4 ms·

Mostly agree, but AMD's TDP estimates are slightly optimistic and in reality Ryzen's power usage is much closer to its Intel counterparts.

by yohui 9y ago

loeg 9y ago
Do you have a source for that?
- paulmd 9y ago
  https://www.pcper.com/reviews/Processors/AMD-Ryzen-7-1800X-Review-Now-and-Zen/Power-Consumption-and-Conclusions https://www.pcper.com/reviews/Processors/AMD-Ryzen-7-1800X-R... https://www.bit-tech.net/hardware/2017/04/06/amd-ryzen-7-1700x-review/6 https://www.bit-tech.net/hardware/2017/04/06/amd-ryzen-7-170... http://www.tweaktown.com/reviews/8072/amd-ryzen-7-1800x-cpu-review-intel-battle-ready/index12.html http://www.tweaktown.com/reviews/8072/amd-ryzen-7-1800x-cpu-... http://hothardware.com/reviews/amd-ryzen-7-1800x-1700x-1700-benchmarks-and-review?page=10 http://hothardware.com/reviews/amd-ryzen-7-1800x-1700x-1700-... (bear in mind the 5820K/6800K/6850K are 6C, the 5960X/6900K are 8C, and the 6950X is 10C) Essentially AMD is rating their TDP under a very light load - like a single thread at full boost clocks. Under an all-core load they are running roughly 30% over TDP. Intel's TDP rating is much more generous - they aren't quite rating at full, nonstop AVX load (eg Prime95) but typical full-load consumption is 15% under their TDPs which leaves some headroom for intermittent AVX usage. Anecdotal evidence: my 10C20T Haswell-E Xeon pulls ~60% of its rated TDP during a Handbrake x264 encoding run. My 6C12T 5820K Haswell-E gaming chip (overclocked to 4.13 GHz all-core at stock voltages) pulls 75% of its rated TDP during a similar Handbrake x264 run. And supposedly x264 is smart enough to use AVX, so that should be a realistic-but-high figure. My OC'd 5820K pulls 90% of its rated TDP during a Prime95 SmallFFT run. It's still a huge gain compared to Bulldozer - but not quite as rosy a picture as AMD is trying to paint. They are equal with Haswell-E/Broadwell-E in most respects, not 40% ahead like AMD is rating them. AMD is playing the same game in the GPU market too. AMD's RX 480/580 are rated for "GPU only" TDP (which they define to exclude memory/etc) but NVIDIA rates the whole card. In practice, AMD is ~30% above their official TDPs while NVIDIA is pretty much spot on. https://www.techpowerup.com/reviews/Sapphire/RX_580_Nitro_Plus/28.html https://www.techpowerup.com/reviews/Sapphire/RX_580_Nitro_Pl... (RX 580 is rated at 185W TDP, GTX 1060 is rated at 120W TDP)
  - loeg 9y ago
    Thanks!
  - dom0 9y ago
    TDP is and always was an expressly manufacturer- and product-specific number, though. It mainly exists for OEMs to match parts up with the correctly sized heatsink.
    - paulmd 9y ago
      That argument stops holding water when the manufacturer starts marketing around the superior power efficiency of their processors based on that same number.
- yohui 9y ago
  paulmd already put up a list, so I won't repeat it. Ars Technica sums it up: https://arstechnica.com/gadgets/2017/03/amd-ryzen-review/ https://arstechnica.com/gadgets/2017/03/amd-ryzen-review/ > Indeed, Ryzen's 95W TDP looks impressive compared to the 140W TDP of Intel's 6900K, a similar 8C/16T processor based on the Broadwell-E architecture. But in practice, Ryzen pulls just as much wattage from the wall as the 6900K, in some cases slightly more. Considering just how power-hungry Bulldozer was, that remains a significant achievement, even if it means that overclocking headroom isn't as high as some might have hoped.
leonlee 9y ago
Unfortunately AMD and Intel don't use the same definition for TDP (look up AMD APC). AMD's values are more related to average power consumption during whatever they consider to be typical use. Intel's values on the other hand are more about the upper limit of power usage, "how high can it go" sort of thing. Both the average power and maximum power are useful bits of info, but it's obviously confusing and opaque for the end-user. The only reason that I can think of why AMD hasn't reverted back is because it would highlight how their processors are relatively power inefficient compared to their competitor's. http://www.amd.com/Documents/43761A_ACP_WPv7.pdf http://www.amd.com/Documents/43761A_ACP_WPv7.pdf
- mtgx 9y ago
  Didn't Intel try to do the same thing with SDP?

Mostly agree, but AMD's TDP estimates are slightly optimistic and in reality Ryzen's power usage is much closer to its Intel counterparts.

by yohui 9y ago

Mostly agree, but AMD's TDP estimates are slightly optimistic and in reality Ryzen's power usage is much closer to its Intel counterparts.

loeg 9y ago

Do you have a source for that?

paulmd 9y ago
https://www.pcper.com/reviews/Processors/AMD-Ryzen-7-1800X-Review-Now-and-Zen/Power-Consumption-and-Conclusions https://www.pcper.com/reviews/Processors/AMD-Ryzen-7-1800X-R... https://www.bit-tech.net/hardware/2017/04/06/amd-ryzen-7-1700x-review/6 https://www.bit-tech.net/hardware/2017/04/06/amd-ryzen-7-170... http://www.tweaktown.com/reviews/8072/amd-ryzen-7-1800x-cpu-review-intel-battle-ready/index12.html http://www.tweaktown.com/reviews/8072/amd-ryzen-7-1800x-cpu-... http://hothardware.com/reviews/amd-ryzen-7-1800x-1700x-1700-benchmarks-and-review?page=10 http://hothardware.com/reviews/amd-ryzen-7-1800x-1700x-1700-... (bear in mind the 5820K/6800K/6850K are 6C, the 5960X/6900K are 8C, and the 6950X is 10C) Essentially AMD is rating their TDP under a very light load - like a single thread at full boost clocks. Under an all-core load they are running roughly 30% over TDP. Intel's TDP rating is much more generous - they aren't quite rating at full, nonstop AVX load (eg Prime95) but typical full-load consumption is 15% under their TDPs which leaves some headroom for intermittent AVX usage. Anecdotal evidence: my 10C20T Haswell-E Xeon pulls ~60% of its rated TDP during a Handbrake x264 encoding run. My 6C12T 5820K Haswell-E gaming chip (overclocked to 4.13 GHz all-core at stock voltages) pulls 75% of its rated TDP during a similar Handbrake x264 run. And supposedly x264 is smart enough to use AVX, so that should be a realistic-but-high figure. My OC'd 5820K pulls 90% of its rated TDP during a Prime95 SmallFFT run. It's still a huge gain compared to Bulldozer - but not quite as rosy a picture as AMD is trying to paint. They are equal with Haswell-E/Broadwell-E in most respects, not 40% ahead like AMD is rating them. AMD is playing the same game in the GPU market too. AMD's RX 480/580 are rated for "GPU only" TDP (which they define to exclude memory/etc) but NVIDIA rates the whole card. In practice, AMD is ~30% above their official TDPs while NVIDIA is pretty much spot on. https://www.techpowerup.com/reviews/Sapphire/RX_580_Nitro_Plus/28.html https://www.techpowerup.com/reviews/Sapphire/RX_580_Nitro_Pl... (RX 580 is rated at 185W TDP, GTX 1060 is rated at 120W TDP)
- loeg 9y ago
  Thanks!
- dom0 9y ago
  TDP is and always was an expressly manufacturer- and product-specific number, though. It mainly exists for OEMs to match parts up with the correctly sized heatsink.
  - paulmd 9y ago
    That argument stops holding water when the manufacturer starts marketing around the superior power efficiency of their processors based on that same number.
yohui 9y ago
paulmd already put up a list, so I won't repeat it. Ars Technica sums it up: https://arstechnica.com/gadgets/2017/03/amd-ryzen-review/ https://arstechnica.com/gadgets/2017/03/amd-ryzen-review/ > Indeed, Ryzen's 95W TDP looks impressive compared to the 140W TDP of Intel's 6900K, a similar 8C/16T processor based on the Broadwell-E architecture. But in practice, Ryzen pulls just as much wattage from the wall as the 6900K, in some cases slightly more. Considering just how power-hungry Bulldozer was, that remains a significant achievement, even if it means that overclocking headroom isn't as high as some might have hoped.

leonlee 9y ago

Unfortunately AMD and Intel don't use the same definition for TDP (look up AMD APC). AMD's values are more related to average power consumption during whatever they consider to be typical use. Intel's values on the other hand are more about the upper limit of power usage, "how high can it go" sort of thing. Both the average power and maximum power are useful bits of info, but it's obviously confusing and opaque for the end-user. The only reason that I can think of why AMD hasn't reverted back is because it would highlight how their processors are relatively power inefficient compared to their competitor's. http://www.amd.com/Documents/43761A_ACP_WPv7.pdf http://www.amd.com/Documents/43761A_ACP_WPv7.pdf

mtgx 9y ago
Didn't Intel try to do the same thing with SDP?