Anands Core-X power consumption results unique on the Internet

[Topweasel]

I think the discrepancy is really easy.

The first issue is TDP hasn't been TDP for a long while. While they try to stick on an ultra conservative closer to max power usage rating, but really it is cooling recommendation rating and is based on power usage of the CPU at full usage with rated cooler (I know that sounds kind if weird). It started back in the Pentium 4 days when they made slight change to what TDP meant and started using 90% of the power usage because a consistent 100% usage over a long period of time was seen as unlikely scenario. AMD countered with I believe it was ADP or Average Design Power. Basically a nearly similar assessment of power usage at given cpu usage rating (they had a test shop of tasks and took average power rating).

The second issue is that the Power rating for Intel's CPU's don't include AVX2 or AVX-512 and most of the tests here don't use it extensively. Want to pop Intel over its power rating. Plug in a AVX2 heavy task. If you want to show general TDP like (like the rating system Intel and AMD use that they call TDP) then go AVX lite. If you want to show full power potential (or by knowing that AMD's CPU's don't fluctuate nearly as much when doing AVX) then use an AVX heavy workload.

 

[ehume]

I think the discrepancy is really easy.

The first issue is TDP hasn't been TDP for a long while. While they try to stick on an ultra conservative closer to max power usage rating, but really it is cooling recommendation rating and is based on power usage of the CPU at full usage with rated cooler (I know that sounds kind if weird). It started back in the Pentium 4 days when they made slight change to what TDP meant and started using 90% of the power usage because a consistent 100% usage over a long period of time was seen as unlikely scenario. AMD countered with I believe it was ADP or Average Design Power. Basically a nearly similar assessment of power usage at given cpu usage rating (they had a test shop of tasks and took average power rating).

The second issue is that the Power rating for Intel's CPU's don't include AVX2 or AVX-512 and most of the tests here don't use it extensively. Want to pop Intel over its power rating. Plug in a AVX2 heavy task. If you want to show general TDP like (like the rating system Intel and AMD use that they call TDP) then go AVX lite. If you want to show full power potential (or by knowing that AMD's CPU's don't fluctuate nearly as much when doing AVX) then use an AVX heavy workload.

Yup. This is why I tested with an i7 4790K running Linpack with AVG2. Doing so can push a heatsink to its limits, even if you use a moderate overclock.

 

[Jan Olšan]

Another review where 1950X ended up with lower power consumption than (stock) Core i9-7900X... it's certainly not uncommon, although it's likely still occuring in less than 50 % of cases.
http://www.phoronix.com/scan.php?page=article&item=amd-tr-1950x&num=9

Here's a look at the AC system power consumption over the span of all the benchmarks carried out for this Threadripper Linux review. The Ryzen Threadripper 1950X yielded an average power draw of 171 Watts while the Core i9 7900X average was actually at 179 Watts. The peak was also slightly lower on the 1950X.

Based on the graphs, it seems that good portion of the tests weasn't actually well threaded at all, which is interesting, because that should have been giving advantage to Skylake-X. Intel has lower idle power and lower turbo voltages and hence single core power consumption in single core turbo, compared to Threadripper/Zens.