Medfield Benchmarks are in: The x86 power myth is finished.

Laptops, PDAs, Cell Phones, and all other tech that you carry with you.

Moderators: mac_h8r1, Nelliesboo, David

Medfield Benchmarks are in: The x86 power myth is finished.

Postposted on Tue Jan 10, 2012 9:26 pm

Well, Anandtech just went live with hard numbers about Medfield: http://www.anandtech.com/show/5365/inte ... rtphones/1

A few interesting notes:
1. Medfield wins the CPU benchmarks including those against dual core ARMs. This review is directed to the phone implementations so quad-core Tegra 3's are not included since they are only in tablets right now.

2. Medfield has about 1/2 the standby power usage of the iPhone 4S(!) The 3G browsing power usage is also lower than the iPhone 4S and a Galaxy S II. The talk power usage is between the Iphone 4S and the Galaxy SII. The Medfield does trail the Iphone & Galaxy II during video playback but is *still* well under 1 watt (which has been called impossible by many).

3. The rumors that Medfield uses 2.6 watts when sitting idle are complete Barbara Streisand (those of us with a clue knew that already), but to have only ~1/2 the standby power of an iPhone 4S (and to even edge out the Galaxy II) while being faster is *very* impressive.
4770K @ 4.7 GHz; 32GB DDR3-2133; GTX-770; 512GB 840 Pro (2x); Fractal Define XL-R2; NZXT Kraken-X60
--Many thanks to the TR Forum for advice in getting it built.
chuckula
Gerbil Elite
Gold subscriber
 
 
Posts: 562
Joined: Wed Jan 23, 2008 9:18 pm
Location: Probably where I don't belong.

Re: Medfield Benchmarks are in: The x86 power myth is finish

Postposted on Tue Jan 10, 2012 10:00 pm

Color me impressed.
Z68XP-UD4 | 2700K @ 4.4 GHz | 16 GB | 770 | PCP&C Silencer 950 | XSPC RX360 | Heatkiller R3 | D5 + RP-452X2 | HAF 932 | 1 TB WD Black w/ SRT
Waco
Gerbil Elite
 
Posts: 746
Joined: Tue Jan 20, 2009 4:14 pm

Re: Medfield Benchmarks are in: The x86 power myth is finish

Postposted on Tue Jan 10, 2012 10:12 pm

Where is NeelyCam when you need him?
i7 860 - GA-P55-USB3 - 8GiB - HD7850 - SSD - 3.64TB HDD - Xonar D1 - U2410 - Win7 Pro x64.
DancinJack
Minister of Gerbil Affairs
 
Posts: 2045
Joined: Sat Nov 25, 2006 3:21 pm
Location: Austin, TX

Re: Medfield Benchmarks are in: The x86 power myth is finish

Postposted on Tue Jan 10, 2012 10:14 pm

TheWacoKid wrote:Color me impressed.

Yes Yes Yes.. but what color is Krogoth when he's impressed?
That was a rhetorical question... Krogoth is NEVER impressed. :wink:
4770K @ 4.7 GHz; 32GB DDR3-2133; GTX-770; 512GB 840 Pro (2x); Fractal Define XL-R2; NZXT Kraken-X60
--Many thanks to the TR Forum for advice in getting it built.
chuckula
Gerbil Elite
Gold subscriber
 
 
Posts: 562
Joined: Wed Jan 23, 2008 9:18 pm
Location: Probably where I don't belong.

Re: Medfield Benchmarks are in: The x86 power myth is finish

Postposted on Tue Jan 10, 2012 10:23 pm

DancinJack wrote:Where is NeelyCam when you need him?

He should be all dancing in the aisles like Joliet Jake at a charasmatic church, seeing as how this will break up the "ARM monopoly" he's always on about.
I do not understand what I do. For what I want to do, I do not do. But what I hate, I do.
derFunkenstein
Gerbil God
Gold subscriber
 
 
Posts: 21337
Joined: Fri Feb 21, 2003 9:13 pm
Location: WHAT?

Re: Medfield Benchmarks are in: The x86 power myth is finish

Postposted on Tue Jan 10, 2012 10:48 pm

Is that a cherry picked engineering sample?
ish718
Gerbil First Class
 
Posts: 183
Joined: Fri Feb 22, 2008 1:03 am

Re: Medfield Benchmarks are in: The x86 power myth is finish

Postposted on Tue Jan 10, 2012 11:06 pm

ish718 wrote:Is that a cherry picked engineering sample?

It's a reference platform. They're ALL cherry-picked engineering samples. (that's probably not true, but it will be bound to make someone sleep better at night)

FWIW I'm with Anandtech on this one - very hopeful that it's possible, but still in doubt until we (they) have a real, live retail unit.
I do not understand what I do. For what I want to do, I do not do. But what I hate, I do.
derFunkenstein
Gerbil God
Gold subscriber
 
 
Posts: 21337
Joined: Fri Feb 21, 2003 9:13 pm
Location: WHAT?

Re: Medfield Benchmarks are in: The x86 power myth is finish

Postposted on Tue Jan 10, 2012 11:47 pm

chuckula wrote:
TheWacoKid wrote:Color me impressed.

Yes Yes Yes.. but what color is Krogoth when he's impressed?
That was a rhetorical question... Krogoth is NEVER impressed. :wink:


It's not a rhetorical question, the answer is:

Krogoth is BLACK because that is the absence of all color.
MadManOriginal
Graphmaster Gerbil
 
Posts: 1424
Joined: Wed Jan 30, 2002 7:00 pm
Location: In my head...

Re: Medfield Benchmarks are in: The x86 power myth is finish

Postposted on Wed Jan 11, 2012 1:23 am

MadManOriginal wrote:
chuckula wrote:
TheWacoKid wrote:Color me impressed.

Yes Yes Yes.. but what color is Krogoth when he's impressed?
That was a rhetorical question... Krogoth is NEVER impressed. :wink:


It's not a rhetorical question, the answer is:

Krogoth is BLACK because that is the absence of all color.


The irony is that Americans call black people "man of color". Oh the humanity!!!!1111
nVidia video drivers FAIL, click for more info
Disclaimer: All answers and suggestions are provided by an enthusiastic amateur and are therefore without warranty either explicit or implicit. Basically you use my suggestions at your own risk.
Arclight
Gerbil Elite
 
Posts: 696
Joined: Tue Feb 01, 2011 3:50 am

Re: Medfield Benchmarks are in: The x86 power myth is finish

Postposted on Wed Jan 11, 2012 9:46 am

However impressive these numbers are, I'd say they're more of a testament to Intel's best-in-the-world manufacture process, than the viability of x86 in low power computing scenarios. At 32nm LP, I don't see how Intel should ever be losing the power consumption fight with any ARM chip at all.

If Intel released a 45nm SoC, we'd have a better discussion of whether x86 can compete with ARM, but as it stands now, Intel's advantage in their process tech will leave that question wide open.

What is answered now is whether or not they can technically compete: Yes they can.
2011 15" MacBook Pro Ci7 2GHz, 8GB DDR3-1066, Intel 510 256GB SSD
Da_Boss
Gerbil Team Leader
 
Posts: 242
Joined: Tue Mar 14, 2006 5:27 pm

Re: Medfield Benchmarks are in: The x86 power myth is finish

Postposted on Wed Jan 11, 2012 10:39 am

Da_Boss wrote:However impressive these numbers are, I'd say they're more of a testament to Intel's best-in-the-world manufacture process, than the viability of x86 in low power computing scenarios. At 32nm LP, I don't see how Intel should ever be losing the power consumption fight with any ARM chip at all.

If Intel released a 45nm SoC, we'd have a better discussion of whether x86 can compete with ARM, but as it stands now, Intel's advantage in their process tech will leave that question wide open.

What is answered now is whether or not they can technically compete: Yes they can.

Um, no. If it were maufacturing process that determined everything then by that logic the 2600K would use no more power than the lowest end processor on the same process. Clearly ridiculous. The circuit design clearly needs to get credit where it is due.
flip-mode
Gerbil Khan
Silver subscriber
 
 
Posts: 9084
Joined: Thu May 08, 2003 12:42 pm
Location: Cincinnati, OH

Re: Medfield Benchmarks are in: The x86 power myth is finish

Postposted on Wed Jan 11, 2012 10:58 am

A sidenote to this is that while this sample is using the SGX540 (a higher clocked one than the Galaxy Nexus), Intel plans to put the iPhone 4S GPU, the SGX 543MP2, into the SOCs for the latter part of this year.

This only matches Apple in terms of GPU performance but far exceeds any other Android including Krait.



Speaking of Krait, it's a A15-like design that's supposed to have 50% better performance per clock over A9. So while this accomplishment is extremely impressive, I'll still hold off judgement until Krait comes out later this year (which is when Medfield also comes out).
ChronoReverse
Gerbil Elite
 
Posts: 736
Joined: Wed Dec 12, 2007 4:20 pm

Re: Medfield Benchmarks are in: The x86 power myth is finish

Postposted on Wed Jan 11, 2012 11:01 am

chuckula wrote:Yes Yes Yes.. but what color is Krogoth when he's impressed?
That was a rhetorical question... Krogoth is NEVER impressed. :wink:


Pink. A very pretty, pretty pink. :)

Arclight wrote:The irony is that Americans call black people "man of color". Oh the humanity!!!!1111


That may come from paint mixing. White is the absence of pigments,and if enough pigments are added, black is created. Well, a very dark brown really.
Flatland_Spider
Gerbil Elite
 
Posts: 852
Joined: Mon Sep 13, 2004 8:33 pm
Location: The 918/539

Re: Medfield Benchmarks are in: The x86 power myth is finish

Postposted on Wed Jan 11, 2012 11:59 am

Them be some nice numbers. Pity we've still got nearly a year before production devices hit the market, but since Motorola is signed on, I assume we'll be seeing some slick new Android handsets right out of the gate.
He who laughs last, laughs first next time.
ludi
Gerbil Elder
 
Posts: 5439
Joined: Fri Jun 21, 2002 10:47 pm
Location: Sunny Colorado front range

Re: Medfield Benchmarks are in: The x86 power myth is finish

Postposted on Wed Jan 11, 2012 1:40 pm

I don't get it. By "x86 myth finished" you meant what exactly? You meant that the myth was that x86 was slower than ARM when they are on the same power envelope and this article proves the contrary?
nVidia video drivers FAIL, click for more info
Disclaimer: All answers and suggestions are provided by an enthusiastic amateur and are therefore without warranty either explicit or implicit. Basically you use my suggestions at your own risk.
Arclight
Gerbil Elite
 
Posts: 696
Joined: Tue Feb 01, 2011 3:50 am

Re: Medfield Benchmarks are in: The x86 power myth is finish

Postposted on Wed Jan 11, 2012 1:58 pm

flip-mode wrote:
Da_Boss wrote:However impressive these numbers are, I'd say they're more of a testament to Intel's best-in-the-world manufacture process, than the viability of x86 in low power computing scenarios. At 32nm LP, I don't see how Intel should ever be losing the power consumption fight with any ARM chip at all.

If Intel released a 45nm SoC, we'd have a better discussion of whether x86 can compete with ARM, but as it stands now, Intel's advantage in their process tech will leave that question wide open.

What is answered now is whether or not they can technically compete: Yes they can.

Um, no. If it were manufacturing process that determined everything then by that logic the 2600K would use no more power than the lowest end processor on the same process. Clearly ridiculous. The circuit design clearly needs to get credit where it is due.


I'm not saying manufacturing determines everything, I'm just saying that this article does nothing to prove that x86 is as suitable for low power SoCs as ARM is. It could be, but this article doesn't quite sway me.

According to the article, Intel's process advantage gives them "43% lower dynamic power or 37% higher frequency at the same power level" than any ARM chip shipping. It's kinda hard to look at this and be convinced that Intel's superior x86 architecture is officially as good as ARM, especially since it runs at a frequency faster than anything shipping in a phone today.
2011 15" MacBook Pro Ci7 2GHz, 8GB DDR3-1066, Intel 510 256GB SSD
Da_Boss
Gerbil Team Leader
 
Posts: 242
Joined: Tue Mar 14, 2006 5:27 pm

Re: Medfield Benchmarks are in: The x86 power myth is finish

Postposted on Wed Jan 11, 2012 1:59 pm

Arclight wrote:I don't get it. By "x86 myth finished" you meant what exactly? You meant that the myth was that x86 was slower than ARM when they are on the same power envelope and this article proves the contrary?


The implication that all the logic - that many pundits call baggage - that makes x86, x86 was too large and by extention power hungry to migrate into a low power field like ARM.
"Welcome back my friends to the show that never ends. We're so glad you could attend. Come inside! Come inside!"
Ryu Connor
Global Moderator
Gold subscriber
 
 
Posts: 3540
Joined: Thu Dec 27, 2001 7:00 pm
Location: Marietta, GA

Re: Medfield Benchmarks are in: The x86 power myth is finish

Postposted on Wed Jan 11, 2012 2:21 pm

Da_Boss wrote:I'm not saying manufacturing determines everything, I'm just saying that this article does nothing to prove that x86 is as suitable for low power SoCs as ARM is. It could be, but this article doesn't quite sway me.

According to the article, Intel's process advantage gives them "43% lower dynamic power or 37% higher frequency at the same power level" than any ARM chip shipping. It's kinda hard to look at this and be convinced that Intel's superior x86 architecture is officially as good as ARM, especially since it runs at a frequency faster than anything shipping in a phone today.

So I've been wondering for a while: why do people care what ISAs are powering their mobile devices? Who cares whether an ISA is "suitable" as long as your performance and power needs are met as a consumer? As a developer, I can see you might care if you've been doing low-level asm programming on the architecture, but the vast majority of "app" writers don't have nearly that level of technical competency so why is there so much nerdraging about ARM being better than x86 at blah blah blah or vice-versa?

Sorry Da_Boss, I was just quoting you as an example of a lot posts I see whenever this topic comes up and people ardently defending both ISAs despite the fact that as consumers it shouldn't really matter.
Zoomastigophora
Gerbil Elite
 
Posts: 613
Joined: Tue Nov 11, 2008 7:10 pm

Re: Medfield Benchmarks are in: The x86 power myth is finish

Postposted on Wed Jan 11, 2012 2:27 pm

For that matter, who even cares what process they're on and whether Intel is "better" because they have the process advantage.

The only thing that matters is if Intel can get this out early enough to matter. If this comes out at the end of the year still with the SGX540 and is up against Krait from Qualcomm, then it'll look bad. If Intel gets this out before the halfway point of the year then it's competitive. If Intel gets this out within a few months and even updates to the SGX543MP2, then it's absolutely killer.
ChronoReverse
Gerbil Elite
 
Posts: 736
Joined: Wed Dec 12, 2007 4:20 pm

Re: Medfield Benchmarks are in: The x86 power myth is finish

Postposted on Wed Jan 11, 2012 2:33 pm

Manufacturing process makes a very large difference on power consumption. I have been reading a lot of papers recently on subthreshold FFTs and they almost always normalize power consumption to take process into account. There's a bigger difference between processes than just feature size (e.g. 32 nm). Intel has the capability to manufacture very advanced processes due to their investments in the desktop CPU business.

The argument that ARM is more efficient than x86 is probably a correct one. x86 has a lot of bulk (instructions that are hardly used, but must be supported) from years of backwards compatibility. There's a reason that research papers rarely focus on x86. It is a dinosaur which is kept alive for backwards compatibility reasons. While Intel may be able to nudge its way into the smart phone market, x86 doesn't really have any logical place there.
"TORTURIS EXUVIAS EUNT"
Phenom X6 1090T @ 3.2 GHz
Sapphire Radeon 6950
TurtlePerson2
Graphmaster Gerbil
 
Posts: 1097
Joined: Sat May 26, 2007 9:08 am
Location: Plano, Texas

Re: Medfield Benchmarks are in: The x86 power myth is finish

Postposted on Wed Jan 11, 2012 2:52 pm

And it doesn't matter because Intel is the only one who can build a CPU on a smaller process at each point in time. Who cares about the ISA (it should be pretty obvious ARM has that advantage in terms of energy efficiency) as long as the actual solution is superior in the same point of time? The academics can go talk about how ARM might be better when it hits the same process but Intel will have already moved on.


Furthermore, the decode for x86 nowadays is like 5% of the non-cache transistors. It's small enough that it's not that big a deal.
ChronoReverse
Gerbil Elite
 
Posts: 736
Joined: Wed Dec 12, 2007 4:20 pm

Re: Medfield Benchmarks are in: The x86 power myth is finish

Postposted on Wed Jan 11, 2012 4:40 pm

Zoomastigophora wrote:So I've been wondering for a while: why do people care what ISAs are powering their mobile devices? Who cares whether an ISA is "suitable" as long as your performance and power needs are met as a consumer? As a developer, I can see you might care if you've been doing low-level asm programming on the architecture, but the vast majority of "app" writers don't have nearly that level of technical competency so why is there so much nerdraging about ARM being better than x86 at blah blah blah or vice-versa?

Sorry Da_Boss, I was just quoting you as an example of a lot posts I see whenever this topic comes up and people ardently defending both ISAs despite the fact that as consumers it shouldn't really matter.

Let's be more blunt: people don't care what ISA is powering their mobile devices, nor do they care what process technology it is manufactured on. The only thing that matters is whether it works as well or better than their current mobile device, and how long the battery lasts, relative to the physical size of the device.

Moreover AFAIK on Android (no clue about iOS) the underlying hardware doesn't even matter that much to app developers beyond not thrashing the available system resources, since the apps are Java and the hardware translation is handled by Android's VM framework.

Intel got execution capabilities and power envelope well within the thresholds of existing devices, and secured design wins from both Motorola and Lenovo. That's huge.
He who laughs last, laughs first next time.
ludi
Gerbil Elder
 
Posts: 5439
Joined: Fri Jun 21, 2002 10:47 pm
Location: Sunny Colorado front range


Return to Mobile Tech

Who is online

Users browsing this forum: No registered users and 2 guests