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

[blackened23]

lol

 

[Exophase]

Exo,

I can believe the bolded sentence. Krait isn't that much faster than Atom, except in FP. General performance doesn't seem to be hugely than the regular A9 cores.

I think it's more nuanced than that, but it's really hard to gain expectations of a chip where so little information is available. One other impression I got from benches is that raw memory performance was substantially better.

According to ARM though, the 1.4GHz A9 gets SpecINT2K score of ~455. The original Silverthorne based core at 1.6GHz gets 653. 1.8GHz Pineview D525 gets 725.

Where has ARM ever delivered SpecInt2K scores for Cortex-A9? Benchmarks like Spec are influenced by more than the CPU core, especially in the case of Cortex-A9 where even the L2 cache isn't part of the core. Of course ARM could do it on an engineering chip that uses their L2 controller (like everyone does AFAIK) and their memory controller (which not everyone does)..

There's a paper by VIA comparing SpecInt2K scores between Atom and Nano: http://www.via.com.tw/en/downloads/whitepapers/processors/NanoX2_whitepaper_201107.pdf There's an important additional point in there: on Atom ICC produces scores 25% better than GCC. It also produces scores 20% better on Nano, just to show that this isn't all due to superior scheduling for an in-order processor that's more sensitive to it. Of course Intel is using ICC when reporting Spec numbers but no one is using ICC on Android and I doubt a lot are using it on Windows 8, especially if also targeting WinRT. This was with GCC 4.5.1, so a little old, but typical x86 performance hasn't improved that much since then (typical ARM performance has improved more)

Part of the advantage from ICC is due to it being a legitimately better compiler, but another part is probably due to Intel specifically tuning for Spec (although given the real world nature of Spec tests it's hard to separate the two)

Phoronix tests are same version of GCC vs GCC so I find them more even and useful for hardware comparison. Hopefully someone gets Linux on a Clovertrail tablet so they can test there, although they could probably do a lot of the tests on a stock Win8 machine.

Acer W510 benchmark shows an advantage of 45-65% in TouchXPRT benchmark. That corresponds to the approximate difference in performance for SpecInt2K.

That difference reduces to 20-60% on the WebXPRT benchmark, which can be explained by having the application multi-threaded to some degree at least.

Those are some data points. Another one is the Word boot where the Tegra 3 machine seems to beat out the Clovertrail. That could be due to storage speeds instead, although the big CPU usage spikes would seem to suggest it isn't I/O bound that whole time.

I asked in the comments for CPU utilization graphs so we could make real analyses about threading vs performance instead of just guessing. This should have been easy for Anand to do, and I hope this gains traction.

It's going to take more than "multi-threaded to some degree" to really make a difference here, when comparing 4C vs 2C/4T on a system that gets a large boost from HT. You'd probably need over 2.5x threading gain to start making a difference on Tegra 3. I personally don't think four synchronously clocked/same voltage cores is a great setup for client loads. That's why I think HT, turbo boost, Qualcomm's asynchronous domains, and big.LITTLE (but much less so Tegra's "companion cores") are very nice features for this space.

Atom's Hyperthreading brings ~35% in SpecIntRate. That with 60% single threaded advantage translates into ~2.2x gain, with probably quad core in Tegra bringing not double, but 1.7-1.8x. That makes sense too.

1.7-1.8x on what, very parallel stuff that maxes both cores without adding a lot of overhead doing so? If so, why? L2 cache competition, bandwidth starving? The latter is going to be more of a Tegra 3 problem than a Cortex-A9 problem, given everyone else is using dual channel controllers..

Regarding memory bandwidth some more.. Anand didn't say what type of memory is used on the tablets (doesn't seem to say it on the W510 review either). Ark says CloverTrail supports up to 400MHz LPDDR2 so we can go with that (http://us.acer.com/ac/en/US/content/model-datasheet/NT.L0KAA.001 at least confirms LPDDR2 and going with < 400MHz seems crazy to me) vs 750MHz DDR3L in Surface RT. That means the raw bandwidth is not that different. But this has implications for power consumption. Even though it's one channel vs two, power consumption of memory doesn't scale linearly with frequency, it's much worse than that (just like with processors). So Tegra 3 takes a big hit running this high frequency RAM. It also takes a hit for being DDR3L instead of LPDDR2, especially in idle.

Memory power consumption will tend to couple pretty closely with CPU power consumption. I don't know if the rails measured are where the memory draws from, but even if they aren't, the memory controller will also use more power for the much higher frequency even w/lower width (64-bit memory controller probably doesn't use anywhere close to twice what a 32-bit memory controller does because a lot of the commanding is redundant).

Anand doesn't mention memory at all, just attributes everything to the SoC which is IMO sloppy.

This is also a major factor in the Cortex-A15 Chromebook power consumption tests that I rarely see mentioned, where dual channel DDR3L 800MHz is used.

A15 is indeed much faster, with score of ~900 for 1.5GHz using Spec.

Source please.. is this from ARM's recent claims regarding A57?

 

[ShintaiDK]

You clearly have no idea what your talking about. Samsung dont get "retail" revenue. How much do you really think they are making per phone?

16GB NAND costs about 2$ and 32GB costs about 4$ and 64GB costs about 30$

So ask yourself why doesnt EVERY phone maker just put 32GB inside? it would give them an advantage over the scroungy 16GB in the iPhone 5 or S3.

Every $ counts thats why. When selling millions of phones 2$ makes a huge difference.

We know the BOM, we know the retail price. Why dont you fill in the blanks when you seem to know more? Or is that a problem?

Samsung only sells around 60million smartphones per year. Yet they doubled their profit quarterly profit to 7.3B due to those smartphones.

 

[Ben90]

If i ever see someone trying to do "real work" on a tablet then quite honestly ill laugh and point at them before feeling sorry for them.

about the only good thing a tablet is for work is for showing people pictures or taking tick box surveys. They are less useful than a pen and note pad.

My work has an iPad and it is nearly infinitely more useful than a desktop for what it was bought for. Not every computer needs to crunch excel to be useful in a business environment. But in your context, of using it instead of a mouse/keyboard, i'll agree: is dumb.

 

[Idontcare]

If i ever see someone trying to do "real work" on a tablet then quite honestly ill laugh and point at them before feeling sorry for them.

about the only good thing a tablet is for work is for showing people pictures or taking tick box surveys. They are less useful than a pen and note pad.

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 )

 

[Homeles]

Simple calculators aren't useful, because they don't have the computational power of a desktop.

I laugh at anyone who tries to do "real work" on them.

Q.E.D.

Anyways, what the hell is "real work?" Fallacious logic abound.
http://en.wikipedia.org/wiki/No_true_Scotsman

 

[blackened23]

Simple calculators aren't useful, because they don't have the computational power of a desktop.

I laugh at anyone who tries to do "real work" on them.

Q.E.D.

Anyways, what the hell is "real work?" Fallacious logic abound.
http://en.wikipedia.org/wiki/No_true_Scotsman

I have a laptop for work and I have an ipad -- I don't game on the laptop, I use it mainly for productivity. As I already mentioned, the ipad is a fun device but only for media consumption.

If I could combine the best of both worlds onto a tablet sized device with ultrabook functionality and x86 compatibility, it would be a great thing indeed. I'm pretty excited about the surface pro.

 

[Ferzerp]

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 )

Maybe by "real work" he means creation instead of consumption. Tablets are wonderful consumption devices, but very poor creation devices.

I've swapped to a tablet for travel (a surface rt) and can do a large number of tasks with it, but when I need to send large amounts of data from myself in to the computing environment, it's pretty poorly suited to that task (though if I really need to do something in depth, I can hook up a usb or bluetooth keyboard and mouse). When I need to retrieve large amounts of data though, it works just fine.

Note: The bulk of the "real work" that I do on my tablet is using it as a terminal for my desktop sitting back in my office.

 

[Homeles]

I have a laptop for work and I have an ipad -- I don't game on the laptop, I use it mainly for productivity. As I already mentioned, the ipad is a fun device but only for media consumption.

If I could combine the best of both worlds onto a tablet sized device with ultrabook functionality and x86 compatibility, it would be a great thing indeed. I'm pretty excited about the surface pro.

Tablets are great "portable document displays." They're great for exploiting the productivity benefits of going paperless.

If I never had to print out another waiver, and could just swing out a tablet and a stylus for the customer to sign — that would be fantastic. And when I needed to find any of those documents again in the future, I don't have to worry about them getting physically lost or having to sort through them all.

Another thing that tablets are great at is battery life, although eventually laptop PCs will catch up in this regard thanks to nudging from Intel.

 

[Idontcare]

Maybe by "real work" he means creation instead of consumption. Tablets are wonderful consumption devices, but very poor creation devices.

I've swapped to a tablet for travel (a surface rt) and can do a large number of tasks with it, but when I need to send large amounts of data from myself in to the computing environment, it's pretty poorly suited to that task (though if I really need to do something in depth, I can hook up a usb or bluetooth keyboard and mouse). When I need to retrieve large amounts of data though, it works just fine.

Note: The bulk of the "real work" that I do on my tablet is using it as a terminal for my desktop sitting back in my office.

If we are willing to confine the definition of "real work" and "creation" to that of the engineering/science facing professions then there is no question that anything which doesn't look/walk/talk like an HPC facsimile is going to fall far short of the end-user needs for that limited segment of the professional sector.

There is a big world out there, and I fear that some folks aren't interested in standing up and looking over the top of their cubicle walls to realize engineering and science is not the end-all be-all of the working world, and thus their views on what makes or breaks a computing solution as being viable versus useless is not going to be well reflected in the reality of shifting market volumes and segments.

But it is the most likely demographic to be present in this type of a forum, so naturally that is going to be the sort of mindset we tend to gravitate towards in these discussions. So I can't fault them, they only know what they know and that is the foot they are going to lead with in just about any debate.

But the world is big, and engineering/science professions are not the sole authority on what is defined as real work or the creation of work versus the consumption thereof. It is easy to pigeon hole ourselves with the close-mindedness that comes from lack of perspective, I wouldn't think about it myself if I weren't surrounded by working professionals in my family who are not engineers themselves.

Makes it hard to ignore the reality of what they use their tablets and smartphones to accomplish, and it is a reality that we don't really correctly fathom here in our discussions on this forum IMO.

 

[Ferzerp]

If we are willing to confine the definition of "real work" and "creation" to that of the engineering/science facing professions then there is no question that anything which doesn't look/walk/talk like an HPC facsimile is going to fall far short of the end-user needs for that limited segment of the professional sector.

There is a big world out there, and I fear that some folks aren't interested in standing up and looking over the top of their cubicle walls to realize engineering and science is not the end-all be-all of the working world, and thus their views on what makes or breaks a computing solution as being viable versus useless is not going to be well reflected in the reality of shifting market volumes and segments.

But it is the most likely demographic to be present in this type of a forum, so naturally that is going to be the sort of mindset we tend to gravitate towards in these discussions. So I can't fault them, they only know what they know and that is the foot they are going to lead with in just about any debate.

But the world is big, and engineering/science professions are not the sole authority on what is defined as real work or the creation of work versus the consumption thereof. It is easy to pigeon hole ourselves with the close-mindedness that comes from lack of perspective, I wouldn't think about it myself if I weren't surrounded by working professionals in my family who are not engineers themselves.

Makes it hard to ignore the reality of what they use their tablets and smartphones to accomplish, and it is a reality that we don't really correctly fathom here in our discussions on this forum IMO.

Tablets have grown on me over the years, but that's because my role has morphed from hard core always transfering information out of my head on to a computer to more design, discussion, etc. I still have those long sessions of furious typing and mousing, but they are not 80+% of my time any more. As that change has happened, tablets have gotten a place in my world.

I use them for what they're suitable for, and I don't try to shoe-horn them in to a complete traditional PC replacement role. For me, they're a supplimental device. I'd prefer one that is x86 compatible, and may take a look at some of the upcoming atom tablets for this purpose.

However, I do recognize that there are a rather large percentage of people who they work fine for. Desktops are cheap enough that I sort of think that most people still would be benefitted from having one of those around as well, but if you're a web browser and mail reader, and can only afford one device, there are plenty of tablets that fit that role.

Of course, at the same time, I also realize that the procs we use in tablets aren't suited for anything more intensive than tablets as well. And even then, I think my tablet is rather too slow for comfort in certain areas (which is why it often just takes the thin-client role).

 

[Idontcare]

Tablets have grown on me over the years, but that's because my role has morphed from hard core always transfering information out of my head on to a computer to more design, discussion, etc. I still have those long sessions of furious typing and mousing, but they are not 80+% of my time any more. As that change has happened, tablets have gotten a place in my world.

I use them for what they're suitable for, and I don't try to shoe-horn them in to a complete traditional PC replacement role. For me, they're a supplimental device. I'd prefer one that is x86 compatible, and may take a look at some of the upcoming atom tablets for this purpose.

However, I do recognize that there are a rather large percentage of people who they work fine for. Desktops are cheap enough that I sort of think that most people still would be benefitted from having one of those around as well, but if you're a web browser and mail reader, and can only afford one device, there are plenty of tablets that fit that role.

Of course, at the same time, I also realize that the procs we use in tablets aren't suited for anything more intensive than tablets as well. And even then, I think my tablet is rather too slow for comfort in certain areas (which is why it often just takes the thin-client role).

I have yet to find a tablet or smartphone that is a suitable computing device for my computing needs. But when I look at what I do with my computers (desktops and laptops) I realize I am definitely a member of the computing minority.

But when I look at what other people do with their tablets and smartphones, not speaking of consumers but speaking of business professionals, I recognize that these new form-factors have enabled an entirely new level of functioning efficiency for them that was simply never going to be achieved with traditional form factors regardless the performance or battery life.

I could never give up my keyboard and mouse, and I don't think anyone in the sciences or engineering will really ever have that luxury. But managers, teachers, social services, health care professionals, distribution/inventory management, etc do have the luxury of taking advantage of it...which is probably why there is such an astonishing adoption rate of those form factors across those professions.

 

[Homeles]

Of course, at the same time, I also realize that the procs we use in tablets aren't suited for anything more intensive than tablets as well. And even then, I think my tablet is rather too slow for comfort in certain areas (which is why it often just takes the thin-client role).

That will be changing within the next year or two.

 

[SickBeast]

I really think that Intel is going to have to swallow their pride and license ARM and then manufacture it on a smaller node than anyone else. They will make a killing if they do this. It might even make sense for them to buy a company like nvidia to give them a leg up.

 

[Homeles]

I really think that Intel is going to have to swallow their pride and license ARM and then manufacture it on a smaller node than anyone else. They will make a killing if they do this. It might even make sense for them to buy a company like nvidia to give them a leg up.

Why would Intel admit defeat? Right now, they're the ones winning the ARM war.

 

[blackened23]

I'm not sure why everyone is so caught up in the semantics of what constitutes "proper work" on a mobile computing device. I know for my needs, a tablet isn't really suited to what I want.

Yes, some industries can definitely make use of tablets. If you fit in this category, good on you. But they are ill suited for many tasks -- My entire point is that the surface pro is more versatile, while still being in a tablet form factor. If you want to use it as a tablet, go for it. If you want to create powerpoints, you can do that the traditional way as well. You can do all of the fun stuff that one can do on an ipad, while still having ultrabook x86 functionality if you want.

That was my entire point; this is why i'm excited about the surface pro. Versatility. That is all.

 

[Grooveriding]

Tablets are used where I work, but mostly in meetings by people I get the impression are using them more out of an affectation than utility. I cannot see how they are serving them better than a laptop.

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.

Can someone ever type as efficiently using a tablet as they can with a proper keyboard ? I don't know. I can see them having a niche place in the right workplace, but never superseding a traditional keyboard/mouse machine overall. I think education is one of the few areas off the top of my head where they could become dominant, but even then more so amongst students.

I think a really near perfect implementation of dictation software with flawless text to speech, grammar and formatting features could go a long way to making tablets more useful than they are for workplaces though. Also just more raw horsepower than they have currently of course. I'd also think a move to tablets based around Windows will be needed so they can interface nicely with a given company's network and be familiar to employees and offer traditional apps they know and use; word, excel, pp, outlook etc.

 

[Ferzerp]

That was my entire point; this is why i'm excited about the surface pro. Versatility. That is all, move along

I'm not so sure. The battery life issue is going to make the 3rd party atom based tablets more attractive for many people.

 

[Arachnotronic]

I really think that Intel is going to have to swallow their pride and license ARM and then manufacture it on a smaller node than anyone else. They will make a killing if they do this. It might even make sense for them to buy a company like nvidia to give them a leg up.

Why? You realize this makes no technical or business sense, right?

Here are just a few reasons why this is absolutely silly:

- Intel gives up the ability to define new instruction set features, hampering potential performance improvements until the "mighty" ARM decides to finally update its instruction set
- Do you mean Intel should just license an ARM core? Because the leader in the field -- Qualcomm -- along with Apple, doesn't do that. So do you mean Intel should build a micro-architecture based on the ARM design? Are you even aware of this distinction?

I'm sorry, but anybody who thinks that there is ANY advantage to Intel becoming a commodity ARM designer (besides perhaps compatibility with iOS, which Intel wouldn't get anyway) and throwing away the advantages that it will have in Windows 8 tablets (FULL COMPATIBILITY) in addition to the ability to control its own destiny AND not have to pay licensing fees, does not "get it" and drinks way too much of the ARM kool-aid.

Intel's problem is that its Atom design is old and crusty. It needs a nice, radical redesign that will take advantage of the 22nm process tech (while others will still be fiddling around on 28nm/20n planar). "Silvermont" is that redesign.

ARM is great for companies that do not have the clout to develop their own instruction set architectures and/or CPU cores. It's not appropriate for a company that dominates almost every aspect of the compute continuum with its own ISA and designs.

 

[SickBeast]

Why? You realize this makes no technical or business sense, right?

Because of the app ecosystem that exists. x86 has no place on phones for the foreseeable future.

 

[SickBeast]

Why would Intel admit defeat? Right now, they're the ones winning the ARM war.

How are they winning when 99% of phones have an arm CPU?

 

[Homeles]

How are they winning when 99% of phones have an arm CPU?

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 &#8212; ULV Ivy Bridge and a 4 year old Atom &#8212; are just stopgaps.

 

[IntelUser2000]

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

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.

 

[SickBeast]

N/m

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

 

[Arachnotronic]

Because of the app ecosystem that exists. x86 has no place on phones for the foreseeable future.

Most Android apps are not written in C/C++ and compiled down into ARM assembler. The ones that are native, Intel will probably write the checks to make sure are recompiled to X86. These are the kinds of investments a company can make when it runs very profitable businesses.