Announcement

Collapse
No announcement yet.

AMD Ryzen 9 3900XT vs. Intel Core i9 10900K Linux Gaming Performance

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

  • AMD Ryzen 9 3900XT vs. Intel Core i9 10900K Linux Gaming Performance

    Phoronix: AMD Ryzen 9 3900XT vs. Intel Core i9 10900K Linux Gaming Performance

    Following the 130+ benchmarks of the AMD Ryzen 3000XT series earlier in the week looking at the CPU/system performance on Ubuntu Linux, here is our first look at the Linux gaming performance with putting the Ryzen 9 3900XT up head-to-head against the Intel Core i9 10900K.

    Phoronix, Linux Hardware Reviews, Linux hardware benchmarks, Linux server benchmarks, Linux benchmarking, Desktop Linux, Linux performance, Open Source graphics, Linux How To, Ubuntu benchmarks, Ubuntu hardware, Phoronix Test Suite

  • #2
    They are both $499 CPUs, but what about the motherboard price? That should be counted in.

    Comment


    • #3
      Not used to seeing AMD lead in performance/watt. Looking good with 105W for 12 cores that are clocked to 3.8Ghz or higher.

      I hope that AMD will improve their firmware. I have to break IOMMU to make the 3900XT work in my motherboard. If it wasn't for that I would have bought one already.

      Comment


      • #4
        Originally posted by lucrus View Post
        They are both $499 CPUs, but what about the motherboard price? That should be counted in.
        You should also consider the amortized cost over multiple years. AMD typically produces sockets that are compatible between 3 to 4 CPU generations. Intel usually pairs one CPU gen with a single mobo gen. Even if the parts were be physically compatible, they would cripple the firmware to make them incompatible. After few years of use, a large part of Intel performance is also lost thanks to all the side channel exploit mitigations.

        Comment


        • #5
          The Noctua NH-U9S looks like a inadequate cooler for the likes of the 3900XT and 10900K, especially for review purposes. Other reviewers usually apply water cooler, or at least a beefy tower cooler with a 120mm fan at least, on CPU's above 100W of TDP.

          Comment


          • #6
            IPC Matters

            In the simplest form, the time it takes to run a task depends on three things:

            1. Instruction count for the task
            2. Clock speed
            3. Instruction per cycle (instruction throughput), or IPC

            There's a lot of numbers that people talk about, but those oftentimes make a lot less impact than IPC. Why don't people talk about IPC? Well, because it changes (sometimes dramatically) between workloads; it is not a spec. A number crunching workload is going to behave quite differently from a game, who in turn will behave quite differently from a database.

            Unlike nanometers (which have very little meaning these days) or memory frequency (which matters a lot less than you'd think because processors use a complex cache hierarchy), the IPC is not a spec. It's way easier to bitch about nm than to understand IPC.

            So let me say it again: IPC matters

            Comment


            • #7
              Originally posted by vladpetric View Post
              IPC Matters

              In the simplest form, the time it takes to run a task depends on three things:

              1. Instruction count for the task
              2. Clock speed
              3. Instruction per cycle (instruction throughput), or IPC

              There's a lot of numbers that people talk about, but those oftentimes make a lot less impact than IPC. Why don't people talk about IPC? Well, because it changes (sometimes dramatically) between workloads; it is not a spec. A number crunching workload is going to behave quite differently from a game, who in turn will behave quite differently from a database.

              Unlike nanometers (which have very little meaning these days) or memory frequency (which matters a lot less than you'd think because processors use a complex cache hierarchy), the IPC is not a spec. It's way easier to bitch about nm than to understand IPC.

              So let me say it again: IPC matters
              Beside the captain-obvious statement, was IPC importance contended?
              Or am I missing something?

              Comment


              • #8
                Originally posted by milkylainen View Post

                Beside the captain-obvious statement, was IPC importance contended?
                Or am I missing something?
                No. This is not directed at either Michael or you, who understand what it is. More like nm AMD fanboys.

                Comment


                • #9
                  Originally posted by vladpetric View Post

                  So let me say it again: IPC matters
                  Sure it does, similar to like max clock rate for badly parallelized code.

                  Comment


                  • #10
                    Originally posted by oleid View Post

                    Sure it does, similar to like max clock rate for badly parallelized code.
                    Well, when the binary code is the same, it is the microarchitecture of the processor that mostly affects IPC. And in this respect Intel seems to be doing better.

                    Comment

                    Working...
                    X