and i7s ULVs do they sell compared to i3, Pentium and Celeron ULV?

I have no idea to be honest but I am interested in finding out more. But I do think everything points to Haswell and successors being Intel's best bet long-term as performance is Intel's clearest differentiator and the larger die size (presumably) can be supported by Intel's fab capacity.

Because many of the other variables - power draw of the rest of the device, continuously improved battery technology, the static nature of the "all day battery life" constraint, the low proportion of the SOC in the final device price, the fact devices are subsidized by telco's et al - are the same for everyone and thus minimize Haswell's disadvantages (slightly higher power draw and higher price) while leaving its advantages intact.
 

pablo87

Senior member
Nov 5, 2012
374
0
0
Yep, and it is amazing to me Intel is doing just that with high power/high leakage process tech for "Core" (Basically a node ahead of ARM, but with greater static power losses than what they could achieve with the SOC/smartphone process.)

I just wonder how much power saving improvement the ULT/ULX chips would see on the low power process (ie, 14nm LP vs. 14nm HP)? <---No doubt, the max turbo clock would take a hit, but the lower clock speed range stands to be more efficient. But maybe Intel doesn't care about improving the low (i3 and below) parts if the more expensive i5 and i7s ULVs ended up needing to carry a less impressive turbo boost spec?

I guess it depends on Intel's volumes? I wonder how many i5s and i7s ULVs do they sell compared to i3, Pentium and Celeron ULV?

Once Haswell and Kabini have more info, it will be interesting to compare the two across the range of power draw (Kabini/temash suppose can go from 3.6W to 25W), how its achieved, the advantages and disadvantages, both theoretical and actual devices. As they do represent different power approach and different fabs.
 

Dadofamunky

Platinum Member
Jan 4, 2005
2,184
0
0
People are going to keep buying tablets. But they are in a fad phase right now. They are not going to replace real computers for anyone who needs a real computer, just for Grandma and little Johnny, who only had a real computer because that's all that was available.

That's what they said two years ago when the iPad came out. But saying "people are going to keep buying tablets" and then immediately saying "they are in a fad phase" isn't logical. A direct contradiction in two sentences. I doubt this is what you meant to say.

The next "killer app" will come eventually, and people will care about more than just battery life once again. It's inevitable.

Gee... How long has it been now? Sorry, I don't see any new killer apps coming from the desktop. That hasn't happened since the Web became popular. The current 'killer app' is the cloud. Sure, you can't run Photoshop and Illustrator on a tablet... yet. Wait a couple years. The Surface Pro can actually do that now! This is why I have no interest in Android. I have an iPad 4 and am extremely happy with it. Best content consumption device I have ever had. A mere toy does not do what this device does.

As for tablets, given the choice between a tablet that costs $600 and runs standard Windows applications, and one that costs $400 and doesn't, plenty of people will buy the latter, but plenty will also buy the former. In five years, though, those tablets will probably be selling for the same price, and that's when the toys will be in real trouble.

We may be paying roughly the same amount, but what we get for our money will be far greater.
 

Nemesis 1

Lifer
Dec 30, 2006
11,366
2
0
I don't know. I've read that with all other things equal to IVB, that haswell has 3 times the battery life. I'm seeing figures of 14 hours continuous battery life, and all day battery life with power optimizations through the operating system. This will be absolutely a huge deal because 14 hour continuous battery life exceeds that of any current gen ARM tablet.

Like I said - intel seems very serious about Haswell being their best work in the history of intel in terms of efficiency gains. If that is even remotely true, I really see that as a blow to ARM devices in the high end tablet market.
It will be killer in the server market also . Along with silvermont for servers
 

ShintaiDK

Lifer
Apr 22, 2012
20,378
145
106
Atoms, ARM etc have and always will be a joke in servers.

They account for what? 0.001%? And their purpose is minimalistic at best. Specially with virtualization technologies. Microservers are actually very costly in terms of TCO and performance/$. Unless you run super light applications that cant be virtualized.
 

Nemesis 1

Lifer
Dec 30, 2006
11,366
2
0
Agreed, but even if the battery life is phenomenal there is the cost issue. An ARM chip costs what, a few dollars, while Haswell costs what, 200.00 or more. Personally, I am so frustrated with the poor performance of my ARM tablet and the annoying Android OS that I would pay the difference, but a lot of people seem satisfied with it.

Somehow I think intel Haswelll 2 core tablet/ Ultra using Soc SOiX isn't going to cost as much as people want to believe
 

ShintaiDK

Lifer
Apr 22, 2012
20,378
145
106
Agreed, but even if the battery life is phenomenal there is the cost issue. An ARM chip costs what, a few dollars, while Haswell costs what, 200.00 or more. Personally, I am so frustrated with the poor performance of my ARM tablet and the annoying Android OS that I would pay the difference, but a lot of people seem satisfied with it.

You can get 40$ x86 CPUs today. And ARM usually cost around 20-30$. Some numbers below are bare production cost without any profit. For example 17½$ for the iPhone 5 CPU. Had it been sold on a regular market, its expected to cost 28$ or more.
 
Last edited:

Nemesis 1

Lifer
Dec 30, 2006
11,366
2
0
Maybe around 10nm. Depending on how much Intel wants for their processors. It doesn't make any sense to put a $200 CPU into a tablet that sells for $199.

What I can envision though, is for example an ipad (Since iOS is a subset of OS X, it can already run on x86) that acts like a tablet, until you get home and plug it into your monitor, keyboard, mouse etc via a single Thunderbolt cable.

Now guys I will wager that Silvermont will do fine in tablets and great on phones . Silvermont gives intel a lower cost solution in this market , In phones Silvermont should kick butt. Airmont I hope its as good as I have read .