ARM: How does it compare, exactly?

cheesehead

Lifer
Aug 11, 2000
10,079
0
0
Does anyone here have any information on the pros and cons of the ARM architecture? I'm curious how it compares to X86, POWER, and the various other options on the market today. The 600mhz G3 in an eight-year-old iBook is fast enough (just barely) for use as a real computer despite only pulling three watts in the latest revision - is the ARM really that much more efficient?
 

BD231

Lifer
Feb 26, 2001
10,568
138
106
Back in the day risc was known for floating point performance that used to trounce the x86 platform (depending on the application), but that simply isn't true anymore. Apple made you pay through the teeth for it which made the down sides (lack of compatibility, programing) that much more apparent.

Floating Point was their bread and butter, but if there's no programing or widespread acceptance like AMD/Intel have garnered it's really not something an end user can take advantage of for anything other than niche computing.

Dose mac even sell mac books with risc processors any more? Personally, I find it a little odd risc computing didn't garner more attention; but over the years it's always found a niche for itself.
 

thesmokingman

Platinum Member
May 6, 2010
2,307
231
106
I don't think risc vs cisc arguments matter anymore in this post risc era especially with how they've evolved gaining characteristics of both sides.
 

evolucion8

Platinum Member
Jun 17, 2005
2,867
3
81
RISC and its iterations like ARM offers a nice balance in terms of power/performance ratio compared to the current x86 architecture, but in terms of pure raw power, ARM is far behind to be used in big computers, it fares much better in small devices.
 

EarthwormJim

Diamond Member
Oct 15, 2003
3,239
0
76
Brightsideofnews did a pretty neat comparison between an Atom system and an Cortex A8 system running Linux. Keep in mind that compilers make a big difference, so it's not a perfect comparison.

http://www.brightsideofnews.com/news/2010/4/7/the-coming-war-arm-versus-x86.aspx

Basically right now, integer performance is comparable at similar clock speeds between ARM (Cortex a8) and Atom (x86), while floating point performance is no where near x86 levels.

Both are in-order designs, so they're really strapped for memory bandwidth. Throw a 64-bit memory bus on a Cortex SoC, and I'm sure performance would go up a lot (not to mention cost and power though).


I would not bother comparing ARM to other x86 designs, Atom is the only comparison, and even then Atom uses a lot more power.
 
Last edited:

Vesku

Diamond Member
Aug 25, 2005
3,743
28
86
Essentially the ARM cores do extremely well at integer ops for their power consumption, not so much floating point ops. There are many apps that don't use floating point much if at all. Which is why there has been mention of ARM possibly making it's way into the server space. Edit: Since rewriting or recompiling for different architectures happens much more often for server farm applications than consumer applications.

Personally, I would like to see more consumer interest in ARM + Linux devices. For some reason it has been a bit tough for them to break out of the phone space so far.
 
Last edited:

Hacp

Lifer
Jun 8, 2005
13,923
2
81
In google's best practices guide for android development, it says to avoid floating whenever possible.
 

cbn

Lifer
Mar 27, 2009
12,968
221
106
Does anyone here have any information on the pros and cons of the ARM architecture? I'm curious how it compares to X86, POWER, and the various other options on the market today. The 600mhz G3 in an eight-year-old iBook is fast enough (just barely) for use as a real computer despite only pulling three watts in the latest revision - is the ARM really that much more efficient?

http://www.semiaccurate.com/2010/04/30/amd-wants-slice-slate-market-fusion/

ARM sounds pretty strong in this article.

In google's best practices guide for android development, it says to avoid floating whenever possible.

Thanks.

The GPU portion of AMD Fusion would use "floating point" math?
 
Last edited:

Fox5

Diamond Member
Jan 31, 2005
5,957
7
81
http://www.semiaccurate.com/2010/04/30/amd-wants-slice-slate-market-fusion/

ARM sounds pretty strong in this article.



Thanks.

The GPU portion of AMD Fusion would use "floating point" math?

You could have a fixed point (integer) gpu, and those used to exist, but it's not well suited to the job of graphics rendering.

AMD doesn't make ARM chips though, so it's not really related to ARM. Though all ARM socs already include a gpu, and many current implementations are opencl compatible, they just lack the supporting software. I expect the new iphone will start to make use of opencl, but it will likely be years (if ever) before you see other phone platforms use it.

The BSN article is interesting, but imo it just shows that while ARM can match atom clock per clock in some things, there's quite a few applications that it trails far behind. Cortex A9 will likely be a closer match up. It's interesting to see AMD's ancient Barton core fairing so well against modern low power designs, especially given than an 800mhz barton was already a sub 20W cpu, and that was at 130nm. AMD's atom competitor could very well hit the sweet spot for power versus performance. It'll likely come in higher in power consumption, but it could potentially achieve a much better performance/power ratio.
 

EarthwormJim

Diamond Member
Oct 15, 2003
3,239
0
76
You could have a fixed point (integer) gpu, and those used to exist, but it's not well suited to the job of graphics rendering.

AMD doesn't make ARM chips though, so it's not really related to ARM. Though all ARM socs already include a gpu, and many current implementations are opencl compatible, they just lack the supporting software. I expect the new iphone will start to make use of opencl, but it will likely be years (if ever) before you see other phone platforms use it.

The BSN article is interesting, but imo it just shows that while ARM can match atom clock per clock in some things, there's quite a few applications that it trails far behind. Cortex A9 will likely be a closer match up. It's interesting to see AMD's ancient Barton core fairing so well against modern low power designs, especially given than an 800mhz barton was already a sub 20W cpu, and that was at 130nm. AMD's atom competitor could very well hit the sweet spot for power versus performance. It'll likely come in higher in power consumption, but it could potentially achieve a much better performance/power ratio.

The problem is there's a limit to the kind of power that a small device can draw. So even if a Barton has a better performance/power ratio, if no matter what, it can't get below say the 1watt range, it's no good for cell phones or small devices.
 

Nathelion

Senior member
Jan 30, 2006
697
1
0
Honestly, we already have a higher-than-atom-performance-but-still-low-power x86 CPU out there. It's called the VIA Nano. Now if nVidia could just give it some Ion love...
 

A_Dying_Wren

Member
Apr 30, 2010
98
0
0
Honestly, we already have a higher-than-atom-performance-but-still-low-power x86 CPU out there. It's called the VIA Nano. Now if nVidia could just give it some Ion love...

+1. also if intel would actually let such a thing happen. I don't doubt intel's willing to risk an anti-trust lawsuit to keep via out.
 

Vesku

Diamond Member
Aug 25, 2005
3,743
28
86
The thing is, as EarthWormJim stated, there is a bit of a gap between the mass of devices that need low single or sub watt power draw and the 15+W of netbooks and laptops. Atom and Nano are good x86 chips for their power envelopes but just won't last long with the type of batteries cellphones and MIDs have. The Atom Anandtech reviewed recently is almost there but still a bit of an ugly duckling with it's platform having half the battery life of current ARM solutions.
 

cheesehead

Lifer
Aug 11, 2000
10,079
0
0
You'll pardon my asking, but are there any inherent advantages as to the ARM architecture for low cost/low power designs? I've read that the use of a RISC architecture allows for a much simpler processor than a comparably powerful CISC design, which in turn makes them cheaper to build and more power efficient. How accurate is this statement? Is it less an issue of ARM's superiority as the designs implementing it being optimized for this sort of application for many, many years?

Yes, GPUs are fancy floating point processors. Although, the line between gpu and cpu continues to blur.

So, lemme see if I have this right:

1. The major disadvantages of ARM at the moment are poor floating-point performance and occasionally having to stop and wait due to the in-order design.
2. A multicore ARM would greatly reduce the disadvantages of an in-order design.
3. Adding a highly integrated GPU to the chipset gives the processor access to what is, essentially, a really fancy FPU.
4. Nvidia is building a dual-core ARM chipset with a GPU.

Anyone else think Ion 2 looks pretty awesome?

I should probably explain that I do tech support for little old ladies from time to time. If someone sold a $150 desktop computer based around a simplistic OS (Android) and solid-state hardware with a minimal chance of failure (Ion 2), I'd be very happy recommending it to them. Also, as a linux user, the proliferation of chipsets compatible with Ubuntu throughout consumer electronics makes me quite happy.
 
Last edited:

sonoran

Member
May 9, 2002
174
0
0
Brightsideofnews did a pretty neat comparison between an Atom system and an Cortex A8 system running Linux. Keep in mind that compilers make a big difference, so it's not a perfect comparison.

http://www.brightsideofnews.com/news/2010/4/7/the-coming-war-arm-versus-x86.aspx

Basically right now, integer performance is comparable at similar clock speeds between ARM (Cortex a8) and Atom (x86), while floating point performance is no where near x86 levels.

Both are in-order designs, so they're really strapped for memory bandwidth. Throw a 64-bit memory bus on a Cortex SoC, and I'm sure performance would go up a lot (not to mention cost and power though).


I would not bother comparing ARM to other x86 designs, Atom is the only comparison, and even then Atom uses a lot more power.

You might be surprised at how much Intel has managed to reduce ATOM power consumption on the upcoming chips: http://www.intel.com/pressroom/archive/releases/2010/20100504comp.htm

Collectively these new chips deliver significantly lower power including >50x reduction in idle power, >20x reduction in audio power, and 2-3x reductions across browsing and video scenarios – all at the platform level when compared to Intel's previous-generation product1. These power savings translate into >10 days of standby, up to 2 days of audio playback and 4-5 hours of browsing and video battery life3. When combined with 1.5-3x higher compute performance, 2-4x richer graphics, >4x higher JavaScript performance, and support for full HD 1080p high-profile video decoding and 720p HD video recording, these low-power innovations bring a rich, PC-like visual experience to powerful handheld computers4.
 

EarthwormJim

Diamond Member
Oct 15, 2003
3,239
0
76
You might be surprised at how much Intel has managed to reduce ATOM power consumption on the upcoming chips: http://www.intel.com/pressroom/archive/releases/2010/20100504comp.htm

What we don't have are power specs for when those 2-4x richer compute capabilities are in use. Basically the video savings and audio savings are when Atom is idling, and the other hardware of the SoC is in use.

2 days of audio is no joke, so Intel has chosen wisely in the hardware that compliments their relatively power hungry Atom.

Atom does have the benefit of being quite a bit faster than anything from ARM, so it can potentially finish its tasks faster and return to an idle state. Such computing work isn't always possible though.

Atom is the only x86 ARM competitor, and that says a lot, I said it above too. But it still is not quite there, almost, but not quite. In one more generation it will at the very least be on the same level that current ARM platforms are at.
 
Last edited:

Fox5

Diamond Member
Jan 31, 2005
5,957
7
81
You'll pardon my asking, but are there any inherent advantages as to the ARM architecture for low cost/low power designs? I've read that the use of a RISC architecture allows for a much simpler processor than a comparably powerful CISC design, which in turn makes them cheaper to build and more power efficient. How accurate is this statement? Is it less an issue of ARM's superiority as the designs implementing it being optimized for this sort of application for many, many years?



So, lemme see if I have this right:

1. The major disadvantages of ARM at the moment are poor floating-point performance and occasionally having to stop and wait due to the in-order design.
2. A multicore ARM would greatly reduce the disadvantages of an in-order design.
3. Adding a highly integrated GPU to the chipset gives the processor access to what is, essentially, a really fancy FPU.
4. Nvidia is building a dual-core ARM chipset with a GPU.

Anyone else think Ion 2 looks pretty awesome?

I should probably explain that I do tech support for little old ladies from time to time. If someone sold a $150 desktop computer based around a simplistic OS (Android) and solid-state hardware with a minimal chance of failure (Ion 2), I'd be very happy recommending it to them. Also, as a linux user, the proliferation of chipsets compatible with Ubuntu throughout consumer electronics makes me quite happy.

Most of the problems with x86/CISC have been designed around and are such minimal costs on modern fabrication processes. However, on a tiny cell phone soc that has to have everything integrated onto a single chip, the hardware solutions to those problems being us back to the early 90s RISC versus CISC situation, where you simple don't have the space/power budget for it. Intel is relying heavily on software optimizations to make atom competitive.
In a few generations, it's likely intel's advanced manufacturing processes could make x86 very competitive (or even superior), but now that Global Foundries is throwing its weight into the field, intel may not get enough of a process advantage to make that come true.

1. ARM's poor floating point performance isn't that poor. It's power optimized, and current ARM socs provide more floating point performance than the xbox 1 cpu. It just so happens that the athlon was a beast at floating point back in the day, and atom and nano are very floating point heavy. I think the use of dedicated dsps and coprocessors largely makes up for ARM's comparatively weak floating point. Video and 3d graphics are already offloaded to coprocessors/dsps, what does ARM need better floating point for when most platforms are closed and limit what can be used?

The in-order design is less of a problem than you think. A very clean RISC design can benefit from a lot of compiler optimizations, and will see less benefit from going out of order than an x86 core. x86 has a lot of implicitly parallel instructions, which makes it harder to design a good compiler for, but means a lot of performance can be extracted from out of order execution. In fact, I think it's almost crazy to design an in-order x86 core at this point in time, but intel made it work better than I would have thought, that it achieves comparable per clock performance to other in order designs is amazing. ARM has very few instructions that will benefit from OOE, but they are adding more with each new instruciton set revision.


2. A multicore arm would require multithreaded code, which is virtually non-existent on the platform. Also, I don't really think ARM cores have reached the point of complexity where dual cores are necessary. I think higher clock speeds and more complex cores are a very valid option for increasing performance still.

3. Once again, opencl is available for the powervr graphics processors, just no platform supports it yet.

4. Tegra 2 is nice, but the SOC market moves so fast that it may get left behind. Tegra 2 has very few design wins at the moment, and the lack of an integrated cellular modem makes it a tough sell for cells.

You can already get ARM based netbooks/smartbooks running ubuntu and android. No big brands yet though, and android is going to hit mainstream first.
 

JFAMD

Senior member
May 16, 2009
565
0
0
I should probably explain that I do tech support for little old ladies from time to time. If someone sold a $150 desktop computer based around a simplistic OS (Android) and solid-state hardware with a minimal chance of failure (Ion 2), I'd be very happy recommending it to them. Also, as a linux user, the proliferation of chipsets compatible with Ubuntu throughout consumer electronics makes me quite happy.

ARM is very interesting.

However the point about the $150 desktop does have a small hole. Software.

Netbooks have a much higher return rate than other computers. People love the size/weight and battery life, but they get frustrated by the performance.

A $150 computer sounds good on paper, but when the little old lady wants to use skype to do a video conference with her grand daughter or watch a video on youtube of her grandson's first little league home run, frustration ensues. You'd actually be amazed at what some of them do.

When we were traveling in japan last summer my wife borrowed her brother's netbook. Needless to say the performance was less than stellar and she ended up using my work notebook whenever she could.

Just bought her a new toshiba (4lbs) and I hope that will have the right balance of size and performance. My fujitsu S2210 is great for me, I would love something smaller with all of my travel, but I need some baseline performance and it is hard to find that.

I really recommend an older refurb system with a decent amount of RAM over a brand new $150 desktop. You'll end up with a better experience.
 

EarthwormJim

Diamond Member
Oct 15, 2003
3,239
0
76
Netbooks have a much higher return rate than other computers. People love the size/weight and battery life, but they get frustrated by the performance.

I've spoken to a lot of fellow college students who have tried out netbooks and returned them. None of them complained about the speed, their complaints centered on the the screen being too small and the keyboard too cramped (you wanted a small laptop dummies..).

People say they want speed, but they tend to not really know what a slow computer is. If you look at the average person's computer, it's so cluttered up with garbage and junk, that it's no faster than a netbook. So yes netbooks are slow, but they don't seem any worse than the average person's computer. That has been my experience dealing with and fixing hundreds of local people's computers.

I think the issue is that people don't really know what they want in a laptop. They want it to be portable and small, but when they actually use it they sort of expect a desktop sized keyboard and screen.
 

CTho9305

Elite Member
Jul 26, 2000
9,214
1
81
Honestly, we already have a higher-than-atom-performance-but-still-low-power x86 CPU out there. It's called the VIA Nano. Now if nVidia could just give it some Ion love...

Has anyone compared perf/power of VIA Nano to the Athlon Neo processors? Last I checked, the single/dual core low-power AMD processors were pretty efficient. Based on this article, the Nano is consuming on the order of 20 watts, but it looks like you can get a 1.7GHz AMD processor for 15W (or stepping up to 18W, 2 cores at 1.6 GHz). Nano is still single core only, right?

ARM is very interesting.

However the point about the $150 desktop does have a small hole. Software.

Netbooks have a much higher return rate than other computers. People love the size/weight and battery life, but they get frustrated by the performance.

A $150 computer sounds good on paper, but when the little old lady wants to use skype to do a video conference with her grand daughter or watch a video on youtube of her grandson's first little league home run, frustration ensues. You'd actually be amazed at what some of them do.

When we were traveling in japan last summer my wife borrowed her brother's netbook. Needless to say the performance was less than stellar and she ended up using my work notebook whenever she could.

Just bought her a new toshiba (4lbs) and I hope that will have the right balance of size and performance. My fujitsu S2210 is great for me, I would love something smaller with all of my travel, but I need some baseline performance and it is hard to find that.

I really recommend an older refurb system with a decent amount of RAM over a brand new $150 desktop. You'll end up with a better experience.

(This part of my reply reads like an iPad ad... but I just think it's just an interesting example of what can be done with the ARM platform when somebody bothers to put some effort into it. It's clearly not right for everyone. IMO, it costs waaaaay too much.)

It will be interesting to see how the ARM Cortex A9 stacks up. As for software, it looks like a lot of it won't be a problem for long. The iPad* plays fullscreen video smoothly on a 1024x768 screen (which is similar to, or only slightly less than, most netbooks) with an A8-based core. Even if Adobe is not going to do hardware-accelerated flash video decode on ARM, the HTML5 <video> tag means that browsers can provide native, accelerated video playback as soon as one of the manufacturers bothers to write the software. The libraries for accelerated video playback on ARM SOCs clearly already exist... I would think somebody like Nokia could put together a WebKit + hardware-accelerated h.264 video playback pretty quickly. I'm not sure any of the x86 netbook OEMs would be willing to put the effort into it though... the netbooks I've seen so far seem to be pretty squarely in the "save every penny we can" bucket.

Skype has an ARM iPhone client, and Skype is also included in ExpressGate instant-on x86 Linux, so the software apparently already exists (or is easily portable) for grandma's webcam needs. We'll see if next year's iPad adds a front-facing camera.

Speaking of software for grandma, it's not like Windows+x86 is actually all that good anyway. I'm sure we all do more family tech support than we'd like to... family members who could probably have gotten by with something like an iPad + keyboard dock for email (or, an iPad-like device with aged-eyes font sizes ). I wonder how many people would choose a "this device doesn't have an antivirus app, PDF reader, Java runtime, etc that constantly pester you for updates, and doesn't ask you questions you'll have to call your son to answer" over a "this device is a little faster".

*What's really interesting is that the iPad does this in an effectively sealed environment. I wonder what you could get out of its hardware if you allowed for the thickness of an x86 laptop's cooling solution. If you want to compare it to a refurb desktop, you have room for plenty of cooling, or even something crazy like a small peltier element... a 20W peltier would probably cool an ARM enough for massive frequency uplift and still put the system at sub-x86 power consumption levels.

edit: just remembered that I do a lot of my work on an ancient Fujitsu laptop with a ~1.5GHz Athlon XP processor in it. I bet A9 will outperform that laptop.
 
Last edited:

beginner99

Diamond Member
Jun 2, 2009
5,223
1,598
136
I've spoken to a lot of fellow college students who have tried out netbooks and returned them. None of them complained about the speed, their complaints centered on the the screen being too small and the keyboard too cramped (you wanted a small laptop dummies..).

People say they want speed, but they tend to not really know what a slow computer is. If you look at the average person's computer, it's so cluttered up with garbage and junk, that it's no faster than a netbook. So yes netbooks are slow, but they don't seem any worse than the average person's computer. That has been my experience dealing with and fixing hundreds of local people's computers.

I think the issue is that people don't really know what they want in a laptop. They want it to be portable and small, but when they actually use it they sort of expect a desktop sized keyboard and screen.

*sign*

I own a T91 (=z520 Atom, slowest atom). Only problem is hd-video playback. I have no issue with skype and video call.
For everyday use the limiting factor is IMHO mostly the harddisk. so most users would probably say an atom + ssd "feels" faster than an i7 + standard hdd.
 

joygame

Banned
May 20, 2010
6
0
0
there's a limit to the kind of power that a small device can draw. So even if a Barton has a better performance/power ratio, if no matter what, it can't get below say the 1watt range, it's no good for cell phones or small devices.
 
sale-70-410-exam    | Exam-200-125-pdf    | we-sale-70-410-exam    | hot-sale-70-410-exam    | Latest-exam-700-603-Dumps    | Dumps-98-363-exams-date    | Certs-200-125-date    | Dumps-300-075-exams-date    | hot-sale-book-C8010-726-book    | Hot-Sale-200-310-Exam    | Exam-Description-200-310-dumps?    | hot-sale-book-200-125-book    | Latest-Updated-300-209-Exam    | Dumps-210-260-exams-date    | Download-200-125-Exam-PDF    | Exam-Description-300-101-dumps    | Certs-300-101-date    | Hot-Sale-300-075-Exam    | Latest-exam-200-125-Dumps    | Exam-Description-200-125-dumps    | Latest-Updated-300-075-Exam    | hot-sale-book-210-260-book    | Dumps-200-901-exams-date    | Certs-200-901-date    | Latest-exam-1Z0-062-Dumps    | Hot-Sale-1Z0-062-Exam    | Certs-CSSLP-date    | 100%-Pass-70-383-Exams    | Latest-JN0-360-real-exam-questions    | 100%-Pass-4A0-100-Real-Exam-Questions    | Dumps-300-135-exams-date    | Passed-200-105-Tech-Exams    | Latest-Updated-200-310-Exam    | Download-300-070-Exam-PDF    | Hot-Sale-JN0-360-Exam    | 100%-Pass-JN0-360-Exams    | 100%-Pass-JN0-360-Real-Exam-Questions    | Dumps-JN0-360-exams-date    | Exam-Description-1Z0-876-dumps    | Latest-exam-1Z0-876-Dumps    | Dumps-HPE0-Y53-exams-date    | 2017-Latest-HPE0-Y53-Exam    | 100%-Pass-HPE0-Y53-Real-Exam-Questions    | Pass-4A0-100-Exam    | Latest-4A0-100-Questions    | Dumps-98-365-exams-date    | 2017-Latest-98-365-Exam    | 100%-Pass-VCS-254-Exams    | 2017-Latest-VCS-273-Exam    | Dumps-200-355-exams-date    | 2017-Latest-300-320-Exam    | Pass-300-101-Exam    | 100%-Pass-300-115-Exams    |
http://www.portvapes.co.uk/    | http://www.portvapes.co.uk/    |