flat assembler
Message board for the users of flat assembler.
 Home   FAQ   Search   Register 
 Profile   Log in to check your private messages   Log in 
flat assembler > Non-x86 architectures > ARM 2.0GHz Multi-Core (5x faster than Intel Atom)

Goto page 1, 2  Next
Author
Thread Post new topic Reply to topic
r22



Joined: 27 Dec 2004
Posts: 805
ARM 2.0GHz Multi-Core (5x faster than Intel Atom)
www.arm.com/products/CPUs/ARMCortex-A9_MPCore.html

Supposedly 5x faster with the same power requirements as an Intel Atom.
Although I couldn't find specific comparisons/benchmarks.

Intel would be in trouble if this thing could only run Windows 7.
Post 16 Sep 2009, 17:01
View user's profile Send private message AIM Address Yahoo Messenger Reply with quote
Borsuc



Joined: 29 Dec 2005
Posts: 2468
Location: Bucharest, Romania
How do they know it's 5x faster if you can't even run the same apps on them (different architectures) Question
Post 16 Sep 2009, 17:15
View user's profile Send private message Reply with quote
mattst88



Joined: 12 May 2006
Posts: 260
Location: South Carolina
They took some code and compiled it before running it?
Post 20 Sep 2009, 18:23
View user's profile Send private message Visit poster's website Reply with quote
Borsuc



Joined: 29 Dec 2005
Posts: 2468
Location: Bucharest, Romania
The very same processor can give two "performance" results depending on the compiler used you know. So an Intel Atom can be faster than an Intel Atom sometimes.

wait did I just say that? Laughing

_________________
Previously known as The_Grey_Beast
Post 20 Sep 2009, 18:46
View user's profile Send private message Reply with quote
revolution
When all else fails, read the source


Joined: 24 Aug 2004
Posts: 15096
Location: The Unicomplex
My experience with the ARM cores is that the current TI OMAP versions with a v7 CPU and DDR memory interface are about 30% faster per clock than an Intel Core2. The MP core versions (not yet available commercially) are not likely to change that much except for the upgrade to a DDR2 memory interface.

For power consumption there is no competition. sub-1W compared to 120W for a Core2.

I don't know how the ARM compares the Atom. Since the Atom is such a ridiculous chip when compared to the ARM that I have not even bothered to buy any for testing. They are something like 6W power usage and physically enormous, they just don't fit the type of application that the ARM chips are aimed at. So any comparison is not really valid.
Post 22 Sep 2009, 09:55
View user's profile Send private message Visit poster's website Reply with quote
Borsuc



Joined: 29 Dec 2005
Posts: 2468
Location: Bucharest, Romania

revolution wrote:
For power consumption there is no competition. sub-1W compared to 120W for a Core2.

Say what? Not even my (best Core2 Quad) takes 120W, at full load it uses 90W.

I actually find what you said hard to believe. Sub 1W with Ghz of speed?

I wonder why Nvidia & Intel & AMD are so dipshits then. My guess is, you probably counted minimum power consumption for ARM.

_________________
Previously known as The_Grey_Beast
Post 22 Sep 2009, 15:33
View user's profile Send private message Reply with quote
revolution
When all else fails, read the source


Joined: 24 Aug 2004
Posts: 15096
Location: The Unicomplex

Borsuc wrote:
I actually find what you said hard to believe. Sub 1W with Ghz of speed?

I wonder why Nvidia & Intel & AMD are so dipshits then. My guess is, you probably counted minimum power consumption for ARM.

Our board with the TI OMAP 3503 (ARMv7) uses a maximum power of 900mW at full performance, running optimised code that uses all possible interfaces together (i.e. a stress test). This power measurement includes all the onboard devices (PMIC, SDRAM, MMC, UART, I2C, etc.) not just the CPU. I don't know the isolated CPU power usage, but I do know it is less than 900mW.

Minimum consumption is sub-10mW in standby mode, and ~1mW in sleep mode.
Post 22 Sep 2009, 16:02
View user's profile Send private message Visit poster's website Reply with quote
r22



Joined: 27 Dec 2004
Posts: 805
I would love to have a SmartPhone sized Linux box using a new ARM chip and IGP.
- DVI port
- Headphone jack
- 4 USB slots
- 128 Gb SSD

Since ARM implemented multi-core and SIMD there's really no saving grace for x86. Kind of sad, even though there's incremental enhancements the "'majority' of the" industry still pushes for the status quo.
Post 22 Sep 2009, 20:56
View user's profile Send private message AIM Address Yahoo Messenger Reply with quote
Borsuc



Joined: 29 Dec 2005
Posts: 2468
Location: Bucharest, Romania
128GB SSD???????? how expensive is it there? it's INSANELY expensive here.


r22 wrote:
Since ARM implemented multi-core and SIMD there's really no saving grace for x86.

Nah Intel will win this.

_________________
Previously known as The_Grey_Beast
Post 22 Sep 2009, 21:44
View user's profile Send private message Reply with quote
revolution
When all else fails, read the source


Joined: 24 Aug 2004
Posts: 15096
Location: The Unicomplex
Borsuc:

r22 wrote 128 Gb.

And the idea of Intel (or ARM) winning anything is not valid. They are not in competition with each other. They produce entirely different product lines. Indeed Intel used to make the things.

The Atom is not suitable for handheld devices. It is too power hungry, too physically large and requires complex (power hungry and large) support interface IC's. The ARM is mostly designed for handheld devices.
Post 22 Sep 2009, 23:45
View user's profile Send private message Visit poster's website Reply with quote
Borsuc



Joined: 29 Dec 2005
Posts: 2468
Location: Bucharest, Romania
ah, my bad. I'm used to bytes.

When I said Intel will win, I meant in the long run, not with their "Atom" series. Smile

_________________
Previously known as The_Grey_Beast
Post 23 Sep 2009, 00:30
View user's profile Send private message Reply with quote
revolution
When all else fails, read the source


Joined: 24 Aug 2004
Posts: 15096
Location: The Unicomplex

Borsuc wrote:
When I said Intel will win, I meant in the long run, not with their "Atom" series. Smile

Intel will win what?
Post 23 Sep 2009, 02:02
View user's profile Send private message Visit poster's website Reply with quote
Raedwulf



Joined: 13 Jul 2005
Posts: 375
Location: United Kingdom
A cookie.
If worse comes to worse, they will just invest into another architecture, albeit RISC or some new massively parallel CISC... or whatever.

Even still, legacy applications will always remain -- their dependency of x86 would mean x86 will have a market for a long time yet; and while the x86 isn't getting slower, unless ARM's chip becomes several magnitudes faster than x86, the balance in the market won't suddenly tip.

Having said that, I want one of those ARM cpus...
revolution buy me one! Now!

_________________
Raedwulf
Post 23 Sep 2009, 10:48
View user's profile Send private message MSN Messenger Reply with quote
revolution
When all else fails, read the source


Joined: 24 Aug 2004
Posts: 15096
Location: The Unicomplex

Raedwulf wrote:
Having said that, I want one of those ARM cpus...
revolution buy me one! Now!

No problem. Just go here and get some samples. But you will also need a good board assembly place to do the BGA mounting, and a good PCB fab to make the HDI boards.
Post 23 Sep 2009, 13:22
View user's profile Send private message Visit poster's website Reply with quote
Borsuc



Joined: 29 Dec 2005
Posts: 2468
Location: Bucharest, Romania

revolution wrote:
Intel will win what?

Market share, of course. Smile

Why do so many people think RISC is so good? Like I said before, this depends on the application, not on some imaginary or specific "performance test". Last time I checked, ARM didn't have a division instruction (and I could be wrong, revolution correct me on this, since ARM don't publish their stuff). Now imagine that in a division-intensive application...

RISC also wastes memory/cache, and with micro-opcodes the CISC parallelism "problem" is nullified. RISC sucks.

_________________
Previously known as The_Grey_Beast
Post 23 Sep 2009, 13:38
View user's profile Send private message Reply with quote
revolution
When all else fails, read the source


Joined: 24 Aug 2004
Posts: 15096
Location: The Unicomplex

Borsuc wrote:
Why do so many people think RISC is so good? Like I said before, this depends on the application, not on some imaginary or specific "performance test". Last time I checked, ARM didn't have a division instruction (and I could be wrong, revolution correct me on this, since ARM don't publish their stuff). Now imagine that in a division-intensive application...

The M profile of the ARMv7 instruction set has DIV. But even in x86 if you have a "division-intensive application" then it is still rather silly to actually use the DIV instruction. There are usually much better and faster ways to do it. DIV sucks no matter whether the hardware supports it or not.

Borsuc wrote:
RISC also wastes memory/cache, and with micro-opcodes the CISC parallelism "problem" is nullified. RISC sucks.

Oh really? I can see that you have not really programmed much in RISC. Even the Intel docs recommend not using the CISCy instructions too heavily because performance is enhanced by using the lower level RISCy instructions. Razz
Post 23 Sep 2009, 13:53
View user's profile Send private message Visit poster's website Reply with quote
Borsuc



Joined: 29 Dec 2005
Posts: 2468
Location: Bucharest, Romania
That's due to lazyness, but they improve. The bit scan instructions were horribly slow and now they improved them and some people even recommend using them in algos.

RISC sucks because true RISC would only have one instruction... Wink

(and yes you would use div if you needed precise results, like in some scientific app or something)

_________________
Previously known as The_Grey_Beast
Post 23 Sep 2009, 15:05
View user's profile Send private message Reply with quote
revolution
When all else fails, read the source


Joined: 24 Aug 2004
Posts: 15096
Location: The Unicomplex

Borsuc wrote:
RISC sucks because true RISC would only have one instruction... Wink

http://en.wikipedia.org/wiki/Reductio_ad_absurdum
Post 24 Sep 2009, 04:50
View user's profile Send private message Visit poster's website Reply with quote
Tomasz Grysztar
Assembly Artist


Joined: 16 Jun 2003
Posts: 6518
Location: Kraków, Poland

Borsuc wrote:
RISC sucks because true RISC would only have one instruction... Wink


That would be an URISC, not just RISC.
http://en.wikipedia.org/wiki/One_instruction_set_computer
Post 24 Sep 2009, 09:00
View user's profile Send private message Visit poster's website Reply with quote
revolution
When all else fails, read the source


Joined: 24 Aug 2004
Posts: 15096
Location: The Unicomplex
Cortex-A72 at 2.5GHz is said to be 3.5 times the performance of A15 at 1.6GHz. That would appear to work out to 2.24 times the performance per clock cycle, but the graph in the article does not show anything above 1.5 times relative performance increase per clock cycle. Where did it go wrong? Did some marketing-droid mess up the figures again?

http://www.anandtech.com/Show/Index/9184
Post 24 Apr 2015, 15:19
View user's profile Send private message Visit poster's website Reply with quote
Display posts from previous:
Post new topic Reply to topic

Jump to:  
Goto page 1, 2  Next

< Last Thread | Next Thread >

Forum Rules:
You cannot post new topics in this forum
You cannot reply to topics in this forum
You cannot edit your posts in this forum
You cannot delete your posts in this forum
You cannot vote in polls in this forum
You cannot attach files in this forum
You can download files in this forum


Powered by phpBB © 2001-2005 phpBB Group.

Main index   Download   Documentation   Examples   Message board
Copyright © 2004-2016, Tomasz Grysztar.