I mean, that’s fair, but the 285k has a TDP of 250 watts compared to 120. In practice, I think that people gaming or doing workstation tasks would use more power on the 285k, because typically, that’s what these will be used for anyway.
For comparison’s sake, the EPYC 7702 is a 64 core (hyperthreaded) server CPU, and it has a lower TDP. The 285k’s TDP is absolutely bonkers.
In practice, I think that people gaming or doing workstation tasks would use more power on the 285k, because typically, that’s what these will be used for anyway.
People doing “workstation” tasks or gaming with desktop CPUs also spend a lot of time in editor tools or basic desktop tasks, and seldom turn off their systems in between tasks. (Some don’t even turn off their systems when they leave work or go to sleep at night.) And very few games will load up all the CPU cores at once. So idle power draw remains significant. Also, many/most real-world tasks have CPU cores constantly varying their power draw, not pegged at the package’s rated TDP. So using TDP as a proxy for overall power usage is unrealistic.
It might be interesting to study the proportion of powered-on time spent at low load vs. high load in a typical week, across a decently large sample size.
I mean, that’s fair, but the 285k has a TDP of 250 watts compared to 120. In practice, I think that people gaming or doing workstation tasks would use more power on the 285k, because typically, that’s what these will be used for anyway.
For comparison’s sake, the EPYC 7702 is a 64 core (hyperthreaded) server CPU, and it has a lower TDP. The 285k’s TDP is absolutely bonkers.
People doing “workstation” tasks or gaming with desktop CPUs also spend a lot of time in editor tools or basic desktop tasks, and seldom turn off their systems in between tasks. (Some don’t even turn off their systems when they leave work or go to sleep at night.) And very few games will load up all the CPU cores at once. So idle power draw remains significant. Also, many/most real-world tasks have CPU cores constantly varying their power draw, not pegged at the package’s rated TDP. So using TDP as a proxy for overall power usage is unrealistic.
It might be interesting to study the proportion of powered-on time spent at low load vs. high load in a typical week, across a decently large sample size.