Power Consumption

For our power consumption readings, we run a Prime 95 load and slowly ramp up the number of threads in play, taking power data from the internal CPU registers that report for when turbo modes or thermal modes should activate. Depending on the CPU access, we can get data that varies from the full package down to individual cores, uncore, integrated graphics and DRAM controllers.

For the Ryzen CPUs, the API pulls out the total package power consumption first.

 Power: Total Package (Idle) Power: Total Package (1T)Power: Total Package (Full Load)

At idle, all the CPUs are pretty much equivalent. The cores are fully idle here, leaving the rest of the chip active enough for tick-over. As we ramp up the load, the higher-frequency Ryzen CPUs move towards their 65W TDP, with the Ryzen 3 1300X almost being spot on at 64.2W. The Intel CPUs are clocked higher, but only have two cores to contend with.  The Ryzen 3 1200 is clocked lower than the Ryzen 3 1300X, hitting a better efficiency point in the Zen design. This ultimately bodes well for upcoming quad-core SKUs in laptops.

Power: Cores Only (1T Load)Power: Cores Only (Full Load)

One of the odd things about the power consumption of the Ryzen 3 CPUs is the difference between how much power the cores internally measure compared to the full power consumption of the package measured as a whole, including the Infinity Fabric, DRAM controllers, IO and such. For the Ryzen 3 1200 for example, at full load the package has a power consumption of 40.43 W total, but the cores only count for 23.05W, leaving 17.38W on the table for the non-core elements in the chip. If we compare that to the Ryzen 5 1500X, we have 68.79W for the package and 49.69W for the cores, a 19.1W difference. For the Ryzen 7 1700X, it becomes 81.51W for the package and 62.10W for the cores, a 19.4W difference.

On the high-end chips, the difference is a smaller portion of the full power consumption, but on the Ryzen 3 processors the cores are only 57% of the power consumption, leaving 43% for the rest of the chip. A lot of this power could be the inter-CCX infinity fabric path, which means that packages like EPYC are giving away a lot of power to IF. If it is more other features, it could spell a number of problems for upcoming mobile chips, limiting the lower bounds of the power consumption. Naturally, I actually want to get my hands on to an EPYC processor in my own lab to see if we can pin down what is happening here.

Gaming Performance: Grand Theft Auto (1080p, 4K) Conclusion and Performance Per Dollar Graphs
Comments Locked

140 Comments

View All Comments

  • CSM 2 - Friday, July 28, 2017 - link

    1200 with A320 + GTX 1050ti perfect family pc or sff media pc I'm very impressed. Thanks AMD for some competition
  • slickr - Sunday, July 30, 2017 - link

    They actually are releasing their APU's at the same time, so if you want cpu+igp you can buy their new APU's.
  • MajGenRelativity - Thursday, July 27, 2017 - link

    Could you add an overclocking section? I've been seeing the 1200 overclock well, and that could definitely close the gap with the i3's
  • Ian Cutress - Thursday, July 27, 2017 - link

    In the plans, maybe Pt 2 next week after finishing the 1300X.
  • T1beriu - Thursday, July 27, 2017 - link

    Yey!
  • AndrewJacksonZA - Thursday, July 27, 2017 - link

    "M." You're missing an "m" at the end of your username, T1beriu. ;-)
  • MajGenRelativity - Thursday, July 27, 2017 - link

    Awesome! I think overclocking is one of Ryzen's strong points, and it doesn't seem to drive up power consumption too much, because you can hit good clocks on stock voltage
  • Oxford Guy - Thursday, July 27, 2017 - link

    Or, maybe we could see what happens without being kneecapped by the absurd JEDEC RAM setting.
  • venkatsrin - Thursday, July 27, 2017 - link

    Thanks, can only second this request! Pls, pretty pls, do include the overclocked benchmarks for gaming not only with high-end cards like GTX1080 but also the RX470 that you have in there currently.
  • QinX - Thursday, July 27, 2017 - link

    The Blender graph seems to be wrong for the 1300X, beating the 1500x by 500 seconds seems excessive.

Log in

Don't have an account? Sign up now