Announcement

Collapse
No announcement yet.

DDR3-800MHz To DDR3-2133MHz Memory Testing With AMD's Kaveri

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

  • #16
    Originally posted by UraniumDeer View Post
    It's pretty interresting how many of the performance results scales about 1:1 with the increase of memory speed.
    ...
    I did a few calculations on the first test, and while it doesn't remain 1:1, but drops more and more the closer you get to 2133Mhz, the numbers seems rather predictive. I think you could make a pretty accurate mathematical projection of the performance effect of an increase in memory speed, if one was so inclined.



    Next test speed / current test speed * current framerate result = rough estimate of next test
    actual result of next test / rough estimate of next test * 100 = actual performance compared to rough estimate of next test in percentage
    1066/800*17,53=23,36
    23,59/23,36*100=100,1%
    1333/1066*23,59=29,50
    29,19/29,50*100=98,95%
    1600/1333*29,19=35,04
    34,29/35,04*100=97,86%
    1866/1600*34,29=39,99
    39,18/39,99*100=97,97%
    2133/1866*39,18=44,77
    42,50/44,77*100=94,93%

    Comment


    • #17
      Originally posted by Article
      The timings on the system memory were maintained the same at all frequencies so it's solely a comparison of the memory bandwidth as it impacts the A10-7850K Kaveri APU at its stock frequencies.
      You mean you kept the timing artificialy high for the lower frequencies instead of using the profiles? That makes the benchmarks (especialy the CPU ones that care more for latency than bandwidth) complete bulshit.

      Comment


      • #18
        Originally posted by TemplarGR View Post
        The truth is, Intels are far better in fpu performance and their memory controller is more efficient. They also consume less power.

        AMD is just cheaper/price competitive on low end and with a better igpu, though it is quite bandwidth starved and you don't get to see its real performance.

        Also HSA and Mantle are vaporware, especially on Linux. By the time AMD brings the software that exploits them, Carizzo or its succesor will be available...
        AMD's Bulldozer CPUs have half the number of FPU per core, so of course that Intel is twice better in floating point operations than AMD.
        Carizzo will also support HSA and Mantle...

        Comment


        • #19
          Originally posted by tomtomme View Post
          learn to read the article before posting in the forum ... please! This is going to many forum-posters out there - not only you AJSB...
          I usually do but was in a hurry going to do some work, so, i simply did a quick look to the charts and didn't read the text

          Comment


          • #20
            if they need short ways for the gddr memory they could place the Memory on the backside of the mainboard. Cases would have to adapt different possibilies of removing memory and/or heat dissipation. You could for instance make some jelly-like pillows on the RAM modules to push against the case without having electrical contact to use the case as a massive heatsink. there are ways. and i think that AMD should make some innovative and risky jumps in order not to lose completely to Intel.

            Comment


            • #21
              Originally posted by Слободан View Post
              AMD's Bulldozer CPUs have half the number of FPU per core, so of course that Intel is twice better in floating point operations than AMD.
              Carizzo will also support HSA and Mantle...
              Anything that uses OpenCL already supports HSA. If you have Catalyst drivers with a Kaveri APU, you should already be getting HSA benefits in OpenCL programs.

              Comment


              • #22
                Originally posted by TemplarGR View Post
                The truth is, Intels are far better in fpu performance and their memory controller is more efficient. They also consume less power.

                AMD is just cheaper/price competitive on low end and with a better igpu, though it is quite bandwidth starved and you don't get to see its real performance.

                Also HSA and Mantle are vaporware, especially on Linux. By the time AMD brings the software that exploits them, Carizzo or its succesor will be available...
                AMD cpus (modules) are a bit FPU choked by design (double int single fpu), now the actual problems are memory controllers about a gen lower ipc and performance per watt. However your foolish to think AMD cpu's are just cheaper. My 8350 will stomp i7's in some bench's especially when optimized for BD arch. Does it mean it's a better CPU? Not really but I got it for reasons that make it way more appealing than Intels offerings anywhere near the price range of it. Kaveri will beat an i7 with onboard graphics, does that make it a better chip? Not really, unless that's the reason you're buying the fucking thing especially being 100US less.

                Comment


                • #23
                  @mmstick: Yes, I'm comparing apples to oranges. But the fx-8350 is quite less performant than i5-4670 (here).

                  @TemplarGR: Yes, good point. However, the difference is 70 euro. Some benchmarks indicate significant differences. I'm planning on using this new computer for as long as possible. The last one lasted 7 years.

                  @nightmarex: Yes, this benchmark has exposed some limitations. However, I would like to purchase a new CPU/Mobo/Memory *now*. I don't really care about the improvements later on. However, if AMD releases a new FX series this year, I could wait. However, I have read this is probably not going to happen anymore.

                  Finally, I don't understand the skeptisism regarding to benchmark sites. (What about Phoronix?). It seems, to me, that clock speed has lost it's value as a reliable speed indicator. I prefer to look at benchmarks and CPU extensions.

                  If I were to look at mHz value's the choice would have been an AMD FX-8350 (8 'questionable'(?) CPU cores) + 1866 mhz memory. However, this setup does not seem perform that well as the clockspeed would indicate. You see this on Phoronix and on other benchmark sites.

                  Last but not least, a fair competitor of the i7 4770 would be the fx-8350 according to nightmarex (cpu-world) says the i7 is significantly faster. Especially in single threaded performance.

                  Comment


                  • #24
                    There's quite a power hit to GDDR5 compared to DDR3. One stick of DDR3 consumes on the order of 1W, while a GDDR5 chip of the same capacity would take 10x more, ~10W. At those wattages you need active cooling, while ddr3 chips don't even have heatsinks usually.

                    Then there's the price difference too. I guess that shows why no embedded gddr5 yet.

                    Comment


                    • #25
                      Interesting data! Thanks for running the specs. I was actually just recently looking for this exact kind of research, except for Intel integrated graphics. I was curious how much of an effect there is as the eventual release of Skylake is supposed to bring DDR4 to mainstream hardware.

                      Comment


                      • #26
                        Originally posted by mmstick View Post
                        Anything that uses OpenCL already supports HSA. If you have Catalyst drivers with a Kaveri APU, you should already be getting HSA benefits in OpenCL programs.
                        Then how do you explain this?

                        Comment


                        • #27
                          @Rexillion

                          The only real difference between Core i5 4670 and Core i7 4770k is Core i5 doesn't have hyper-threading. IIRC the i7 also has a littlie more l3 cache, but that won't affect regular desktop performance that much.

                          I don't believe you will need more than 4 threads in the near future. And even if you need those, H/T will not make that much of a difference, since those aren't real 8 threads...

                          Also, and that advise is for everyone, you should never buy a pc system to last for many years. It is better to buy a cheaper system now, and buy again in 3-4 years.

                          It is not only performance you need, it is compliance with recent standards for cpus and gpus, drivers etc...

                          Also you get to have 2 systems for the cost of 1...

                          Depending on the things you do with your computer, you might not need such an expensive cpu. I am a professional developer and even i don't need a high end cpu... Don't be fooled by marketing. Buy what you need for your needs right now, do not spend more than necessary, and plan only for 3-4 years, then make a new system.

                          Frequent cheap upgrades are far better than an expensive system every 7 years...

                          Comment


                          • #28
                            The cost of ram is so high compared to 2011 :/, you can buy a low end nvidia like gt 640.

                            Comment


                            • #29
                              Originally posted by curaga View Post
                              There's quite a power hit to GDDR5 compared to DDR3. One stick of DDR3 consumes on the order of 1W, while a GDDR5 chip of the same capacity would take 10x more, ~10W. At those wattages you need active cooling, while ddr3 chips don't even have heatsinks usually.

                              Then there's the price difference too. I guess that shows why no embedded gddr5 yet.
                              Indeed, all negative points for GDDR5...and i don't like ANY embedded memory solution.

                              I would prefer at least TRIPLE Channel DDR3 in a mITX MoBo...not completely impossible to do so:

                              Just needs a OEM with b***s to remove the PCIeX16 connector, remove speaker connector , reduce SATA connectors to 2-3,etc. and you get the PCB space for one more channel.

                              That board would also be a "technological statment" : We don't need a freaking dGPU.
                              Last edited by AJSB; 01-21-2014, 07:52 AM.

                              Comment


                              • #30
                                Originally posted by Слободан View Post
                                Then how do you explain this?
                                <graph showing SW vs OpenCL vs HSA>
                                I don't think OpenCL automatically uses the HSA paths by default yet, pending further testing & tuning.

                                Comment

                                Working...
                                X