Announcement

Collapse
No announcement yet.

AMD FX-8150 Bulldozer On Ubuntu Linux

Collapse
X
  • Filter
  • Time
  • Show
Clear All
new posts

  • #61
    Originally posted by Kano View Post
    The problem with bulldozer is that it is only a 8 core for apps that dont use the fpu, but a 4 core for apps that do (well it always shows 8 cores, but internally). 7zip seems to use the fpu much, so it does not scale much, doing simpler stuff like parallel compiling you can see some better results. Also amd did the false decision in advertising 8 full cores instead of 4 cores with ht to the os, so you would need a different sheduler as workasround, for win you will need version 8 to gain a little bit. well it will not be the case that the cpu will fly away then in benchmarks for me the whole design is crap, they should never have build that 8/4 core mix thing just to get a 8 core for the masses.
    I guess they were aiming to get a title: World's first 8 core desktop processor, and not actually aiming for the consumer to have the best performance. They should've just die shrunk the phenom ii and added more instruction sets. Anyways, what are some apps that don't use the fpu?

    Comment


    • #62
      Btw. for the new i7-3960X cpu intel even DISABLED 2 cores (the chip itself has got 8 when called Xeon). They wanted more turbo boost time, thats clear... If they wanted they could have sold it as 8 core as well, but compared to AMD they test before what solution is faster: higher clock speed or more active cores...

      Comment


      • #63
        That so many people lambast the Bulldozer architecture and think that AMD "should have tweaked Phenom II" goes to show how short-sighted readers are.
        Readers of computer technology websites at that, so you'd expect them to have some modicum of literacy in the matter. Scary... Worst I've seen is probably HardOCP.

        1. Software: if software can't use the CPU properly, said CPU is neutered
        2. Fusion: that's the long term plan, with FPU-intensive, massively parallel workloads taken care of by the GPU part of the silicon

        Comment


        • #64
          Originally posted by Kano View Post
          Btw. for the new i7-3960X cpu intel even DISABLED 2 cores (the chip itself has got 8 when called Xeon). They wanted more turbo boost time, thats clear... If they wanted they could have sold it as 8 core as well, but compared to AMD they test before what solution is faster: higher clock speed or more active cores...
          yeah exactly, intel knows that 8 cores wouldn't be beneficial atm, so they didn't make an octacore processor. if they wanted to, they could've implemented it into 3930k and 3960x easily.

          Comment


          • #65
            Originally posted by PsynoKhi0 View Post
            That so many people lambast the Bulldozer architecture and think that AMD "should have tweaked Phenom II" goes to show how short-sighted readers are.
            Readers of computer technology websites at that, so you'd expect them to have some modicum of literacy in the matter. Scary... Worst I've seen is probably HardOCP.

            1. Software: if software can't use the CPU properly, said CPU is neutered
            2. Fusion: that's the long term plan, with FPU-intensive, massively parallel workloads taken care of by the GPU part of the silicon
            hmm. i sure would love to see the gpu part to rape fpu tasks. also, if the software can't use the cpu properly, they shouldn't have released it in the first place, save it for later on, keep optimizing it, keep an ace up the sleeve which gets rid of the competition when the time is right. just keep baking it till its cooked. not half. and atm, if they did release a tweaked phenom, consumers would be much happier.

            Comment


            • #66
              I don't think Intel is disabling cores on SB-E because it's a better chip that way, but rather because it creates a better fit with their yield management and market segmentation strategy. How else would they charge twice as much for the Xeon E5-2687W? Similarly, why else would 3930K only have 12MB of L3 cache?

              Comment


              • #67
                Originally posted by raj7095 View Post
                also, if the software can't use the cpu properly, they shouldn't have released it in the first place, save it for later on, keep optimizing it, keep an ace up the sleeve which gets rid of the competition when the time is right. just keep baking it till its cooked. not half. and atm, if they did release a tweaked phenom, consumers would be much happier.
                Or they release it publicly so any software developer has time to play around with the architecture before the Fusion part.
                And again: the marketing surrounding BD has been complete and utter trash. I'm first in line saying that.
                Re-read all reviews, replacing all instances of "n cores" by "n/2 modules". Suddenly you're comparing 2/3/4 module Bulldozer to 4-6 core Phenom II. And performance looks much better. What difference the wording makes!

                Comment


                • #68
                  I've had my FX-8120 for about a week now. It is extremely fast, regardless of whatever spin Intel and their minions try to put on it. Let's discuss how and why:

                  Virtualbox: I can practically simulate an entire virtual city, I'm limited more by "only" having 16GB of RAM than I am by Bulldozer. If you think Sandy Bridge can run as many VMs simultaneously then you are clearly delusional, unless you're rationale is that SB wins some bizarre floating point benchmark in a VM. Given that no sane person would choose to virtualize a number-crunching/rendering/compiling type app, that is a really a poor way to benchmark it. 99% of people virtualize low to medium intensity servers and give them one real core each, where Bulldozer wins by default, by having more cores. Assigning more than 1 VM to a core on an Intel HT CPU is just asking for trouble.

                  Truecrypt: 2GB/s AES encryption, and 200+MB/s AES-Twofish-Serpent encryption using the built in benchmark. This also translates into a server than easily saturate 2 or more gigabit RJ45 ports with SSH encrypted data using scp without slowing down anything else that's running.

                  Multitasking/Apps: I have yet to hit the threshold where anything starts slowing down, I can run numerous VMs, Netbeans, Firefox and Banshee all at the same time, and new apps still launch instantaneously.

                  I've also used Sandy Bridge in real life(where most of you haven't used Bulldozer), and Sandy certainly doesn't feel any faster in single threaded loads(because 10-30% isn't noticeable in real life), but it definitely feels slower with more parallel workloads.

                  Comment


                  • #69
                    I would not say that the average VBox user will use more than 2 vms the same time. But when you really need more than 4 vms then you could be right that a few more cores would not hurt.

                    Comment


                    • #70
                      Originally posted by Kano View Post
                      I would not say that the average VBox user will use more than 2 vms the same time. But when you really need more than 4 vms then you could be right that a few more cores would not hurt.
                      Well, my purpose in buying it was to run an elaborate Virtualbox setup(prototyping networks/servers), so it was definitely the best purchase for me. The average user also doesn't need more than a dual-core Core2 or Phenom II, since the average user doesn't do anything too demanding, and the CPU will spend 99% of it's lifetime at 800mhz, and the average user won't even notice if something takes a few extra seconds. I guarantee if you set anybody in front of a Bulldozer computer, and tell them it's a Sandy Bridge-E, they won't be able to tell the difference, nobody is going to perceive it as being slow in real life.

                      Comment


                      • #71
                        I'm looking at it on the virtualization front as well

                        For me I don't give a rats about the gaming or desktop app performance which is what keeps getting benchmarked. How does it handle 4 VM's, and 3 of them have 2 vCPU's? For that type of load, it seems this chip is perfect for home server usage. Going forward I think this is the right direction in architecture since everything (in business, and some at home too) is going to virtual, from phones, tablets, and desktops, while new "servers" are already almost all virtual already. The big issue is this is a version 1 part, and it is still server tuned rather than desktop. AMD's strategy *should* work well LONG term, but short term? Yeah, not good. Intel is still betting on single threaded performance, and that could backfire really quick if everything *finally* shifts to multithreaded now that the writing is firmly on the wall... Time will tell whether Intel's current strategy or AMD's current strategy will succeed or backfire. I'd prefer that AMD keep going and innovating, since Intel hasn't proven itself capable of doing that in the absence of proper competition.

                        Comment


                        • #72
                          Originally posted by admiraljkb View Post
                          Intel hasn't proven itself capable of doing that in the absence of proper competition.
                          That applies to any company, really.

                          Comment


                          • #73
                            @admiraljkb

                            if intel wanted to sell an 8 core desktop cpu they would have done that. i7-39xx cpus are internally 8 cores with only 6 enabled. the new xeon single socket series will feature 8 cores at lower clockspeed. i dont think that amd has got a superior strategy in the long run, amd is just currently faster with some bulldozer server (opteron) benchmarks until intel will get the new xeons ready.

                            the main reason why the fastest intel desktop cpu does not use 8 cores is that turbo boost will work much better when more cores are idle so that the max tdp is not maxed out already at a lower clock speed. that means with fewer cores the cpu can run longer with higher turbo modes. with a shrinked layout that problem could be fixed most likely but will take a while. often amd has got production problems with its chips, at least more than intel when they want to shrink em. the splitted foundary is not even able to build trinity and gfx chips so tsmc needs to build those.

                            Comment

                            Working...
                            X