Announcement

Collapse
No announcement yet.

Intel Core i7 4770K "Haswell" Benchmarks On Ubuntu Linux

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

  • #16
    Originally posted by YAFU View Post
    Do you have performed the benchmark with Hyper-Threading enabled from the BIOS?
    I've read somewhere that applications are not prepared for multiple threads go faster if you have the Hyper-Threading disabled. I think the "gamers" usually disable it.
    Anyway, you could add that to the benchmark to see if there are differences with the Hyper-Threading enabled or disabled.
    No the reason some applications go faster is not due to them not being prepared for multiple threads but due to the nature of SMT. Basically the way SMT works is that there are 2 (Can be more in non-Intel implementations) threads queued up in any particular core at any one time, and so when one thread stalls out or sleeps it switches over to the other thread, now this is all nice and fine for workloads that are heavily threaded but are not processor intensive, but it is detrimental in cases where the threads are processor intensive, because instead of just having to wait for it to finish what it's doing now it stalls out and the other thread steps in and it has to wait for that thread to stall out before it can start processing again, thus resulting in a loss of performance.

    It's important to point out that although the front end is currently a massive bottleneck (which will be fixed by steamroller) the Bulldozer architecture and derivs don't have this problem because instead of a processing queue they are doing processing on the threads. That said AMD does have a more interesting problem because of this in terms of how to handle scheduling.

    Comment


    • #17
      Originally posted by Tgui View Post
      What article were you reading?
      The one where an ~8 month old chip beat a 0 day old chip in some tests despite being a hundred and fifty dollars cheaper? Not to mention how many tests where only 20-40% faster when the next amd core is estimated at 30% faster per clock. Also unlike intel AMD is boosting clocks with each generation 5800k vs 8350 vs 6800k

      Comment


      • #18
        @ Luke_Wolf, thanks for the explanation.
        ===
        By the way, would have been nice to have also the comparison of power consumption and heat generation with the FX-8350, in any of the other applications that use intensively the CPU.

        Comment


        • #19
          Originally posted by chris200x9 View Post
          The one where an ~8 month old chip beat a 0 day old chip in some tests despite being a hundred and fifty dollars cheaper? Not to mention how many tests where only 20-40% faster when the next amd core is estimated at 30% faster per clock. Also unlike intel AMD is boosting clocks with each generation 5800k vs 8350 vs 6800k
          Yes, AMD has come back considerably lately (which I love to see). But keep in mind that the FX lines doesnt have integrated graphics, so you need to buy some graphic card to have something comparable. Plus, the Intel chip has a lot more OC'ing potential because it has a much saner TDP. In terms of graphics, you need to compare with an amd APU, and those are a lot slower for CPU usage (even the i5's beat them)

          Cheers!

          Comment


          • #20
            Originally posted by Luke_Wolf View Post
            No the reason some applications go faster is not due to them not being prepared for multiple threads but due to the nature of SMT. Basically the way SMT works is that there are 2 (Can be more in non-Intel implementations) threads queued up in any particular core at any one time, and so when one thread stalls out or sleeps it switches over to the other thread, now this is all nice and fine for workloads that are heavily threaded but are not processor intensive, but it is detrimental in cases where the threads are processor intensive, because instead of just having to wait for it to finish what it's doing now it stalls out and the other thread steps in and it has to wait for that thread to stall out before it can start processing again, thus resulting in a loss of performance.

            It's important to point out that although the front end is currently a massive bottleneck (which will be fixed by steamroller) the Bulldozer architecture and derivs don't have this problem because instead of a processing queue they are doing processing on the threads. That said AMD does have a more interesting problem because of this in terms of how to handle scheduling.
            Wierd. I do scientific heavy threaded stuff and my testing showed 20% performance improvement on core i7 with HT enabled over disabled. And yes it is very cpu heavy and I burn up core i7 laptops (the RAM isn't throttled and causes the hard crash).

            I am curious how amd's new architecture might handle things but since we're so heavy on double precision I'm not confident we would get good performance. Our software running on core i7 runs circles cpu wise around the pre bulldozer opterons (dual socket sandy-e easily outruns quad socket opteron, about 2x faster on cpu heavy loads).

            I noticed on other reviews how the 3930k still outruns this new haswell on heavily threaded loads.

            Comment


            • #21
              Originally posted by chris200x9 View Post
              The one where an ~8 month old chip beat a 0 day old chip in some tests despite being a hundred and fifty dollars cheaper? Not to mention how many tests where only 20-40% faster when the next amd core is estimated at 30% faster per clock. Also unlike intel AMD is boosting clocks with each generation 5800k vs 8350 vs 6800k
              If you like heaters...
              Few years ago Intel was the one boosting clocks and we all know how it ended. High clock frequency does not necessarily mean high performance and definitely not efficiency.

              Comment


              • #22
                Originally posted by bnolsen View Post
                Wierd. I do scientific heavy threaded stuff and my testing showed 20% performance improvement on core i7 with HT enabled over disabled. And yes it is very cpu heavy and I burn up core i7 laptops (the RAM isn't throttled and causes the hard crash).

                I am curious how amd's new architecture might handle things but since we're so heavy on double precision I'm not confident we would get good performance. Our software running on core i7 runs circles cpu wise around the pre bulldozer opterons (dual socket sandy-e easily outruns quad socket opteron, about 2x faster on cpu heavy loads).

                I noticed on other reviews how the 3930k still outruns this new haswell on heavily threaded loads.
                Well, that's basically got to do with how your science software is programmed, because you have to understand that SMT is basically a bet on whether the CPU has more idle time or more active time while processing and betting on the idle time, and that if it's idle that it can slot in this other thread without effecting the performance of the thread that was running as all it really amounts to is a time-sharing trick. Which while that's true for most workloads it's not true for all workloads, gaming and virtualization are your more typical examples.

                Also if you're having hard crashes like that you might want to rethink the platforms you're using to compute this on, even if it's just remoting the workload out to a server and having it report back.
                Last edited by Luke_Wolf; 06-04-2013, 06:35 PM.

                Comment


                • #23
                  Originally posted by bnolsen View Post
                  Wierd. I do scientific heavy threaded stuff and my testing showed 20% performance improvement on core i7 with HT enabled over disabled
                  Anyway I also read that from the third generation the problem is not present with HT enabled. It would have been interesting to see in this benchmarks tests with HT enabled and disabled on a single thread applications.

                  Comment


                  • #24
                    I look forward to Steamroller coming out in a few months, never mind Excavator which will make Intel's products look like they've crapped the bed in all benchmarks.

                    Comment


                    • #25
                      Originally posted by mendieta View Post
                      Yes, AMD has come back considerably lately (which I love to see). But keep in mind that the FX lines doesnt have integrated graphics, so you need to buy some graphic card to have something comparable. Plus, the Intel chip has a lot more OC'ing potential because it has a much saner TDP. In terms of graphics, you need to compare with an amd APU, and those are a lot slower for CPU usage (even the i5's beat them)

                      Cheers!
                      Excavator ends that concern. It is the future of APU design that Intel will get thumped on.

                      Comment


                      • #26
                        Originally posted by Marc Driftmeyer View Post
                        I look forward to Steamroller coming out in a few months, never mind Excavator which will make Intel's products look like they've crapped the bed in all benchmarks.
                        I said the ssame thing about prior AMD products and had to eat shit later when it didnt happen. I've learned not to over estimate AMD. Barcelona and later Bulldozer were heavily hyped and neither met the expectations that were made.

                        Comment


                        • #27
                          Why is the FX8350 mysteriously missing from some of the benchmarks?
                          nevermind, I'm retarded.

                          Comment


                          • #28
                            Originally posted by Marc Driftmeyer View Post
                            Excavator ends that concern. It is the future of APU design that Intel will get thumped on.
                            Not sure about excavator, but yes, the focus is no longer CPU power, but thermal power. Most people have had powerful enough computers for a while. Now they want more portable, lighter, more silent, less bulky. That means boxed size PCs, Lappies, Tablets, even phones, heck. AMD had a great twist when buying AMD, in that sense. And they do have the upper hand, still. Not sure for how much longer. It still holds a lead (and it's still a much better value), but Intel is closing the gap.

                            Anyways, I do agree that the focus is the APU and lower power, rather than achieving gazillion MFlops at insane freqs with burning hot CPUs. Also, software is progressively parallelized, so we'll soon have chips with hundreds of cpu + gpu cores, each using a tiny amount of power, but collectively giving you a massive computational power.

                            There goes my crystal ball, my friends.

                            Comment


                            • #29
                              Originally posted by Luke_Wolf View Post
                              Well, that's basically got to do with how your science software is programmed, because you have to understand that SMT is basically a bet on whether the CPU has more idle time or more active time while processing and betting on the idle time, and that if it's idle that it can slot in this other thread without effecting the performance of the thread that was running as all it really amounts to is a time-sharing trick. Which while that's true for most workloads it's not true for all workloads, gaming and virtualization are your more typical examples.

                              Also if you're having hard crashes like that you might want to rethink the platforms you're using to compute this on, even if it's just remoting the workload out to a server and having it report back.
                              Usually desktops or servers. It's nice to have portables occasionally. Core i7 notebooks are pretty useless and worthless. That cpu should never see the inside of a laptop even in mobile form.

                              I typically "overstuff" the threads to make sure they stay pretty busy. There's always some small chance for something to go idle. And I haven't quantitatively seen anything more than a couple % difference when playing with 1.5x 2x and 3x on the overstuffing.
                              Last edited by bnolsen; 06-04-2013, 07:39 PM.

                              Comment


                              • #30
                                Originally posted by flecht View Post
                                If you like heaters...
                                Few years ago Intel was the one boosting clocks and we all know how it ended. High clock frequency does not necessarily mean high performance and definitely not efficiency.
                                Yes heat is going up like a rocket! /sarcasm

                                the a10-5800k is 3.8 ghz base with a TDP of 100w the a10-6800k is going to be 4.1 ghz base with a TDP of 100W

                                So how is heat going up from boosting clocks?

                                Seriously the 8350 is only 125w the 4770k is 84 if you can make a GOOD argument about why 41 watts is a HUGE difference on a high end desktop please let me know. (And please don't say it'll save money)

                                With leaving a PC on 24/7 the cost difference due to the CPU power draw is 31.25 a year, let's see 150 / 31.25 = 4.8. Congratulations your over priced intel crap will pay for itself in 4.8 YEARS!

                                Comment

                                Working...
                                X