Barcelona trumps fastest quad-core Xeons, says AMD

Not content with claiming to have broken the teraFLOPS performance barrier with its next-generation R600 graphics processor, AMD has done some additional boasting on the microprocessor front. As EE Times reports, the company has asserted that Barcelona will offer 42% higher floating-point performance than Intel’s current fastest quad-core offering, the 2.66GHz Xeon X5355. AMD is a little quieter when discussing integer performance, but Mario Rivas, general manager of the company’s microprocessor group, nevertheless claims that Barcelona will offer a “double-digit leap in integer performance” over the quad-core Xeon. AMD chief of sales and marketing Henri Richard also says Barcelona will have a “significant integer performance lead” over Intel’s quad-core processors.

In related news, Anand Lal Shimpi over at Anandtech has put together an article that examines Barcelona’s key architectural enhancements and compares them to the improvements Intel and AMD have made to their chips over the past few years. The article looks at Barcelona’s 128-bit SSE execution width, improved branch prediction unit, Sideband Stack Optimizer, out-of-order load improvements, speedier memory controller, new prefetcher, L3 cache, virtualization tweaks, and other enhancements. Anand says Barcelona will initially launch at speeds of 2.1-2.3GHz, but that it will climb to the 2.7-2.9GHz range on the desktop by the end of the year.

Comments closed
    • JoshMST
    • 13 years ago

    Admitedly, I was first not terribly excited about the quad core products from AMD. The more I am reading though, the more interested I am becoming. The meat of the market will be Kuma vs. C2D, but when programs like Supreme Commander are bogging down everyone’s machine, and things like Alan Wake are going to pop up, then quad core is starting to look more and more appealing.

    As for AMD running taskmanager and simulators… in every other previous processor release AMD hasn’t done much more than show taskmanager to the press. It is either that, or they run something that does not specifically show the performance of the chip. I remember 9 months before the Opteron release they were showing working silicon, but wouldn’t let anyone touch it, or see how it was performing in the applications it was running at AMD’s hotel room during IDF. Don’t get me wrong, I don’t think AMD’s marketing guys can walk on water, but I do think that they have historically shown that they do not exagerate. I think mainly upper management does its best to keep their mouths shut as much as possible. Perhaps with Intel doing what they did last year with C2D has changed the rules a bit, but I don’t think that AMD marketing are chock full of liars.

      • shank15217
      • 13 years ago

      you are right, AMD has to release some information and possibly the increase in performance they are highlighting is their best case but even if they were able to bring performance up to 5% of core 2 in one major revision thats quite a feat. Lets give them the benefit of the doubt.

    • Fighterpilot
    • 13 years ago

    #44 hehe that WAS a good one.Seriously tho,it will be cool to see the quad core Barcelona do its stuff but I reckon most people here are mainly waiting for the Agena dual core V C2D showdown.
    Hardly anyone I know of has bought Kentsfields for home use and I find it hard to get too interested in the whole quad core thing at this point.
    I think TR better have some extra bandwidth on standby for that article’s front page comments (0-0)

    • Jigar
    • 13 years ago

    This is all BS. Intel’s C2D and Xeon CPU will never loss against the pathetic slow processors of AMD.. What do u say Shintai and Proesterchen.. I am with you this time. ๐Ÿ˜‰

    EDIT: Oh did i forgot to say .. I bet C2D @ 800MHZ can kick A$$ Barcelona @ 4.5GHZ

      • Proesterchen
      • 13 years ago

      q[

        • Jigar
        • 13 years ago

        Why do u feel like that honey?? ๐Ÿ™

          • Proesterchen
          • 13 years ago

          Mostly because my “I’m with Idiot ->”-Shirt is at the cleaner’s. ๐Ÿ˜‰

            • Jigar
            • 13 years ago

            ๐Ÿ˜‰ Good one.. I was expecting something like this.. hahaha

    • Wintermane
    • 13 years ago

    It is a proven opinion that I ajm exactly 42% more cudly and 47% more wudly then all of you.

    … Cudly numbertaken after bribing neice with chocolate.. not valid if sugar rush ensues also not valid arter 5 min from …NOW!

    …. Wudly number may vary depending on wether ive earen chese and as such am stinky or not… Valid only before midnight and after use of mouthwash. Real world wodlyness may vary.. Not valid when im nroke.

    • albundy
    • 13 years ago

    says AMD? that doesnt matter. what matters is what TR says in the reviews.

    • melvz90
    • 13 years ago

    of course thats what sales and marketing do.. they’re paid to claim…

    AMD chief of sales and marketing Henri Richard also says Barcelona will have a “significant integer performance lead” over Intel’s quad-core processors.

    bring out the real benchmarks.. not just hot air..

    • Usacomp2k3
    • 13 years ago

    So a chip released at the end of the year is faster than a chip that was released at the beginning of the year. *gasp*

      • Ardrid
      • 13 years ago

      While part of me understands statements like this, they really don’t make any sense (two major examples are NV30 and Prescott). It’s not like you can simply wait for the competition to do something and then change your product to one-up them. Such an ideology doesn’t work in the semiconductor industry because of the time necessary to complete a design, tape it out, test it, and possibly respin it. None of those steps can be done in a quick enough time frame to allow you to trump your competition. The only exception to that rule is GPU refreshes and even then that only applies if your competition delays their product countless numbers of times. Case in point: we all know the 8900 is going to be faster than the 8800 and we also know that NVIDIA is going to release it just in time for R600 if at all possible.

        • Buub
        • 13 years ago

        Interesting. Two examples that support the same point from the opposite ends of the spectrum.

        The Prescott was a string of poor judgements all added together to make one very underwhelming product.

        The NV30 on the other hand was a matter of over-reaching. Trying to go too far too soon. They were trying to move to the newest process when it was truly bleeding-edge, they were coming out with their first FP-based core, and they were trying to make it as flexible and powerful as possible. The latter was what made it not as good at gaming as the ATI alternative, which was more adept at gaming but not as generalizable. The process thing was what really bit them the hardest — it was too hard and didn’t ramp up fast enough.

        Either way, yeah, two examples of something that comes out later that isn’t necessarily “better” than something that came before.

        Another would be the P4 itself. For the first year or so, many tasks still ran better on a PIII, and ran way better on an Athlon, than they did on the brand-spanking new P4.

        In spite of those examples, though, the common case still prevails where evolutionary changes improve on the things they replace.

          • Ardrid
          • 13 years ago

          The general premise certainly holds the majority of the time if we’re talking about products from the same company/family; however, I don’t think you can necessarily make that same assumption if you’re talking about competing products because you have no way of knowing how good your competition is going to be. You could have a number of architectural changes on the board that would clearly bolster your performance but if your competition decided to go a step further, and still gets out ahead of time, they’re going to outperform you despite the fact that you arrive after them.

            • Buub
            • 13 years ago

            Agreed. I didn’t mean to imply across companies, because that really wouldn’t be evolutionary changes. It would be a mutation. ๐Ÿ™‚

            Now that I look at my post, I guess I DID imply across companies. Well I guess that works in a more general sense. But as you point out, in a more focused perspective, you would want to concentrate on iterations of the same or similar products, which would imply from the same company.

    • flip-mode
    • 13 years ago

    I just want to say that if these claims do turn out to be true, I will “ROFL” at Shin and (especially) Proest on a daily baisis. I might have to start a special thread for it. Fanboys…. ๐Ÿ™„

      • Fighterpilot
      • 13 years ago

      y[

      • Shintai
      • 13 years ago

      Before you say anything more, perhaps you should visit ยง[<http://www.spec.org<]ยง and gain some basic knowledge about the numbers there. Specially how 4x single cores own 2x dualcore systems. And today AMD is about 20% ahead in specfp_rate. And Intels quadcore is "penalized" in that benchmark. 4x singlecore 2.6Ghz DDR1-333 ProLiant BL45p (AMD Opteron (TM) 852) 4 cores, 4 chips, 1 core/chip 83.5 92.6 2x Dualcore 2.6ghz DDR2-667 ProLiant BL465c (AMD Opteron (TM) 2218) 4 cores, 2 chips, 2 cores/chip 82.7 89.5 2.66ghz Xeon quadcore: FB-DIMM 667 ProLiant BL460c (2.66 GHz, Intel Xeon processor X5355) 4 cores, 1 chip, 4 cores/chip 60.7 65.7 Ye..40% must be pretty amazing...*cough cough*. You got fooled by a PR spin. So "ROFL" at you flippie.

      • Proesterchen
      • 13 years ago

      q[

      • flip-mode
      • 13 years ago

      Wow, a rank and file reply from the Trinity! Weeee!

        • Shintai
        • 13 years ago

        Facts is a biatch flippie ๐Ÿ˜‰

        And guess who started the crap o.O

        I hope it will support your MB whenever it comes out. Would suck if the VRM didnt….then you would just keep hugging then X2 3600 for a very long time.

        Anyway, you should reply to the spec numbers and facts I posted instead of trolling together with Jigar. I know this is the part where it gets very tricky and hard for you. But try with some effort and see what you can come up with. Again..start at ยง[<http://www.spec.org<]ยง Also in terms of your post #18. Again you just got fooled by marketing spins. Their number is surely perfectly fine in that synthethic bench. Hey I can post some Sandra SSE numbers and claim what, 250-500% better performance than a K8? And again, K8 is already beating Intel in this bench.

        • Jigar
        • 13 years ago

        me too amazed … :popeyes:

    • flip-mode
    • 13 years ago

    Woot, me hugs my AM2 mobo. Of course, what is the point of upgrading when an X2-3600, even at stock speed, is faster than anything I need.

    On a more serious note, I’ll say a similar thing that I said in the frenzied run up to the c2d release: it would be pure stupidity for AMD to make these claims and then fail to deliver on them. (Actually, my statement regarding Intel at the time during which many were doubting the promises they were making was more like, “there is NO WAY Intel would make these claims and then not deliver them.” Even though I root for AMD’s continued success, for some reason I don’t trust their performance claims as much as I trusted Intel’s)

      • d2brothe
      • 13 years ago

      Gonna have to agree…we still *need* AMD to exist otherwise we wouldn’t have C2D greatness…or P4 crappiness…but either way, they’re important…but I don’t trust their boasting as much as intel’s…not cause one company is inherently more trustable, but because intel is more likely to have the R&D to back it up, plus we already saw the improved performance on the Core 1.

        • green
        • 13 years ago

        it’s actually an interesting one
        if we had a lack of AMD would it mean intel would have stuck with pIII?

        one of the main reasons for the netburst architecture was the gigahertz race against AMD
        so with AMD out of the picture the next rival would have been ibm’s powerpc architecture i suppose
        intel’s only motivation to brng out anything better then would be to make sure it beat powerpc
        so all they’d ever have to do is tweak the p6 architecture every now and then and give a speed bump
        which is essentially where we are today given everyone says c2d is just an improved pIII

        but i guess we’ll never know….

          • SGWB
          • 13 years ago

          Considering that before AMD released the K7 and Intel looked like they would be the king of the deskop CPU for a long while, Intel was talking about slowing down cpu advancement. They had set a goal of 1GHz around 2005 or so. Then K7 hit and suddenly everybody and their brother had a 1GHz or better CPU.

          I have to believe that without AMD we would be using somthing very similar to the PII/PIII, on the desktop, but nowhere near the performance of the Core 2 duo. You also probably wouldn’t have x86-64 desktops, because Intel was trying to push Itanium for 64 bit computing. Dual core desktop CPU’s are, I think, also questionable without AMD mixing things up.

      • blastdoor
      • 13 years ago

      I’m also a little more skeptical of AMD’s claims than Intel. But, I doubt that they are flat out lying, probably more likely they’re just being selective in their benchmarks. Nothing wrong with that exactly. I just find it hard to believe that K8L is going to beat contemporaneously shipping Xeons on a majority of benchmarks.

      If nothing else, Intel appears to have a lot more room to crank up clock speeds, even at 65nm, than AMD. I’m still fairly amazed that my new 1.8 GHz Allendale is running so effortlessly at 2.7 GHz without any extra voltage and at 40 degrees C after both cores have been running at 100% all day long. If Intel had to, I bet they could get even quad Core2 chips to 3 GHz and higher.

    • Buub
    • 13 years ago

    Once again, none of this actually means very much until we have production silicon in the hands of trusted testers. Until then, it’s just posturing and speculation.

      • JoshMST
      • 13 years ago

      But I thought that posturing and speculating was the draw to any technical based forum? How much fun is it to just talk about released products?

    • Ardrid
    • 13 years ago

    While I still have a hard time swallowing the 42% performance improvement claims, especially after what C2D has already managed to do, I will say that I have a bit more faith in those figures now that we have some solid, concrete facts about Barcelona’s improvements, courtesy of Anand, rather than rumors and assumptions spilling from the mouths of fanboys and the Inq.

    • Smurfer2
    • 13 years ago

    I seriously find this very interesting and hope those figures are correct… but a company saying something is “x%” faster is meaningless.

      • packfan_dave
      • 13 years ago

      Yup. If they’re saying Barcelona @ x GHz is y% faster than Clovertown @ z GHz in a specific benchmark, that’s something that can be tested (although the relevancy could be questioned). If they’re just saying Barcelona is x % faster than Intel, it’s meaningless.

    • PRIME1
    • 13 years ago

    I’m better looking than everyone, says PRIME1

    • JoshMST
    • 13 years ago

    Oh man, first Shin then Proe!

    At this point, I am feeling a bit better about AMD’s ability to compete in terms of performance. I think that on a per clock basis the upcoming chips will beat C2D’s and C2Q’s by a good 5% to 10% in most applications. I think the thing to worry about is AMD’s ability to produce these chips, and win back some mind share before the 45 nm hammer comes down from Intel. It also sounds like at least on the quad core side, it is going to be a slow ramp for AMD. I am curious if that will apply to the Kuma products?

      • SVB
      • 13 years ago

      Prepare thyself, here come the fanbois.

      • Beomagi
      • 13 years ago

      Unfortunately, I can’t have that kind of faith. Intel made a huge jump because they already had working silicon to base another design off of. For barcelona, Each tweak is minimal in effect, but I REALLY hope their combined effect is enough to at least make them competitive again.

      is am2 dead to Barcelona? I know amd stated am3 chips will work on socket am2, but are they going to screw am2 users?

        • JoshMST
        • 13 years ago

        Well, it certainly sounds like AMD went through the core with a fine toothed comb. I think it will be a nicely performing part. When you see how they widened pretty much everything, I think the overall effect will be pretty good. 40% overall… heh, AMD shareholders only wish. But still faster than C2D on a per clock basis.

        AM2 boards will support Kuma/Agena. Barcelona is the socket F variant for servers (well… and Agena FX).

        • Shintai
        • 13 years ago

        AM3 chips will not work in AM2/AM2+ sockets. It was an early rumour when people thought(hoped) one chip could have both DDR2 and DDR3. Also the serial VIDs makes it impossible for any AM2/AM2+ chip to work and the other way around for AM3.

        (Barcelona is an AM2+/F+(?) chip)

    • Krogoth
    • 13 years ago

    Who cares I just want to see the silicon in action.

    Intel and AMD are both infamous for exaggerating the performance of their products prior to their official depute.

    For the looks of Anand’s analysis on Barcelona. I suspect the design to be quite the power house.

      • packfan_dave
      • 13 years ago

      I hope I’m not being to repetitive when I say Barcelona/Agena/Kuma looks pretty good on paper, but the existing Core chips have tons of clockspeed headroom, Tigerton (Intel’s first iteration of quad core with shared L3 cache) should be out later this year, rumors of forthcoming 1600MHz FSB and/or 1666 MHz FSB server chips are flying all over the place, and Pernyn and its 45nm cousins should be out at the end of the year.

        • Krogoth
        • 13 years ago

        65nm C2D can obiviously go higher, but their TPD skyrockets as well. Intel would like to avoid doing that unless it was necessary to keep the performance crown. 45nm Pernyns also look good on paper, but the final silicon isn’t done yet.

        Even a sight voltage boost can greatly increase the thermal output of a C2D.

          • Shintai
          • 13 years ago

          Thats not entirely right. Intel showed a bunch of 45nm systems to the press with running applications and games. AMDs numbers are estimated from a simulator.

            • JoshMST
            • 13 years ago

            AMD has working silicon, and have since around September/October of last year. Those results are likely not from a simulator.

            While I think the green team’s marketing folks are probably too optimistic in this area, it is far too early to say that it is 100% PR spin.

            • Shintai
            • 13 years ago

            Thats why they write estimated performance? Even in late december they could only show a Windows with a taskmgr. Unlike Intel who 6 weeks later showed working systems with games and applications.

            Nopes, I think we stay at the simulated performance marker.

            • Krogoth
            • 13 years ago

            Intel marketing drones did the same crap with C2D before its depute and moreso, back in the Netburst days.

            • Shintai
            • 13 years ago

            Atleast with the C2D they said overall 20% that hold water. It was Anand that made the other mistake. But yes, its comon practice.

            But the only product on the list known is the Xeon quadcore, not even specs on the K8 was there or any other info. Looks applelish.

            It could justaswell be 80% faster than a 2Ghz K8, and the Barcelona used could have been a 2.9Ghz for that matter. It darn useless.

            • Krogoth
            • 13 years ago

            They were using A1 batch silicon which is far from the final tape. It is impressive that they even work, however it doesn’t elimiate the possiblity that the silicon had hidden problems (cueing the infamous P5 bug).

            I am not doubting the potential of 45nm process, but knowning Intel’s past behavior when there were the top dog. I wouldn’t be surprise if retail 45nm parts show up Q1 2008 unless Barcelona provides surprising results.

    • Proesterchen
    • 13 years ago

    This was first reported in late January.

    How much more often will AMD repeat their “42% faster than Clovertown in SPECfp_rate200x” claim? Or rather, how much more often will this be “news”?

    And I’m going to say right now that with 2.6 GHz Barcelonas more than a year away, Agena won’t be coming anywhere near those older targets this year.

    • Shintai
    • 13 years ago

    Just a shame its specFP_rate that they already beat Intel in.

    • BoBzeBuilder
    • 13 years ago

    Time to buy AMD stocks. In ye I trust.

    • DrDillyBar
    • 13 years ago

    yeah, the processor that we’ve not released yet is going to own Xeon…. really. :S

    • Dposcorp
    • 13 years ago

    Cue the poster who always writes “cue the fanbois.” ๐Ÿ™‚

    • Shinare
    • 13 years ago

    y[

Pin It on Pinterest

Share This