Medfield phones look promising

[Vesku]

I'm liking what I see in the Anandtech preview. If the price is reasonable I'll definitely be willing to early adopt an x86 phone. Intel should really push a unified x86 application marketplace, apps that are functional from all your x86 devices.

 

[Nemesis 1]

Ya, Looks real good for intel ; The next step is the 2 core medfield with 543gpu. Will it be enough to compete with A15 until 2013 when 22nm atom wipes the floor with arm. ARM/HAMMER

 

[Khato]

Will it be enough to compete with A15 until 2013 when 22nm atom wipes the floor with arm.

It's more a question of how it compares with Qualcomm's S4 family which will be showing up in products around the same time frame no? Whereas the most optimistic dates I've seen for A15 based designs is 2H'12 rolling into 1Q'13 - I'm not expecting to see A15 based products until 2013. The real question now is exactly what sort of performance improvements will Qualcomm's S4 family/ARM's A15 core design actually offer over the current A9... and will it be at the cost of increased power consumption.

Medfield's been a long time coming for Intel, but if they make the same kind of improvements that they've been consistently doing on their primary core architecture... well, ARM designs may find themselves in the same position AMD's been for awhile now - looking back at their brief glory days before they roused the beast from its slumber.

 

[piesquared]

Nemesis, have you seen TSMC's 28nm or GLOBALFOUNDRIES'? Don't know what's going on at GloFo, but TSMC's node looks pretty good and that's not the low power process.

 

[drizek]

Makes it seem like intel needs to be at 32nm to build something that can compete with ARM at 45/40nm.

Still impressive though. I just feel kind of... dirty... getting an x86 phone.

 

[cbn]

Intel should really push a unified x86 application marketplace, apps that are functional from all your x86 devices.

Sounds like they are trying to move in that direction...

http://www.anandtech.com/show/5371/...ppup-on-ivy-bridge-ultrabook-reference-design

Next up was a demo with Intel's AppUp store which aims to be a one-stop shop for PC software, and then M-GO, which is a movie and TV show marketplace component of AppUp. The process of purchasing a movie, paying with an NFC-enabled smartphone, and then playing back the move on an LG WiFi display was demoed on-stage.

 

[cbn]

It's more a question of how it compares with Qualcomm's S4 family which will be showing up in products around the same time frame no?

The real question now is exactly what sort of performance improvements will Qualcomm's S4 family/ARM's A15 core design actually offer over the current A9... and will it be at the cost of increased power consumption.

Well the CPUs are one thing, (Qualcomm will no doubt be very strong) but I am really wondering about how all the other tech on the Snapdragon Platform compares to Intel?

How will Intel's radios and other integration compare to what will be equipped on Qualcomm's SOC?

 

[jhu]

But does it run Windows?

 

[Khato]

Well the CPUs are one thing, (Qualcomm will no doubt be very strong) but I am really wondering about how all the other tech on the Snapdragon Platform compares to Intel?

How will Intel's radios and other integration compare to what will be equipped on Qualcomm's SOC?

I think it's pretty much a given that currently Qualcomm's going to have the lead when it comes to radio integration... they lead all the other SoC manufacturers in that area too after all. They're also going to have an advantage in graphics for the time being as the platform development time for Medfield past silicon meant that it's a bit behind the curve in that regard.

As stated in the Anandtech article though, Intel's going to be on par with Qualcomm (and quite possibly ahead of other SoC manufacturers) on radio integration sometime in 2013. Combined with that atom performance roadmap, well, 2013 promises to be quite the interesting year in the mobile SoC space.

 

[drizek]

But does it run Windows?

That question became irrelevant about two years ago. Nobody cares anymore. Besides, windows 8 will run just as well on Tegra 3.

 

[sm625]

Given all the problems I have with my ipoop, I would gladly use a x86 phone. It's not like the battery life could get any worse. I was playing tiny wings for 20 minutes yesterday and it drained my battery 25%. It's too bad the whole standby/sleep/resume/hibernate thing is a mess on windows.

 

[Chiropteran]

Looks interesting, but since it's a different architecture it'll have to run it's own version of android. I could see that being a big deal-breaker among the hacker types who like to root their phone and try alternate OS versions, since all the ARM based images will be off limits. I don't know enough about how android market apps are built, do they run in a VM like java or something or will the ARM apps need to be recompiled to work in the x86 phones?

Other than that, I wonder just how useful the benchmark results are. It's a single core atom compared to several dual core ARM chips, and while it wins it doesn't win that much. Could it be that the shown benchmarks simply favor a strong single core and don't take advantage of the second core in the ARM phones? Also, it's not rare for the latest and newest phone to beat all the existing phones, by the time it's released the newest ARM phones may be just as fast or faster.

It's good to have another option, but I'm not buying all the hype just yet.

 

[CLite]

I am curious how this binary interception business will work, if games built for ARM are sluggish because of it then intel is going to have a tough time gaining the interest of a lot of consumers. Regarding the power consumption going as low as ~18mW at idle, and being very comparable to other platforms, I am quite curious how intel has defeated the law of physics.

http://forums.anandtech.com/showpost.php?p=32744767&postcount=78

It's not a question of wisdom and foresight. It's a question of physics. The theoretical lower limit of the power needed by a decoder is VERY HIGH.

Either you defeat physics law, or drop x86.

In all seriousness, it seems intel using the 256KB low power SRAM and long pipelines have really conquered some of the issues that the decoder presents. The article doesn't seem to go into it too deeply so if there are any cpu wizards I hope we could hear a more indepth/accurate explanation.

More competition is good, I just hope this business (and AMD's capitulation) doesn't distract intel from creating cutting edge power hungry desktop behemoths.

 

[IntelUser2000]

In all seriousness, it seems intel using the 256KB low power SRAM and long pipelines have really conquered some of the issues that the decoder presents.

http://semiaccurate.com/2012/01/10/intel-finally-gets-a-phone/

-Gshare branch predictor doubled from 4K to 8K per thread
-Improved memory copy performance
-Microcode and scheduling bottlenecks

The things above should improve both performance and performance per watt, but the most interesting thing is this:

Dynamic power has also been addressed in many areas, large and small. Loop Stream Detection has been added to Saltwell, and ‘significant’ PLL power savings have been found.

First featured in Nehalem, that'll go a long way in reducing decoder's footprint basically.

Compared to the predecessor Moorestown, it should perform about 10% faster per clock while consuming 1/2 the power at top frequency. More integration, and better software optimizations allow further reduction at the platform level. Perhaps Moorestown was fated to never being in Smartphones.

 

[Lotheron]

Based on the data Intel shared with us as well, the x86 power problem appears to be a myth - at least when it comes to Medfield.

I know this has been touched on, but this is huge. I know Intel is behind ARM, but as Anand pointed out, only about a year. If Intel can get into a faster product cycle, we could really see a mobile CPU war in the next 2 years or so.

Now if only we could have it so the mobile GPUs don't require a table to see their specs and what tech they support, or not.

That being said, is it just me, but does x86 in a smartphone feel wrong and dirty?

 

[epidemis]

I'm not that optismistic, it only shows medfield is a little better than the competition, but they would need a knockout to cause wide adaption. There is very little reason to switch from ARM to x86 currently.

 

[Grooveriding]

Not surprised with the results at all. Saw some speculation that Intel would struggle here and boggled. This is just their first entry product as well, and apart from GPU performance, it's the best thing out there.

A year or less/more and they will have the best chips out there in all ways for smart phone applications.

The only thing I did not see mentioned that is of interest is what is the pricing on these and how does it compare to offerings from competitors like Qualcomm and nvidia. Intel could simply charge more as they do elsewhere for having the best product available or are they going to attempt to be price competitive and sell volume of these chips.

 

[Ferzerp]

This is kind of amusing, in regards to the responses here (and to the article, and other responses I've seen about this). You have the same kind of people who think arm is magic fairy dust and only needs to become "fast enough" to commpletely displace x86 everywhere suddenly changing their tune when Intel's first real effort in this area shows that there is no magical arm fairy dust (what we've been trying to tell you all along). Magically, *now* inertia is important, when talking about arm's interia....

 

[blckgrffn]

Isn't this exactly what Dirk told the BoD @ AMD they couldn't do?

At least if Intel gets broad x86 adoption in the mobile space in the next 12-24 months, AMD will have the time to get a product out.

You know, one that probably won't be quite as slick, but cheaper, and might have better graphics

*not intended to derail, I am pretty impressed to. Considering how sluggish the atom feels in Windows, it makes you wonder how much we have been trained to accept smart phone performance being slow. Is windows really that big of a hog (I am doubtful) or our expectations for really mobile devices are just really stinking low.

How much of this power saving magic will make it's way into mobile Haswell? And how awesome will that be?

 

[TuxDave]

I am quite curious how intel has defeated the law of physics

That actually really cracks me up. You know, even after all this time I still sometimes think the CPU is full of voodoo black magic and the process technology is full of fairy dust. But hey if it works, it works!

 

[sm625]

I am curious how this binary interception business will work, if games built for ARM are sluggish because of it then intel is going to have a tough time gaining the interest of a lot of consumers.

Games shouldnt feel sluggish due to binary translation. It should happen once, at load time. Once it is loaded it should run just fine. The real work would be done by the gpu. Assuming it actually works at all. I would expect a lot of apps to just crash at loading. They already crash all the time anyway so Intel is on the right track.

 

[Vesku]

I think one of the reasons mobile apps are dominated by dinky games is because of current mobile processing speeds. The graphics are actually pretty speedy for the screen resolutions phones currently have. Hence a plethora of games that play to the graphics strengths and work within the CPU restrictions. The astronomy app I have is quite sluggish in comparison because it needs more on the CPU side than Angry Castle *cough* I mean Birds.

*not intended to derail, I am pretty impressed to. Considering how sluggish the atom feels in Windows, it makes you wonder how much we have been trained to accept smart phone performance being slow. Is windows really that big of a hog (I am doubtful) or our expectations for really mobile devices are just really stinking low.

 

[cbn]

A couple of questions I have about Medfield and Android:

1. Will Medfield's Power VR GPU driver suffer the same problems running Android that other Intel Power VR drivers suffer from when trying to run Windows?

2. How will Intel Medfield compare to other ARM based phones when it comes to security on Android?

 

[HopJokey]

I think that even though Medfield won't "dominate" the market this year it shows that Intel is on the right track to being very competitive, capturing market share, and even being a leader in 2013/2014 for smartphones. Once they move to the new Atom uArch, integrate the baseband, and use the bleeding edge manufacturing process (right now 22nm) things should be quite interesting to say the least.

 

[Concillian]

That actually really cracks me up. You know, even after all this time I still sometimes think the CPU is full of voodoo black magic and the process technology is full of fairy dust. But hey if it works, it works!

Well, hard drives have "pixie dust" in them, so why not?