backtop


Print 40 comment(s) - last by bug77.. on Sep 26 at 5:34 PM


"Steam Roller" and "Jaguar" cores will drive AMD's 2013 next-generation successors to "Llano" and "Brazos", supposedly.  (Source: India Kev/Free Extras)
You've heard of "Krishna" and "Trinity", but how about "Hondo", "Kaverl", "Kabini", and "Samara"?

German hardware site 3DCenter has published an intriguing series of slides, date July 2011, which reveal CPU and graphics chipmaker Advanced Micro Devices, Inc.'s (AMD)  "APU" plans for 2011 through 2013.  

I. Background

In January of this year, the situation looked bleak for AMD.  Rival CPU-maker Intel Corp. (INTC) had just dropped Sandy Bridge, a brand new 32 nm architecture.  AMD's own Phenom II high-end CPU lineup was still sitting at 45 nm and despite aggressive pricing was fast becoming badly dated in peformance.

But while Intel had conquered the high-end, its low end solutions were less attractive. Sandy Bridge was relatively expensive for budget laptops and netbooks, and the Intel Atom platform was looking dated graphically, still relying on Intel GMA 31xx graphics.

AMD pounced deploying Brazos "advanced processing units" (APUs), chips that combined its Evergreen graphics cores with its new low-power Bobcat cores.  The chips stacked up similar to Atom in CPU performance, but they offered superior graphics and the ability to use power-saving DDR3 memory (Atom only supported DDR2).  Branded the C-Series and E-Series, the new chips were a sales hit, driving an AMD resurgence and profits.

AMD followed up with a beefier set of APUsLlano, branded the A-Series.  These chips swapped the Bobcat cores for a 32 nm die shrink of the K10/Stars architecture (dubbed Husky cores).  The number of shaders in the GPU was also bumped by a factor of 2 to 5 times.  Whereas Brazos primarily gunned for Atom's market share, the new chips went head-to-head with low-end Sandy Bridge chips.

Of course Sandy Bridge was an APU of sorts too, with a surprisingly peppy on-die GPU -- the HD 3000.  But members of the A-Series managed to scrape a narrow win over Sandy Bridge in graphics, which helped distract from Sandy Bridge's superior CPU performance.  

In testing by AnandTech the i5 chip handily beats AMD in CPU-centric tests, but the AMD chip pulls off an equally impressive win in gaming.  And thanks to attractive pricing, AMD clearly is doing a good job getting OEMs onboard and moving product into the channel -- Newegg lists 21 laptops powered by A-Series CPUs and 24 powered by E-Series or C-Series chips.

II. What's Next: The Known

Turning to the leaked roadmap, here it is:


Fig. 1: The leaked roadmap gives us a refresher on 2012 -- an a peak at AMD's 2013 Fusion plans.

And here's the code-name guide from the slides:

 

Fig. 2: AMD draws its codenames from wild cats, cities, rivers, and even Indian provinces.

It shows that AMD likely has already shipped engineering samples of its Llano refresh "Trinity".  Trinity prospect launch date is January 2012.  Its chips will be branded E2s, A4s, A6s, or A8s.  It will come with a graphics refresh ("London"/Northern Islands) and will swap the "Husky" cores for "Piledriver" cores -- "enhanced" versions of AMD's new 32 nm Bulldozer architecture.

Confusingly, Trinity's platform has yet two more code-name -- "Comal" for laptops and "Virgo" for desktops -- and the slides point to it being sold alongside Llano for Q1 2012, with Llano being phased out by Q2 2012.  This may create some market confusion as both the new and the old chips will share the same A-Series branding.

The Zacate chips will be replaced by Wichita, keeping the E-Series branding.  The Ontario chips will be replaced by Krishna, retaining the C-Series branding.  Both lineups will feature a die size will shrink to 28 nm.  New "London" graphics will be dropped in.  And the Bobcat cores will be "enhanced" variants (though this may refer simply to lower power consumption via the die shrink).  

The slides show Wichita and Krishna collectively referred to as Deccan.  They are slotted for a Q2 2012 launch, and will again sell along-side the previous models for one quarter, sharing the same brand name.  By Q3 2012 Brazos will be out, Deccan will be full in place.

 

Fig. 3: Q1 and Q2 2012 should be confusing as it appears that two generations of chips will be sold under the same brand name during these periods.

AMD's 6 W Desna chip, which is geared at tablets, but has seen minimal OEM interest, will be replaced with Hondo, a 4.5 watt design, disappointingly still built on the 40 nm node.

Interestingly, Deccan and Desna will be getting brand new chipsets ("Yuba" and "Hudson", respectively), but Comal is going to use virtually the same chipset as Llano (Llano used FS1, Comal uses FS1r2, a minor update).

Of course most of this information has been pretty widely publicized in bits and pieces, but it's useful to consolidate it and recap the latest info on the timing of the chip drops.

III. 2013 Brings New Cores to the Fusion Lineup

The slides indicate an intriguing battle shaping up in the APU sector.  Remember, Intel is planning to bring DirectX 11 goodness to its iGPUs with Ivy Bridge at the 22 nm node in 2012.  With 22 nm Atom processors planned for 2013, it's likely that they will get some DirectX 11 TLC as well, and it's already been confirmed that the 2013 Atom will be getting the power-sipping tri-gate transistors that year as well.

In short, AMD has better bring it, if it hopes to hang on to its short-lived low-to-mid-range lead.

AMD is pinning its hopes on two new core designs -- "Steamroller", the successor to "Bulldozer", and "Jaguar", the successor to "Bobcat".  As with before Steamroller will occupy the higher-power, higher-performance side of things (think Llano), Jaguar will occupy the low-power, lower-performance niche.

Indus (named after a river in India) will replace Comal (named for a river in Texas), which in turn replaced Llano (named for another river in Texas ... catch the trend?).  The Indus APU will be known as Kaveri (there appears to be a typo in AMD's slides, which refers to them as Trinity, the code-name of the last-gen Comal platform's APUs).

On the low-power end Kerala (named for a south-west Indian state) will replace Deccan (named for a southern Indian plateau), which in turned replaced Brazos (named for a river/county in Texas... whoops).  

On the tablet-geared ultra-low voltage end, will be Samara (named after a Russian river) will replace Hondo (likely named for New Mexico's Rio Hondo River), which replaced Desna (named for another Russian river).

Indus will receive a new chipset, "Bolton".  Kerala will get a new chipset named "Yangtze" (more rivers!).  And Samara will get a chipset named Salton (code-named in honor of the salty Californian lake).

  

Fig. 4: New chipsets will air in 2013 for all Fusion platforms

Little is known about what improvements these chipsets will bring.  And the slides revealed no details about what GPUs might be included on die as upgrades to the Northern Islands designs found in the previous generation's dies.

But whatever AMD is cooking up in these categories, it better bring its A-game or Intel's 22 nm Atoms could be eating their lunch.  2013 should be an interesting year indeed.


Comments     Threshold


This article is over a month old, voting and posting comments is disabled

It really doesn't matter
By Beenthere on 9/24/2011 10:07:13 PM , Rating: 5
AMD is doing just fine. It doesn't matter if they have the fastest chips in any or all market segments because they have the best performance/value and that is what mainstream consumers buy, not over-priced extreme series CPUs. From all indications Bulldozer will do just fine against Intel price-to-price and there is more coming right behind Bulldozer just like with Llano and Trinity.

The fact that Intel can no longer illegal suppress AMD sales means that AMD is gaining a huge increase in industry support which is good for everyone except Intel. Llano has proven the benefit of having a better APU and Bulldozer will deliver similar results for discrete graphics use.




RE: It really doesn't matter
By frozentundra123456 on 9/24/2011 10:19:51 PM , Rating: 1
You have to be kidding. AMD is barely surviving, and when they do have a competitive (at best) product, they dont have the foundary capacity to produce it in sufficient volume. The products coming out seem interesting and perhaps at least competitive, but if the bulldozer debacle is any indication, who knows if anything will come out on time. And if bulldozer is disappointing, they will be in a world of hurt, because then they will have two inferior CPU architectures and no time to develop an alternative.


RE: It really doesn't matter
By evo slevven on 9/25/2011 1:38:44 AM , Rating: 3
That was closer to the truth 2 years ago, I'd say even 16/18 months. The thing that changed was the move to fusion and then the new chipsets. The fusion chipsets were driving most of the $500 and under laptops which, given the price difference between sandy bridge and fusion/phenom classes, customers were buying more AMD's than expected via laptops. For over a year sub-$500 laptops for Intel were Pentiums while AMD competed with Athlon and Turions @ sub-$300 and Phenom-class/Fusions @ $400-500.

The bulldozer hasn't been a debacle to the extent that most enthusiast desktop makers don't sell in groves and even builders don't really outnumber general consumers who head to a local or online store just to buy a laptop.

Intel has the larger market share but I'm not arguing that, what I am saying is AMD's volume sold, share as a percentage of the market and profitability have all increased.

Currently Intel has faced a few problems and if you think Bulldozer was a debacle, Intel's atom was a bigger one that lead to firings at Intel and wasted R&D regarding how to approach mobile computing. They have no answer for nVidia and its a market that AMD is absent in and doesn't compete.

I did build my own desktop and I do use an Intel but I'm not a blind fanboy. Also when you think about profitability, you have to also wonder about Apple; AMD removed nVidia for being the chipset in the Macbook Pro's so your claims are really baseless.

And for the record AMD's strategy works if you target the right customer base and work up from there. The fact they are taking away customers in the business and consumer section kind of point to contradict that high performance means everything. I'm sure the person whose just browsing the web and using MS office doesn't care about not running core i5 or i7 but only care about the initial purchase price.


RE: It really doesn't matter
By someguy123 on 9/25/11, Rating: 0
RE: It really doesn't matter
By Solandri on 9/25/2011 3:13:21 AM , Rating: 2
Looking ahead to the future though, performance is going to become less and less important. By my estimates, a Brazos CPU/GPU has enough computing power for 75% of computer buyers. A Llano or i3 has enough for 90%. Intel may reign in the i5 and faster category, but only about 10% of people need that much power. And as processors improve and dies shrink, that percentage is going to get smaller.

AMD called it right IMHO. The competition for the majority of the market is going to be at the low end, and with a low-end CPU combined with decent graphics (for gaming and video). In the next few years, I think what happens with ARM processors is going to be more important for the PC market than Intel's top-performing CPU. The low-end low-power ARM CPUs are going to be entering this same low-end CPU/GPU segment of the market where the majority of computer buyers shop.

That's not to say Intel's performance lead is irrelevant. If they can leverage that performance lead into low-power and low-cost at the low end, they can still outmaneuver AMD and ARM at the low end. I went with an i5 for my new file server because at idle it used barely more power than an E-350 or Atom, while having serious horsepower at its beck and call when needed.


RE: It really doesn't matter
By bug77 on 9/25/2011 6:52:18 AM , Rating: 1
quote:
Looking ahead to the future though, performance is going to become less and less important. By my estimates, a Brazos CPU/GPU has enough computing power for 75% of computer buyers.


If you're looking to the future, you may want to consider that the software of tomorrow may (and probably will) require quite a bit more horse-power than the software of today.


RE: It really doesn't matter
By dark matter on 9/25/2011 6:55:38 PM , Rating: 4
???

Word is going to require 8 cores is it?

IE 12, 16 cores???

Ok....


RE: It really doesn't matter
By vignyan on 9/26/2011 12:34:58 PM , Rating: 2
yeah... Web browsers can exploit parallelism to a large extent... If internet evolves, you will need more cores and more powerful ones at that... Your statement is like making this statement in 1994...
Windows needs a GB of memory?


RE: It really doesn't matter
By bug77 on 9/26/2011 5:34:12 PM , Rating: 2
Imagine speech recognition as a generalized input method.


RE: It really doesn't matter
By Taft12 on 9/25/2011 9:27:05 PM , Rating: 2
quote:
If you're looking to the future, you may want to consider that the software of tomorrow may (and probably will) require quite a bit more horse-power than the software of today.


MORE horsepower? Windows 8 is going to run on ARM!


RE: It really doesn't matter
By Gondor on 9/25/2011 3:36:26 PM , Rating: 3
You went with quad core processor for your file server yet you estimate Brazos to be sufficiently powerful for 3/4 of computer users. This doesn't compute.


RE: It really doesn't matter
By Taft12 on 9/25/2011 9:29:02 PM , Rating: 3
A lot of posters here have more money than brains.

Actually, not necessarily a lot of money, just an addiction to buying new hardware whether they need it or not.


RE: It really doesn't matter
By someguy123 on 9/25/2011 9:57:36 PM , Rating: 2
You can argue all day and night about what this ethereal "average" consumer needs, but the fact of the matter is that less progress is never ideal. ARM is actually pretty powerful considering its power draw, and a lot of it is thanks to not having to deal with the bloated x86 instruction set. If we stifled progress based on what we needed ten years ago none of these fancy smartphones or tablets and the sleek OSs attached to them would even exist.


RE: It really doesn't matter
By someguy123 on 9/25/2011 5:18:32 PM , Rating: 1
AMD has the right idea in terms of practicality/performance, but the category of "average" computer is dominated by intel's IGP. The performance on that thing incredibly poor, yet it suits the needs of the majority of users because they simply don't use high demanding software or play complicated video games. Intel's current IGP is good enough to play the sims or WoW, and they seem to be taking steps to push it further according to Carmack.

Fact of the matter is that they're chasing a market that is incredibly low margin. The high performance market and industrial markets are the most lucrative. If bulldozer can't compete then they'll leave intel with a year or two window where there is no competition in industrial computing. They may not need a king, but they need something comparable at least. They're basically marginalizing themselves if they're dipping out of the performance race.


By frozentundra123456 on 9/25/2011 1:57:45 PM , Rating: 2
If you are referring to me in the "intel fanboy" catogory, I beg to disagree with you. A friend I work with was very anti AMD, and I defended AMD and their ATI acquisition for a long time. At time has gone on though, I have become much more disappionted with AMD. Fusion was hampered by an outdated CPU architecture and graphics still just marginal. Maybe Bulldozer will be a sucess, but no one really knows. And it has been plagued by delay after delay.

I am not a business person, and I could not find the record of earnings for the past few years for AMD. But I know they operated at aloss for a long time and now are only barely making a profit. And I dont think a company that is selling at 6.00 per share, down from a high of 40.00 per share is "doing quite well".


RE: It really doesn't matter
By unimatrix725 on 9/26/2011 11:59:16 AM , Rating: 2
"They have no answer for nVidia and its a market that AMD is absent in and doesn't compete."
Thats a bit odd and confusing. AMD and nvidia both have gpus and cpus.


RE: It really doesn't matter
By someguy123 on 9/24/2011 10:44:03 PM , Rating: 1
Not sure where people get this idea that "mainstream" consumers are buying AMD processors by the droves. What's their market share, 12%?

I think you're confusing mainstream consumers with enthusiast gamers. Those consumers may find better value in the lower priced amd processors due to developers leaning on GPU, but tasks that actually require decent CPU performance are heavily hampered by most AMD processors, outside of the octacore phenom.


RE: It really doesn't matter
By dark matter on 9/25/2011 6:58:19 PM , Rating: 2
They could be selling in droves, but still be at 12% market share. Do you know why?

You do realise that market share consists of the entire product range sold right up until today.

Do you know how many computers there are out there?

Do you need a maths lesson?


RE: It really doesn't matter
By dark matter on 9/25/2011 7:00:44 PM , Rating: 2
What average consumer is going to be performing heavy duty CPU cycles?

I imagine the "average" consumer isn't going to require a "semi-pro", i7...

Never in a million years.

Watching DVDs, Ebaying, Facebooking, and probably some light spreadsheet work.

Even a spot of light video editing at all will be fine.

If you need heavy duty CPU work, you're NOT a consumer, you're most likely semi-pro or pro.

Think REAL!


RE: It really doesn't matter
By someguy123 on 9/25/2011 7:20:46 PM , Rating: 2
Those types of consumers aren't going to see much benefit from mid spec CPUs or APUs, either. Whats your point? The mainstream market is dominated by intel IGP, which performs pretty poorly but gets the job done in those tasks you listed.

Marketshare consisting of an entire product range just makes 12% even more anemic. If a decent percentile aren't purchasing new CPUs, where are they going to get income from? It may seem like a large population to the average person, but looking at AMD's earnings they aren't making much money.


RE: It really doesn't matter
By Motoman on 9/24/2011 11:03:40 PM , Rating: 1
"Mainstream" consumers buy what the TV tells them to.


RE: It really doesn't matter
By Flunk on 9/24/2011 11:24:59 PM , Rating: 2
You forgot price, that's #1 for mainstream.


RE: It really doesn't matter
By msroadkill612 on 9/25/2011 11:12:46 AM , Rating: 2
which I guess is a TV


RE: It really doesn't matter
By bug77 on 9/25/2011 5:55:34 AM , Rating: 1
"...they have the best performance/value and that is what mainstream consumers buy, not over-priced extreme series CPUs."

Funny, I paid under $200 for my 3.3GHz quad-core beast that will handle anything I care to throw at it. Is that overpriced?


RE: It really doesn't matter
By bigboxes on 9/25/2011 9:10:32 AM , Rating: 4
Ah, there's my little troll. You do realize that most Intel users respect AMD. You need not make it your life's work in protecting their honor. Heh. You forgot to say, "AMD rocks!"


"It's okay. The scenarios aren't that clear. But it's good looking. [Steve Jobs] does good design, and [the iPad] is absolutely a good example of that." -- Bill Gates on the Apple iPad














botimage
Copyright 2014 DailyTech LLC. - RSS Feed | Advertise | About Us | Ethics | FAQ | Terms, Conditions & Privacy Information | Kristopher Kubicki