Performance Per Watt isn't really that important to Joe Consumer's desktop PC. It's much more important for mobile/battery-powered devices (smartphones, tablets, notebook computers) and people whose operations include a substantial electricity bill as a major part of their recurring costs.
The latter why ARM CPUs are making great strides in datacenter/HPC environments. It also explains why cryptocurrency miners favor chips that aren't x86.
For the average PC gamer, it really doesn't matter if one desktop PC's CPU uses $5 more electricity per month than a less powerful CPU.
As for benchmarks, it's really shortsighted to focus on synthetic benchmarks or individual statistical measurements like GPU TFLOPs. There's much more to the overall gaming performance than just the CPU and GPU.
When I started my first builds last summer, I decided to go with AMD despite that Intel CPUs historically had better single-core performance. AMD has crushed Intel for years in multi-threaded performance but most games don't take advantage of that.
I went with AMD because PCIe 4.0 was supported by the CPU and the newest motherboard chipsets. Most importantly this meant blazingly fast SSD performance. Even today, there is no consumer-grade Intel-based motherboard that supports PCIe 4.0. They are coming in the near future with Rocket Lake.
All three of my motherboards have two m.2 PCIe 4.0 Gen 4 slots apiece. With AMD's Zen 3 CPUs, their single-core IPC performance now crushes anything Intel currently has on the market. Whether or not Intel can leapfrog past Zen 3 with Rocket Lake remains to be seen.
I'm populating my three builds with Sabrent Rocket m.2 drives as boot drives and relying on less expensive m.2 drives for my media libraries (mostly games) where read performance is far more important than write performance. If I need more storage space, I'll add a SATA SSD but that's usually for stuff like older/smaller games and video.
A lot of people who build gaming PCs seem to fixate on one or two measurements and don't have the ability to see the big picture. Someone will rattle off how great Radeon GPUs perform in raw rasterization but ignore features like ray traced reflections.
The latter why ARM CPUs are making great strides in datacenter/HPC environments. It also explains why cryptocurrency miners favor chips that aren't x86.
For the average PC gamer, it really doesn't matter if one desktop PC's CPU uses $5 more electricity per month than a less powerful CPU.
As for benchmarks, it's really shortsighted to focus on synthetic benchmarks or individual statistical measurements like GPU TFLOPs. There's much more to the overall gaming performance than just the CPU and GPU.
When I started my first builds last summer, I decided to go with AMD despite that Intel CPUs historically had better single-core performance. AMD has crushed Intel for years in multi-threaded performance but most games don't take advantage of that.
I went with AMD because PCIe 4.0 was supported by the CPU and the newest motherboard chipsets. Most importantly this meant blazingly fast SSD performance. Even today, there is no consumer-grade Intel-based motherboard that supports PCIe 4.0. They are coming in the near future with Rocket Lake.
All three of my motherboards have two m.2 PCIe 4.0 Gen 4 slots apiece. With AMD's Zen 3 CPUs, their single-core IPC performance now crushes anything Intel currently has on the market. Whether or not Intel can leapfrog past Zen 3 with Rocket Lake remains to be seen.
I'm populating my three builds with Sabrent Rocket m.2 drives as boot drives and relying on less expensive m.2 drives for my media libraries (mostly games) where read performance is far more important than write performance. If I need more storage space, I'll add a SATA SSD but that's usually for stuff like older/smaller games and video.
A lot of people who build gaming PCs seem to fixate on one or two measurements and don't have the ability to see the big picture. Someone will rattle off how great Radeon GPUs perform in raw rasterization but ignore features like ray traced reflections.
Last edited: