AMD to launch quad-core K8L platform in Q3 2007

Sledgehammer70Sledgehammer70 California Icrontian
edited October 2006 in Science & Tech
AMD has indicated to motherboard makers that it will advance the launch of its quad-core K8L desktop platform to the third quarter of 2007, according to sources at the motherboard makers. Previously, AMD had reportedly scheduled the launch of the K8L platform for 2008.
AMD's quad-core desktop processor will be named Altair, and will be manufactured on a 65-nm node, with clock speeds between 2.7-2.9GHz, the sources indicated. Each core will have 512KB of L2 cache, and all cores will share a 2MB L3 cache, the sources added.

The quad-core series that AMD will launch in the third quarter next year include Athlon 64FX and Athlon 64X4 parts, the sources indicated.

However, AMD's branch office in Taiwan today declined to comment on its planning for the quad-core platform.
Source: TG Daily

Comments

  • lemonlimelemonlime Canada Member
    edited October 2006
    AMD will have to do a little better than late 2007 if they want to stay in the 'performance' arena. I seriously hope that there are other optimizations in store other than shared L3 cache. There are already Intel ES quad core chips in circulation based on the Conroe (Kentsfield) architecture.
  • Sledgehammer70Sledgehammer70 California Icrontian
    edited October 2006
    Well 4x4 will be the big thing for AMD now... which is slated for next month. From what I here it provides leaps in gaming performance.
  • LeonardoLeonardo Wake up and smell the glaciers Eagle River, Alaska Icrontian
    edited October 2006
    Previously, AMD had reportedly scheduled the launch of the K8L platform for 2008.
    ...until until actually released working quad core models for testers and developers and AMD got the bejeezus scared out of them. Right now, it's all hands on deck at AMD. Intel is no longer the bumbling, fumbling giant.
    AMD will have to do a little better than late 2007 if they want to stay in the 'performance' arena.
    I agree. It took years for AMD to lose the perception that AMD CPUs were unstable. It took the advent of Nvidia chipsets to shoot down that notion. AMD needs to act fast, or they will have a new perception - 'AMD, the choice for value and budget; Intel - when performance matters!' I certainly don't want to see that happen. But it will happen if AMD doesn't pull some rabbits out of the hat.
  • OrianeOriane Turn around.
    edited October 2006
    Leonardo wrote:
    AMD needs to act fast, or they will have a new perception - 'AMD, the choice for value and budget; Intel - when performance matters!' I certainly don't want to see that happen...
    ... again.

    My first PC was was a bargain AMD XP when Intel did P4C. But I'm not shedding too many tears for AMD- because it took forever for their ridiculous dual core prices to come down until Conroe showed up. But now I have a bargain X2! :D

    But you're right. If Conroe can be considered "The Empire Strikes Back", AMD better come out with "Return of the Jedi"- both sooner and better. Intel seems bent on beating AMD at whatever right now.

    Not matter- it's looking good for folders all over. :)
  • LeonardoLeonardo Wake up and smell the glaciers Eagle River, Alaska Icrontian
    edited October 2006
    My first PC was was a bargain AMD XP when Intel did P4C.
    The P4Cs were not more powerful than their AMD counterparts. They only had higher operating frequencies, which looked good on advertisements and boxes on store shelves. But as such, the perception continued for a while that AMD was inferior. AMD64 turned that perception on it's head when the gaming and do-it-yourself community started switching nearly en masse to AMD.

    (P4C versus AMD of that time frame.) The hyperthreading engineering did work well, but only under limited circumstances, such as intense multitasking. My former AMD 2800+ system at stock speed was nearly as fast as my highly overclocked P4C 2800, except in Folding@Home and in full-on multitasking.

    What is more critical now for AMD, is that if they don't keep up their progress and innovation, and dare I say accelerate it, that 'perception' of being second class will be a true perception. At low end and mid-level, AMD is still a strong competitor. They have lost the high end, period.
  • lemonlimelemonlime Canada Member
    edited October 2006
    4x4 certainly looks like a big boost on paper, but I seriously doubt it will do anything for 95% of gamers and enthusiasts. There are just not enough multi-threaded applications that can take advantage of two cores, let alone four. I could see it as an advantage for some workstation type of tools, but not for gaming. I do think it's cool to have SMP outside of the Opteron realm, but it will definitely not be a saving grace for AMD. I think its just a way for them to get four cores into a system as quickly as possible.

    AMD really needs to concentrate on improving the performance of each single core. A shared cache architecture and improved core efficiency (from things like better 'out of order' processing) are what is truly needed. The Quad core is great, and will definitely happen in due time but things like that shared L3 is of utmost importance if they want to get core performance anywhere near Conroe. That is one reason that Conroe smokes the K8 in number crunching benchmarks like SuperPI.. they are single threaded apps that can utilize all of the L2 available for both cores. Just imagine an FX57 with 4MB of cache--I'm guessing you'd see some rather impressive SuperPI scores (even without the Conroe-like efficiency improvements).

    AMD saw this coming for a very long time and should have been better prepared. It simply does not make sense to wait until the quad core K8L a year from now to revamp the architecture. Those changes are needed immediately, in the dual-core models if they want to remain competitive. The funny thing is that the K8 architecture is still far superior to Intel's 'Core' architecture in many respects. It loses in 'raw' performance due to much weaker caching and overall processing efficiency. The Hyper Transport is still far superior to Intel's FSB (even the latest 1066MHz FSB). Lets not forget the superior on-die memory controller, which has direct and unrestricted access to the main memory--memory requests do not even have to traverse the HT bus. So you have a very wide 2000MHz (vs 1066MHz) bus that does not even need to have rather saturating memory traffic traversing it. With that being said, just imagine how an AMD based processor with 4MB shared L2 and 'conroe like' efficiency improvements would perform. It would be the P4 vs K8 all over again. Even the current AM2 chips can really take advantage of high performance DDR2 in ways that Conroe can not. The Conroe seems impacted little by the timings or frequency of the DDR2 used. If Conroe had the same non-shared cache configuration as the K8, you'd see much higher overall memory latency.

    I just find it surprising that AMD was willing to let the performance crown go so easily and cut all of their prices by 50% overnight. They worked so hard to become a respected 'quality and performance' brand. It just seems odd that they would allow themselves to become the budget brand once again in the blink of an eye. I totally agree with your last point, Leo. They have such an awesome platform to build upon, and I do expect big improvements from AMD in the near future. It just seems logical to me.

    I'm finished with my 2c rant.. back to my Dill Pickle chips :)
  • LeonardoLeonardo Wake up and smell the glaciers Eagle River, Alaska Icrontian
    edited October 2006
    AMD really needs to concentrate on improving the performance of each single core.
    That won't happen in any significant way until 65nm production is standard...maybe not until 45nm. Intel is way of ahead of AMD in that respect. AMD ambushed Intel with the release of 64bit desktop processing and true dual core processors, but the 65 and 45nm architecture are perhaps even more critical for future performance, especially with respect to instructions per cycle and energy efficiency.
    I just find it surprising that AMD was willing to let the performance crown go so easily and cut all of their prices by 50% overnight.
    The didn't "let" the performance crown get taken away. It was wrested from AMD's hand despite their best efforts. What do you expect? Anyone who knew anything about CPUs and real performance was laughing at Intel for years. Did anyone really believe that Intel wouldn't wake up eventually and get serious about their engineering and long term prospects. It was only a matter of time before forward looking people and engineers at Intel prevailed over the marketing departments. As for the 50% price drops, AMD absolutely had to do that. The last thing they needed was to lose market share on top of losing prestige and their technological lead. Market share is critically important if they wish maintain sales and convince OEMs to bulk purchase. In other words, success begets success.
  • lemonlimelemonlime Canada Member
    edited October 2006
    Leonardo wrote:
    That won't happen in any significant way until 65nm production is standard...maybe not until 45nm. Intel is way of ahead of AMD in that respect. AMD ambushed Intel with the release of 64bit desktop processing and true dual core processors, but the 65 and 45nm architecture are perhaps even more critical for future performance, especially with respect to instructions per cycle and energy efficiency.

    Totally agree with you, Leo. 90nm has been stretched as far as it will go. IIRC, AMD already has 65nm ready to go. I just hope that they'll do more than just re-release the existing K8 AM2 architecture platform on 65nm. It may be quite a while until AMD gets any 45nm fabs producing. Although being behind in the process technology, they usually do seem to make the best of what they have to work with.
    Leonardo wrote:
    The didn't "let" the performance crown get taken away. It was wrested from AMD's hand despite their best efforts. What do you expect? Anyone who knew anything about CPUs and real performance was laughing at Intel for years. Did anyone really believe that Intel wouldn't wake up eventually and get serious about their engineering and long term prospects. It was only a matter of time before forward looking people and engineers at Intel prevailed over the marketing departments. As for the 50% price drops, AMD absolutely had to do that. The last thing they needed was to lose market share on top of losing prestige and their technological lead. Market share is critically important if they wish maintain sales and convince OEMs to bulk purchase. In other words, success begets success.

    There was no question that intel would strike back in full force and I have no doubt that AMD knew that better than anyone. Obviously, there are a lot of 'business' aspects driving the industry that I do not fully comprehend but the K8 has not changed a whole lot since 2003. They developed a groundbreaking platform, capitalized on it and made some refinements along the way. Things like dual-core are significant, but the underlying platform has remained the same. I'm just a little surprised that in almost four years, they did not have significant core enhancements prepared for deployment in light of the inevitable. I do agree with your point though, Leo. I often have to remind myself that AMD does not have the massive financial and physical resources at it's disposal that Intel does. Perhaps they really did do the best they could in the last four years. I suppose its difficult to know for sure, but it is not unrealistic to think that they could not 'one up' or match Intel's inevitable comeback. As for the price drops, I know it is definitely a necessity. I just find it interesting that AMD and Intel basically traded places in the price/performance arena within a matter of a couple weeks.
  • LeonardoLeonardo Wake up and smell the glaciers Eagle River, Alaska Icrontian
    edited October 2006
    Perhaps they really did do the best they could in the last four years.
    Oh, they did! If they hadn't they would have been buried by now. Intel holds an 8:1 capital advantage over AMD and a corresponding FAB capacity against them. Intel can shut down fabs at their leisure and start up completely now lines when they are good and ready. AMD essentially has to engineer, experiment, develop, and roll out now products without shutting down fabs. AMD has to do everything on the fly. In that respect, they are an amazingly agile, dynamic company.
  • TheSmJTheSmJ
    edited October 2006
    Keep in mind the 4x4 processors will all be Socket F, which is AMD's server socket - meaning you'll need an expensive new server motherboard to use it.

    AMD won't have anything really new for AM2/"AM2+" users until Barcelona's release in Q2 of '07.

    Source: http://www.firingsquad.com/hardware/cpu_gpu_roadmap_07/page2.asp
  • vaiovaio England
    edited October 2006
    As late as that, sigh.
    Guess I'll hafta give Kentsfield a try then
  • TheSmJTheSmJ
    edited October 2006
    There's only a 4-6 month difference between Intel's first Quad core offering and AMDs, furthermore AMD's will be true quad core, and will run much faster than Intel's product, which is only two dual core processors slapped onto the same dye (and will need to share one FSB).

    I think AMD's offering will be worth the wait. Besides, outside of Alan Wake, there won't be any applications that will benefit from quad cores, and who knows when that will actually be released.
  • vaiovaio England
    edited October 2006
    Who or what is Alan Wake?.......quad core folding!!!!;)
  • EnverexEnverex Worcester, UK Icrontian
    edited October 2006
    TheSmJ wrote:
    I think AMD's offering will be worth the wait. Besides, outside of Alan Wake, there won't be any applications that will benefit from quad cores, and who knows when that will actually be released.

    Not true. On a Linux system where you do any compiling you're likely to see an almost 400% performance increase while compiling (compared to a single core) and in general the system will be faster due to being able to dump any new processes on a free processor (I've noticed I tend to load the system a lot more under *nix). Not to mention anything multithreaded will benefit.
  • TheSmJTheSmJ
    edited October 2006
    Fact is not that many programs are multithreaded (yet). True, the more cores the better for heavy multitasking, but not many people play Half Life 2 while recording TV and encoding DiVX at the same time.

    Alan Wake is the new game coming from Remedy, the guys who made the Max Payne series. It will be multi threaded, one process handeling audio, another feeding the video card, one processing physics, one generating terrain, and the last doing the typical system overhead. It's the game Intel has been using to advertise their upcoming quad core proc at trade shows.

    These in game screenshots speak for themselves...


    Alan_Wake_2006_002.jpg



    Alan_Wake_2006_003.jpg

    More info here: http://www.alanwake.com/
  • LeonardoLeonardo Wake up and smell the glaciers Eagle River, Alaska Icrontian
    edited October 2006
    There's only a 4-6 month difference between Intel's first Quad core offering and AMDs, furthermore AMD's will be true quad core, and will run much faster than Intel's product, which is only two dual core processors slapped onto the same dye (and will need to share one FSB).
    Your post is part fact and part assumption. You have correctly described Intel's architecture and AMD's intended architecture. What is conjecture is that you've implied that Intel will just sit still after they launch the first quad series and wait for AMD to best them. Were this situation two years ago, I'd say that's a fair assumption. This time around, I would assume that Intel will be working on the next generation of quad cores before the first set even makes it to the distribution channels.
  • TheSmJTheSmJ
    edited October 2006
    As a matter of fact I believe Intel is going to be coming out with true quad core CPU -Yorkfield- right around Q3, as stated in the article I linked above.
  • LeonardoLeonardo Wake up and smell the glaciers Eagle River, Alaska Icrontian
    edited October 2006
    Nonetheless, it should be an interesting tech battle between Chipzilla and Chimpzilla. I know I'll enjoy watching it. It will be interesting to see what benefits quad brings to the desktop. Personally, I can't think of what I'd use it for other than running four instances of Folding@Home simultaneously. But then, I hadn't really taken dual core seriously until I bought my first one off eBay a year ago. After that, I was hooked, even though it was a lowly Intel D820.
  • ThraxThrax 🐌 Austin, TX Icrontian
    edited October 2006
    Intel is vastly ahead of AMD in architecture with the release of the Merom, Conroe and Allendale. With nothing on the table except Barcelona and Altair, both much later than Kentsfield and Intel's 45nm die shrink + revisions, it's going to be a dark time for AMD again.
  • OrianeOriane Turn around.
    edited October 2006
    Leonardo wrote:
    The P4Cs were not more powerful than their AMD counterparts...

    >sigh<

    Whether perception or reality - Intel has made AMD CPUs affordable for me ... again. My point is that AMD is no stranger to being viewed as a budget, second class CPU. And now that Intel has spent some time there as well, it works for us.

    While someone may be bummed about AMD or Intel getting slapped around for a time, I'm finding that this one-up competition has made for much better bang-for-the-buck CPUs and that performance dominance only created bloated prices and $frustration$ for people like me.

    I'm also not having issues with multi-cores and whether they will be used or not. I also don't think things are going to stop with 4 cores. I think this is going to be the world we're going to have- people may have more CPU cores than they have programs to run on them (>giggle< - but you get the idea).

    I think that's where business wants us. The future has DCPs (Distributed Computer Projects) like Folding@Home in it. Money making companies are looking for computer power- why not when you have all that cheap computer resource available by selling multi-core processors that people will only partially use? While non-profit and humanity-significant projects like FAH would still be Gratis to me, I think we should be able to sell CPU power by the TFLOP to those who can capitalize on it.

    Why not? It costs us time and money to do DCPs- why not charge for it if money-making companies are requesting it? We pay for the machine, the parts, the power, the maintenance- why shouldn't we be able to sell it? I think there is opportunity here.

    For now, though, I may never use the full capability of a 48-core processor but I'll have no problem about having them while FAH is around.
  • vaiovaio England
    edited October 2006
    Yup, the competition is good for the end user :thumbsup:
Sign In or Register to comment.