ARM vs. Intel vs. AMD

[DockMac]

What's holding ARM processors back from overtaking Intel and AMD? They seem more powerful with less energy (yes, Intel and AMD are more powerful overall.) What about the ARM architecture makes it possible to use less energy as opposed to Intel and AMD processors.

 

[smartpatrol]

http://arstechnica.com/business/news/2011/02/nvidia-30-and-the-riscification-of-x86.ars

Read this first. (Skip to the last section on the first page if you want to skip the Nvidia-specific stuff).

 

[DockMac]

Great read. Thanks.

 

[Blitzvogel]

Simply: x86 IPC eats ARM for lunch while actual performance and power usage will scale together. That is why ARM currently has no real business competing against x86. I think it will eventually happen, but not that soon. x86 is too ingrained with servers and practical larger system computing to be replaced all of a sudden, and I actually see x86 taking some of the mobile ground in tablets (thanks to Atom and Bobcat) and if Intel get's it's very low power usage chips out, some of the phone market too.

 

[Deanodarlo]

As mentioned above - ARM may indeed take the portable market by storm, which is starting to take over from desktop use, but x86 isn't going to die anytime soon.

If it were, Intel and AMD wouldn't still invest so heavily in it! A battle is indeed on however, which keeps things interesting.

 

[quest55720]

As mentioned above - ARM may indeed take the portable market by storm, which is starting to take over from desktop use, but x86 isn't going to die anytime soon.

If it were, Intel and AMD wouldn't still invest so heavily in it! A battle is indeed on however, which keeps things interesting.

I think eventually Arm will just be in phones. I see the bobcat and the sucessor to the atom taking over the tablet market once W8 comes out. I think it would be a waste for AMD to go after the phone market. The margins are just way to small to invest money into. ARM and NVIDIA have no other real choice other than to chase those low margin markets.

 

[Ferzerp]

Oh look, this thread again.

 

[Genx87]

I think eventually Arm will just be in phones. I see the bobcat and the sucessor to the atom taking over the tablet market once W8 comes out. I think it would be a waste for AMD to go after the phone market. The margins are just way to small to invest money into. ARM and NVIDIA have no other real choice other than to chase those low margin markets.

Microsoft doesnt seem to agree with that assessment. If they did, Windows 8 wouldnt be coming to ARM.

The margins may be thin. But, the growth potential is astronomical compared to desktop\laptop space. The x86 desktop\laptop space is expected to be stagnant over the next decade. Mobile and ultra mobile markets are set to explode.

 

[quest55720]

I would say the ARM version of W8 will be mostly for phones. It will be the next version of WP.

It would take a very long time for AMD to be competive in the phone space. They would have to design new GPU from the ground up. They are years behind NVIDIA and other companies in that space. That RD money would be better spent on their APU lines. Getting next gen GPU cores on board with better memory controllers should be the priority. Gathering support for open CL is also a must. If they can't compete with intel on the CPU half move the focus to the GPU part where they destroy intel.

 

[Voo]

Microsoft doesnt seem to agree with that assessment. If they did, Windows 8 wouldnt be coming to ARM.

Yep it's not as if MS completely focused on tablets for Windows 8, something where the current low power, low performance ARM designs do actually make sense.

The simple fact is, that performance scaling increases power quite a bit and so far we haven't seen any ARM CPU anywhere in the same league as a modern x86 CPU. Large experienced teams across the whole industry have cocked up high performance CPU designs once or twice - even with all their experience a CPU that looks great on paper doesn't necessarily work in reality (did someone just mention bulldozer? or let's see: pentium4, power6, original itanium,..). So they first need a CPU in the same performance league as say SB and then it's more than questionable if the power efficiency will be any better - especially considering intels process lead.

Also it's not as if the ARM ISA were the second coming (different instruction modes with different lengths, predicates, instructions setting flags,..) - they try to get rid of some of these things though with thumb, but still has its problems too.

 

[-Slacker-]

Let's see...

-ARM is trying to get into laptops and, eventually desktops, while x86 is trying to get into ultra mobile (tablets and phones) score: 0:0

-ARM has failed to get into said market, while x86 can be found in tablets. score 0:1

-ARM dominates a market where operating systems are becoming more and more complex with an ever increasing selection of apps, something that x86 excels at, but still lacks the power efficiency in order to consolidate itself in that market, for now. To put it simply, software for ultra mobile devices is becoming more like the software for laptops and desktops, and x86 is all like "c'mon baby, grandma needs a new pair of shoes!", while ARM is like "Hey man, slow down brah, I can't keep up...".
Score: 0:1



Overall score ARM 0:2 x86.


Looks like x86 is here to stay, from where I'm standing.

 

[Aikouka]

Looks like x86 is here to stay, from where I'm standing.

I'm actually wondering if we'll ever see a revamped ISA.

I know someone is going to read this and bring up the "Itanic" (Itanium), but let's just forget about that... shall we?

 

[CTho9305]

Let's see...

-ARM is trying to get into laptops and, eventually desktops, while x86 is trying to get into ultra mobile (tablets and phones) score: 0:0

-ARM has failed to get into said market, while x86 can be found in tablets. score 0:1

Which x86 tablets don't kinda suck?

-ARM dominates a market where operating systems are becoming more and more complex with an ever increasing selection of apps, something that x86 excels at, but still lacks the power efficiency in order to consolidate itself in that market, for now. To put it simply, software for ultra mobile devices is becoming more like the software for laptops and desktops, and x86 is all like "c'mon baby, grandma needs a new pair of shoes!", while ARM is like "Hey man, slow down brah, I can't keep up...".
Score: 0:1

Overall score ARM 0:2 x86.

Huh? The complexity of software on modern phones and tablets is rapidly reaching the point of "does everything I want", and I'm having trouble imagining how many of my apps will benefit from the speed of the Apple A5 or other dual-core A9 processors. As for power efficiency, I haven't seen data showing an x86 advantage.

Looks like x86 is here to stay, from where I'm standing.

Can you clarify where you're standing? Or provide actual data?

 

[Idontcare]

I'm actually wondering if we'll ever see a revamped ISA.

I know someone is going to read this and bring up the "Itanic" (Itanium), but let's just forget about that... shall we?

I'm sure we will at some point see a revamped ISA. That said, I'm not well versed enough in this stuff to know why we need to see a revamped ISA. Hell, I still can't figure out why the Itanium ISA gets such a bad rap. The HPC stuff I was doing at the time Itanium came out (computational chemistry) showed Itanium to be a real beast.

 

[IntelUser2000]

I'm sure we will at some point see a revamped ISA. That said, I'm not well versed enough in this stuff to know why we need to see a revamped ISA. Hell, I still can't figure out why the Itanium ISA gets such a bad rap. The HPC stuff I was doing at the time Itanium came out (computational chemistry) showed Itanium to be a real beast.

Itanium got a bad rep because excluding Itanium 2 and the 0.13u shrink, every other version got serious delay and was defeatured.

We'll see what happens on Poulson, but after being burned so many times, I can see why the reception is rather poor.

 

[Aikouka]

I'm sure we will at some point see a revamped ISA. That said, I'm not well versed enough in this stuff to know why we need to see a revamped ISA. Hell, I still can't figure out why the Itanium ISA gets such a bad rap. The HPC stuff I was doing at the time Itanium came out (computational chemistry) showed Itanium to be a real beast.

The biggest problem associated with an ISA is that it defines the lowest level of commands that can control a processor. If you redefine these, every application that you were previously using (under x86) will no longer run. Itanium helped alleviate this by providing an x86 interpreter/emulator (not sure exactly what it did processing-wise), but it was too slow.

Adoption of a new ISA has to be one of the hardest things to facilitate. AMD64 (x86-64) pretty much worked because it provided new instructions and features ("64-bit registers", which I believe are actually around 48 bits?) but was backwards compatible with x86.

 

[IntelUser2000]

Itanium helped alleviate this by providing an x86 interpreter/emulator (not sure exactly what it did processing-wise), but it was too slow.

They should have gone with the IA32-EL with the first Itanium. Crappy implementation on hardware is worse than good one on software.

 

[-Slacker-]

Which x86 tablets don't kinda suck?

The point is they exist and they sell, to whatever extent.

Huh? The complexity of software on modern phones and tablets is rapidly reaching the point of "does everything I want", and I'm having trouble imagining how many of my apps will benefit from the speed of the Apple A5 or other dual-core A9 processors. As for power efficiency, I haven't seen data showing an x86 advantage.

Is there any reason to assume there is a sweet spot for the "complexity of software on modern phones" that won't shift forward, despite this being the case with everything else in the software industry since it's inception?

Also, how do you know that what you want from a mobile phone is what everyone does?

Can you clarify where you're standing?

What do you mean? My stance is that x86 has more potential in the overall chip market than ARM. I thought I made that obvious.

Or provide actual data?

Data of what?

 

[Blitzvogel]

Huh? The complexity of software on modern phones and tablets is rapidly reaching the point of "does everything I want", and I'm having trouble imagining how many of my apps will benefit from the speed of the Apple A5 or other dual-core A9 processors. As for power efficiency, I haven't seen data showing an x86 advantage.

"Does everything I want" may mean it does everything you expect a phone to do, but not necessarily what you expect a desktop PC to do. Imagine how long video editing and encoding would take, even on Kal-El , then you have the crappy, narrow 2 x 32 bit memory buses, slow memory, and lack of decent FPU performance without bolted on Neon FPUs, etc. Content creation on mobile ARM CPUs will require a new generation of hardware and software to make the most of that hardware. On x86, it's already here, and systems can be properly configured to do the job, especially when mobility isn't an issue. Even crappy dual core Atoms or low end APUs actually have some real capacity for video editing if you're willing to be patient.

 

[CTho9305]

"Does everything I want" may mean it does everything you expect a phone to do, but not necessarily what you expect a desktop PC to do. Imagine how long video editing and encoding would take, even on Kal-El , then you have the crappy, narrow 2 x 32 bit memory buses, slow memory, and lack of decent FPU performance without bolted on Neon FPUs, etc. Content creation on mobile ARM CPUs will require a new generation of hardware and software to make the most of that hardware. On x86, it's already here, and systems can be properly configured to do the job, especially when mobility isn't an issue. Even crappy dual core Atoms or low end APUs actually have some real capacity for video editing if you're willing to be patient.

I guess I expect that anyone doing a real ARM system will do what every other ARM SOC has done so far, and use acceleration hardware for video. It's an x86-ism (or maybe even a Intel-ism, if you look at Zacate vs. i3) to run media en/decoding on power-hungry out of order cores rather than dedicated hardware.

The memory limitations are, in my view, purely a result of power tradeoffs made at the SOC level; if someone built an ARM SOC targeting 18W (i.e. comparable to Ontario/Zacate), I'd expect them to put a fast, wide DDR3 interface on it; if Intel builds an Atom SOC targeting a phone, I'd expect them to use the same gimped memory controllers you see on the low-power ARM SOCs.

That ARM offers a vfp/neon-free option doesn't mean you have to take it. Up through A9, even with VFP & NEON the floating point performance is pretty low, but my understanding is that A15's VFP/NEON performance is much.

I think it will be very interesting to see what people do with the next 2-3 generations of ARM cores.

 

[Genx87]

Yep it's not as if MS completely focused on tablets for Windows 8, something where the current low power, low performance ARM designs do actually make sense.

The simple fact is, that performance scaling increases power quite a bit and so far we haven't seen any ARM CPU anywhere in the same league as a modern x86 CPU. Large experienced teams across the whole industry have cocked up high performance CPU designs once or twice - even with all their experience a CPU that looks great on paper doesn't necessarily work in reality (did someone just mention bulldozer? or let's see: pentium4, power6, original itanium,..). So they first need a CPU in the same performance league as say SB and then it's more than questionable if the power efficiency will be any better - especially considering intels process lead.

Also it's not as if the ARM ISA were the second coming (different instruction modes with different lengths, predicates, instructions setting flags,..) - they try to get rid of some of these things though with thumb, but still has its problems too.

Are you under the impression ARM based Windows 8 computers will be in the high end? Expect ARM to be in Win8 Tablets, and possibly start to get into low end laptops.

ARM doesnt require the performance crown to be a force.

 

[bryanW1995]

Yeah, I think that ARM is a lot more of a threat to intel than AMD is. I randomly speculated the other day about how much $$$ intel would pay to buy ARM and then just, um, kill it. Is $10 billion too much? What would it take to buy them, anyway?

 

[Genx87]

The point is they exist and they sell, to whatever extent.

Well with what market share? x86 tablets have been on the market for years. But the sales numbers were pathetic. The biggest single brand for Tablets is the iPad\iPad2. That doesnt use an x86 chip. The rest of the top 5-10 is litered with Android based designs not using an x86 chip. So it is hardly a win for x86 if they own a miniscule % of the market. As the market continues to expand, x86 is not the markets leader in any shape or form. Microsoft knows this and built Windows8 so it can run on ARM designs.

Here is an idea where x86 is in the tablet market.

Q2 marketshare data. iPad2 =68.2, Android = 26.8%. That leaves 4.9% for windows based x86 designs.

 

[blckgrffn]

Well with what market share? x86 tablets have been on the market for years. But the sales numbers were pathetic. The biggest single brand for Tablets is the iPad\iPad2. That doesnt use an x86 chip. The rest of the top 5-10 is litered with Android based designs not using an x86 chip. So it is hardly a win for x86 if they own a miniscule % of the market. As the market continues to expand, x86 is not the markets leader in any shape or form. Microsoft knows this and built Windows8 so it can run on ARM designs.

Here is an idea where x86 is in the tablet market.

Q2 marketshare data. iPad2 =68.2, Android = 26.8%. That leaves 4.9% for windows based x86 designs.

Haha, don't forget the firesale Touchpad

 

[quest55720]

I guess I expect that anyone doing a real ARM system will do what every other ARM SOC has done so far, and use acceleration hardware for video. It's an x86-ism (or maybe even a Intel-ism, if you look at Zacate vs. i3) to run media en/decoding on power-hungry out of order cores rather than dedicated hardware.

The memory limitations are, in my view, purely a result of power tradeoffs made at the SOC level; if someone built an ARM SOC targeting 18W (i.e. comparable to Ontario/Zacate), I'd expect them to put a fast, wide DDR3 interface on it; if Intel builds an Atom SOC targeting a phone, I'd expect them to use the same gimped memory controllers you see on the low-power ARM SOCs.

That ARM offers a vfp/neon-free option doesn't mean you have to take it. Up through A9, even with VFP & NEON the floating point performance is pretty low, but my understanding is that A15's VFP/NEON performance is much.

I think it will be very interesting to see what people do with the next 2-3 generations of ARM cores.

FYI amd has a tablet version of zacate that runs at 5.9 watts for tablets the Z-01. Next gen zacate should be even much better at 32nm next year.

An arm tablet can not even run a full fledged web browser let alone real programs. Apps are ok but I want real programs on my tablet. I have played with a few android tablets that the ARM could not even run the basic OS smooth.

I don't see any market for a ARM based desktop any time soon. The current Zacate is cheap and many times more powerfull.

Here is preview of the world of hurt Android and the Ipad tablets are in for next year once W8 hits.

http://www.techspot.com/review/441-msi-windpad-110w-windows-8/page5.html