cbn
Lifer
- Mar 27, 2009
- 12,968
- 221
- 106
The ARM market is flooded with competition.
AMD can find a niche.
Maybe they could field a ARMv8 Server chip that can also be used on Desktop? (Surely they can handle this)
The ARM market is flooded with competition.
Assume they're secretly working on a ARM APU/CPU, they sort of need windows to push the market in that direction and be able to enter a market where one competitor doesn't have a retarded unfair process advantage ?
So they sort of need ARM to be pushed into that area by microsoft, don't they?
AMD Changes Netbook Plans for 2012: Set to Introduce Brazos 2.0.
AMD Postpones Major Refresh of ULV Platform Due to Issues
[11/21/2011 09:34 PM]
by Anton Shilov
Advanced Micro Devices has decided to delay or even cancel its code-named Deccan ultra low-voltage (ULV) platform for netbooks, notebooks and nettops and introduce much less advanced Brazos 2.0 platform for inexpensive PCs in 2012. Although the new platform will boost performance slightly compared to existing Brazos 1.0, it will not be as revolutionary as the Deccan.
According to a source with knowledge of AMD's plans, the company ran into serious problems with its code-named Wichita accelerated processing unit for low-power low-cost personal computers. While it is not completely clear whether the production problems were conditioned by design issues (Wichita system-on-chip was supposed to have up to four x86 cores and integrated input/output controller) or by Globalfoundries' 28nm fabrication technology issues. Nonetheless, at present the Deccan is absent from AMD's 2012 roadmap and the Brazos 2.0 is supposed to substitute it.
The Brazos 2.0 platform is based on accelerated processing unit with up to two Bobcat-class x86 cores, next-generation Radeon HD 7000 graphics adapters and single-channel DDR3 memory controller. The new A68 Fusion controller hub (FCH) input/output controller (Hudson D3L) will bring support for USB 3.0 as well as Serial ATA-600 to AMD's ULV platform.
The Brazos 2.0 APU is supposed to be pin-to-pin compatible with FT1 infrastructure, but since A68 FCH utilizes 656-pin BGA package (instead of 605-pin that A45 uses), manufacturers will have to slightly redesign their existing products for Brazos 2.0. On the one hand, AMD will relatively easily upgrade available entry-level Fusion-based machines, but on the other hand Deccan platform would enable thinner and sleeker designs and would improve competitive positions of AMD.
At present AMD readies several flavours of Brazos 2.0 offerings for netbooks, notebooks and nettops. For example, models E1-1200 (two cores at 1.40GHz, 1MB cache, Radeon HD 7310 graphics engine with 80 cores at 500MHz, 18W, etc.) and E2-1800 (two cores at 1.70GHz, 1MB cache, Radeon HD 7340 graphics engine with 80 cores at 680/523MHz, 18W, etc.) will target nettops and low-power notebooks.
The source expects AMD Brazos 2.0 to hit production stage by mid-February, 2012. Since the APU will be made using "good-old" TSMC's 40nm process technology, the ramp up should be fairly quick and the new chips will be launched in late Q1 or early Q2, 2012.
The Deccan/Wichita and Deccan/Krishna platforms are not the first major new introductions that AMD decided to scrap for 2012. Previously, the company cancelled its Corona platform with next-gen Comodo processors and decided to introduce much less progressive Volan platform with Vishera CPUs.
AMD can find a niche.
Maybe they could field a ARMv8 Server chip that can also be used on Desktop? (Surely they can handle this)
Some "information" on Brazos 2.0.
Some nice upgrades with the HD7xxxx graphics (will this be GCN?), USB 3.0 and SATA 3 are mentioned.
I just wonder what the idle power will be like?
http://www.xbitlabs.com/news/mobile...razos_2_0_as_Krishna_Wichita_Get_Delayed.html
Although they really need to roll out a true next gen Brazos within a few months of Ivy Bridge netbookish CPUs being sold.
That simple huh? I guess thats it then. No more x86 for AMD, they will find a niche in ARM.
Judging from the sudden appearance of low end SB notebooks right around when Llano actually appeared in store, I would say Intel is willing to shave off a bit of profit in the low end to save marketshare. IMO, price is only getting AMD half way there. Llano is just a bit better for the price point right now for the low to mid-range media consuming user.
Guess we will see Q3-Q4 of 2012 whether AMD has a coherent strategy and if it will be better executed than the first iteration of Fusion and Bulldozer.
AMD could always compete on price
AMD has plenty of time for ARMv8.
With compression occurring in the x86 area, they could use Bulldozer APU for Windows x86 Notebook/Desktop.
ARM would replace Bobcat in Windows machines and would expand AMD's marketshare to Android devices (and any other emerging OS). Ideally, the ARMv8 SOC would also double as a server part.
And that's been such a winning strategy for AMD over the past five years.
I think the new CEO is focused on better margins (per the BOD).
Apple has never used an AMD cpu.
I disagree that AMD would be at the bottom of the food chain. Their brand recognition alone would place them in a desirable position.
AMD can find a niche.
Maybe they could field a ARMv8 Server chip that can also be used on Desktop? (Surely they can handle this)
I kinda laughed at this, but not in a good way.
I've owned AMD CPU's going all the way to my very first 286 system. K6-2, Athlons, XP's, etc. Always liked them for what they were worth, which at many times they were worth more than anything Intel had in the market.
But I have never, not at any time and not on even a single occasion, had any of my friends or family, ever, say anything postive about AMD on those few rare occasions where they could remember such a company even existed.
IMO, in North America at least (family and friends spread from coast to coast), AMD's "brand" is to their detriment. They have associated themselves with low-cost because of low-performance and lowered expectations on behalf of the consumer.
If anything, AMD needs to divorce itself from its brand and attempt to create a new brand. Too much time has passed since they had anything good to associate with their brand.
Now the same was said of Toyota in the 70's and look what they did to build their credibility, AMD is not hopeless, but suggesting they currently have a brand recognition that is leveragable in consumer markets is sadly not true in my experience.
They are on the verge of becoming the next Cyrix IMO. That's not the kind of brand recognition anyone wants for themselves, and that's not the kind of brand recognition that gets you traction in a market.
Right now there is no way I would attempt to convince any of my family to buy an AMD system because I know my family is already, for whatever their reasons, jaded against AMD because AMD has made itself synonomous with low-cost and that means low-quality in the minds of the consumers that my friends and family have turned out to be.
The shrink to 22nm and 3D transistors (FinFET) almost represents a two-node process technology jump, so we expect performance at various power levels to increase quite a bit.
Then if Intel can actually go establish a x86 Smartphone/Tablet market they can hitch a ride on their success.
Here is a chart (I found on another forum posted by username "bladerash") showing process technology leads over the years:
Intel
250nm_____january 1998_____Deschutes
180nm_____25 october 1999_____Coppermine
130nm_____july 2001_____Tualatin
90nm_____february 2004_____Prescott
65nm_____january 2006_____Cedar Mill
45nm_____january 2008_____Wolfdale
32nm_____7 january 2010_____Clarkdale
22nm_____january 2012??_____Ivy Bridge
AMD
250nm_____6 january 1998_____K6 ''Little Foot''
180nm_____23 june 1999_____Athlon ''original''
130nm_____10 june 2002_____Thoroughbred
90nm_____14 october 2004_____Winchester
65nm_____5 december 2006_____Brisbane
45nm_____8 january 2009_____Deneb
32nm_____30 june 2011_____Llano
If this chart is correct at one time AMD and Intel were on par with each other. (with AMD actually beating Intel to the 180mm node).
But look at things today? AMD is 18 month behind Intel on the 32nm node. Furthermore, Intel is about to release 22nm Finfet process technology, which will probably result in the gap widening even more.
http://www.anandtech.com/show/4318/intel-roadmap-ivy-bridge-panther-point-ssds/1
It doesn't look good for AMD in the x86 market.
If they actually work on fixing Bobcat's weaknesses, rather then just shrinking it (though, in the short term, that's about all they can do), they could probably have a winner, with it (unless Intel comes out some kickass Atoms). It's not like there aren't obvious ways to improve Bobcat:AMD going all ARM sounds like HP wanting to exit the PC business with #1 market and revenue share. Sure it might be a low margin business, but switching to something entirely different doesn't guarantee anything. And you lose whatever you have now, which for both companies is BIG.
If Bulldozer architecture goes nowhere they might just go all Bobcat. At least the die is small, so perhaps it can be made affordably enough.
Well, not a big surprise to see anandtech shilling and salivating over intel's claims of 22nm. Proof is in the puding though and so far it doesn't look very good despite intel's 'friendly' review sites. Looks like 22nm won't see the light of day until at least the end of April 2012, and would be no surpise if it was delayed further yet lol. Of course there'll be the damage control crowd that try to plant the seed of it being intentional because they have no competition, so why release it earlier. For those that believe that FUD, I LMFAO!
If they actually work on fixing Bobcat's weaknesses, rather then just shrinking it (though, in the short term, that's about all they can do), they could probably have a winner, with it (unless Intel comes out some kickass Atoms). It's not like there aren't obvious ways to improve Bobcat:
1. Improved caches. Any one of the following I would think could benefit Bobcat:
1A. Bigger fast L1 caches. This would probably require some fairly major changes to the rest of the CPU.
1B. Add separated big and fast L2 caches, so L2I and L2D can have different eviction policies, or at least tailored eviction algorithms (algo tweaks if no L3); leaving L1 mostly alone.
1C. Add a shared L2 cache, that's big and fast, exclusive with L3 (L1 LRU, L2 LFU (fast, takes victims from L1), L3 LFU (dense, takes victims from L2)?).
A dense/slow last level of cache makes sense for such a small cheap processor, but the thing performs like it is maimed, sometimes, and I'm going to put the blame on I$ and TLB misses. Shared caches with policies tuned to loopy code that worries itself mostly with data misses (SPEC, games, etc.) tend to be poor when I$ and ITLB misses start occurring often, as it is common that you may want LRU for instructions when LFU for data, and vice versa. As such, shared caches can end up evicting some of what you'll need. Adding a middle cache with high speed and different eviction counters and rules can mitigate that problem, while keeping that common case fast.
2. Make all int ops use the int units. Int divides and mods may not be common if you look at a spreadsheet of common instructions, but they do matter. I'm sure it was a size/cost trade-off.
3. Gradually make FPU and SIMD better. These were clearly sacrificed for the sake of size and/or cost for the first gen Atom-killer models. Good first-gen trade-off, but something to work on as time goes by.
2-issue ALU is probably good enough. Maybe 3 would perform better, but whether it would be worth it over other means to decrease stalls and latencies is something that really only AMD knows for certain. Given how it seems to just bottom out, sometimes, I doubt the ALU width is a big bottleneck.
And now for something completely different:
IMO, they also need to think about replacing the Geode, crappy MIPS boards, and all those Celeron-Ms out there in embedded appliances, too, where CPUs need quality software support (x86 has this due to history--people trusts major x86 compilers and libs), good non-loopy memory performance matters (there's good reason well-tested x86, MIPS, and SH are still popular, despite the mass market ARM take-over, and this is one--if mispredicted branches per second were as common a spec as MIPS or MFLOPS...), and their Radeons would have value (not for network appliances, but ATMs, slot machines, billboards, control interfaces, etc. could use next-gen APU GPUs well enough). The best part being that with a couple accelerators here and there (SSL offloading might be handy, FI), and maybe ECC support for some niches (which was likely in the pipe, already), they could make such chips just by fusing features of mass-market chips off, and binning them for low speed operation. The thing that gets me about such niches is that x86 is already fairly strong, even with old crappy CPUs, Intel doesn't care (they want consumers to want Intel; behind-the-curtain markets don't matter so much), and yet only AMD's Radeon division really wants it.
The pervasiveness of Pentium-class or worse x86 CPUs, and long-dead mobile CPUs, in boxes that don't tell you what's inside, might surprise you. What's also really surprising is how much better even they tend to be, than everyone's favorite bay these days, ARM. Code size, code testing, and optimizing CPUs for ugly memory operations, are edges that x86 still has, even with old processors in little systems. If AMD doesn't work on keeping those markets, though, ARM will replace them in the next few years. AMD's general want to put chips out there and let others decide their success, is going to allow ARM vendors, with good management and PR people, to eat their lunch, as ARM SoCs with decent performance begin coming out (2012-2014). These aren't flashy markets for consumers, but I have a hard time believing they couldn't offer decent revenue to AMD, if AMD were to tailor mass-market chips for them (to some degree, they do this now), and actively promote them (this, they hardly even try).
They need to take the good technology they have, improve it, and promote it where it can be superior (even with delays and not performing as hyped, Zacate and Llano are good examples of this!). Forget trying to beat Intel; they need to beat their historic selves. I think IDC's Toyota analogy is quite apt.