A lot of mystery surrounds roadmaps of technology companies due to their constant intention to find the rivals completely unaware of the plans. But if competitors usually are alert about the plans of each other thanks to espionage, the analysts and market observers have hard times understanding plans of companies like Advanced Micro Devices.
At the most recent conference call with financial analysts AMD’s president and chief operating officer Dirk Meyer said that the next-generation micro-architecture and processors on its base code-named Bulldozer were in development with first samples due in 2009. What Mr. Meyer did not say is when the final central processing units (CPUs) were scheduled to arrive, a piece of information that both analysts and investors are curious to know.
“The Bulldozer core is in development in 45nm [process] technology and we will be sampling that in 2009,” Mr. Meyer told the audience.
Any transition to a new-generation micro-architecture and processor design is closely tied to transition to a new process technology. Chipmakers want to improve fabrication technology as much as possible before starting to make brand-new CPUs, therefore, manufacturing technology conversion happens before transition to new chip designs.
The same applies to 45nm manufacturing process and Bulldozer processors. It was originally expected that AMD will be in position to start volume production using 45nm process technology by mid-2008, but the company is going to ship 45nm CPUs in volume only in Q4 2008. But will it affect the Bulldozer lineup. It may, or may not.
Based on track record of AMD’s transitions to new to new process technologies and micro-architectures in the last six years (see table below), the following can be noticed:
- It takes AMD from 3 to 6 quarters from commercial introduction of process technology to introduction of a radically new chip design based on this technology.
- It takes AMD from 4 to 5 quarters to shrink/improve a bit a CPU using a new process technology from the initial introduction of the chip.
- AMD’s K7 lived for 4 years, AMD’s K8 lived for 4 years, but was meant to live for 3.5 years.
Considering AMD’s recent history, if AMD succeeds in initiating volume shipments of 45nm processors code-named Shanghai, Deneb and others in Q4 2008, the Bulldozer has some chances to be unveiled in late 2009 (if it takes 3 to 4 quarters to make it after AMD initializes 45nm shipments on time), but that is not likely, as in that case K10 will live for only 1.5 years. In a little worse case scenario, the chip will be released in Q1 2010 or Q2 2010 (if it takes AMD 5 to 6 quarters to create Bulldozer after 45nm intro), inline with previous transition histories. However, it Bulldozer does not fit into typical pattern of AMD CPU introduction from the production technology point of view, the chip may be unleashed only in late 2010 or even 2011, which is inline with micro-architecture transition plans.

Bulldozer is the next-generation micro-architecture and processor design developed from the ground up by AMD. It is expected that the next-generation micro-processors will offer considerably higher performance than current-generation chips. AMD Bulldozer CPUs will feature SSE5 instruction set.
Comments currently: 20
Discussion started: 04/22/08 11:37:10 AM
Latest comment: 06/28/08 03:43:03 AM
Expand all threads |
Collapse all threads
[1-9]
1.
*gets ready for the impending bashing*
I'll get people started:
"AMD sucks" -and the rest of the derivatives of
"Intel rules" -and the rest of the derivatives of
That should cover much of it.
*goes inside the Anti-Shit Tank for protection against brainless morons"
Now that's out of the way, I can say that this looks decent; if it comes out correctly.
[Posted by: RtFusion | Date: 04/22/08 11:37:10 AM]
+ expand thread (1 answer)
- collapse thread
Silly Canadian.
[Posted by: RtFusion=AMD fanboy | Date: 04/28/08 09:01:15 AM]
2.
Quote from article:
" AMD’s K8 lived for 4 years, but was meant to live for 3.5 years. "
I hope AMD doesn't expect Phenom to live 3.5 to 4 years. They are only kidding themselves (and their investors).
Intel's tick-tock would have 2 new architectures released over a 4 year time period.
Intel will be long past Nehalem, and onto 32nm Gesher by 2011. Will Bulldozer be able to compete with that schedule?
It's no wonder AMD is struggling.
[Posted by: EndPCNoise | Date: 04/22/08 11:41:35 AM]
+ expand thread (1 answer)
- collapse thread
You should look at history: what currently has fashionable name "tick-tock", has been working for years at both Intel and, to some extent, AMD.
[Posted by: Anton | Date: 04/22/08 12:23:49 PM]
3.
It wont feature SSE5 because AMD cannot make SSE instructions
[Posted by: 1234 | Date: 04/22/08 01:40:44 PM]
+ expand thread (8 answers)
- collapse thread
AMD released SSE4, why Intel had to call there SSE4 4.1. AMD can use SSE name. It was months ago when AMD talked about SSE5, how it will combine reorder instructions so you can do more instrustions pre cycle. If I remeber correctly Intels SSE5 was to focus on speeding up reading/writing/converting text/ascii to binary format. Both SSE5 are going to be awesome!
[Posted by: Hokum22 | Date: 04/23/08 05:31:17 AM]
about sse(3) instructions...
One advantage of 3DNow! is that it is possible to add or multiply the two numbers that are stored in the same register. With SSE, each number can only be combined with a number in the same position in another register. This capability, known as horizontal in Intel terminology, was the major addition to the SSE3 instruction set.
http://en.wikipedia.org/wiki/3DNow%21
how greatly innovative is intel in matter of new instructions..
[Posted by: abw | Date: 04/24/08 05:26:39 PM]
Do you know who invent the AGP? PCI Express? High-bandwidth Digital Content Protection (HDCP) ?
I don't think AMD was capable of invent anything like that... lalala~ :)
[Posted by: V-tanium | Date: 04/24/08 06:32:32 PM]
agp and pci express are innovative?...lol..
agp is an extended pci bus with increased frequency..
and pci express is nothing else in fact that extending the agp like bus to the
other slots....you call innovation evolution willingly created to screw users and
make them change their hardware..
but surely that the hypertransport bus used by
amd is not as innovative..in fact so retarded that intel
use it as well with a different protocol under the name "quickpath"..to better solutions?..))
[Posted by: abw | Date: 04/25/08 07:58:00 AM]
AMD innovative?!?
Remember AMD started as an x86 cloner that copied everything from Intel!
[Posted by: AMD Phanboy | Date: 04/26/08 05:23:20 PM]
AMD created HyperTransport. It is innovative in a way that components designed for it, can get an increase performance through DMA and a straight shot to the processor. Also it is serial and network bus, so the components can be further from the processor or any where and co-processors can be used to provide extra performance for the main processor. It is also an open bus unlike Intel's CSI or QuickPath.I do not see QuickPath better than HyperTransport.
AMD's 64-bit instruction set for 80x86 was won over Intel's 64-bit instruction set for 80x86 because it provides a lot of features for now and the future. Intel reverse engineer AMD64 and named it EM64T.
AMD has hit two blows at Intel since K8 processor was introduced. Unfortunately, they struggled after they acquire ATI and the TLB bugs with their K10 core. Any company struggles when things go wrong. Intel struggled to get where they are now.
Both 80x86 processor makers are innovators in their own way. It depends how well they sale their produces matters who will win.
[Posted by: linuxnerd | Date: 04/26/08 06:41:03 PM]
AMD did not create hypertransport. look it up. They are just using the technology.
[Posted by: Know your facts. | Date: 04/28/08 08:42:27 AM]
AMD began developing the HyperTransport™ I/O link architecture in 1997.
HyperTransport technology, formerly codenamed Lightning Data Transport (LDT), was developed at AMD with the help of industry partners.
AMD has disclosed HyperTransport technology specifications under non-disclosure agreement (NDA) to over 170 companies interested in building products that incorporate this technology.
AMD is releasing the specifications to an industry-supported non-profit trade association in the fall of 2001.
The HyperTransport Consortium will manage and refine the specifications, and promote the adoption and deployment of HyperTransport technology.
San Jose, Calif., July 24, 2001 -- A coalition of high-tech industry leaders today announced the formation of the HyperTransport™ Technology Consortium, a nonprofit corporation that supports the future development and adoption of AMD's HyperTransport I/O Link specification.
[Posted by: JcK | Date: 04/29/08 02:17:52 PM]
4.
Thank you for the clarifications AMD!!!
It's now official!! AMD will continue to screw up for the rest of the Phenom life!!
COOL
[Posted by: AMD Suckers United | Date: 04/22/08 06:12:56 PM]
5.
AMD is only good at making roadmap presentations these days!!! They will continue to have crappy products for a couple more years!!!
[Posted by: anti-"AMD Fanboy" | Date: 04/22/08 06:20:34 PM]
6.
AMD again? Are they not bankroupt yet???
[Posted by: Tavix | Date: 04/22/08 09:33:34 PM]
+ expand thread (1 answer)
- collapse thread
Yeah then you start to pay Q9600 for $800..Very good deeds Doh!!!
By the way I'm AMD hatred but dont want Intel to stand alone.
[Posted by: Uncle Hussein | Date: 04/27/08 05:05:11 AM]
7.
Intel is fighting domestic-market saturation...in other words, if they don't release something next year that massively outruns what they released last year, their sales stagnate and all they get is replacement business. Of course, developers and game designers play their part...gotta have a new system to do/play X.
AMD, frankly, is a specialty producer that whores themselves in retail to keep volume to support their server products. Sometimes enthusiasts find value with them, sometimes not. But they are a long, long way from having any real control in the X86 market.
Look at it this way, Intel could without much effort, supply 100% of the PC market within 60-90 days if AMD closed shop. The reverse would simply not happen.
Intel of 2009 must have a substantial win against Intel 2007...their only real competition is their previous self. Prices will never again be in the $1000's for modest processors...unless a new 'killer app' comes along that everyone's gotta have (and current hardware chokes on).
[Posted by: mark1 | Date: 04/27/08 09:25:52 PM]
8.
This article overlooks some important aspects:
AMD does not have the same production setup they used to. The reason their transitions used to take so long was because they had to do the shrinks and the architecture revisions from the same fab. Mow they are opening a second fab right next to their old one (Fab 38 joins Fab 36) and they will now be able to also implement a tick-tock cycle.
Because of this, AMD will not let Phenom live another 4 years. That should be quite clear.
[Posted by: Manifesto | Date: 04/28/08 08:57:10 AM]
9.
Sounds good, I hope the schedule can be brought forward and that we will see a GPU integrated with the CPU, in the same way that a memory controller currently is today.
[Posted by: Garth | Date: 06/28/08 03:43:03 AM]
[1-9]