The Intel Atom Thread

[Homeles]

No kidding. Has there even been a device announced with Merrifield?

 

[Arachnotronic]

But that doesn't change the fact that the Silvermont CPU architecture is competitive with Cyclone on integer workloads while providing all the encryption and floating point performance that's needed in the intended markets. As well, Silvermont provides that performance at a fraction of the power consumption.

Fair, but Intel shareholders need to ask the following tough question: when did A7 show up in a shipping device (or Snapdragon 800), and when did Merrifield?

Snapdragon 800/801 has won >=150 designs, including the Galaxy Note III, Galaxy S5, Sony Xperia Z2, etc. What has Merrifield won? Why did none of Intel's partners have a single Merrifield device available for show at MWC?

Even if Merrifield is great and awesome on a perf/watt basis, this doesn't change the fact that both Merrifield and Moorefield will be facing some serious competition in the form of Apple's A8 and potentially a 20nm Qualcomm part by year's end.

I expect Goldmont based devices will be much better, but again, we all thought Silvermont would be the CPU core (and Merrifield/Bay Trail the SoCs) that finally made Intel competitive. Unfortunately, since these parts were late, late, late, they're not competitive.

Intel can do better. I know it can. It just needs to stop being afraid of its own shadow (i.e. afraid of cannibalizing Core) and start playing to win. With the PC market on the decline and with server growth seriously anemic, Intel can't afford to miss the mark again with Broxton without seeing a pretty dramatic decline in its earnings/share price.

 

[tarlinian]

I really don't get the comparisons to Apple's cores. They are not competition for Intel. Unless Intel opens up an unimaginably wide gap over them (like PPC vs. x86) they aren't going to switch to an Intel design regardless of how good Intel designs are. And no one expects Intel to be orders of magnitude better than their competitors. In comparison to Qualcomm, the only future product that's been announced is 805, which is in exactly the same situation as Merrifield. No shipping designs or even talk of design wins. And it's CPU is no better than 800/801. As for Tegra, I don't think anyone thinks that A15 is a better core for mobile than Silvermont, and the same applies to Exynos. On the graphics side of things Intel is behind, but for the market Merrifield/Moorefield is targeted for it's the best SoC for CPU and graphics performance with baseband integration being it's only failing. That alone is what is keeping it from winning sockets.

 

[R0H1T]

I really don't get the comparisons to Apple's cores. They are not competition for Intel. Unless Intel opens up an unimaginably wide gap over them (like PPC vs. x86) they aren't going to switch to an Intel design regardless of how good Intel designs are. And no one expects Intel to be orders of magnitude better than their competitors. In comparison to Qualcomm, the only future product that's been announced is 805, which is in exactly the same situation as Merrifield. No shipping designs or even talk of design wins. And it's CPU is no better than 800/801. As for Tegra, I don't think anyone thinks that A15 is a better core for mobile than Silvermont, and the same applies to Exynos. On the graphics side of things Intel is behind, but for the market Merrifield/Moorefield is targeted for it's the best SoC for CPU and graphics performance with baseband integration being it's only failing. That alone is what is keeping it from winning sockets.

Did you ever, or anyone else on this thread, consider the possibility that only Apple & Samsung can afford to pay that premium price for Intel products, since they have ~90% of the sector's profits, & that other OEM/ODM's don't wanna get tied to the Intel ecosystem, which pretty much killed the entire PC sector D:

 

[frozentundra123456]

Dont want to divert from this interesting discussion of phones, but didnt want to start a new thread either.

If anyone is interested, Best Buy this week has the Lenovo mix tablet on sale for 199.00. I saw it in the store a few days ago and was impressed. At that price, I am seriously thinking of picking one up. Has anyone had any experience with it, and does it compare favorably to the Dell 8 inch windows tablet?

BTW, in regards to phones, I think Intel is going to have a hard time breaking into the market no matter how good their SOC is. Android is just too firmly entrenched, and is a good enough OS for a phone, even for me, and I basically hate android, but would settle for it on a phone.

In a tablet or larger form factor though, after having an android device, I would definitely go with windows next time, and would not even consider android, unless they open them up to easy installation of a "real" OS like linux.

 

[Homeles]

Did you ever, or anyone else on this thread, consider the possibility that only Apple & Samsung can afford to pay that premium price for Intel products, since they have ~90% of the sector's profits, & that other OEM/ODM's don't wanna get tied to the Intel ecosystem, which pretty much killed the entire PC sector D:

Have you ever considered that Intel's Atom prices are actually competitive in these markets?

BTW, in regards to phones, I think Intel is going to have a hard time breaking into the market no matter how good their SOC is. Android is just too firmly entrenched, and is a good enough OS for a phone, even for me, and I basically hate android, but would settle for it on a phone.

In a tablet or larger form factor though, after having an android device, I would definitely go with windows next time, and would not even consider android, unless they open them up to easy installation of a "real" OS like linux.

Huh? Atom can run Android fine.

 

[R0H1T]

Have you ever considered that Intel's Atom prices are actually competitive in these markets?

Huh? Atom can run Android fine.

For now but how long do you think they(Intel) can delay the premium that we're so accustomed to(on our desktops/notebooks) & if I'm having this thought then other vendors are too cause there's no telling when Intel withdraws their incentives, as a lot of PC manufacturers have found out the hard way !

 

[witeken]

I see. When it's a positive Intel article, it's quite good, and when it says, "hey, wait a minute...why does Apple's A7 show a markedly different result in Geekbench than in Intel-sponsored WebXPRT?" it's "poor quality"?

Not really. I can give you some examples of good articles: Intel's Tablet 'Giveaway' Explained, Does Intel Spend More Than Qualcomm In Mobile?, Intel: Competitive Analysis Of The Upcoming 'Merrifield' System-On-Chip, Intel Vindicated, Very Competitive With Apple's A7.

As you see, he already compared Silvermont to other CPUs. He came to the conclusion that Silvermont's single threaded performance is slower, but other chips consume significantly more power. In his recent article "Apple Outguns Intel", he compared Silvermont again to A7 and, unsurprisingly, Silvermont isn't faster single threaded (and because Merrifield has only 2 cores, it isn't faster in multithreaded workloads either). But after he came to that conclusion, he completely ignored power consumption. If power consumption didn't matter, he could have also compared it to a Core CPU.

Now, let's take a look at one of his articles that I consider "less good": Intel's ARM Problem Intensifies.

Basically, Ashraf here realizes that other companies also have new products scheduled. How surprising! So based on some statements from ARM about one of their new products (that won't be available any time soon), he comes to the conclusion that Intel suddenly isn't competitive anymore, and even worse: that Intel is too slow.

My own conclusion is that his expectations are too high and Intel won't have a free launch. According to what I've heard, Intel's working very hard and they've done a lot in just a few years. Now it will take them another few years to get where they want to be (my own expectation is that ARM will have become irrelevant by 2017).

Ashraf Eassa publishes a huge amount of articles, some are informative or even excellent, some have interesting views, some are decent and some might be "poor" while others contra-dict themselves.

Edit: And yes, it isn't too unsurprising that I think his bearish are generally worse; my own well-informed conclusion is that Intel will succeed in mobile.

 

[Khato]

Fair, unfortunately, as somebody who is personally invested *very heavily* in Intel stock, I have to ask you the following question: when did A7 show up in a shipping device (or Snapdragon 800), and when did Merrifield?

Oh don't get me wrong, I'm as annoyed by Intel's sluggish pace in this area as anyone else. And that's a perfectly valid argument to make. Whereas the article in question... yup, not a single mention of device wins. It's purely attacking Merrifield on the grounds of lacking CPU performance compared to Apple's A7 when a previous article by the same author gave the opposite impression by actually digging into the very same benchmark results.

Also if going into the design win realm... Isn't it the case that you can pretty much summarize Intel's lack of success in that area to two factors? The first being certain uneducated international markets where an SoC using four or eight ARM A7 cores is believed to be superior to a dual core Merrifield purely on merits of core count? And the second being those markets where the first design choice is the modem to ensure it has the checkmark capabilities? Which is to say that Qualcomm has a lock on these markets because no one else has caught up to them on the modem side yet. And that lock is strong enough that Samsung is forced into using Qualcomm SoCs for these markets rather than their Exynos SoCs. (It's pretty obvious that they want to use Exynos when possible given that they go through the expense of designing and manufacturing a second version of many of their phones using Exynos for those markets where a non-Qualcomm modem is acceptable.)

 

[SiliconWars]

Not really. I can give you some examples of good articles: Intel's Tablet 'Giveaway' Explained, Does Intel Spend More Than Qualcomm In Mobile?, Intel: Competitive Analysis Of The Upcoming 'Merrifield' System-On-Chip, Intel Vindicated, Very Competitive With Apple's A7.

As you see, he already compared Silvermont to other CPUs. He came to the conclusion that Silvermont's single threaded performance is slower, but other chips consume significantly more power. In his recent article "Apple Outguns Intel", he compared Silvermont again to A7 and, unsurprisingly, Silvermont isn't faster single threaded (and because Merrifield has only 2 cores, it isn't faster in multithreaded workloads either). But after he came to that conclusion, he completely ignored power consumption. If power consumption didn't matter, he could have also compared it to a Core CPU.

Now, let's take a look at one of his articles that I consider "less good": Intel's ARM Problem Intensifies.

Basically, Ashraf here realizes that other companies also have new products scheduled. How surprising! So based on some statements from ARM about one of their new products (that won't be available any time soon), he comes to the conclusion that Intel suddenly isn't competitive anymore, and even worse: that Intel is too slow.

My own conclusion is that his expectations are too high and Intel won't have a free launch. According to what I've heard, Intel's working very hard and they've done a lot in just a few years. Now it will take them another few years to get where they want to be (my own expectation is that ARM will have become irrelevant by 2017).

Ashraf Eassa publishes a huge amount of articles, some are informative or even excellent, some have interesting views, some are decent and some might be "poor" while others contra-dict themselves.

Edit: And yes, it isn't too unsurprising that I think his bearish are generally worse; my own well-informed conclusion is that Intel will succeed in mobile.

You are aware that Intel17 *is* Ashraf right? And every time you link one of his posts and somebody clicks it, he gets that bit closer to his 10 bucks per 1000 views?

 

[Homeles]

For now but how long do you think they(Intel) can delay the premium that we're so accustomed to(on our desktops/notebooks) & if I'm having this thought then other vendors are too cause there's no telling when Intel withdraws their incentives, as a lot of PC manufacturers have found out the hard way !

With their process advantage? Indefinitely.

 

[witeken]

Fair, unfortunately, as somebody who is personally invested *very heavily* in Intel stock, I have to ask you the following question: when did A7 show up in a shipping device (or Snapdragon 800), and when did Merrifield?

Just a few years after Intel became serious about mobile, which is actually more than a decade shorter than Qualcomm.

Snapdragon 800/801 has won >=150 designs, including the Galaxy Note III, Galaxy S5, Sony Xperia Z2, etc. What has Merrifield won? Why did none of Intel's partners have a single Merrifield device available for show at MWC?

Great question, I wish I knew the answer. But remember that x64 Android is only ready since January.

Even if Merrifield is great and awesome on a perf/watt basis, this doesn't change the fact that both Merrifield and Moorefield will be facing some serious competition in the form of Apple's A8 and potentially a 20nm Qualcomm part by year's end.

Apple's A8 doesn't directly compete with Silvermont. I don't think we'll see a 20nm Qualcomm part in 2014, or at least one that will be ready for the phones that will be announced at IFA. Qualcomm's main advantage is that their designs are just improvements from last year's parts, so OEMs can simply take the updated parts, without the need for a new company, let alone ISA. Intel's new in this market and has a lot of things to prove.

I expect Goldmont based devices will be much better, but again, we all thought Silvermont would be the CPU core (and Merrifield/Bay Trail the SoCs) that finally made Intel competitive. Unfortunately, since these parts were late, late, late, they're not competitive.

They're still competitive, which possibly won't even change until TSMC has 20nm FinFET.

Intel can do better. I know it can. It just needs to stop being afraid of its own shadow (i.e. afraid of cannibalizing Core) and start playing to win. With the PC market on the decline and with server growth seriously anemic, Intel can't afford to miss the mark again with Broxton without seeing a pretty dramatic decline in its earnings/share price.

Then I think Brian Krzanich is the CEO that you wanted:
"We have amazing assets, tremendous talent, and an unmatched legacy of innovation and execution. I look forward to working with our leadership team and employees worldwide to continue our proud legacy, while moving even faster into ultra-mobility, to lead Intel into the next era."

 

[witeken]

You are aware that Intel17 *is* Ashraf right? And every time you link one of his posts and somebody clicks it, he gets that bit closer to his 10 bucks per 1000 views?

*confusion*

 

[Arachnotronic]

Just a few years after Intel became serious about mobile, which is actually more than a decade shorter than Qualcomm.

Need I remind you that Intel has been trying to "get into mobile" since at least 1999?

http://news.cnet.com/Intel-to-buy-DSP-Communications/2100-12_3-265455.html

"Combining DSPC's cellular expertise with Intel's semiconductor and data capabilities will create a leading provider of cellular voice products, as well as establish voice and data solutions for the future," said Davidi Gilo, chairman and chief executive officer of DSPC.

Great question, I wish I knew the answer. But remember that x64 Android is only ready since January.

Did Merrifield need an x64 version of Android? Also, I believe only the kernel is x64 capable; everything else is pretty much X86.

Apple's A8 doesn't directly compete with Silvermont. I don't think we'll see a 20nm Qualcomm part in 2014, or at least one that will be ready for the phones that will be announced at IFA. Qualcomm's main advantage is that their designs are just improvements from last year's parts, so OEMs can simply take the updated parts, without the need for a new company, let alone ISA. Intel's new in this market and has a lot of things to prove.

Indeed. This is why Intel can't afford to just do "good" parts; it needs to show OEMs that they'd be making a grave mistake *not* working with Intel. I don't think Merrifield has really done that, nor do I suspect Moorefield will.

They're still competitive, which possibly won't even change until TSMC has 20nm FinFET.

Even at 20nm planar, TSMC has a pretty substantial density advantage over Intel 22nm (at least in logic, which is driven primarily by metal pitch). On the flip side, Intel's 22nm process has a sizable performance/transistor lead over anything TSMC will be shipping at 28nm or 20nm.

Then I think Brian Krzanich is the CEO that you wanted:

BK sounds like he knows what he's doing and he's cleaned up house since he took the CEO post. I look forward to Broxton.

 

[witeken]

Need I remind you that Intel has been trying to "get into mobile" since at least 1999?

http://news.cnet.com/Intel-to-buy-DSP-Communications/2100-12_3-265455.html

Then why did the same CEO that purchased DSPC say that they didn't pay much attention to the smartphone market with an average selling price of 5$? I also said "serious"; they sort of tried with Saltwell, but it's obvious that they wouldn't really compete with a 5 year old microarchitecture on an n-1 node in the year were quadcores became mainstream in phones.

Did Merrifield need an x64 version of Android? Also, I believe only the kernel is x64 capable; everything else is pretty much X86.

I don't know. I suppose that OEMs want to pair it with a 64-bit OS. Maybe no one simply had a place for Merrifield in their short-term roadmaps.

Indeed. This is why Intel can't afford to just do "good" parts; it needs to show OEMs that they'd be making a grave mistake *not* working with Intel. I don't think Merrifield has really done that, nor do I suspect Moorefield will.

I think it has, at least on the GPU and CPU side. And honestly, who won't get excited by leading edge, tick-tock roadmap? I think it is indeed a mistake to ignore a company who will be 2 years ahead of any one else. I guess Silvermont acts as the proof that they're serious about this.

Even at 20nm planar, TSMC has a pretty substantial density advantage over Intel 22nm (at least in logic, which is driven primarily by metal pitch). On the flip side, Intel's 22nm process has a sizable performance/transistor lead over anything TSMC will be shipping at 28nm or 20nm.

...which Intel will take back within a short time for smartphones and TSMC won't even get for tablets. And it will keep getting worse. By 2017, while TSMC is trying to sell their 20nm node with the 16nm FinFET Plus transistor, Intel will be 2 full nodes ahead of TSMC in density for a full year.

 

[Nothingness]

Funny how some find Ashraf Eassa articles good when they are pro Intel and bad when they are not, and vice versa. This tells a lot about people being biased

 

[Homeles]

Funny how some find Ashraf Eassa articles good when they are pro Intel and bad when they are not, and vice versa. This tells a lot about people being biased

Your need to point it out is equally telling.

If this conversation isn't interesting to you, then stay out of it. Hit and run mudslinging is cowardly and isn't exactly constructive.

It's rather amusing that you weren't able to pick up on the more obvious giveaways: avatars... usernames...

Literally every person in the world has their biases. Get over it. Judge someone by their ability to process information, not by what information they find most interesting.

 

[witeken]

avatars... usernames...

Not sure who you're looking at :sneaky:.

Judge someone by their ability to process information, not by what information they find most interesting.

Beautifully formulated.

 

[bullzz]

@Intel17 - as mentioned earlier, Intel competes with qualcomm and not apple. so lets try to compare it against s801

baytrail was initially planned for netbook segment and it would have taken them a couple of steppings for them to get it working for low end tablet space. I fully expect merrifield to be competetive with S801 but I dont expect there will be design wins in the U.S. In markets where intergrated LTE is not that important, merrifield may get into a few designs

 

[Homeles]

baytrail was initially planned for netbook segment.

I'm doubtful of the veracity of this statement.

I fully expect merrifield to be competetive with S801 but I dont expect there will be design wins in the U.S. In markets where intergrated LTE is not that important, merrifield may get into a few designs

I expect it to be competitive as well, but it's just not going to be making it into very many devices.

 

[jpiniero]

SoFIA might be Intel's best bet to crack the market, not Merrifield.

 

[tarlinian]

Did you ever, or anyone else on this thread, consider the possibility that only Apple & Samsung can afford to pay that premium price for Intel products, since they have ~90% of the sector's profits, & that other OEM/ODM's don't wanna get tied to the Intel ecosystem, which pretty much killed the entire PC sector D:

Except there is no vendor lock-in anymore. Without a huge install base of Windows software there's nothing to keep folks from switching between Intel and an ARM vendor of their choice. And Intel isn't charging premium prices. Unfortunately for them they need to charge less than the competitors due to the lack of an integrated baseband, which drives up PCB costs and increases time to market. I don't think anyone really has a shot in the phone market until they can compete with Qualcomm on basebands, which may happen soon, but that's a much longer and more difficult slog than making a competitive CPU and GPU.

 

[Nothingness]

Your need to point it out is equally telling.

If this conversation isn't interesting to you, then stay out of it. Hit and run mudslinging is cowardly and isn't exactly constructive.

It's rather amusing that you weren't able to pick up on the more obvious giveaways: avatars... usernames...

In fact I was not thinking about people in this forum: I was reading the comments on Ashraf articles where you see some people of both camps saying he's right then wrong depending on the article orientation, which I find quite funny.

BTW I posted numbers in this very thread for 32 vs 64-bit on Z3770. So accusing me of not being constructive or not interested in the discussion is obviously wrong.

Literally every person in the world has their biases. Get over it. Judge someone by their ability to process information, not by what information they find most interesting.

Sorry, but when someone is obviously too biased to recognize when he's wrong, no matter how well he processes information, discussion is useless (and to be clear, you're certainly not one of these ).

 

[frozentundra123456]

Have you ever considered that Intel's Atom prices are actually competitive in these markets?

Huh? Atom can run Android fine.

Well, it has been delayed, and as far as I know no currently on sale devices with atom run android. In any case, the primary advantage of atom to me is the ability to run x86 apps. Not sure why anyone running android would want to switch to intel, unless their price/performance/watt is clearly superior, which right now I dont think it is.

 

[Homeles]

In fact I was not thinking about people in this forum: I was reading the comments on Ashraf articles where you see some people of both camps saying he's right then wrong depending on the article orientation, which I find quite funny.

Okay, thanks for that clarification.

BTW I posted numbers in this very thread for 32 vs 64-bit on Z3770. So accusing me of not being constructive or not interested in the discussion is obviously wrong.

I know that. The issue was that you seemed to be making a passive-aggressive comment directed at forum users here, hence my response.

Sorry, but when someone is obviously too biased to recognize when he's wrong, no matter how well he processes information, discussion is useless (and to be clear, you're certainly not one of these ).

I totally agree. Unfortunately, the number of people that can step back and view things objectively, and see the big picture, are few and far between both on SA, here, the rest of the internet, and so on.

Nobody's perfect, but damn, a lot of people are far, far from it.

Well, it has been delayed, and as far as I know no currently on sale devices with atom run android. In any case, the primary advantage of atom to me is the ability to run x86 apps. Not sure why anyone running android would want to switch to intel, unless their price/performance/watt is clearly superior, which right now I dont think it is.

It's actually really competitive on those metrics. The issue is that it's so late to the game. Bay Trail should beat out a comparable Tegra 4 or Snapdragon 800 device, or at least come very close and trade blows, but as you've mentioned, you can't get Bay Trail on Android yet. When you can later this year, there basically won't be any real con to getting an Intel device vs. an ARM one -- it won't be any different than a Nvidia SoC vs. a Qualcomm SoC.