Zen 4 CPU prices fall dramatically, Ryzen 9 7950X is now down to $554

[Avro Arrow]

It helps to view the Blender figures as a total energy consumption to complete the test:
View attachment 88708

Ok, so what is the point of this graph then?

https://static.techspot.com/articles-info/2535/bench/Power-p.webp

The figures I used are based on the actual graph published by Steve Walton. That graph you're showing is something that I've never seen before and it doesn't appear to agree with the graph from the review.

The 7600X system required 195 kJ of energy to do the test, whereas the 5950X system needed 105 kJ. At face value, this suggests that the 7600X system has half the efficiency of the 5950X system, but it's 6C/12T CPU versus a 16C/32T one. So the latter can use three times more threads to complete the test - so it does it quicker, thus using less energy in total.

Ok, that makes sense but it doesn't mean that Zen-4 is more efficient because the R9-7950X still required 61% more juice to produce only 42% more performance compared to the R9-5950X.

Another thing to bear in mind is that different motherboard manufacturers set different power limits, but Zen 4 CPUs do have higher power limits than Zen 3 ones - the TDP/PL1 of the 7600X is 105W, whereas the 5600X was just 65W. In the case of the 5950X, it may also be the case that when all cores are fully loaded, the BIOS may be limiting the power.

Regardless, the R9-5950X is still producing more performance-per-watt than the R9-7950X. If the BIOS limiting power was the issue, the R9-5950X wouldn't have the performance level that we see. The fact that the R9-7950X uses 61% more juice to produce only 42% more performance means that it's less efficient, period. It's like AMD and Intel have jacked up the power use to make it appear that these parts are a better upgrade than they actually are. I don't see an advantage for Zen-4 here, I see the exact opposite. Sure, they're faster but the cost of that speed isn't necessarily worth it.

Speed should come because of efficiency, not at the cost of it.

 

[neeyik]

The figures I used are based on the actual graph published by Steve Walton. That graph you're showing is something that I've never seen before and it doesn't appear to agree with the graph from the review.

The graph was made using Steve’s Blender time results and system energy figures - power is simply energy x time, so the graph shows how much system energy was used, in total, to do the Blender test.

It’s sometimes better to see the data in this format because it gives a potentially clearer picture to how efficient or inefficient a given system is. In the case of the 7600X it needed around 20kJ more energy to do the same test as the 5600X system.

Does that make Zen 4 less efficient than Zen 3? Well there’s more than just a different CPU involved, as the motherboards and RAM are different. There may also be different power limits and associated times in the BIOSes too. Zen 4 could well be less efficient as a system compared to Zen 3 but it requires a deeper analysis than time permitted to get the review done.

 

[Avro Arrow]

The graph was made using Steve’s Blender time results and system energy figures - power is simply energy x time, so the graph shows how much system energy was used, in total, to do the Blender test.

It’s sometimes better to see the data in this format because it gives a potentially clearer picture to how efficient or inefficient a given system is. In the case of the 7600X it needed around 20kJ more energy to do the same test as the 5600X system.

Does that make Zen 4 less efficient than Zen 3? Well there’s more than just a different CPU involved, as the motherboards and RAM are different. There may also be different power limits and associated times in the BIOSes too. Zen 4 could well be less efficient as a system compared to Zen 3 but it requires a deeper analysis than time permitted to get the review done.

Ok, that makes sense. However, at this point in time, all we know is that AM5 as a system is less power efficient than AM4. That includes motherboards (and everything on them), RAM and CPUs. If the AM5 platform itself is less efficient (which would also make no sense but here we are), then the efficiency of the Zen-4 becomes irrelevant as it is only one part of the AM5 platform and must be used with the other parts. Something in AM5 is less efficient than AM4 and even if it's not the CPU itself, the end result is the same because you can't use a Zen-4 CPU without also using the AM5 platform.

 

[neeyik]

Efficiency is a word that does need to be used with care. Ultimately, it's a ratio -- the minimum amount of energy required to carry out a task against the actual amount of energy used to do it.

The 7600X system is using more energy to do the Blender task than the 5600X system. However, where the 5600X system used 179 kJ and took 1204 seconds to do the test, the 7600X system used 195 kJ and took 862 seconds, so naturally, the power measurement will be higher (power = energy/time). But for 9% more energy, the test time was reduced by 40%.

So on that basis, you're getting a notable improvement in processing speed for a relatively small increase in energy consumption. It's not the same as 'efficiency' but I suspect a lot of Blender users would be more than willing to make that trade.

 

[Avro Arrow]

Efficiency is a word that does need to be used with care. Ultimately, it's a ratio -- the minimum amount of energy required to carry out a task against the actual amount of energy used to do it.

The 7600X system is using more energy to do the Blender task than the 5600X system. However, where the 5600X system used 179 kJ and took 1204 seconds to do the test, the 7600X system used 195 kJ and took 862 seconds, so naturally, the power measurement will be higher (power = energy/time). But for 9% more energy, the test time was reduced by 40%.

So on that basis, you're getting a notable improvement in processing speed for a relatively small increase in energy consumption. It's not the same as 'efficiency' but I suspect a lot of Blender users would be more than willing to make that trade.

Ok, here's Steve from Gamers Nexus showing the power consumption difference and he uses specialised equipment to read only the CPU draw for 5 minutes of Blender. Then he shows blender performance and then he shows efficiency in which the R9-5950X just destroys everything else:

Note how he describes the R9-5950X as "shockingly efficient".

Do you believe me now?

 

[neeyik]

The 5950X has 16C/32T and Blender can split the rendering workload across all 32 threads. The 7600X has a third the thread count so, clock for clock, it will take 3 times longer to render the test frame, thus consuming more energy in the process. This is why in Steve’s testing the 5950X system used 105 kJ compared to the 7600X system’s 195 kJ.

It’s akin to comparing two factories, one with 32 parallel production lines and the other with 12, then criticising the smaller factory for taking longer and using more energy to manufacture the same quantity of goods.