AMD and ARM Team Up 108
Vigile writes "Today AMD is making an announcement that is the first step in a drastic transition for the company by integrating an ARM Cortex A5 processor on the same die with upcoming Fusion APUs. Starting in late 2013, all AMD APUs (processors that are combinations of x86 cores and Radeon SIMD arrays) will also integrate an ARM Cortex A5 processor to handle security for online transactions, banking, identity protection and DRM integration. The A5 is the smallest Cortex processor available, and that would make sense to use it in a full APU so it will not take up more than 10-15 square mm of die space. This marks the first time AMD has licensed ARM technology and while many people were speculating a pure ARM+Radeon hybrid, this move today is being described as the 'first step' for AMD down a new road of dexterity as an IP-focused technology company with their GPU technology as 'the crown jewel.' So while today's announcement might focus on using ARM processors for security purposes, the future likely holds much more these two partners."
Fan-fucking-tastic. (Score:5, Insightful)
Thanks for nothing, guys.
Re:Fan-fucking-tastic. (Score:4, Insightful)
So AMD and ARM team up, and the product of their blissful union is an on-die TPM?
Thanks for nothing, guys.
Basically. The only ones who will ever make use of it are DRM assholes.
They make it sound like a feature by talking about "security for online transactions", "banking", and "identity protection", but no one will ever use it for that.
It's dead silicon until Windows is updated to recognizes it and allow DRM schemes to tap into it.
Comment removed (Score:4, Insightful)
Re: (Score:2)
From AMD's point of view, OpenCL was so yesterday, today they are pushing HSA [slashdot.org]...
Re:Fan-fucking-tastic. (Score:4, Interesting)
While I won't be happy to see it used for more DRM, it actually could prove quite beneficial to have increased general access to a chip which should be usable as a crypto coprocessor.
Crypto offloading cards are pretty expensive, and this would allow for modules to be rewritten on AMD+Linux boxes to dedicate SSL and like functions without increasing the general processor load.
Re: (Score:3)
You'll be able to program the A5 for any use via OpenCL, so yes, you could use it for dedicated crypto for your own programs.
Re: (Score:2)
While I won't be happy to see it used for more DRM, it actually could prove quite beneficial to have increased general access to a chip which should be usable as a crypto coprocessor.
Crypto offloading cards are pretty expensive, and this would allow for modules to be rewritten on AMD+Linux boxes to dedicate SSL and like functions without increasing the general processor load.
Or you could use the onboard GPU, which actually serves another integral purpose and therefore isn't a waste of die space.
Re: (Score:2)
On typical systems needing such a card, the acceleration functions of an onboard GPU have no integral purpose.
Re: (Score:2)
On typical systems needing such a card, the acceleration functions of an onboard GPU have no integral purpose.
You'd have to go completely headless for that to be true, and then you'd have to ignore the ability of the GPU to do what the ARM shit does, AND do it much faster.
Adding the ARM block is really fucking pointless because the chip without it already does everything the chip with it does, and better.
You could make an argument for a CPU with the ARM shit but without the GPU. But since AMD sells their "APU"s all over the place there is absolutely zero way such an implementation would win out on manufacturing co
Re: (Score:2)
No, you only need the relatively limited transistor count necessary for 2D output. All those extra transistors for the acceleration functions are a complete and total waste of space.
Anyway, back on-point, current ARM processors are designed architecturally with explicit security considerations in mind which are fundamentally impossible to implement on x86 cores, so no, you actually can't just "use the GPU."
Re: (Score:2)
No, you only need the relatively limited transistor count necessary for 2D output. All those extra transistors for the acceleration functions are a complete and total waste of space.
Anyway, back on-point, current ARM processors are designed architecturally with explicit security considerations in mind which are fundamentally impossible to implement on x86 cores, so no, you actually can't just "use the GPU."
I've explained this fully and clearly already. You don't get it. So now you get the following:
lolbro you dumb.
Re: (Score:2)
What makes you so sure this isnt the next iteration of accelerated encryption a la AES-NI?
Re:Fan-fucking-tastic. (Score:5, Insightful)
Re: (Score:2)
Re: (Score:2)
But if it's not locked down like a TPM, you've lost the trust and it becomes pointless. A TPM requires storage to store private keys in memory that can't be accessed. Even if someone melts the top of the package off and attaches wires directly to the flash/eeprom part of the die.
A SEM might do this.
Re: (Score:2)
If you're hoping for an ARM SoC with a radeon gpu, it's possible they have a don't-compete agreement with Qualcomm - who bought the Adreno off ATI.
Re:Fan-fucking-tastic. (Score:4, Informative)
It's not a mere TPM solution. It's a Cortex A5 core. AMD has been pushing their Fusion Systems Architecture for heterogeneous SoCs, so it's likely that the Cortex A5 will be programmable by developers through OpenCL. The TPM thing is just what the marketing people envision as one of its uses, but it could be used for anything... it's left up to developers imaginations to find something more worthwhile than TPM.
You could use it as a dedicated audio decoder and DSP, for example, as the OpenCL vector math functions will map directly to the NEON SIMD instructions.
I don't like the sound of this. (Score:5, Insightful)
Re:I don't like the sound of this. (Score:5, Interesting)
Gotta start somewhere.
So today they may be writing programs to use the extra ARM core for DRM, but I don't see where these are limited to just DRM.
Re:I don't like the sound of this. (Score:5, Informative)
That's pretty much why you would have a separate DRM module at all, when you already have a perfectly good x86 core to work with...
It is interesting that AMD appears to be throwing their hat in with this ARM stuff, rather than the 'Trusted Computing Group's TPM, available from a number of vendors on x86s already; but the expected use cases are every bit as malignant...
Re: (Score:1)
Actually, looks like it has a lot more purposes than just DRM, but they have to market to their target audiences, which in most cases, isn't you.
Hell, if these target audiences really wanted, they could do some extreme DRM with GPUs or various other pieces of coprocessing equipment readily available on a modern computer system. It's a small ARM coprocessor. I suspect what will come from this in the long run is a good asynchronous multiprocessing system. Shut off the x86 core when not very busy, save a lot o
Re: (Score:2)
Knowing AMD, the ARM Cortex A5 will be openly available for general use to developers through OpenCL. It's not something that will be restricted to AMD's proprietary applications and drivers, but will be open for any use. The problem here is that the people reporting on AMD's hardware announcements in article aren't developers themselves, and so stuff like this gets lost in translation.
They should be called AAA. (Score:5, Funny)
Why not rename the whole business to AAA, for ARM, AMD, ATI?
This would also make them the first chip maker in the phone book.
Re:They should be called AAA. (Score:5, Funny)
Because then people would be calling them for a tow truck
Re:They should be called AAA. (Score:5, Funny)
Not quite the bulldozer you were looking for..
Re: (Score:2)
Hey that was me up there with the bulldozer suggestion :) Thanks for letting me know it made others laugh!
Re:They should be called AAA. (Score:5, Funny)
Is tow truck after piledriver?
Re: (Score:2)
Is tow truck after piledriver?
No. The Tow Truck architecture is not slated until 2017. there are five versions between Pile Driver, and Tow Truck.
The AMD road map reads: PileDriver, Qbertitecture (Q2, 2013), RadStep (Q2, 2015), then Tow Truck (Q1, 2017).
Re:They should be called AAA. (Score:5, Funny)
what's a phone book?
Re:They should be called AAA. (Score:5, Insightful)
what's a phone book?
It's that lump of wood pulp that is left on your doorstep once a year. It is completely filled with advertisements and phone numbers, but no reviews at all.
Re: (Score:1, Funny)
What's the use of phone numbers without the notice "for a good time"?
Re: (Score:2)
Why so many A's Why not just simplify to AA that's much less confusing.
Re: (Score:2)
We try to avoid AAA while flying, the shrapnel reduces our chance of mission completion.
When you are in the flak, you may be nearing the target.
Re: (Score:1)
They should be called AA (Score:2)
By that standard they should have been calling themselves "AA".
But then of course people would be calling for meetings all the time...
It's AMAZING (Score:2)
So AMD is outsourcing DRM with an ARM core... somehow I don't think this is the utopian fairytale nirvana that the fanboys were trolling about when they started rumoring that AMD would go ARM.
OMG TPM (Score:5, Informative)
Before you start flaming about DRM and TPM taking over your computer and all, please remember that all TPM chips currently available allow you to install your own keys. This hardware root of trust allows you to verify that your Linux installation has not been tampered with. It also is a good place to store hard disk encryption keys, because the TPM chip makes it extremely difficult to do brute force attacks on your password. I simply can not imagine why anybody would intentionally buy a modern computer without these wonderful capabilities.
Re:OMG TPM (Score:4, Insightful)
Before you start flaming about DRM and TPM taking over your computer and all, please remember that all TPM chips currently available allow you to install your own keys.
And once you're tied to using them, they'll stop allowing you to install your own keys.
Vendor lockin FTW.
Re: (Score:1)
That isn't really how a TPM works.
First, most machines don't possess one, so DRM can't really assume it is there.
Second, it is shipped off and disabled. A user has to explicitly flip the TPM on in BIOS setup, then allow the OS to take ownership.
Of course, this technology is a double-edged sword. Look at the PS3 is an example.
However, on laptops, it provides an additional security boost, especially with a full disk encryption utility like BitLocker. The enhanced security it provides (allowing the OS to bo
Re: (Score:2)
That isn't really how a TPM works.
First, most machines don't possess one, so DRM can't really assume it is there.
Second, it is shipped off and disabled. A user has to explicitly flip the TPM on in BIOS setup, then allow the OS to take ownership.
Of course, this technology is a double-edged sword. Look at the PS3 is an example.
However, on laptops, it provides an additional security boost, especially with a full disk encryption utility like BitLocker. The enhanced security it provides (allowing the OS to boot without a password needed, as well as protecting against "evil maid" attacks) is a help.
Like an Au Pair in the middle attack.
Re:OMG TPM (Score:5, Insightful)
I simply can not imagine why anybody would intentionally buy a modern computer without these wonderful capabilities.
Ever since TPM was created, we're always just a few bits and bytes away from having it leveraged against us, by them.
And by "us" I mean "the computer users."
By "them" I mean "the hardware manufacturers and software/media companies."
Example: The newest motherboards don't need the ability to disable trusted boot. Heck, it'd have been easier to not include it!
We're more or less at the mercy of a small number of companies and their design decisions.
Worse, we have no real power other than social pressure.
Re:OMG TPM (Score:4, Insightful)
You have the power to not purchase locked down motherboards.
Of course once Windows requires that all systems are locked down before it runs, there'll be a remarkable lack of affordable motherboards that will run any other operating system. Nor will you be able to buy a Windows machine and install another operating system instead.
Anyone who thinks this isn't intended to create vendor lockin is incredibly naive.
OMG antitrust (Score:2)
Re: (Score:2)
Interchangeable motherboards for laptops? (Score:2)
You have the power to not purchase locked down motherboards.
One ordinarily does not buy a "motherboard" of a laptop computer separately from the rest of the package.
Re: (Score:1)
It also is a good place to store hard disk encryption keys, because the TPM chip makes it extremely difficult to do brute force attacks on your password. I simply can not imagine why anybody would intentionally buy a modern computer without these wonderful capabilities.
Actually, a TPM makes it easier to steal your keys if the thief has physical access to your computer, because the thief knows exactly where the keys are stored. See the famous paper at
http://citp.princeton.edu/research/memory [princeton.edu]
No Internet without TPM (Score:4, Insightful)
please remember that all TPM chips currently available allow you to install your own keys.
Which won't help if both the cable company and the DSL company start using Trusted Network Connect [slashdot.org] to control home customers' access to their networks. In such a case, you wouldn't be able to get Internet service with your own key on the TPM.
Re: (Score:2)
If you follow the chain of links, youll see predictions from 2005 that "within 5-10 years 50% of computers will support TNC", which just isnt true. It also assumes that TPM will ship on by default, which isnt the case with any computer I've seen.
Maybe that was once a threat, but right now its just fearmongering.
Re: (Score:2)
It also assumes that TPM will ship on by default, which isnt the case with any computer I've seen.
Heh. In my machine the TPM module wasn't shipped "on" as in "on the motherboard". There's just an empty spot on the board with some solder points.
Will they be calling it... (Score:1)
Order of magnitude problem (Score:2)
Re: (Score:2)
Actually A5 has a very decent performance. Check Qualcomm MSM7227A (A5-based 1GHz SoC) benchmarks for example. Many new phones are based on this chip.
Re: (Score:2)
Qualcomm does not use ARM's reference Cortex designs.
Re: (Score:2)
Re: (Score:3)
I think it's a bad idea for many reasons. The main one being that the swiss army knife approach means mediocre subcomponents.
While a SoC solution may be good for embedded devices, it just doesn't seem a good idea for a desktop device; it's like replacing a discrete component stereo system with a receiver - it's bigger, lower quality, and you can't upgrade the one component that's not up to par - you have to toss the whole thing. Yay for consumerism.
Yes, having components on the chip can have advantages, w
Re: (Score:2)
Re: (Score:2)
Re: (Score:2)
Think of what you could do with an ARM subcore to handle transactions for math equations, or even graphics. Don't get hung up just because you saw 'DRM' in the article. I think this is a neat idea, especially if you have full access to the ARM core from your x86 OS
What could you do with the DRM logic block that you couldn't do with the many cpu cores or the fucking gpu that's also on-die?
Re: (Score:2)
Re: (Score:2)
Think of what you could do with an ARM subcore to handle transactions for math equations
Oh! Oh! And I know just what to call it! The AMD8087!
Re: (Score:2)
what (Score:3)
Seriously, you go through all that trouble to cram an ARM core in there, and you use it for exactly what it's *worst* at?
Crypto is best done by specialized, single-purpose hardware. Intel has special units on their chips just for certain common crypto algorithms. Doing it in software, on a core that's underpowered compared to the x86 cores next to it is retarded.
The strengths of ARM is low power. Doing what the Wii did would be a wonderful idea - they had a small ARM core on the northbridge, used to do online updates and such while in sleep mode. Imagine if your computer could keep your emails and RSS feeds synched and run updates while in sleep mode. Yes, it would need some OS-level support, and could probably be done better with an ultra-weak x86 core just for better compatibility (just take an old K6 core, shrink it down to 32nm and trim the cache - you don't need power, you just need small). Maybe it wouldn't be a killer feature, it would probably go unused by most users, but it's something that would actually *work*.
Re: (Score:2)
Near as I could figure from the limited information, they're doing it this way so that the crypto-subsystem can be software-compatible with tablets and smartphones.
Re:what (Score:4, Informative)
Now that I've read more, I think I was only partially correct. It looks more like the main reason to go with ARM is that there was a fully-baked hardware/software solution already in place around ARM's implementation of TPM. So AMD could sort of glom on to the whole thing.
Playstation 4? (Score:5, Interesting)
Re: (Score:2)
Sounds like a console chip to me.
We have a winner!!!!
A5 is the smallest Cortex? (Score:4, Informative)
The A5 is the smallest Cortex processor available
Really? I figured that the Cortex-M0 would be smaller. The M0 doesn't even have a cache. Indeed, ARM's Cortex-M0 product page [arm.com] agrees, saying:
The ARM Cortex(tm)-M0 processor is the smallest ARM processor available.
so it's not clear why the article is calling the A5 the smallest?
Re:A5 is the smallest Cortex? (Score:5, Informative)
It is the smallest core with TrustZone support. The article is wrong.
Re: (Score:2)
The A5 is commonly used in embedded applications too, as it's quite compact. It makes perfect sense here to leverage all the security applications developed on top of ARM TrustZone system. It's small, puny (compared to the x86 cores
Re: (Score:2)
You kids with your fancy GHz CPU's always forgetting about us small embedded microcontroller guys. Get off my lawn!
Gramps you don't even have your own account anymore... that isn't your lawn, it is a public park. And if I see you napping on the bench, I'm calling Homeland Security.
Or.... (Score:2)
Re: (Score:2)
I'm not sure where they got the 10-15sq mm from. The Cortex A5 takes up less than 1sq mm.
So does AMD multicore mean... (Score:2)
So will an AMD 3 core or a 6 core APU imply that 1 or 2 cores is Fusion, another 1 or 2 cores is Radeon and another 1 or 2 cores is Cortex?
But I'm not getting the logic. The only reason ARM consumes less power is its implementation more than its instruction set. If a multicore processor includes ARM, that's not gonna make it consume less power, since the other non-ARM cores will still be there for that. And is there a problem w/ lack of ARM compatibility the same way there is w/ x86 compatibility?
How
Re: (Score:2)
That is the stupidest idea i've heard today (Score:2)
For security transactions? How often are those occurring and what cant you handle them in software or on a separate chip?. Why put it on-die? I bought a zacate and now its in the closet because its slow.
We need to figure out what to replace AMD with so that we have a decent competitor for Intel.
Re: (Score:2)
My first thought was that this would be really great for Point Of Sale (POS) systems. Even if somebody manages to boot a flash device and get a virus in, you could still have your credit card validation secured.
There is already a linux kernel module for ARM's TrustZone, at least some of it. Seems like the people who are afraid are the ones who are already vendor-locked. lolol
If you can't beat them... conjoin them! (Score:2)
Suboptimal soln at best from the blurb.
H.
The smallest ARM Cortex? (Score:2)
What this is really for. (Mod this up) (Score:2, Interesting)
ARM's processors aren't powerful enough to run Win8 Boxes.
And lets be honest here, when Android, IPhone, and all the other big OS players in the mobile market were developing their OS's they were developing them to small Motorola, Arm, Apple and Intel embedded non-x86 processors. So that means VERY limited app functionality from their standpoint and to get cross platform compatability you need java and browser support because NOBODY is going to spend a fortune developing an app just for one vendor's OS.
Wha
Re: (Score:2)
Funny you mention android, if they expose the cortex A chip to user programs then one could develop on the amd64 and run the emulator on the ARM core.
Re: (Score:2)
Cell signaling is not relevant here, for a cellular solution you need RF and DSP functions that are just not there. And don't expect AMD to go into the cellular market, it's a specialist area they don't know about and already crowded. By the way, there's no specific acceleration on the ARM cores for cellular. The part that run on the ARM is the protocol stack and it's regular software. Neon is not even used, it's good for media but for cellular DSP processing you need mu
*facepalms* (Score:2)
AMD, I am going to help you here. ARM is nice, and yes, Windows RT might possibly prove every analyst wrong and come out a winner, with a tremendous demand for ARM devices.
But, as a long time advocate of your processors, and (now) your video cards, I'd like to take a moment, and ask if you wouldn't mind listening to what I, personally, would like to see from you in...16 months. Here's what I'd like: you to fix your Bulldozer design / work with Microsoft & the open source community to patch their code /