Comments Locked

58 Comments

Back to Article

  • biassj - Saturday, October 29, 2011 - link

    Still can't beat an I5 single threaded. AMD fan here =(::
  • zondas30 - Saturday, October 29, 2011 - link

    oh this time its not only 1 vore left alive from 8 cores but 2 cores :D what a record, try it with all 8 cores to prove that amd is truly oc worth cpu
  • Kristian Vättö - Saturday, October 29, 2011 - link

    Each Bulldozer module consists of two cores, so two cores is most likely the minimum you can use.
  • Shining Arcanine - Saturday, October 29, 2011 - link

    It is 1 core. These so called modules are AMD marketing terminology for actual cores. The two logical cores are from AMD's SMT implementation. Unfortunately, it does not seem like it is possible to turn that off to get 1 logical core.
  • SlyNine - Saturday, October 29, 2011 - link

    That is a grossly misinterpretation of the truth. Each module still has 2 integer cores and many components are doubled. It is DEF not SMT, Sure the FPU is shared but even then its a 4 FPU core CPU. No excuse for benchmarks below a Phenom core per core.
  • silverblue - Saturday, October 29, 2011 - link

    CMT for integer workloads, SMT for floating point.
  • dgingeri - Saturday, October 29, 2011 - link

    "CMT" Like Country Music Television?

    Sorry, SMT is Symetric Multi Threading, useful for integer or FP multithreading.

    A single Bulldozer module will fully multithread integer caluclations and up to 128 bit SIMD (Single Instruction, Multiple Data) and FP instructions. (Standard FP calculations are 64 or 80 bit.) The only point where it will only handle one thread per module is 256 bit SIMD instructions. The big downside is that the x86 instruction converter/scheduler only handles 4 instructions per clock. That's primarily holding back the BD performance. If they could get that up to 8, it would match the current Phenom II architecture. If they could get it up to 12 and shorten the execution pipeline, I bet it might even match SB.

    Never heard of CMT other than the cable channel.
  • silverblue - Saturday, October 29, 2011 - link

    CMT = cluster multi-threading.
  • rns.sr71 - Saturday, October 29, 2011 - link

    conroe, wolfdale, nehalem and sandy bridge all decode a max of 4 instructions per clock, per core. nehalem and SB share that between two threads per core.
    K-8, K-10 and K-10.5 can decode 3 instrctions per clock, per core. no sharing.
    buldozer decodes 4 instructions per module shared between 2 integer cores. it doesn't need an 8 wide decoder- it would be HUGE! bulldozer also has the same pipline length(roughly) as nehalem/SB. so thats fine.
    the issue is in the cache speed/latency, OoO buffers(too small compared to nehalem and SB) and i would surmise the the 'fusion ops' isn't as good as what intel implements- thats huge.
    a 3rd alu added per integer core would help, though. but thats going to increase the size, too and create executions resources that would go unused in some instances.
    silverblue is right- 'cluster multi-threading'. sun micro-systems(now oracle) used it on the sparc and ultra-sparc processors.
  • JKflipflop98 - Monday, October 31, 2011 - link

    So that means Intel 990X chips have 12 cores then? Because i7's have the same stuff.
  • Wierdo - Monday, October 31, 2011 - link

    No they don't, Hyperthreading uses the same core to semi-process two threads, it tries to fit a core's pipeline with the second thread when there's a hiccup in processing the first one more or less, basically taking advantage of unused processing windows when possible.

    Bulldozor uses two cores, with their own integer pipelines and the related per-core goodies to do so. So it's two cores in the true sense of the word from a technical point of view.

    Performance of those cores, on the other hands, is another matter.
  • Wierdo - Monday, October 31, 2011 - link

    Thought this was an interesting page, showing performance of Bulldozer with 1core/module vs 2core/module settings:

    http://www.hardware.fr/articles/842-9/efficacite-c...

    Seems that vs 1core/module the Bulldozer actually loses some performance in games, about %5 usually. But in well threaded applications the Bulldozer gets up to ~%80 of the benefit of having the second core.

    Also compared performance gains from Hyperthreading vs Bulldozer's 2cores/module approach, in HT providing a gain of %5-%36 vs %36-%80 respectively.

    I'd suggest putting the page through google translator since it's in French, but the charts are easy to understand without translation.
  • joe4324 - Saturday, October 29, 2011 - link

    I know its purely academic, but this bodes well for first gen scalability correct? Can we get a benchmark of this chip at say 6ghz? somehow just to see how it stacks up? This is the reason for the long pipe design right?
  • Aphelion02 - Saturday, October 29, 2011 - link

    Funny thing is, it will still lose to a stock i7 2600 in benches.
  • Obsoleet - Saturday, October 29, 2011 - link

    No, a stock i7 2600 wouldn't beat an 8ghz Bulldozer at anything.
  • piroroadkill - Saturday, October 29, 2011 - link

    Actually, with only one module enabled, it would have just less than half of the resources a 2600 starts with.

    I'm not sure having it at 8GHz would offset that enough.
  • Obsoleet - Saturday, October 29, 2011 - link

    He said an 8ghz Bulldozer. Not with the setup used in this article.

    Countering over the top, idiotic slander that "Haha, even an 8ghz Bulldozer would lose to a stock 2600K" is a necessity.
  • neotiger - Saturday, October 29, 2011 - link

    The only 8GHz Bulldozer in existence is one that only has 1 module enabled.

    So when someone says "8GHz Bulldozer" it is understood to mean the real 1-module CPU instead of the imaginary 4-module 8GHz Bulldozer that does NOT exist that you seemed to have in your mind.
  • squizz - Sunday, October 30, 2011 - link

    Nope. It does exist:
    http://www.xtremesystems.org/forums/showthread.php...
    In fact disabling 3 module gives you upmost 50 MHz gain. The reason why there are low nomber of 8c 8000MHz results is because everyone wants that +50 MHz gain.
  • Aphelion02 - Sunday, October 30, 2011 - link

    The one slandering is you. I said 'it", which can only be the chip mentioned in this article. YOU are the only one who mentioned this hypothetical 8ghz bulldozer.
  • Obsoleet - Sunday, October 30, 2011 - link

    Nice try, but stop saying arrogant, stupid shit. Either way, it does exist.
  • seamonkey79 - Sunday, October 30, 2011 - link

    Wow... this CPU would lose to a 2600... which makes him right. How ignorant are you?

    "It" very very very simply denotes the CPU in THIS post, not all of them in the world. Reading comprehension and critical thought for the win.
  • Obsoleet - Sunday, October 30, 2011 - link

    Wrong again. Any 8ghz Bulldozer would beat a stock 2600k. Meaning this, or any other.
  • formulav8 - Monday, October 31, 2011 - link

    The guy said an 8ghz Buldozer. Not a 1 Module version. So if Ob took it to mean a full moduled BD than thats completely fair..
  • hsew - Sunday, October 30, 2011 - link

    Ummm.... Power Consumption? Heat? Stability? General ability to not be a flop?

    Fine, I'll leave....
  • Beenthere - Saturday, October 29, 2011 - link

    BD might be up to 20% faster with a few tweaks and Piledriver could build on that.

    http://techreport.com/articles.x/21865

    http://www.pcstats.com/articleview.cfm?articleID=2...
  • undermined - Saturday, October 29, 2011 - link

    The problem being that Intel can launch IvyBridge at any performance level they want to. and AMD cannot with the current steppings of Bulldozer.

    Even if AMD can tweak for 20-30% improvements to help it beat SandyBridge, IvyBridge will launch beating them shortly after all the while SandyBridge can drop in price to block any $ to performance advantage AMD might try to exploit.

    Overclock ability to higher levels also is using power like a pentium4 so AMD would have to address power use, performance and price all in the next stepping or revision were normally one could work on 1 or two of those aspects.

    I was really hoping AMD had a good competitor but instead they are going to have to fall back on possible updates to fix the issues instead of Bulldozer really helping them in any way at all currently.
  • Beenthere - Saturday, October 29, 2011 - link

    Consumers buy what makes them happy. Few people need or are willing to pay for bleeding edge performance.

    Some folks even have a moral compass that prevents them from buying products from a company that has illegally done all they can to put AMD out of business so that a market monopoly exists where consumers have no choice.

    The good news is people are voting with their wallet and demand for AMD CPUs and APUs far exceeds supply. FX works well for heavy load apps. With Win 8 it will see a ~5% performance gain. Trinity will be a winner no doubt and available soon. Piledriver could be a nice upgrade for desktop? We'll just have to wait and see.
  • silverblue - Saturday, October 29, 2011 - link

    I'd wish you'd stop going on about this. I dislike Intel's business practices, but if I was looking for a new setup and they offered the best performance for the price, you can be sure I'd go for them regardless of whatever they've done.
  • Eridanus - Saturday, October 29, 2011 - link

    It's being shortsighted, you know that.
  • menting - Sunday, October 30, 2011 - link

    i'm sure you bought Rambus then back in the days when it was like 5x the price for 1% more performance over DDR then? Just to be farsighted?

    you should buy based on what's the best performace to ratio, and only decide on brand loyalty if prices were at par. Companies who do not have a loyalty towards their consumer base should not be rewarded with consumer loyalty.
  • frozentundra123456 - Saturday, October 29, 2011 - link

    I am so tired of all the people that think AMD is some white knight that is out to help the consumer, while Intel is some evil corporation only bent on making money. They are both businesses. AMD wasnt so morally upright as to not charge outrageous prices when they had the performance advantage with the original Athlon X2 chips.

    My "moral compass" tells me to buy the best product, since I dont really believe either AMD or Intel is more moral than the other. So I support the company that makes the best product, not rewarding the company that makes an inferior one.

    And if intel wanted to, with the resources they have, and the cash on hand, they cut cut prices even further and decimate AMD.
  • Eridanus - Sunday, October 30, 2011 - link

    Increasing prices when the market lets you do so and being illegal knowingly is still two very different things. Avarage people are rewarded by jail sentences in the latter case.
  • Eridanus - Sunday, October 30, 2011 - link

    And not to forget, a weak AMD is good for Intel, a bankrupt AMD is not, because then they are going to get separeted into two companies.
  • kilkennycat - Sunday, October 30, 2011 - link

    Wishful thinking wins over logical reason.....?
    And demand exceeding supply? Sure, since Global Foundries has major 28nm. yield problems, and in the case of Bulldozer not exactly helped by the wasteful silicon-.inefficiency of the design. AMD is currently flushing $$ down the drain with every Bulldozer they sell. For the part to be even acceptably profitable, it would need to be priced ~ 3x the price of the i7-2500.
  • silverblue - Monday, October 31, 2011 - link

    I think you mean TSMC with the yield issues.

    AMD only pay for each acceptable product and not everything GloFo produces, luckily.
  • dgingeri - Saturday, October 29, 2011 - link

    yeah, well, 20% won't get it past the i7-2600k. It's going to need a lot more than that, and the new IB chips will beat it down even further.

    Don't get me wrong, I would like to see AMD succeed. I was really hoping for BD to be a great chip, but they let me down. BD is just garbage, and should be scrapped in favor of keeping around the current 6 core Phenoms, and perhaps expanding that to 8 cores on the desktop. (Server 8 and 12 core chips already exist, but they're too expensive to use on the desktop.) BD might do well on the server end, but they're garbage for desktop or gaming work. Even a 20% boost in performance won't help that.
  • Obsoleet - Saturday, October 29, 2011 - link

    The issue isn't that Intel is faster for most things, they are and are expected to be.

    The issue is that the CPU has taken a back seat to the GPU, and now the APU.
    i5/7s, Phenoms and Bulldozers are all grossly overpowered for the tasks people do.

    The advancements Intel makes are meaningless.
    Where it's at is technologies like CUDA and AMD's Fusion APUs. In fact, Intel's SSD developments are more important than their CPU developments.

    Intel's winning a race against itself, in a market segment that mostly matters only to forum-goers. The real market competition is with ARM and APUs.

    I'm sticking with my Q9450 for a reason, it has more power than I need and I use it more than most people would.

    When I upgrade, it will be more in cost-consideration than anything in regards to a CPU purchase, since all modern CPUs are more than enough.
  • nofumble62 - Saturday, October 29, 2011 - link

    I am sure graphic is important in games and but in a lot of benchmarks, the score was just measured in the time it took to complete a task. And Intel systems just finished it faster.

    If you are doing work, you know time is very important.
  • kilkennycat - Sunday, October 30, 2011 - link

    Grossly overpowered?... nope..certainly not for PC gaming....
    Take a little time studying Battlefield 3's innovative use of **CPU** resources.
  • silverblue - Monday, October 31, 2011 - link

    It doesn't need anything more than a dual core Phenom II.
  • silverblue - Saturday, October 29, 2011 - link

    They've got the bolding on the second link all messed up, awarding some wins to one OS when it's actually a defeat. Interesting links though, thanks.

    Mask 55/AA is a rather intriguing idea.
  • Beenthere - Saturday, October 29, 2011 - link

    AMD already sells 65W Phenom IIs. Intel is just copying AMD's technical insights of what consumers want, as Intel has done for several years now.

    For the record Intel CPUs have had a cold defect since after Celeron and that is why they don't post 8+ GHz. CPU performance results like in this article.
  • silverblue - Saturday, October 29, 2011 - link

    Both companies have offered lower power CPUs for years now.

    Also, didn't AMD have the cold bug until Phenom II?
  • Beenthere - Saturday, October 29, 2011 - link

    Nope, no cold Bug defect in AMD CPUs. That's why they have held the OC'ing record for years.
  • cfaalm - Saturday, October 29, 2011 - link

    While all nice and fancy, this means nothing to a normal user or even a prosumer. I'd like it to be an AMD rig when I get to upgrade and I have the "luxury" to wait and see what Piledriver brings to the table. BD needs some heavy tweaking on the idle and load power as well as the performance. Piledriver should trounce Phenom II X6 with a vengeance in every benchmark to convince me.

    What I see mentioned seldom here is that AMD has repeatedly said that a bulldozer module would do 80% of the work of a normal dual core because of shared resources. If we do a little oversimplyfied math then 8 x 80% = 6.4. I shouldn't suprise anyone that it is'nt always able to blast past the ole Phenom II X6. I think AMD built this thing for the future. The ealry dualcores didn't always outperform the single cores. Now they bring an 8-core or a sort there of that doesn't always outperform a true 6-core. My guess is it will improve when the architecture is refined (IPC, cache latency etc.) and software makes better use of the abundance of cores. It has happened before. The simple fact that there are benchmarks where BD does very well must indicate that AMD have done something right.
  • Locut0s - Saturday, October 29, 2011 - link

    Seems to me this is as much or more a marketing stunt for ASUS' Crosshair V Formula as it is for Bulldozer. I get that they are trying to show BDs scalability and in some ways they are but they are also using very restricted conditions here. Reduced numbers of cores, extreme cooling etc... You can bet that ASUS is going to use this on their packaging / marketing material for the Crosshair V, less so for BD.
  • Sunsmasher - Saturday, October 29, 2011 - link

    What does this have to do with anything?
    Bulldozer is still slower tha Phenom thread-to-thread.
    Only mildly interesting as a science project....
  • frozentundra123456 - Saturday, October 29, 2011 - link

    I totally agree with you. Unfortunately, every CPU article that comes up seems to get taken over by AMD fans spouting the same old justifications for AMD.

    AMD CPUs are good enough
    Intel is the evil empire and only is ahead because they used unfair practices
    Bulldozer doesnt matter on the desktop, it will be great as a serever
    What really matters is graphics
    Oh, and a new one in this thread, Intel has a Cold Bug. Gee, I will remember to be sure to get an AMD cpu for my next liquid nitrogen cooled system.
  • Ronakbhai - Sunday, October 30, 2011 - link

    I guess Intel prices will still sky-rocket due to the lack of competition. Hell, the 1.5 year old 980X is still rocking socks.
  • lukarak - Sunday, October 30, 2011 - link

    Not exactly. It loses to 2600 unless it's complete 6 (or 12) thread load. That design is not as efficient. It's successor will completely trash it, just as the most popular i5-750 gets trashed by the 2500.
  • piroroadkill - Sunday, October 30, 2011 - link

    His point still stands, though. Whoever bought a 980x near launch is still rocking an extremely high end CPU. With a mild overclock, it can keep up with Sandy Bridge, and if something is heavily threaded, it rapes all.

    That's a pretty unique situation, because it really is fairly old now.
  • Roland00Address - Sunday, October 30, 2011 - link

    I repeat, 3 years ago and it is still one of the best cpus out there.
  • johnsmith9875 - Sunday, October 30, 2011 - link

    Release the X6 in 32nm, put two of them on a single AM3+ chip, and sell it as a budget CPU upgrade for existing users.

    Then you will have some time to fix Bulldozer and get it to market.
  • N4g4rok - Sunday, October 30, 2011 - link

    That's not a bad idea. A lower TDP Phenom II would have a lot of potential, considering the clock speeds you could get out of it.
  • silverblue - Monday, October 31, 2011 - link

    Whoa whoa whoa... you're saying release Magny Cours for the desktop?

    That thing needs a quad-channel memory architecture. Interesting idea though.

    A GPU-less Llano would make more sense as long as it came with Turbo Core - 4GHz basic, 4.5 turbo, that sort of thing.
  • mforce - Sunday, October 30, 2011 - link

    This Bulldozer reminds me of the Pentium IV days, Intel was so excited about it but almost everyone else wasn't. It too kept breaking GHz records one after the other only to get a good spanking from the Athlons.
    Bulldozer can break all the records it wants as far as GHz is concerned it's still a fail in my opinion. It's getting its ass kicked by current gen I7 cpus. I'm an AMD fan but I just got a Phenom II X6 and I'm happy with it.
  • lucypinderol - Tuesday, November 1, 2011 - link

    Small company grants are unfortunately becoming harder, and taking longer to access. Now, small business funding is a new alternative. ==== h t t p : / / w w w . f u n d f a c t o r . c o m / s m a l l - b u s i n e s s - f u n d i n g . h t m l ====

Log in

Don't have an account? Sign up now