3 ms·

It's even worse than that - the youtubers don't compare like for like. They should pick a common state for comparison (eg. same power or same performance), inst

It's even worse than that - the youtubers don't compare like for like. They should pick a common state for comparison (eg. same power or same performance), instead it's just a free for all, and whoever picks the lower power cap has better efficiency, whoever picks the higher has better performance. CPU efficiency is so nonlinear at high power levels as to make comparison meaningless if they aren't matched.

LegitShady 4y ago
this is bad because certain market segments, like the most profitable consumer ones, are differentiated by performance not performance per watt. Picking an arbitrary power cap and testing them both there will tell you which is most efficient, which is a bit like telling a lamborghini customer that your honda gets better gas mileage - its not something he really cares about. CPU efficiency is an interesting number to the tech youtubers who focus on it, but is not that market differentiator at the highest end most profitable part of the market, and only matters to them. The AMD 7950x is very fast chip which is much more efficient than its intel counterpart, but at the top end of the market what matters is who performs better, not their fuel mileage. You'd be investigating a number interesting to testers but not consumers likely to buy those products - its a waste of time and effort.
- Panzer04 4y ago
  I agree, if I'm buying a top-end consumer chip I only really care about raw performance - finishing a compilation run 20s faster 20x a day or running my Grand strategy game faster in the laggy endgame :P. I guess from a practical perspective, if you aren't modifying the chip clocks then yeah, they are talking about the right thing. But from a technical perspective, I'm more interested in the underlying architecture performance, which is to say: If the intel chip is set to a Power Limit of 100w, and the AMD chip is set to a Power Limit of 100w, what does performance look like then? Because it's nearly meaningless to run one processor at 250w and the other at 150w, then say the 150w chip is more efficient. Processor power skyrockets at high voltages and clocks, so likely the intel chip can use 50% less power for 10% less performance, if desired - Intel merely made the (correct, imo) calculus that maximal performance is more important than efficiency for this class of chip. I think we pretty much agree - I just think it would be interested to have a more accurate architectural efficiency/performance comparison (which would also help when choosing eg. a laptop where those numbers DO matter)
  - LegitShady 4y ago
    I think that performance per watt measurement ONLY applies to the budget end of the market. I'm not buying underlying architecture, I'm buying a CPU with a stated performance. If you're buying budget parts and pinching pennies for electricity its certainly relevant. If you're choosing between an AMD 7950x and an intel i9-13900k, its stupid and doesn't matter. As I said before, the efficiency per watt is a youtuber number. Nobody who is buying chips really cares, or if they care they watch the youtubers who talk about it nonstop already. It's the youtubers bringing it up on top tier chips who are just out to lunch and not understanding the actual selling to a market and distorting things because of numbers they care about that class customers don't. I don't care what a 7950x does at 100w, because if I wanted that performance I would have bought a cheaper chip. It's a stupid nonsensical number unrelated to why people buy $700 high end CPUs. You don't buy that chip to save $3 worth of electricity.
    - Panzer04 4y ago
      Again, I am mostly interested because it would be representative of architectural efficiency to perform such a test. If the top-spec processor is more efficient at a given power level than another, the same applies to lesser processors as well, in general (in particular, laptops and the like) This comparison is meaningless though if the processors aren't running at the same power, however :(

LegitShady 4y ago

this is bad because certain market segments, like the most profitable consumer ones, are differentiated by performance not performance per watt. Picking an arbitrary power cap and testing them both there will tell you which is most efficient, which is a bit like telling a lamborghini customer that your honda gets better gas mileage - its not something he really cares about. CPU efficiency is an interesting number to the tech youtubers who focus on it, but is not that market differentiator at the highest end most profitable part of the market, and only matters to them. The AMD 7950x is very fast chip which is much more efficient than its intel counterpart, but at the top end of the market what matters is who performs better, not their fuel mileage. You'd be investigating a number interesting to testers but not consumers likely to buy those products - its a waste of time and effort.

Panzer04 4y ago
I agree, if I'm buying a top-end consumer chip I only really care about raw performance - finishing a compilation run 20s faster 20x a day or running my Grand strategy game faster in the laggy endgame :P. I guess from a practical perspective, if you aren't modifying the chip clocks then yeah, they are talking about the right thing. But from a technical perspective, I'm more interested in the underlying architecture performance, which is to say: If the intel chip is set to a Power Limit of 100w, and the AMD chip is set to a Power Limit of 100w, what does performance look like then? Because it's nearly meaningless to run one processor at 250w and the other at 150w, then say the 150w chip is more efficient. Processor power skyrockets at high voltages and clocks, so likely the intel chip can use 50% less power for 10% less performance, if desired - Intel merely made the (correct, imo) calculus that maximal performance is more important than efficiency for this class of chip. I think we pretty much agree - I just think it would be interested to have a more accurate architectural efficiency/performance comparison (which would also help when choosing eg. a laptop where those numbers DO matter)
- LegitShady 4y ago
  I think that performance per watt measurement ONLY applies to the budget end of the market. I'm not buying underlying architecture, I'm buying a CPU with a stated performance. If you're buying budget parts and pinching pennies for electricity its certainly relevant. If you're choosing between an AMD 7950x and an intel i9-13900k, its stupid and doesn't matter. As I said before, the efficiency per watt is a youtuber number. Nobody who is buying chips really cares, or if they care they watch the youtubers who talk about it nonstop already. It's the youtubers bringing it up on top tier chips who are just out to lunch and not understanding the actual selling to a market and distorting things because of numbers they care about that class customers don't. I don't care what a 7950x does at 100w, because if I wanted that performance I would have bought a cheaper chip. It's a stupid nonsensical number unrelated to why people buy $700 high end CPUs. You don't buy that chip to save $3 worth of electricity.
  - Panzer04 4y ago
    Again, I am mostly interested because it would be representative of architectural efficiency to perform such a test. If the top-spec processor is more efficient at a given power level than another, the same applies to lesser processors as well, in general (in particular, laptops and the like) This comparison is meaningless though if the processors aren't running at the same power, however :(