The ARM v.s. Intel Thing - Let's Discuss It

[podspi]

IMHO, dockable/convertible tablets/laptops are going to become huge in the next few years.

Most people will find the need at some point for a keyboard and mouse, while at the same time might not be interested in carting those input devices around when they're just consuming media. Yes, you could always just have multiple devices, but they take up space, you have to sync data between them, etc. From experience, I can say the ability to transform the device you are currently working with into the appropriate form factor is awesome-fantastic.

 

[SickBeast]

http://www.anandtech.com/show/6529/busting-the-x86-power-myth-indepth-clover-trail-power-analysis

It's only a matter of time at this point. It takes a long time to make adjustments in the electrical engineering world, but the hammer comes crushing down beginning next year. The majority of the changes being implemented in Haswell and Atom are made specifically for dealing with the ARM problem. Intel's current products — ULV Ivy Bridge and a 4 year old Atom — are just stopgaps.

You're comparing future Intel chips to current ARM-based CPUs. The new quad cores coming out soon for ARM are going to be very powerful. Project Denver will be a beast. I will really believe it when I see it with Intel. I think that they will stick to x86 and will release some incredible laptops and maybe tablets based on Haswell and Broadwell. I don't see them making any major inroads into phones for the next 3-5 years though, which is really their loss. Like I said, if they would design an ARM chip and manufacture it on a smaller node they would make a killing.

 

[Exophase]

Exophase, Clover Trail supports LPDDR2, but at 800MHz frequency.

What you're referring to is the data transfer rate, which is 800MT/s - double data rate @ 400MHz. Look at the numbers on ark: 6.4GB/s. 400MHz * 2 transfers per cycle * 8 bytes (64-bit interface).

A lot of people refer to that as clock speed, but there's no convention for that. IMO the terminology I used is more accurate. More importantly, my numbers were however consistent - I called Tegra 3 out as using 750MHz DDR3L, in other words, DDR3L-1500. And the underlying point is the same regardless of the terminology: the Tegra 3 solution is offering slightly less peak bandwidth at a much higher operating frequency using a less power efficient memory technology, and I would expect is consuming more power to do so.

I'm sure Intel could push a higher memory clock but are avoiding it in the interest of power savings. It may be one of the things keeping TDP down vs Cedar Trail. As I mentioned earlier, Exynos 5 has a ton of bandwidth for its class (2x Cedar Trail) but pays dearly for it in power consumption when paired with DDR3L. AFAIK it supports LPDDR3.. Tegra 3 doesn't have that luxury, and you're forced to use the much less power efficient solution if you want to get acceptable bandwidth.

And yes, its from the A57 one, but I have seen actual numbers for A9 as well. It's similar to 1 thread portion on the rate bench.

I remember seeing that number from someone (possibly Wilco?), but not from ARM. If you haven't noticed, I'm not that trusting of benchmark numbers provided by a first party company for a product, particularly with zero disclosure of the test conditions, nor am I very enthusiastic about comparing benchmark numbers between separate sources. Intel gets around having to present anything to SPEC by using older suites; they don't even list SPEC2k results on their site so that much is moot. David Kanter said that Intel would probably be happy to disclose the testing details if someone asked and said he would, but I never heard anything more about this.

 

[Arachnotronic]

You're comparing future Intel chips to current ARM-based CPUs. The new quad cores coming out soon for ARM are going to be very powerful. Project Denver will be a beast. I will really believe it when I see it with Intel. I think that they will stick to x86 and will release some incredible laptops and maybe tablets based on Haswell and Broadwell. I don't see them making any major inroads into phones for the next 3-5 years though, which is really their loss. Like I said, if they would design an ARM chip and manufacture it on a smaller node they would make a killing.

Project Denver...from a company that has yet to design a well balanced SoC, let alone processor core...will be a beast?

Did you know Denver is targeted as Tesla applications first and not for smartphones/tablets?

So you would trust Nvidia over Intel...a company that is another "me too" SoC vendor over one that has very tangible cost and technological advantages (as well as cash flow/profitability)?

Intel's profits in a year could buy Nvidia twice over. Think about it.

 

[IntelUser2000]

Exo, you are right.

But for Cedar Trail, it uses 1066MT one, so its not that much higher. I think the higher TDP is mostly due to Intel aiming Tablet and Smartphone parts as premium compared to Netbook's Cedar Trail. Think how Desktop and Laptop Core parts compare in the same TDP for example. The 35/45/65W versions of Desktop chips perform lot less than the mobile variants even at the same TDP!

About the compilers: Back when Oak Trail was being aimed for Tablets, initial benchmarks for Android 3.0 surfaced. It was horrible. It performed 1/3 to 1/4th of A9 cores clocked at 1GHz, and actual usage was really jerky too. But it wasn't on 2.1 of course.

After they announced collaboration with Google, they showed results of optimization with the Atom chips. I wouldn't doubt if the OS itself is top notch optimized for Intel.

 

[SickBeast]

Project Denver...from a company that has yet to design a well balanced SoC, let alone processor core...will be a beast?

Did you know Denver is targeted as Tesla applications first and not for smartphones/tablets?

So you would trust Nvidia over Intel...a company that is another "me too" SoC vendor over one that has very tangible cost and technological advantages (as well as cash flow/profitability)?

Intel's profits in a year could buy Nvidia twice over. Think about it.

Intel has high profits for now while AMD offers them no competition. Going forward their market is eroding as more and more people move to tablets and smartphones. They have no tangible CPUs for these products yet, and they won't for the foreseeable future so long as they stubbornly stay with x86. x86 phones will be a very niche market so long as Android and iOS continue dominating. Even if Intel made an OS they would be relegated to 1% of the market like RIM is getting.

 

[Fox5]

Project Denver...from a company that has yet to design a well balanced SoC, let alone processor core...will be a beast?

Did you know Denver is targeted as Tesla applications first and not for smartphones/tablets?

So you would trust Nvidia over Intel...a company that is another "me too" SoC vendor over one that has very tangible cost and technological advantages (as well as cash flow/profitability)?

Intel's profits in a year could buy Nvidia twice over. Think about it.

My thoughts:
By the end of 2013:
Intel's Silvermont will prove to be a beast and take the high end (performance wise) of the cell phone market, and more so the tablet and netbook/ultrabook/chromebook slim and light category, but won't have very many design wins overall.
Nvidia will gain a ton of low - mid cell phone market share with a low cost Tegra 3 variant. They will have some Tegra 4 tablet / slimbook wins and almost no Tegra 4 presence in smart phones.
Samsung will have switched over almost exclusively to their own processors.
Qualcomm will continue to have the majority of the non-Samsung cell phone design wins.

About the compilers: Back when Oak Trail was being aimed for Tablets, initial benchmarks for Android 3.0 surfaced. It was horrible. It performed 1/3 to 1/4th of A9 cores clocked at 1GHz, and actual usage was really jerky too. But it wasn't on 2.1 of course.

After they announced collaboration with Google, they showed results of optimization with the Atom chips. I wouldn't doubt if the OS itself is top notch optimized for Intel.

At the time, the Dalvik JIT didn't run on x86 or wasn't optimized for it.

 

[CHADBOGA]

Intel has high profits for now while AMD offers them no competition. Going forward their market is eroding as more and more people move to tablets and smartphones. They have no tangible CPUs for these products yet, and they won't for the foreseeable future so long as they stubbornly stay with x86.

Even on an old and uninspiring designed Atom x86 on 32nm, for Windows 8, it is clearly better than Microsoft's ARM(Nvidia's Tegra 3) choice for Surface RT.

What do you think will be the case when the newly designed and much revamped Atom chip ships on 22nm?

x86 phones will be a very niche market so long as Android and iOS continue dominating.

Intel will become a significant player in phones running Android, so Android's existence isn't a problem, nor does Intel need their own O/S.

 

[Genx87]

Wintel's dominance falling is starting to show. I think Intel has a better chance of making it out alive than Microsoft. Microsoft cant do anything in the mobile market correctly. As the PC markets becomes more and more irrelevant so will Microsoft. Intel's fate is debateable as they may be able to compete in the ultra mobile markets. The question is if they can afford the lower margins given all that infrastructure they have to support(fabs).

 

[blackened23]

Intel has high profits for now while AMD offers them no competition. Going forward their market is eroding as more and more people move to tablets and smartphones. They have no tangible CPUs for these products yet, and they won't for the foreseeable future so long as they stubbornly stay with x86. x86 phones will be a very niche market so long as Android and iOS continue dominating. Even if Intel made an OS they would be relegated to 1% of the market like RIM is getting.

What makes you think that intel chips aren't able to run in android environments?

 

[Exophase]

Exo, you are right.

But for Cedar Trail, it uses 1066MT one, so its not that much higher. I think the higher TDP is mostly due to Intel aiming Tablet and Smartphone parts as premium compared to Netbook's Cedar Trail. Think how Desktop and Laptop Core parts compare in the same TDP for example. The 35/45/65W versions of Desktop chips perform lot less than the mobile variants even at the same TDP!

Yeah, but DDR3. I can't even find word that it supports DDR3L, much less would it be required to, therefore it has to take the brunt of 1.5V DDR3. Actually, if cpu-world is to be trusted it also supports DDR2 (1.8V) at the same clocks, which makes the situation even worse. Although the power consumption of the memory itself wouldn't be reflected in the TDP having to drive a higher voltage interface and with less power saving features will hit back on the peak consumption possible on the memory controller, which will impact TDP.

35/45/65W TDP desktop variants grossly underperforming laptop counterparts is news to me. I wasn't even aware of < 65W desktop chips for SB or IB, at least for Core series. Since none of those are unlocked (to the best of my knowledge?) is the difference real and not just artificial binning? Or was that your point?

The thing is, there's no real demand for low TDP on desktop parts, but there's plenty of demand for cheap parts. And if they happen to get a lower TDP after castrating the chip that's just a perk. Now I don't know the pricing for Cedar Trail vs Clover Trail, but on the former there's only a small amount of market segmentation as far as pricing would be concerned, and the slower one is probably not the cheaper one given how much lower its TDP is. And both sets are very much power sensitive parts so there'd be a reason to offer them at the lowest TDP feasible.

About the compilers: Back when Oak Trail was being aimed for Tablets, initial benchmarks for Android 3.0 surfaced. It was horrible. It performed 1/3 to 1/4th of A9 cores clocked at 1GHz, and actual usage was really jerky too. But it wasn't on 2.1 of course.

After they announced collaboration with Google, they showed results of optimization with the Atom chips. I wouldn't doubt if the OS itself is top notch optimized for Intel.

Probably, although I'm not sure what that'd entail. We don't see enough tests confirmed to be Dalvik running. And of course this wouldn't play an issue for anything native. Would be curious to see these poor benchmarks; it'd pretty much have to be down to Dalvik's JVM not yet having good x86 output (this is pretty much the one and only place I'd expect to see ARM given first priority). The rest of the OS isn't going to make a big difference for most benchmarks, even if it can make a big difference for how smooth it feels.

Almost everything tested is Javascript, and Atom was never behind there - in fact people loved to show Atom slaughtering Cortex-A8 and A9 in JS performance back when it was 2-3x worse than it is now. This was true even on those old Honeycomb Atom releases.

 

[Arachnotronic]

What makes you think that intel chips aren't able to run in android environments?

*sigh*...

 

[Arachnotronic]

Wintel's dominance falling is starting to show. I think Intel has a better chance of making it out alive than Microsoft. Microsoft cant do anything in the mobile market correctly. As the PC markets becomes more and more irrelevant so will Microsoft. Intel's fate is debateable as they may be able to compete in the ultra mobile markets. The question is if they can afford the lower margins given all that infrastructure they have to support(fabs).

Common misconception. If you listen to the analyst day presentations, you will hear that Atom MARGINS are quite good. It's that they are lower ASP that is a problem if you start seeing 1-for-1 exchanges of Core for Atom.

If Intel dominates the mobile market, it will do so with 60%+ gross margins. The question is whether Atom sales will be at the expense of Core sales.

 

[Homeles]

You're comparing future Intel chips to current ARM-based CPUs.

And you aren't? Quad core A15s are not on the market yet, and next gen Atom and Haswell are right around the corner as well. They'll be launching only shortly after quad core A15.

Project Denver will be a beast.

Intel will already have a 64 bit smartphone chip by then. Depending on 14nm's time frame, they could very well have 14nm Atom and Broadwell out by then. Then it will be game over.

Like I said, if they would design an ARM chip and manufacture it on a smaller node they would make a killing.

Why would they do that? They'll have next gen Atom out before they could put out an ARM design. Silvermont has been in the oven for quite some time... it won't be a slouch. They also haven't flexed their 22nm FinFET muscle yet.

 

[IntelUser2000]

35/45/65W TDP desktop variants grossly underperforming laptop counterparts is news to me. I wasn't even aware of < 65W desktop chips for SB or IB, at least for Core series. Since none of those are unlocked (to the best of my knowledge?) is the difference real and not just artificial binning? Or was that your point?

That was my point. Core i7 3770T at 45W TDP has lower Base and Turbo clocks compared to 45W mobile 3820QM. S is 65W and T is 45W for desktop. Ok, maybe the difference there isn't drastic, but still its there.

Or that how all Mobile Core chips have GT2 but only certain ones for Desktop.

The difference probably isn't all artificial, but probably plays a non-significant part.

About optimization: They said SSE3 optimizations brought 7x increase in some cases. Various cases were there, while most were nowhere near 7x, it was still significant.

 

[Idontcare]

N/m

I will reply later, this is a pita on my phone.

Ironic post given the topic of the thread

 

[Arachnotronic]

Ironic post given the topic of the thread

The PC is dead!

 

[krumme]

Intel market cap tell it all. There is a reason for such a low market cap despite the absolutely huge competences inside.

If i was a huge player in the mobile space, why should i chose Intel?

I can get an open ARM standard, where i can change supliers as i like, it gives me plenty of the performance i need. I need to manage my suppliers, not be dictated the other way round. No way. Ask Apple if they need more than A7V? Man even a beefed up little in little-big A7-A15 is enough. All for 10usd for the performance variants.

What do i give a s..t if Intel offered me an Atom that does it at twice the speed for only 15 usd?

There is only strategic disadvantages to chose Intel. The hkmg in TSMC 28nm is effective going to shut the door for the new Atom, no mather how good they are. Excactly as cheap 40nm it is now for the technically superior 32hkmg Atom variants.

 

[ShintaiDK]

For how long can you freely choose ARM maker? To me it seems like it will be Samsung And Qualcomm left as the only ARM makers for the smartphoen segment in the near future.

You aint getting a performance ARM chip for 10$.

Even if we imagined that you did get it for 10$, just as we would have to with a 15$ Atom. Then if the Atom offers something that consumers want. You would instantly switch to it to protect the 200% profit margin that higher smartphones usually carry. In that game, 5$ is nothing.

Atom is right now on the path to beat ARM in both performance and power consumption. Two very important factors.

Right now the 32nm Atoms are expanding in the smartphone segment. Quite the opposite of what you claim would show.

 

[sontin]

"Expanding"? How many designs are on the market right now? Two? nVidia has more with Tegra 3.

And how is "Atom is right now on the path to beat ARM in both performance and power consumption", when Intel can only beat a one year old ARM design which is nearly end of life? They are one year to late to the market.

 

[ShintaiDK]

"Expanding"? How many designs are on the market right now? Two? nVidia has more with Tegra 3.

And how is "Atom is right now on the path to beat ARM in both performance and power consumption", when Intel can only beat a one year old ARM design which is nearly end of life? They are one year to late to the market.

There are atleast 7.

The current Atom is an even older design. Getting a complete uarch refresh next year.

 

[paulcheung]

Common misconception. If you listen to the analyst day presentations, you will hear that Atom MARGINS are quite good. It's that they are lower ASP that is a problem if you start seeing 1-for-1 exchanges of Core for Atom.

If Intel dominates the mobile market, it will do so with 60%+ gross margins. The question is whether Atom sales will be at the expense of Core sales.

I think this is the main problem the management of Intel have. Should they worry the sale of Atom will cut into the i-cores sales too much. But I don't think they have too many choices now.

I am wonder if the old Wintel thing will come back and create another Anti-trust issue if Microsoft and Intel team up again use the Atom for the Surface RT and the i-cores for the Surface Pro. Is Surface really work as Microsoft claim? without any problem?

 

[Fx1]

My sister is a special ed teacher in elementary school, the school is setup with ipads and all the students in her class (as well as herself) use ipads for all school work lessons, exams, grading, etc.

To her, the students, the state, and the tax-payers, that is real work.

I could provide more examples, first person ones, but I have a feeling you live in a rather sheltered world that has not made available to you the opportunity to see much of how the professional world operates.

So I would agree, if your life up until now has pretty much involved standing in unemployment lines, working at best buy, or flipping burgers...then from that perspective a tablet would surely seem like a silly toy with little value added to the work environment of a professional.

(no worries, I understand you are like the average AT'er, you make 6+ figure salary, are an MIT grad who designs nuclear submarines, and you know for sure tablets are useless to anyone with an inferior IQ to yours )

Sheltered?

I have worked for some of the worlds biggest companies. Most of which still use Lotus Notes and windows XP. Let alone iPads.

Let me get something straight here. Content consumption by kids is not "work" nor is what a teacher does on it. Its basically a modern day text book.

Most kids who learn on ipads will have the written skills of an idiot by the time they are 18 and be unemployable. You only have to look on facebook today and see the damage already done to 'da idyots hu spk lyk dis' and think this is written communication. Most cant even use a pen now.

Typing is a chore on an ipad it has the worlds worst keyboard. infact doing anything on it is a PITA. i have owned 2 of them now and sold both because they are just terrible.

 

[atomheart]

Part of the problem as well is that they do not integrate well with the network and are not secure so they cannot be used to access and make use of records that contain private/personal information. This is in a health care setting. In our department I don't see a tablet ever being used until it supports the resolutions needed to display imaging results. Even then they will not be of use as you need not just resolution but a screen with a fairly large form factor as well.

That's strange because at my Father's work (health care clinic) the doctors now use iPads to handle the patient files and a lot of the record keeping when going from exam rooms. He said part of the reason they switched was due to it working with the clinics network properly (servers are run offsite by a company called Telemark).

Then again he's just a CFO there so maybe he wasn't 100% correct as to the reason behind the switch.

 

[Grooveriding]

That's strange because at my Father's work (health care clinic) the doctors now use iPads to handle the patient files and a lot of the record keeping when going from exam rooms. He said part of the reason they switched was due to it working with the clinics network properly (servers are run offsite by a company called Telemark).

Then again he's just a CFO there so maybe he wasn't 100% correct as to the reason behind the switch.

I work in a hospital. None of the physicians, nursing staff or techs here are allowed to use their private tablets/laptops (the hospital does not purchase any sort of tablets for use) to access the internal network, only for internet access using wi-fi.

I've seen them used in private offices where GP's work, but I am not aware of any of them being implemented in hospitals in the city I live in.

In a smaller scale facility such as a private clinic where the decisions on to use or not use tablets are going to be guided more by the preference of one or two people, it would make sense. I think this would be true of any smaller scale business though, health care oriented or not. There are situations where a small scale and portable device could bring some utility. If you were speaking to an inpatient, who is more than likely going to be in a bed, then a tablet can bring something to the table for consulting their charts by their bedside. Currently we have computers throughout every area of the hospital and you often use them before speaking to a patient, so I can see some benefit from a tablet allowing you to access records in the midst of being with a patient.

At our hospital it would require a complete rewrite of our software though. The software is custom, search functions and drilling down through records would be a real PITA via a tablet interface.

I think tablets need to be more robust in the way a user can manipulate the interface before they can trump the utility of keyboard/mouse. Just speaking from my personal experience where I work of course.