AMD Barcelona on schedule

Sledgehammer70Sledgehammer70 California Icrontian
edited January 2007 in Science & Tech
AMD has come out of the gates swinging and is bragging about its four-core Barcelona beast, AMD is claiming the new native four core monster will offer a 40% increase over Intel's current Clovertown.

According to Pat Patla, director of AMD's server worksation division, Barcelona is a new processor design, 90 per cent of which is brand new. He reckons it will make AMD the first firm to deliver a native quad-core X86 chip and its introduction will be as "substantial" as that of the original Opteron in 2003.
«1

Comments

  • LeonardoLeonardo Wake up and smell the glaciers Eagle River, Alaska Icrontian
    edited January 2007
    Two questions:

    1. Will it put AMD back in the performance game, beyond just good bang for the buck.

    2. Will Barcelona become affordable to the consumer as quickly as Intel's Conroe did.

    The answer to those two questions will determine whether I think this development is merely good or GREAT. I hope this does turn out to be as significant as the 2003 Opteron release. Intel needs competition and AMD needs to gain fiscal strength.
  • miragemirage
    edited January 2007
    AMD is targeting quad-core Barcelona mainly for server/workstation market to compete with Intel quad-core Xeons. Although Intel is rapidly gaining back the market share from AMD in both markets, loosing market share in the server/workstation space hurts the bottom line more. Because the server market is more profitable than consumer market. Sun has announced their collaboration with Intel this week after having AMD-only products since AMD64 was released.
  • OrianeOriane Turn around.
    edited January 2007
    A 40% jump in performance over the new Intels is really quite a boast, especially considering the thrashing AMD is taking from Conroe and friends. It also sounds like it could be a jump in a lot more than the on-die memory controller.

    If you take a trip over to AMD Zone, there is little shortage of confidence in this delivery. There is also some honesty as to what it may mean to NOT deliver on these claims- or at the very least- contenders to the new status quo.

    For me though the proof will be in the Folding- and it looks like this could be a very exciting year.
  • Sledgehammer70Sledgehammer70 California Icrontian
    edited January 2007
    you must remember the Barcelona has a ton of new items and updates under the hood, so the performance AMD is claiming could be true :) I guess we will find out.
  • ThraxThrax 🐌 Austin, TX Icrontian
    edited January 2007
    I'm saving money now, just in case.
  • Sledgehammer70Sledgehammer70 California Icrontian
    edited January 2007
    I think it has potential but no one knows :) if it is a Conroe killer Intel might just be in a pickle as they have no new architecture to push forward. It will be like when the A64 first came out.
  • miragemirage
    edited January 2007
    Core2 is a superior processor by a wide margin but there is no magic into it. It has just more execution units, shorter pipeline, lower-latency/faster/larger cache, etc., all the known technologies, improved and optimized. So, I think AMD can match and improve on the performance of Core2 with its next generation. IMHO, the real problem with AMD is process technology. While AMD has just started to transition into 65nm, Intel is preparing to start 45nm production. All those architectural improvements increase the cost by increasing the number of transistors and die size. If AMD will not catch Intel in process technology and keep following one generation behind, it should have product performance one generation ahead to be able to compete. This is the only way they can charge the premium price and offset the cost of older process technology, just like their position before Core2 was released. Now, Intel has best of both worlds: process and architecture. Did you guys see AMD's earnings announcement this week? it was in red.
  • Sledgehammer70Sledgehammer70 California Icrontian
    edited January 2007
    Even with AMD being at 65nm they have been able to perfect one fabrication to the T before moving on, making it much easier to work with. I am sure AMD has 45nm processes going and being built on as we speak. But I also believe Die shrinking is not the only way to increase performance. AMD has already proved you can save massive amounts of energy with a 90nm setup let along a 65nm, and they also have some great new on die tech that is sure to beef up what they got. The introduction of L3 cache is one thing I am looking at. What type of performance is to be had with that? Only time will tell.
  • miragemirage
    edited January 2007
    Semiconductor process technologies are developed in three stages: path finding, development, ramp up. While one process is in the stage of ramp up, the next one is in the development stage, and the following is in path finding. Each stage takes approximately 18-24 months (at least for Intel). So, it is for sure AMD is also working on 45 nm technology. But Intel is now in ramp-up stage while AMD is still at development. In the good old days, every new die shrink was released with increase in frequency and decrease in voltage. Not anymore, neither Intel nor AMD are able to increase the frequency with every new generation, they need to add more cores and cache to increase the performance which increases the die size. While the feature size is shrinking, dies do not shrink as much nowadays, which makes it even more important to have the latest process technology. More than half of the space on the die is consumed by cache memory in today's processors and this is expected to increase.
  • Sledgehammer70Sledgehammer70 California Icrontian
    edited January 2007
    But I still think AMD is on the path to get more performance out of one size rather than shrinking all the time.
  • OrianeOriane Turn around.
    edited January 2007
    Intel isn’t going to let AMD have all the thunder anyhow. Intel should be hot on their heels with Penryn and the 45nm Nehalem/SSE4 architecture which brags about further performance increases and power reductions. Penryn is also slated to be Merom’s successor– i.e. mobiles.

    But AMD has not gone to sleep about this either. The Pande Group seems pretty excited about all this too and left us this:
    lindahl(Pande Group) wrote:
    All the SSEX are marketing names from Intel. SSE2 made a big difference from SSE, since it added double precision floating-point support. SSE3 essentially didn't bring any floating-point improvements whatsoever, just optimizations for integer video encoding. SSE4 improves shuffling of data and some summation, but nothing that would help us do the nonbonded forces any faster.

    The big difference with woodcrest is that the CPU can now execute a full SSE instruction per clock instead of splitting them, but that doesn't require any code rewrites. At least for 64-bit code, woodcrest really seems to shine.

    We are manually profiling the code on single assembly instruction level, which is way beyound what Vtune or any other automatic tools do. And Peter is pretty right - we start thinking of optimizations when new instructions are only on the rumor stage. The K8L opterons released in 2008 are looking _very_ promising.

    Cheers,

    Erik
    linky

    So for sure count on AMD to adopt the extensions ASAP with their own version

    All I can say is: Keep the faith. AMD is not going away and AMD vs. Intel is very good for us all. The discussion at this time next year may add how many 64-bit full-extension version cores we can stuff into one system. And that, as far as the FAH project and the quest for cures are concerned, is VERY good news.
  • miragemirage
    edited January 2007
    Cost is everything about profit, especially in this cut-throat price war. The best way to decrease cost is shrinking. If AMD can come up with a ground-breaking technology that can give better performance without increasing number of transistors, yeah, they can keep using the same old 90nm tech (or 65m that they just started). Apparently they are not able to do this, looking at the latest earnings announcement. No one could break Moore's Law yet :)
  • Sledgehammer70Sledgehammer70 California Icrontian
    edited January 2007
    Did we ever think about AMD going from 65nm to 32nm? I mean they could make the jump with a bit more research, and with the new L3 cache this could be very possible. Not saying it would happen, but what if?

    Also with AMD pushing more for a GPU/CPU hybrid, they may be able to cut massive costs on Motherboards for the entry level market. And possibly get into the high end market. There are too many new architecture aspects of AMD's new tech that are unknown or that are in massive speculation to really pin point AMD as out of the race. The improvements that are known to be in the new architecture show much more promise than Intel's architecture, and most of the industry realizes that, and AMD also knows if they don't put a Intel killer out they will keep losing market share, that they worked so hard for.
  • miragemirage
    edited January 2007
    I am also expecting that AMD will come up with a very competitive processor in their next generation. After all, as I wrote earlier, there is no magic in Core2. But, I am afraid, they will be loosing money with their products unless they do either of the following two: catch up with the process technology of Intel which is extremely hard in the short term or release a clearly superior processor architecture that they can charge the premium which is also not easy since the Netburst fiasco of Intel is in the past.
  • miragemirage
    edited January 2007
    Sledgehammer70 wrote:
    Did we ever think about AMD going from 65nm to 32nm?

    It is not possible for AMD. If there is one company which can do this, that is Intel. AMD is not even able to produce all of its processors yet; they work with IBM. AMD is working very hard on increasing capacity at Dresden and building new fabs in up state NY, but these are billion dollar investments that can not happen in short term. Intel's production capability is not even comparable with any company, their standard yield is more than 90%, the best yield in the industry, which also contributes to their bottom line. Competing with Intel in a price war is *brutal*, in one word.
  • Sledgehammer70Sledgehammer70 California Icrontian
    edited January 2007
    As we have seen in the past price wont matter if the performance is there... and it is possible for AMD to drop die size, but I highly doubt they will. They will have to have products pushing the 40-50% increase in performance to be able to win any war. Regardless of 65nm, 45nm, or even 32nm... If they have the performance edge than they will gain the love of the end user. Lower power with more performance seems to be the key.

    If you noticed Intel took a huge hit because of this price war as well... something like 1 billion dollars? If I remember correctly...
  • miragemirage
    edited January 2007
    Sledgehammer70 wrote:
    As we have seen in the past price wont matter if the performance is there... and it is possible for AMD to drop die size, but I highly doubt they will. They will have to have products pushing the 40-50% increase in performance to be able to win any war. Regardless of 65nm, 45nm, or even 32nm... If they have the performance edge than they will gain the love of the end user. Lower power with more performance seems to be the key.

    If you noticed Intel took a huge hit because of this price war as well... something like 1 billion dollars? If I remember correctly...


    I already love them, check my sig. And I just ordered another socket 939. Just recently when AMD processors were more expensive, I was buying Intel processors though ;) And before that, K6, even K5. If they beat Intel in performance and increase prices, Core2 here I come.:D

    I am like an alcoholic when it comes to overclocking, the quantity beats the quality :crazy:
  • Sledgehammer70Sledgehammer70 California Icrontian
    edited January 2007
    I am not much of an overclocker as the overall performance increase for the tasks I do are minimal and not worth the extra cost to dissipate heat from the high power usage. TBH I no long own a desktop.. as I am laptop bound :)

    But I will most likely jump back in the ball game once I see a huge performance increase in the CPU & GPU market.
  • miragemirage
    edited January 2007
    Which laptop do you use? You know... maybe we could overclock it :tongue2:
  • Sledgehammer70Sledgehammer70 California Icrontian
    edited January 2007
    Actually I have done everything in my power to try to OC it, and the bios is locked up like no other, Dang Phoenix and HP...

    But to stay on topic I was thinking and reviewing current AMD and Intel performance specs, and if AMD is claiming a 40% increase over Intel’s current offering that means AMD is planning on a 60% to 70% increase over the current chips they offer currently. As in some cases Intel is tromping on AMD by 30%. That to me is a much greater increase in performance than we could ever expect out of a 90nm to 65nm change… right?
  • miragemirage
    edited January 2007
    How about something like this :eek:
    http://www.notebookforums.com/thread80879.html
  • LincLinc Owner Detroit Icrontian
    edited January 2007
    Some really great content in this thread. Oriane and mirage - if either or both of you are interested in writing for us, please let me know :)
  • OrianeOriane Turn around.
    edited January 2007
    Actually, the upstate NY plant that AMD is building is initially for their 32nm architecture This site, however, is still 5-7 years away from being done, but AMD's current roadmap shows them having 32nm late 2009/early 10. 45nm is mid-2008 only 6-8 months after Intel. If you understand the significance of that- AMD is lopping off almost a year in being behind Intel in process manufacturing. By the time you get to the 22nm process, they could be neck and neck.

    As far as spending billions on the NY fab- that's just business. You have to spend money to make money. As far as having vanishing profits- Intel just had to suffer their share with the weight of their huge overhead + tanking stock and survived. Intel is also still not doing all that well if you've seen their last quarterly. Stock and borrowing are part of that survival process and banks love to lend lots of money if they can see a return and stockholders love buying low to sell high. It's called investment and having what it takes to take the risk. I don't think AMD is going to fold up anytime soon and I'm confident that they will have another good day against Intel ... but they may have to tighten their belt for a while right now.

    >whew< This thread made me do some business homework-hehe!
  • OrianeOriane Turn around.
    edited January 2007
    General Keebler wrote:
    Some really great content in this thread. Oriane and mirage - if either or both of you are interested in writing for us, please let me know :)

    Thank you, General Keebler. :) There are many very bright people like mirage here who really make me think. If I can ever find the time to post something remarkable, I'll be sure to PM you with it. Being caught up in studies, job, and athletics could make that quite a wait :eek: but thank you just the same :thumbsup:
  • miragemirage
    edited January 2007
    General Keebler wrote:
    Some really great content in this thread. Oriane and mirage - if either or both of you are interested in writing for us, please let me know :)

    Thank you sir. I am sure it will take more than throwing my 2c here to write up something comparable with the other valuable writers of SM. It sounds very attractive and I will let you know if I can come up with something worthwhile.

    Oriane, thank you for your kind words, likewise :)
  • profdlpprofdlp The Holy City Of Westlake, Ohio
    edited January 2007
    Is a 40% speed advantage worth waiting a few months for? :eek:
  • LeonardoLeonardo Wake up and smell the glaciers Eagle River, Alaska Icrontian
    edited January 2007
    That to me [Barcelona] is a much greater increase in performance than we could ever expect out of a 90nm to 65nm change… right?
    AMD touts the core to be about 90% completely new design. It is not legacy. It is a jump forward in technology.
  • miragemirage
    edited January 2007
    Google gets back on Intel track

    It was Sun last week and Google this week. The reason of AMD's concerns is becoming even more clear. If they do not deliver the performance lead they claim, we can expect that processor prices will be increasing. I hope they do, but in any case 2007 will be a very difficult year for AMD. If they deliver, it will be a good year for the consumer :)
  • OrianeOriane Turn around.
    edited January 2007
    profdlp wrote:
    Is a 40% speed advantage worth waiting a few months for? :eek:

    Actually, I don't think it will be the 40% increase in performance people will wait for as much as a 40+% decrease in price.($880USD for Q6600 @top). I think this will only happen when Barcelona is on the horizon, and to the cost/performance conscious- it may be worth the wait for both.

    A guy I work with just bought an X2 system this weekend. It was a deal he could afford. If you are a college person living on pocket change & on the road a lot, it may be next year before you bother to think quad-core :( .
  • LeonardoLeonardo Wake up and smell the glaciers Eagle River, Alaska Icrontian
    edited January 2007
    If you are a college person living on pocket change & on the road a lot, it may be next year before you bother to think quad-core
    Or a family man paying the mortgage, utilities, car loan, college tuition, roof repair...

    In some ways, it doesn't get a whole lot easier. :)
Sign In or Register to comment.