AMD Reveals Brazos 2.0 APUs and FCHby Jarred Walton on June 5, 2012 10:23 PM EST
- Posted in
- Trade Shows
- Computex 2012
AMD today announced their Brazos 2.0 APUs, also known as their 2012 AMD E-series APU. Brazos has actually been a major success for AMD, particularly in emerging markets, as it handily beats Intel’s Atom offerings and costs very little to manufacture. AMD has shipped over 30 million units, and there are over 160 different designs using Brazos. So what exactly is new in the world of Brazos for 2012? Not much, actually, other than names and model numbers. Here’s the short list of the new APUs:
|AMD E-Series APU for Essential Notebooks and Desktops|
|APU Model||GPU Model||TDP||CPU Cores||
|L2 Cache||Max DDR3|
|E2-1800||HD 7340||18W||2||1.7GHz||80||680MHz/ 523MHz||1MB||
|E-450||HD 6320||18W||2||1.65GHz||80||600MHz/ 508Mhz||1MB||
If that looks strikingly similar to the current E-series APUs, that’s because “Brazos 2.0” is using the same die. The E2-1800 is the replacement for the current E-450, with CPU clocks that are 50MHz higher, while the E1-1200 is also a dual-core die but with a lower 1400MHz clock—100MHz more than the previous E-300. The GPU gets a few more changes: first, AMD has rebranded the HD 6310/6320 as the HD 7310/7340, and second, the GPU clocks are higher. E-300 clocked the GPU at 488MHz, so the E1-1200 is only 12MHz (2.5%) faster; E-450 had the GPU clock at 508MHz with a max Turbo clock of 600MHz, so the 523/680MHz clocks of the E2-1800 are 3% and 13% higher, respectively. How often you’ll actually hit the higher GPU clocks isn’t exactly clear, but don’t count on being able to play the latest gaming blockbusters regardless.
If you’re a little depressed about the rebranding of the Brazos Zacate as Brazos 2.0, you’re not alone. This looks like a marketing driven move, particularly with the HD 7000 branding of the GPUs. There’s nothing even remotely similar to Southern Islands chips in Brazos, and the 80 core design has its roots in AMD’s 5000 series of GPUs. That still gives you DX11 and OpenCL 1.1 support, and given the CPU performance of Brazos—still substantially slower than any modern laptop CPU other than Intel’s Atom—there’s not really a need for more GPU performance. The 7000 branding essentially carries over from what we’ve seen on the other laptop GPUs, where everything below 7700M is simply a rebranded HD 6000M chip (which in some cases were rebranded HD 5000M chips).
While the new APUs don’t appear to have changed from previous Brazos chips, the platform has seen some updates. The real changes are in the Fusion Controller Hub (FCH), “Hudson-M3L” or A68M, which now supports two USB 3.0 ports as well as native support for SD card readers. AMD also lists support for their Steady Video Technology and Quick Stream Technology, though why those aren’t supported on older Brazos chips isn’t clear. Finally, AMD lists the FCH idle power as 750mW, down from 950mW on the previous A50M FCH.
All told, the changes and tweaks appear to have improved battery life slightly along with adding a few new features—or at least, the process technology is more mature and yields have improved to the point where the latest chips are better than the first models. AMD lists battery life improvements of 5% at idle for E2-1800/E1-1200 compared to E-450/E-300. There’s no indication of expected availability, but other than the changes to the FCH, the new APUs should be drop-in replacements for previous E-series APUs, so we expect to see updated designs sooner rather than later. Pricing as always will be up to the OEMs, and choices of memory, storage, and other components will largely determine how inexpensive Brazos 2012 products will be, but as long as OEMs can continue to push prices down in lieu of more substantial upgrades they likely won't catch too much flak from buyers.
Post Your CommentPlease log in or sign up to comment.
View All Comments
silverblue - Wednesday, June 6, 2012 - linkIt's a shame that AMD cancelled it and opted for slightly speed-bumped versions of their current architecture. It's not too dissimilar to Atom in this regard - release a new albeit pretty much unchanged generation, possibly with a couple of tweaks here and there to reduce power usage. One can only hope there's an E2-2000 or 2200 on the roadmap somewhere, however until next year (and a new process) there's not a real point in AMD throwing out a much faster Zacate.
I did think "Why not just make a single-core K10.5?" but realised that one single K10.5 core is probably larger than the entire Zacate die. Still, it'd crush Zacate, let alone Atom.
krumme - Wednesday, June 6, 2012 - linkYou might not like it, but this is good for the OEM and AMD shareholders as this product remove some of the shortcommings, while keeping cost down. Dropping 28nm conversion was an excellent move. 2013 we will se the new stuff on a mature process with lower cost than now. But i guess this cheap dude, will continue to sell for years, as all the cost is depreciated and 40nm will become even cheaper. x86 for all the world :) 30M is facts that speak for themselves, but it should have been the double. But there is still much left in this platform, its just turning a cash cow.
stmok - Wednesday, June 6, 2012 - link"If you’re a little depressed about the rebranding of the Brazos Zacate as Brazos 2.0, you’re not alone. This looks like a marketing driven move, particularly with the HD 7000 branding of the GPUs."
=> It's not marketing driven. Its strategic driven by the new CEO.
The reason Brazos 2.0 exists is because AMD is using it as a "delaying product" in order to buy their engineers time for the 2013 line: Kabini APU. (It also provides time for their manufacturing providers to transition to the 28nm process.)
* 2 to 4 cores.
* "Jaguar" cores. (Next major evolution to "Bobcat").
* iGPU moves to GCN architecture.
* 28nm manufacturing process.
Originally, they were to release “Krishna” (2 to 4 cores) and “Wichita” (up to 2 cores) APUs in 2012. These were nothing more than 28nm die-shrinks of the existing Brazos platform. They wouldn't have offered anything new from an architectural standpoint. (No "Jaguar" cores. No GCN for iGPU side). Since 28nm process wasn't happening for this low-cost market as scheduled, they cancelled them.
Overall, don't waste your money with 2012 lines from AMD. They are largely superficial improvements to buy time and breathing space for engineers to focus on 2013 lines.
2012 => 2013
Brazos 2.0 => Kabini (Jaguar-based APU)
Trinity => Kaveri (Streamroller-based APU)
The Vishera (Piledriver-based) or 2nd generation FX CPU; does not get replaced in 2013. It actually sticks around with Socket AM3+ in 2013...Because the 3rd generation FX (Steamroller) and its new socket format comes in 2014.
* Mainstream APUs (A-series) will be revision 1 of the new architecture.
* Performance CPUs (FX/Opteron-series) will be revision 2.
* Trinity APU => Piledriver-based (Revision 1)
* Vishera CPU => Piledriver-based (Revision 2)
* Kaveri APU => Steamroller-based (Revision 1)
* Vishera CPU => Piledriver-based (Revision 2)
* A-series APU => Excavator-based (Revision 1)
* FX-series CPU => Steamroller-based (Revision 2)
* A-series APU => Unknown next generation based architecture (Revision 1)
* FX-series CPU => Excavator-based (Revision 2)
This is AMD's "Tick-Tock" model. Instead of die-shrink and new architecture focus, they're doing it with different platforms. Their focus is for the mainstream APU (A-series) to play as the reference or "guinea pig" for the new architecture; while their performance CPUs (FX/Opteron-series) will receive the matured revisions.
JarredWalton - Wednesday, June 6, 2012 - link"It's not marketing driven. Its strategic driven by the new CEO."
What's the difference? In either case, the result is the same: AMD is rereleasing an existing product with a minor clock speed bump and a new name in order to increase interest from buyers. (Why wants to buy "last year's product" this year? Oh, but wait -- E2-1800 is new!)
Naming is also quite odd; E1-1200 seems like it should be E1-1400 and E2-1800 should be E2-1700. I know they're just model names and are arbitrary, but given the model numbers are close to the clocks why not just have them the same? Because that's also marketing: make the E2 sound a lot faster than an E1.
silverblue - Wednesday, June 6, 2012 - linkEvery 200-step is 100MHz in CPU speed. Despite that faster GPU, however, it's not really going to improve too much on the E-450, even if it IS 4x the numbering sequence. ;)
krumme - Wednesday, June 6, 2012 - linkThe product is not released because of marketing. They are released because it makes sense for the OEM business. So yes its strategic driven. The diference is the product is here. If AMD was porting to 28nm as fast as possible it would not have happened. The strategic context is interesting, and should get attention, the naming is the same old story.
Regarding the naming we better get used to the same renaming nonsense as everyone elses practices,- and these names a truly idiotic and manipulating - as probably the oem wants. But the buyers of these product hardly cares of the name. And cares if tis cgn 28 whatever as if that should have any importance at all for that segment.
Shadowmaster625 - Wednesday, June 6, 2012 - linkWhy didnt they jsut put the FCH in the cpu die. It is only a single channel memory bus. Surely they could create a fairly full featured SoC and still keep the pin count under 800.
eanazag - Wednesday, June 6, 2012 - linkAMD should have done more than just rebrand. I guess when they let go of all the marketing folks, they kept the cheaper, lower performing models.
I agree that the 15.6" laptop space is not where these should have shown up unless it was incredibly thin. I believe those first Windows 7 tablets should have came with these instead of Atoms for Dell and HP.
I am glad that AMD is pushing the process to give more performance, but in truth it should have been named an E - 500.
Brazos 2.0 should have been a process/die shrink if the architecture was untouched.
silverblue - Wednesday, June 6, 2012 - linkI'm surprised at AMD keeping at the same process, too. Ordinarily, you'd have a noticable architectural change, or a process change, sometimes even both... but in this case you can't really say they've made either. True, it'll use less power per clock, and they've been able to bump clocks up overall, but the arguments will generally centre around relative performance when compared to a much faster CPU, and the amount of time said CPU spends at idle after completing a certain task in a significantly faster amount of time. If system A is twice the price, but idles at a similar power and completes a task three times faster, it's going to use less power overall, but is it worth the extra cost?
Until we get reviews, there's no real indication of power usage, but it can only be a slight improvement, surely? Would 32nm have been better?
Jumangi - Wednesday, June 6, 2012 - linkWho was going to do the die shrink? TSMC can barely give what AMD and Nvidia wants for their GPU's which are much more profitable then low and stuff like Brazos.