2 Replies Latest reply on Oct 30, 2015 7:49 PM by aaronamduser

    Why is Carrizo much slower than Richland?

    aaronamduser

      I have the AMD FX 8800P in an HP Envy and it is over 10 times slower (5 minutes vs 20 seconds) at completing an identical task in pymol compared to my older AMD A8 5550M APU laptop also an HP Enzy.  I am wondering if there is any reason for this and is AMD doing something to fix it.  For comparison the Richland 5550M completes the task in about the same time as an Intel Xeon E5 CPU.  It's frustrating owning a laptop that was advertised as a power house only for it to be so slow it's almost unusable.    I should also point out that the task in question is fully placed on the CPU and the GPU is completely removed the equation.  Finally the task uses two cores and the Richland and Carrizo show max CPU clocks on two cores.

        • Re: Why is Carrizo much slower than Richland?
          warpfact0r10

          Have you disabled power savings options in Windows and do you have the AC power adapter plugged in when running the task?

           

          If yes to both, it very well may be a question for HP. As Carrizo has a configurable TDP, HP may have simply decided to cap it at 15 watts. I recall others here expressing similar disappointment with HP's treatment of Carrizo, so it wouldn't surprise me if this was the case.

            • Re: Why is Carrizo much slower than Richland?
              aaronamduser

              I was running these tasks with the power option set to High Performance with the AC adapter plugged in.  This laptop is running on 15W, but I just am having a hard time understanding how a 35 W chip running with two cores at max clock speeds (3.1 GHz) is over 10 x faster than a 15W running at close to its max clock speed (~3.2 GHz).  I am no computer expert but I would think AMD chips would be comparable at same clock speeds.  I can also follow up with HP and see if they plan on issuing a BIOS update that allows the chip to use up to 35W, I would assume the power supply is high enough to allow this, but again I could be wrong.