Mobile: Intel Will Overtake Qualcomm In Three Years

[Blandge]

[citation][nom]rpgplayer[/nom]The problem of course is that because Intel can use a much more advanced node it can make the chips extremely cheaper than what ARM chips can be made. Medfield is somewhere around 65mm^2 while the Tegra 3 is somewhere in the neighborhood of 90mm^2. After Intel's SoC moves to a 22nm node the die size should shrink to around 45mm^2. Which means Intel will be able to push these things out for pocket change.[/citation]

You make a very good point. With Intel moving to 450mm Wafers in the next 5 or so years, these tiny Atom chips may yield major margin at low cost because so many chips will fit on each wafer. Economies of scale are what will drive Atom prices down, while keeping it profitable.

So many people are so quick to bash Atom, but consider that it was never even a major focus of Intel's until recently. That's why the Atom architecture is so simple compared to Sandybridge. Atom was recently added to Intel's core business, and you can bet that Intel has many of the world's top engineers are working on it. This means Core, Xeon and Atom now have the same weight behind them in terms of people and $$. Let's consider the market share for these products core products. ~80% (Core), ~95%(Xeon) and ~0%(Atom, smartphones & tablets). While I don't think Intel will have 80% market share any time soon, I think it's safe to be they will be a major player.

 

[americanbrian]

Wrong, wrong and wrong again.....

Intel has a "good enough" approach to GFX, as has been demonstrated again and again.

Intel always aims at exclusivity level pricing.... Even during economic downturns.

Their is no question that they WILL be able to produce a higher "compute" performance solution.... But as for system level design for mobile "truly mobile" platforms, they have 0 experience. It will be like PowerVR all over again, "we didn't realise that users would want 3D video calling on their handset" they will say.... etc.

It will all turn out ok though. they have the money to burn. But I heartily dispute your predictions. Mobile is all about features and balance. Intel do not walk the same tightrope/slackline that the east has shown to be so adept at.

 

[-Fran-]

Very interesting article, but there's something fundamentally wrong with 1 assumption.

You say that "because Qualcomm let go brilliant folks is somewhat doomed". Come on... Their expertise and knowledge might be enormous, but they're not the only ones that might have it or can come close to it. Let alone, the only smart guys; there are tons of smart folks willing to work for Qualcomm.

Now, what Qualcomm is doing with it's 28nm fab could be the real game changer in favor of Intel. Intel cannot be taken lightly in fab processing. In any area but that one IMO. If they license from PowerVR or even Qualcomm itself, they can come up with a superb product. Hell, like you say, they could work with SIRU and make something new and awesome.

At the end, IP, money and process will rule over creativity in this world; sad but true. And Intel has them all.

Cheers!

 

[Blandge]

[citation][nom]americanbrian[/nom]Intel has a "good enough" approach to GFX, as has been demonstrated again and again.Intel always aims at exclusivity level pricing.... Even during economic downturns.Their is no question that they WILL be able to produce a higher "compute" performance solution....[/citation]

So why is "good enough" suitable for Qualcomm's and AMD's CPU and not for Intel's GPU? Isn't the main selling point for AMD/ARM that their CPU is "good enough?"

If all of them can run the Android UI fluidly and play 1080p@60fps then what other use is there for a GPU on a phone? If you're looking to play Skyrim on ultra high graphics on your smartphone then I think the best route would be an external GPU.

 

[Guest]

Intel needs fatter margins. Intel need BIG volume, see previous sentence. Intel won't put different application specific implementations onto their SOCs. Rather, a one size fits all and depend on software for innovation. That might be a good strategy after they capture a high percentage of the market, not a good plan to enter.

Intel is going to have pressure on their traditional markets from below.

 

[Guest]

This article is serious flawed. In between all the imprecise arm and hand waving which isn't worth a bag of beans there is the biggest mistake, TSMC produce Qualcomm's chips (40nm, not 45nm, and 28nm) not Glo-Fo.

http://www.xbitlabs.com/news/other/display/20111024134511_TSMC_Begins_to_Manufacture_28nm_Chips_for_AMD_Nvidia_Qualcomm.html

This article is best summed up by the axiom that a little knowledge is a dangerous thing

 

[Onus]

[citation][nom]DjEaZy[/nom]Intel Will Overtake Qualcomm In Three Years, If Qualcomm sits on hiz balls and do noothing.[/citation]
But it sounds like Qualcomm may have lost their balls.

[citation][nom]rpgplayer[/nom]The problem of course is that because Intel can use a much more advanced node it can make the chips extremely cheaper than what ARM chips can be made. Medfield is somewhere around 65mm^2 while the Tegra 3 is somewhere in the neighborhood of 90mm^2. After Intel's SoC moves to a 22nm node the die size should shrink to around 45mm^2. Which means Intel will be able to push these things out for pocket change.[/citation]
...except I don't think Intel's mindset is ready for competitively low pricing. The article points out that many of the other guys are run by engineers, and it seems that Intel (while heavily influenced by engineers) is run by suits. Now maybe their prices have been kept high in other areas for fear of attracting too much attention from regulators and other parasites; they may be fearful that low pricing (even if economically feasible for the company) would appear to be an attempt to force competitors out of business. This may be a limitation which does not apply here, where Intel is not yet a major player.
All in all, I find much more agreeable in the article than disagreeable. A lot will depend on what Intel wants to do, and the coming global economic meltdown makes that very hard to predict.

 

[loydc1]

If I am not wrong? Intel tried to get into cell phone arana a few years back and sold it off. So much buzz but what happened to WMAX¿¿?? Intel only. Creates hipe not product! To get you early and if it does not work tell sell it.

 

[Guest]

Intel sure has one massive mountain of cash to help people argue its case. Qualcomm had one big problem, namely the use of an also-ran GPU design. It does not matter how fast Adreno was in the early days of ARM SoC design wins. The ARM GPU winners are going to be chosen from MALI, Nvidia, AMD (soon to announce), and PowerVR. Of these 4, at least two are available for Qualcomm to license. Dropping Adreno will gain Qualcomm far more than it loses.

Meanwhile, Intel continues to circle the drain. x86 + PowerVR is not, and never will be ARM + PowerVR. Intel's driver support is putrid, always has been, and always will be. Whatever GPU Intel uses, it is Intel that is 100% responsible for creating the GPU driver. PowerVR on ARM however can use drivers that are mostly universal to all modern ARM SoC designs using PowerVR, allowing a driver maturity and stability that Intel can only dream about.

Intel's x86 designs are often quoted as superior to ARM, but this is a complete nonsense. Where ARM is currently slower, clock-for-clock (usually on internal caches), the slowness is explained by design choices that balance performance with power consumption. The engineers at ARM know how easy it is to design faster units, if greater power consumption is allowed (say,with mains powered servers). Intel, on the other hand, hasn't a clue how to get sufficient performance when the power usage has to be as low as current ARM mobile devices.

The irony is that Intel relies on the heavy computational lifting being moved from Intel's dreadful CPUs, to the GPU, so that future Intel mobile SoC designs can compete. What does a mobile CPU need to do, when MP3 decoding, Video decoding, and 2D rendering (screen output) occurs entirely on dedicated units, or the GPU? The x86 architecture is so bad, Intel believes that almost never using x86 processing on a mobile device will be its saviour.

All Intel can bring to the mobile marketplace is the INTEL TAX- the tax you pay in power consumption, extra die size, extra coding costs, and Intel IP costs, when you use the Intel x86 architecture. Who will win? Those products that include the Intel tax, or those that do not? Of course Intel will win, just like the Z80 and CP/M won- oh wait, the Z80 and CP/M went the way of the dinosaur, despite being the x86 + windows of their day!

 

[Blandge]

[citation][nom]marsavian[/nom]This article is serious flawed. In between all the imprecise arm and hand waving which isn't worth a bag of beans there is the biggest mistake, TSMC produce Qualcomm's chips (40nm, not 45nm, and 28nm) not Glo-Fo.http://www.xbitlabs.com/news/other [...] lcomm.htmlThis article is best summed up by the axiom that a little knowledge is a dangerous thing [/citation]

Except that it doesn't change the fact that Qualcomm isn't using HKMG which was the important idea in the article.

From the link you posted:

[citation]At present, TSMC produces chips using 28HP [high-performance with HKMG], 28HPL [high-performance low-power with HKMG] and 28LP [low-power with SiON][/citation]

28LP isn't going to use HKMG, later in the article:

[citation]The Snapdragon S4 class of processors are manufactured in TSMC’s highly sophisticated 28LP[/citation]

GF may also be producing 28nm for Qualcomm if this article is to be believed

http://www.brightsideofnews.com/news/2011/10/12/qualcomm-announces-its-2012-superchip-28nm-snapdragon-s4.aspx

[citation]Qualcomm is working with both TSMC and GlobalFoundries, although TSMC will produce the first chips.[/citation]

Read more here:

http://semimd.com/blog/2011/02/07/qualcomm-shies-away-from-high-k-at-28nm/

 

[Guest]

There is nothing to stop Qualcomm using HPM or HPL later, it was just a time to market and low risk option for the first chips which will be the first 28nm mobile chips on the market coming out the same time as 32nm Medfield. This whole gate-first switch story is a fantasy, Glo-Fo was only a prototype option, TSMC was always their main production option as shown be the fact they produce the current 40nm Snapdragon. They are also only giving up a 20% power reduction by not waiting for HPL ...

http://www.tsmc.com/advanced_technology_AD/28nm.htm

Qualcomm have also taken out Power VR licenses for the future ...

http://www.eetimes.com/electronics-news/4233204/Qualcomm-becomes-Imagination-Technologies-licensee-

which doesn't leave much of a factual basis left for this article.

 

[clonazepam]

Why is the author only known as "Mark"? Why no last name? Seems odd. Anyway, I enjoyed reading this article.

I didn't read all the comments, but at some point, when is performance enough? If we're still using a touch interface, with a response time of around 200ms, any additional horsepower that completes tasks under that number, seems to be a waste.

To tap into this additional horsepower, beyond the typical smartphone uses, all smartphones should be able to dock to a keyboard/screen device to be used as a netbook/laptop. It'd also be nice to see docking a common occurrence in the everyday automobile as well, beyond just music playlists and the the typical bluetooth phonecalls through a car deck.

I guess they need to focus on India and China markets. That's where the billions of users are.

Anyway, its nice to see more and more performance and efficiency, but with the lag time of a touch interface, it becomes less and less important once you reach a certain point in performance/efficiency. I'm excited to see new innovative ways to utilize smartphones and tablets, but they all really need to be able to dock to a keyboard/screen and/or a typical video game controller to really benefit from the extra juice.

Whoever moves forward in that direction in the biggest way, will probably get my money in 1.5 yrs when my contract is due for an upgrade.

 

[pug_s]

Kind of overly optimistic that Intel will take over ARM in the next few years. But I am not. Intel and MIPS couldn't take over ARM's dominance in the smartphone field. A few years back intel realizes that it could not do that with its StrongARM cpus (remember those) so they sold it off.

Second, so far, Medfield just came out and we have yet seen any benchmarks outside from Intel made ones.

Third is cost. Will Intel sell Medfield for less than $15 per cpu? When Intel realizes that they do not make enough profit on their CPU's, they will quickly abandon the endeavor.

 

[billj214]

I could make the same comment about Intel overtaking Qualcom but never give the detail of this article! I would just say: When it comes to making chips Intel does it better than anyone!

Great article!

 

[i8cookiemonster]

Interesting article, a great read for sure (especially for those of us old enough to remember the BitBoys)!

Overall, I agree with the assessment. After reading the comments a lot of people are still concerned with the economics. A few people have pointed out the upcoming larger wafers and diminutive size of the SoC, but more importantly I think they overlooked that Intel is probably willing to 'buy' itself into the market by accepting smaller profits in the short term to get a foot in the door, but once that foot is in the door and they have the leverage, they can start wringing more $$$ out of partners. Plus, Intel wins either way, as they'll be powering the backend (all those portable devices have to pull their content from somewhere!)

This assumes they will have a competing product in 3 years time. I wholeheartedly agree they will on the CPU side, they are a slow moving company but once they set their sites on a target, look out! (Just ask our friends at AMD).

The elephant in the room in my opinion are the other SoC designers. I think graphics will play an ENORMOUS role in future designs, and I think the companies pushing the best graphics will have the largest advantage. This puts Intel at a disadvantage, but companies like nVidia become wildcards. While the Tegra 3 might not trounce the competition, they the potential for them is there.
I'm not sure how much impact the new BitBoys startup will have in this space only because (and I'm speculating) I doubt Qualcomm gave up the IP it acquired with them, so they'll have to navigate that nasty patent minefield everyone loves so much...

I'd love to hear more thoughts on this subject...

 

[geekapproved]

True, but someone else will surpass qualcomm and Intel.

 

[Tab54o]

This article forgets to mention some points. Intel may have all this expertise with cpu but Qualcomm has paved the way in CDMA technology. They have all the expertise with phones. A phone isn't just about raw CPU power. Also Qualcomm makes more than just the cpu cores, we have a chipset. The chipset includes the PMIC a special IC that controls all power in the phones. Also what does Intel have along the lines of phones technology. There is probably more debugging and testing that goes into just getting all the cell technologies up and running than what goes into testing just a cpu. Technologies like WCDMA and LTE are ridiculously complicated. Then you add in things like beam forming etc. Good luck to Intel on starting from scratch from there.

If Intel goes down the path of design the whole phone they will lose. Yes they could possibly make a better cpu for them but that's only one part of the design.

 

[Tab54o]

BTW I hadn't read the entire article here when I commented.

 

[Tab54o]

Maybe all this will push linux/Andriod based high end desktops for gaming. Id love to be able to do that.

 

[Guest]

In three years, Qualcomm should have just about completed the purchase of Intel.

 

[blazorthon]

Right now Qualcomm's problem seems to be mostly factors outside of their own company, Intel has hurdles to cross both in their own company and among other companies they will need to work with. Intel might dominate Qualcomm but they might not. Three years may not be enough time for Intel to dominate in the market but they probably will get a strong foot hold at some point.

 

[Guest]

Additionally, we all know that graphics hardware is only as good as its supporting software driver. Updated GPU drivers for the Adreno 205 were responsible for an almost-50% performance improvement.[unquote]

Very interesting article... it reminds me when i was waiting for each new demo released by Future crew, those guys were amazing!

When i started as an engineer, we needed to develop an application running on the new VGA Card (successor of the EGA). I started to develop in C and accessed the video card through the standard INT13... i was barely getting 1 frame per 2 sec. So i went directly into the VGA BIOS, but still i was barely getting a couple of frames per second. My last attempt was to get pass all those and access directly the GPU, so i rewrote my own code to drive the VGA card components... i could get over 30 FPS!!!

My point is, and it's to support the above quote from this article, i don't think the hardware is the real issue nowadays, they are all fairly equal (or powerful enough) but the software...far too complex, not optimized and waste of useless code!

 

[A Bad Day]

Money, fab plants and teams of engineers really help. The question is, can Qualcomm win a war of attrition? Intel has lots of resources and wants the mobile market share, Qualcomm's resources are a lot smaller. One of them can afford to make lots of mistakes and recover from it, the other can not or will face a slow painful death.

 

[Guest]

You are missing the paradigm shift that is happening...the idea of the personal computer is changing from a desktop to a device with more power than today's workstations in a form factor that fits in your pocket and runs for weeks on a single charge. For those that need a keyboard, large monitor, mouse, printer, scanner....all those will have wireless interfaces...you can bet that Intel is not going to miss out on this.

 

[americanbrian]

@ Blandge:

"isn't the selling point for AMD and ARM that the CPU is good enough?"

No, the main selling point is that it is AFFORDABLE. You are taking one of my points and disregarding the other.

High end handsets are priced at the limit of what the market will tolerate. Those billions of revenue that Qualcom currently makes are off the back of budget, mid-range and high spec phones. I am willing to bet in terms of revenue (not necessarily profit) the majority is from budget and mid-range.

Intel will not offer enough VALUE is my arguement. And also they will move too slowly when it comes to FEATURES on handsets. Like I say, powerVR all over again. I have confidence that qualcom and its manufacturing partners operate in a way that allows them to move quickly with market trends.

I am confident Intel will try to maintain exclusivity deals with customers etc. in the vain of Apple. It worked for Apple, but I don't think it will for Intel. That's it. My opinion.