Cortex A15 - 5x faster.


3 years away. Let's hope it actually does ship then and not a year later. But Cortex-A15 looks like it could be quite a step up from A9, in terms of IPC even before you get to the clock and core increases.


As I posted on the other forum: Word is that there's a move to a triple-issue front end, 128-bit floating point NEON, and improvements in the memory hierarchy (more than one load or store at a time would be quite nice). Stated goal is a typical 50% better IPC over Cortex-A9.


http://www.electronicsweekly.com/Articles/2010/09/09/49414/Update-ARM-Cortex-A15.htm


I remain unconvinced Intel was ever going to be a player in the modern phone market in the first place. They should be more concerned about AMD's Bobcat kicking them off netbooks.. something I very much hope happens.


By the way, the Register article claims:


"ARM's next processor, the Cortex A9, won't be shipping until the end of this year, with devices expected during 2011."


Which is just nonsense, seeing as how you can already buy Tegra 2 Android netbooks.

lulzfish said:
"as many as 16 cores"


So it might be good for software video decoding and multitasking, but otherwise difficult to make use of.
Click to expand...

16 cores would definitely be targeted at the server market, which they're trying to make a dent in, by touting lots of low power cores. Hence why they've added something to address > 4GB. I think they're going to have a difficult time with this.
 
Last edited by a moderator:
Craigix said:
... I'm not sure right now what we would do with all that power, but something awesome most probably.
Click to expand...

Pandora 2 perhaps?


To 'consume the same energy as today's ARM chips, but will sport as many as 16 cores running at 2.5GHz' sounds implausibly awesome to me.
 
ok so at what point would the possibilities end with an cortex a15? how much better would this be in comparison to what our a8 can do?
 
Last edited by a moderator:
Mr_Loon said:
Pandora 2 perhaps?


To 'consume the same energy as today's ARM chips, but will sport as many as 16 cores running at 2.5GHz' sounds implausibly awesome to me.
Click to expand...

I don't think a 16 core version running at 2.5GHz consumes the same energy one of today's ARM chips (say, 1GHz dual core Cortex-A9) does. Now, based on what we've heard, we should expect Cortex-A15 to come out on Global Foundries' 20nm node, which I'm sure will be quite a leap, but I don't think THAT much of one.

DaMummy said:
how much better would this be in comparison to what our a8 can do?
Click to expand...

A single core Cortex-A15 at 1.5GHz (realistic clock speed for handhelds) would probably typically be a good 4 times faster than Pandora's Cortex-A8 at 600MHz, before comparing things like NEON-heavy applications. Four of them (realistic count for handhelds) could potentially be up to 4 times that (normally nowhere close though)
 
Last edited by a moderator:
Exophase said:
16 cores would definitely be targeted at the server market, which they're trying to make a dent in, by touting lots of low power cores. Hence why they've added something to address > 4GB. I think they're going to have a difficult time with this.
Click to expand...
My personal server's already pretty low-power and if they can't beat Intel's computational or I/O power or whatever in the real server market, who will they sell it to?
 
Why do you think they can't? AMD's Magny Cours is a good 115W with 12 cores, so nearly 10W per core at 2.2GHz. Intel's 8-core Nehalems are bigger consumers. I expect Cortex-A15 at 2.5GHz to be closer to 1W per core (or maybe much less). Of course this is all pretty crude speculation and I could end up being way off, or Intel/AMD might change the game completely when they make it to that node themselves, but ARM is obviously deliberately targeting the server market now so they must think they have a fighting chance.


Understand, in server applications that scale very well with core count thermal envelope/power consumption basically determines overall performance, and therefore, so long as you can get together enough CPU cores in the same die, you'll win out with slower cores if they consume enough less power. When we're looking at farms with thousands of nodes there's really no "good enough" for power consumption, if everything else allows it it can completely make or break you.


Of course, for desktops and workstations where single-core performance is still important at all Intel's x86 offerings will probably continue to dominate.
 
the problem with ARM that lot of people don't seems to be able to understand is that when ARM anounce a new tech, it takes some times, until they effectively deliver it. after that you still need that chip founder eg. TI build a new chip, publish it as engineering sample, device maker start building new device using theses ES, and you finaly have to wait for a full consumer device... I wouldn't be surprised if we'll only see the first real consumer device using cortex A15 in 2014...


it sound like a great core, but I'dd love to see what will Intel have to propose at the same time...
 
"UK chip designer ARM [Note: check out this short history of ARM chips in mobile devices contributed by an anonymous reader] today released the first details of its latest project, codenamed 'Eagle.' It has branded the new design Cortex-A15, which ARM reckons demonstrates the jump in performance from its predecessors, the A8 and A9. ARM's new chip design can scale to 16 cores, clock up to 2.5GHz, and, the company claims, deliver a 5x performance increase over the A8: 'It's like taking a desktop and putting it in your pocket,' said [VP of processor marketing — Eric Schorn], and it was clear that he considers this new design to be a pretty major shot across the bows of Intel and AMD. In case we were in any doubt, he turned the knife further: 'The exciting place for software developer graduates to go and hunt for work is no longer the desktop.'"


I like the part where it says:


"It's like taking a desktop and putting it in your pocket". They must have designed it with the future pandora in mind!?!?


Maybe by then we can have a working Philips CD-i emulator.


Here's the story on slashdot: http://hardware.slashdot.org/story/10/09/10/0312224/ARM-Unveils-Next-Gen-Processor-Claims-5x-Speedup
 
Last edited by a moderator:
elwing said:
the problem with ARM that lot of people don't seems to be able to understand is that when ARM anounce a new tech, it takes some times, until they effectively deliver it. after that you still need that chip founder eg. TI build a new chip, publish it as engineering sample, device maker start building new device using theses ES, and you finaly have to wait for a full consumer device... I wouldn't be surprised if we'll only see the first real consumer device using cortex A15 in 2014...


it sound like a great core, but I'dd love to see what will Intel have to propose at the same time...
Click to expand...

I dont think so that TI takes so much time to develop a ready-to-use SoC with the new A15 core. I can imagine that's not an easy project but 1.) the core part, the CPU (and more) is already developed by ARM. 2.) For example, Nvidia will introduce Terga 3 mid of next year - and each year there will be a new Tegra chip (which doubles the performance, so Nvidia's marketing screens). Means, if TI waits too long for introducing new products, others like Nvidia, Samsung, Qualcomm, Sony, etc. could steal there show. The market situation has changed a bit I would say. On one side there are more competition on ARM Side (Nvidia and Co.) and on the other side the x86 big ones (AMD and Intel) and small ones (VIA) will/could enter the battle of Tab-pcs, sooner or later (f.e. AMD's Bobcat).


If you ask me, we can skip the Cortex A9 based CPUs for a Pandora 2 because its too late to set a product on it - imagine that developing a Pandora 2 could take at least 1 year. And when a Pandora 2 comes out with an A9 based CPU there will be other products at least of same speed. So the speed argument would fail.


Besides you can ruin the sales for the Pandora 1, too, if you introduce a new one too fast. Better take your time - 2-3 years - and make a fine product based on next-gen like A15 core.


An other point is that it is quite possible that the Pandora could get competition. Tab-pc with "open" operating systems (like Android) will float the market (if you read all the announces of the companies) and there can be a company which brings a Tab-pc with gaming controls later. Then we could have a problem; no problem for the Pandora 1 because there are NO similiar products on market, YET and perhaps for 1 year. But for a overhasty Pandora 2 LATER (in 1 or 1.5 year) with an A9 based CPU it could be a problem, because the Tab-pc could even be more powerful. And there wouldnt be major reasons for buying a Pandora 2, anymore.


So we need something unique for the Pandora 2. Just an example, if the Pandora 2 appears in 2-3 years, it would be always fine to be near cutting edge technology. It is always a good argument for a product. So a A15 or similiar would be nice. But we can think of other technology, too. Perhaps 3D displays - in 2-3 years there could be an opportunity of getting these without the need of being a big company..


But I am not a business person nor I am a future teller. This is just my point of view.


PS: I am thinking of something unique for the Pandora 2. Ive mentioned the 3D displays - ok, they are not that unique, the 3DS will include one, but they are not standard.


What could a Pandora 2 get on board - I dont mean "standard wishes" of "everyone has that technology" like 3G (4G) network capabilities, OLED or GSM?


F.e. if we analyze the Pandora 1 unique things are the gaming controls combined with a full keyboard and Linux. And because of these things I will wait and buy a Pandora 1.


So I am restructuring my question:


Imagine that you've paid in advance for a Pandora 2. What technology could let you wait 1-2 years for it ? ;)
 
Last edited by a moderator:
Exophase said:
16 cores would definitely be targeted at the server market, which they're trying to make a dent in, by touting lots of low power cores. Hence why they've added something to address > 4GB. I think they're going to have a difficult time with this.
Click to expand...
They really need a 64-bit instruction set if they want to get into this market. I don't think PAE hacks are going to cut it.
 
IMO, they'd be pretty well off having a hack that has > 32-bits in some/all registers that only the AGU can see. Then you only need 40 or 48-bit registers, you only need to enhance the AGU to handle wider additions, and you only need an instruction to load the top bits (something as simple as: movupper r0, r1 would do it) It'd require that pointer arithmetic to be done in 64-bit, but this is a pretty small price. I think that going full 64-bit just to increase address space is pounding a nail with a sledgehammer, even if it's the move the industry has always made. A really small and low power platform like ARM could benefit from having a more streamlined design.
 
Craigix said:
http://www.theregister.co.uk/2010/09/09/new_arm/


Intel can't be too happy about all this.


We on the other hand can look forward to an amazing future. I'm not sure right now what we would do with all that power, but something awesome most probably.
Click to expand...

In My Humble Opinion Only:


Keep building what you're building through the end of 2011.


During 2011 secretly design a new Pandora II to have:


Retina display (or something else with a resolution you can brag about as being an improvement anyway)


Dual core ARM 1.2Ghz


1GB RAM


8GB NAND


GPS


Accelerometer


802.11b/g/n


Keep the rest (case, nubs, KB, etc... as is. Release April 1, 2012. Have a few thousand units in stock for immediate delivery then take pre-orders past that.


For 2013 do it again for a 2014 release with a quad core.


For 2015 do it again for a 2016 release with a eight core (or quantum computer or whatever is there by then.)


In My Humble Opinion Only. Your Mileage May Vary.
 
lolz quantum computer (I don't know if you were being serious, but there will definitely not be a fundamentally different kind of computer available in 2016, much less for handhelds... and if there were you wouldn't want to use it, the software wouldn't even be close to useful for it)


I have a feeling that come 2016 an 8-core handheld, while technically maybe an option, will not be a very useful thing to have. Most of the sorts of things people are really interested in running on handhelds (and even desktops) don't scale that well. Some games today do, but most don't. We might see more of a shift there, but Amdahl's Law does tend to kick in at some point and it's a bitch.


Now, for all of this to be possible, there have to be SoCs available at these levels and someone has to be willing to sell to them (unless OP grows to be the size of a big enough player like Sony, that can work with companies to get a custom core.. of course they won't)... and I don't think that market has been moving quite fast enough for every 2 years.
 
And just like on the PSP boards, the device barely hits the market and the next few years are spent idealizing the systems successor.


Honestly, I'm not sure there will ever be any big-name companies I would see as a competitor to the Pandora. My personal reasons I think would pale in comparison to the more experienced members around here, but it seems unlikely someone like Sony who's spent the last several years closing down the PSP will soon change tactics by opening up the system and Apple doesn't seem fond of physical controls anymore.


For now, I think the main concerns for OP should be finishing solving the minor issues with the device (screen bezel issue...). That aside, I'm sure with time talks will resume on how to improve on the hardware, I'm still fond of an internal camera but honestly not so concerned. Software additions to my existing Pandora will likely take care of the bulk of my interests.
 
to touch on a couple of statements made here so far:


some things that could be done with that much power would be XBox 360/PS3 emulation to some degree as well as (what is currently considered) high-end media production to name a few things... I would have added Wii emulation, but I don't know the specs for that system at the moment and I'm under the impression that the current device overclocked may do the job (possibly) plus there is already Wii emulation software out there.


the server market would be a nice target for this processor, especially since I have read up on competitions and contests going on where groups are being requested to create supercomputers that can run on a max of 26amps... we are making our way into a conservative age where the "new cool toy" is rediculously powerful, yet runs on almost no energy at all. I'm actually glad to see things following that path... will be interesting to see what we can come up with.


When all said and done... GPH may be the only attempted "competitor" for Pandora in the next few years to come unless another group of DIYers decides to create some direct competition for the Pandora, though aside from power, space and speed, I can't see anyone really improving on this device without creating something that's just odd or unnecessary. The clamshell design is perfect for portability and protection, it has touchscreen (the only thing to be improved upon in that area may be size at some point) keyboard and controls need no real improvement (but with a larger screen, more buttons could be added for a "better keyboard experience" or whatever... and most anything else would end up best as an accessory to the device in itself. But this is strictly my opinion, so don't quote me on any of this unless you take that disclaimer. :D


As far as Intel dominating the desktop... you never know which way that could go... with nettops being moreso the thing than standard desktops (and nettop modification slowly opening up to rival desktop modification) it could either go the way of ARM-based desktops having a big chunk of the game or ARM-based motherboards stepping up to the plate for standard, micro and server-sized desktops... also, the thing with arm being as physically small and low powered, I could picture at least one mobo mfg creating a multi socketed mobo for ARM processors similar to the AMD server board that handled 2 separate socket 1207's.... can you imagine an standard sized desktop mobo with 4 or even 8 A15 sockets onboard? With a total power consumption (with a standard set of PCIx16 -or x32 by this time- devices) running at around 150W??? big dream, I know... but not impossible.


As far as the PAE statement goes, I'm sure that the PAE solution is a "right now" thing while true 64-bit computing is being worked on between now and release time... with an expectation of coming to the table no more than 6 months after official release (or 6 months after the release of the first consumer product running an A15)... any truly innovative business mind would be sitting on a though process of such or similar type. As we all know, being linux people in general, the hacks always come before the full-blown solutions... and in a lot of cases, the hacks are the path followed to achieve the full-blown solution (granted PAE is probably not a path to follow in order to get true 64bit, but I could be wrong...)
 
I've been saying it for a few years now, if Intel don't move to ARM they are going to become irrelevant, they have been struggling like crazy to squeeze the dead horse that is x86 for too long.


ARM is going to brutalize X86 within the next 8 years and during that time we will also see full Windows ARM released and pushed as the next big thing.


I always see X86 like an old out dated uneconomical petrol engine still being shoehorned in to nice looking cars with more and more elaborate modifications to get a few more BHP from it each time. You know it has to stop somewhere.


Also I expect Apple to move their laptops over to ARM very soon, I get the impression they never liked using x86 anyway.
 
You must log in or register to reply here.

Similar threads

H
Please do not use arm cortex-a15 for pandora2!!!!!!!!
2 3
Replies
55
Views
7K
homejogger
H
E
Cortex-A15 Details
Replies
1
Views
4K
DaveC
D
P
Psp 2 2010?
2
Replies
24
Views
6K
Goity
Goity
Phawx
I Can't Believe No One Mentioned It Yet.
Replies
4
Views
793
Enverex
Enverex
sporkbox
The Legend of Zelda: Skyward Sword
2
Replies
36
Views
6K
Jumpman
Jumpman
Back
Top