Announcement

Collapse
No announcement yet.

AMD's New Athlon/Semprons Give Old Phenom CPUs A Big Run For The Money

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

  • AMD's New Athlon/Semprons Give Old Phenom CPUs A Big Run For The Money

    Phoronix: AMD's New Athlon/Semprons Give Old Phenom CPUs A Big Run For The Money

    Curious how AMD's new AM1 platform APUs compare to the original AMD Phenom processors? Wondering myself, I ran some tests showing how the Sempron 2650 and 3850 along with the Athlon 5150 and 5350 compare to the original Phenom 9500 and Phenom II X3 710 processors with RS780/RS880 motherboards. Besides the new APUs being competitive against the old hardware while costing much less than the original Phenom CPUs, their power consumption is also at a fraction of AMD's former high-end processors. Here's a brief but nice look at AMD's processing evolution in going from Phenom CPUs to today's AMD budget APUs.

    http://www.phoronix.com/vr.php?view=20234

  • #2
    Let's see:

    AMD Phenom X4 9500:
    Frequency: 2200MHz
    Cores: 4
    Cache: 4MB
    Introduction date: Nov 19, 2007

    AMD Athlon 5350:
    CPU Frequency: 2050MHz
    Cores: 4
    Cache: 2MB
    Introduction date: March, 2014

    In most tests the old Phenom is a lot faster than the new Athlon (sometimes twice as fast, e.g. in John the Ripper:md5), i.e. AMD's CPUs IPC hasn't really improved for seven years. Meanwhile Intel's CPUs sport almost 50% improvement in IPC.

    At least TDP has improved considerably (25W vs 65W), but that's not a relief when competitor's CPUs run 30-100% faster depending on tests.

    Comment


    • #3
      Originally posted by birdie View Post
      Let's see:

      AMD Phenom X4 9500:
      Frequency: 2200MHz
      Cores: 4
      Cache: 4MB
      Introduction date: Nov 19, 2007

      AMD Athlon 5350:
      CPU Frequency: 2050MHz
      Cores: 4
      Cache: 2MB
      Introduction date: March, 2014

      In most tests the old Phenom is a lot faster than the new Athlon (sometimes twice as fast, e.g. in John the Ripper:md5), i.e. AMD's CPUs IPC hasn't really improved for seven years. Meanwhile Intel's CPUs sport almost 50% improvement in IPC.

      At least TDP has improved considerably (25W vs 65W), but that's not a relief when competitor's CPUs run 30-100% faster depending on tests.
      Phenom's were $250 when they came out. Athlon's are $50.

      Compare the Phenom vs Kaveri 7850K if you want an actual "similar-price-point" comparison, and even then the Phenom was $70 more at launch than the 7850k is.

      Comment


      • #4
        Originally posted by birdie View Post
        Let's see:

        AMD Phenom X4 9500:
        Frequency: 2200MHz
        Cores: 4
        Cache: 4MB
        Introduction date: Nov 19, 2007

        AMD Athlon 5350:
        CPU Frequency: 2050MHz
        Cores: 4
        Cache: 2MB
        Introduction date: March, 2014

        In most tests the old Phenom is a lot faster than the new Athlon (sometimes twice as fast, e.g. in John the Ripper:md5), i.e. AMD's CPUs IPC hasn't really improved for seven years.
        I think you should have spent more time at school doing maths, the difference ids less than 50% which is probably predominantly a memory bandwidth issue.

        If you want to cheery pick benches make a point there are other benches where the Athlon is around 3 times faster than the Phenom such as Unvanquished.

        Comment


        • #5
          Thanks Michael, a very interesting article. Too bad you didn't have any x6s around anymore.

          Comment


          • #6
            Still would love to see those cheap A-series chips pitted against Athlon! You can score an A6-6400k for a mere 65 bucks!

            Comment


            • #7
              Interesting. I have a Phenom II X4, and at the moment it's in an HTPC (overpowered, I know), but in the future I plan to use that in a self-hosting server. Good to know it's still halfway decent at the job of actually computing things.

              Comment


              • #8
                Also note that the 25W of the 5350 includes CPU, GPU, and controllers, while the phenom only the CPU. Phenom GPU and system controllers are on north/south bridges and require additional power.

                Comment


                • #9
                  Why on Earth were the graphics/games benchmarks included? Anyone still using these CPUs (and caring about GPU speed) would almost certainly have a separate graphics card, not relying on the GPU in the chipset.

                  Of real interest with this comparison was CPU integer and floating point performance, and memory speed (including caches). The graphics benchmarks were pointless.

                  Comment


                  • #10
                    Originally posted by birdie View Post
                    Let's see:

                    AMD Phenom X4 9500:
                    Frequency: 2200MHz
                    Cores: 4
                    Cache: 4MB
                    Introduction date: Nov 19, 2007

                    AMD Athlon 5350:
                    CPU Frequency: 2050MHz
                    Cores: 4
                    Cache: 2MB
                    Introduction date: March, 2014

                    In most tests the old Phenom is a lot faster than the new Athlon (sometimes twice as fast, e.g. in John the Ripper:md5), i.e. AMD's CPUs IPC hasn't really improved for seven years. Meanwhile Intel's CPUs sport almost 50% improvement in IPC.

                    At least TDP has improved considerably (25W vs 65W), but that's not a relief when competitor's CPUs run 30-100% faster depending on tests.
                    Ah yes because making comments on AMD's IPC by comparing their Netbook Processors to their Desktop Processors, and pretending they're the desktop processors makes so much sense. Here's a hint for you, if you want to compare modern AMD IPC look at the Kaveri APUs or the Piledriver FX series, this article is to see where the new/cheap netbook processors line up compared to the particularly old Phenom processors

                    Comment


                    • #11
                      Originally posted by Imroy View Post
                      Why on Earth were the graphics/games benchmarks included? Anyone still using these CPUs (and caring about GPU speed) would almost certainly have a separate graphics card, not relying on the GPU in the chipset.

                      Of real interest with this comparison was CPU integer and floating point performance, and memory speed (including caches). The graphics benchmarks were pointless.
                      +1 graphics benchmarks are pointless in this case.

                      And yeah, while I agree that they are cheaper, low power etc etc, a part of me just dies when I look at the performance of these chips.
                      I have a $40 celeron from 2009 that can beat them on stock speeds. When you add a 1ghz overclock the APU is left coughing dust.
                      There's reasonable performance, and then there's "I'm still slower than a core2duo".

                      I wish these low-power designs took a page or two out of ARM's playbook and included one or two bigger cores. 1 big core + 3 of the current cores would be a hell of an improvement.

                      Comment


                      • #12
                        I've been running a server with a Phenom II 965 X4 for the last few years. I keep it pretty up to date, new kernel every six months or so to which I have ported the K8 PHC cpufreq driver. (I merged the last PHC release with the changes to the in-kernel powernow-k8 and acpi drivers, and created a new combined driver phc-k8 driver.)

                        I can attest that at least for this particular "Black Edition" CPU, the PHC potential was Phenomenal! 0.7v at idle/800Mhz and 1.3v at full load @3.4GHz. Not so bad in a HTPC with those specs, and makes for a nice quiet server with some grunt when needed. I wonder how the new CPUs really compare like-for-like, TDP can really be worst case, lowest common denominator. A typical CPU can really be much better than the official rating.

                        Comment


                        • #13
                          Originally posted by s_j_newbury View Post
                          I've been running a server with a Phenom II 965 X4 for the last few years. I keep it pretty up to date, new kernel every six months or so to which I have ported the K8 PHC cpufreq driver. (I merged the last PHC release with the changes to the in-kernel powernow-k8 and acpi drivers, and created a new combined driver phc-k8 driver.)

                          I can attest that at least for this particular "Black Edition" CPU, the PHC potential was Phenomenal! 0.7v at idle/800Mhz and 1.3v at full load @3.4GHz. Not so bad in a HTPC with those specs, and makes for a nice quiet server with some grunt when needed. I wonder how the new CPUs really compare like-for-like, TDP can really be worst case, lowest common denominator. A typical CPU can really be much better than the official rating.
                          That's pretty cool, I didn't know about PHC. Although in my case I would need to underclock the graphics card first and foremost, as it's the one generating most of the heat (and thus noise).

                          Comment


                          • #14
                            Originally posted by s_j_newbury View Post
                            I've been running a server with a Phenom II 965 X4 for the last few years. I keep it pretty up to date, new kernel every six months or so to which I have ported the K8 PHC cpufreq driver. (I merged the last PHC release with the changes to the in-kernel powernow-k8 and acpi drivers, and created a new combined driver phc-k8 driver.)
                            FYI, you don't need to do that at all. K10 CPUs implement p-state switching in hardware. The responsible model-specific registers can easily be reprogrammed from userspace, and the kernel cpufreq governor will still work as before. Tools like TurionPowerControl or k10ctl allow you to do it.

                            Comment


                            • #15
                              The problem with the new CPU's is that they use an entirely new socket. One that isn't AMD3+ or FM2+. What's the point of getting a CPU that doesn't even outperform your existing CPU? Spend money to lower power consumption? While a AMD A4-4000 is $60 and offers future upgrades. Or spend a bit more and get a AMD FX 4300.

                              If I wanted laptop performance, I would use a laptop.

                              Comment

                              Working...
                              X