12 ms·

IMHO, any benchmark should run with Turbo disabled. And come with additional tests about how much turbo brings, and how long it can stay on within a given therm

by Aissen 4y ago

IMHO, any benchmark should run with Turbo disabled. And come with additional tests about how much turbo brings, and how long it can stay on within a given thermal setup. Otherwise all you have is garbage, not data.

brokencode 4y ago
Why? It’s not like turbo is not a real performance enhancement. With enough thermal headroom, there’s no reason why turbo can’t be sustained, as heat is the limiting factor. Good CPU reviews include long render and compile benchmarks that would suffer if the turbo performance couldn’t be sustained. If a CPU can sustain high performance, then it doesn’t matter if it’s using a turbo mode.
- Aissen 4y ago
  It's all about the "With enough thermal headroom". If this isn't specified in the benchmark, then the benchmark is useless because you can't compare with your setup. It's even truer in laptops, but this is true of servers and desktops as well. In addition, recent CPUs will use this thermal headroom to the limit.
sz4kerto 4y ago
Turbo is such an inherent feature of new CPUs that turning them off would make the benchmark completely unrepresentative for real world usage. For example - a 16 core CPU might run a single core at extremely high clock rates during single threaded workloads. What would be the point of turning it off? Complex devices need complex benchmarking, unfortunately. You won't get a simple, single number that shows how powerful a cpu is.
- Aissen 4y ago
  > Complex devices need complex benchmarking, unfortunately. You won't get a simple, single number that shows how powerful a cpu is. This is the gist of my comment (see after the first phrase), so I think we agree.
viraptor 4y ago
You can have both at the same time by publishing a graph of performance over time where you see the ramp up and the time where the turbo can't be sustained. Quite a few reviewers started doing that maybe 2 years ago?
- Aissen 4y ago
  Yeah, I think having both is exactly what should be done (in addition to talking about the thermal setup).
xen2xen1 4y ago
Like hitting the button on my 486sx/16.

by Aissen 4y ago

brokencode 4y ago

Why? It’s not like turbo is not a real performance enhancement. With enough thermal headroom, there’s no reason why turbo can’t be sustained, as heat is the limiting factor. Good CPU reviews include long render and compile benchmarks that would suffer if the turbo performance couldn’t be sustained. If a CPU can sustain high performance, then it doesn’t matter if it’s using a turbo mode.

Aissen 4y ago
It's all about the "With enough thermal headroom". If this isn't specified in the benchmark, then the benchmark is useless because you can't compare with your setup. It's even truer in laptops, but this is true of servers and desktops as well. In addition, recent CPUs will use this thermal headroom to the limit.

sz4kerto 4y ago

Turbo is such an inherent feature of new CPUs that turning them off would make the benchmark completely unrepresentative for real world usage. For example - a 16 core CPU might run a single core at extremely high clock rates during single threaded workloads. What would be the point of turning it off? Complex devices need complex benchmarking, unfortunately. You won't get a simple, single number that shows how powerful a cpu is.

Aissen 4y ago
> Complex devices need complex benchmarking, unfortunately. You won't get a simple, single number that shows how powerful a cpu is. This is the gist of my comment (see after the first phrase), so I think we agree.

viraptor 4y ago

You can have both at the same time by publishing a graph of performance over time where you see the ramp up and the time where the turbo can't be sustained. Quite a few reviewers started doing that maybe 2 years ago?

Aissen 4y ago
Yeah, I think having both is exactly what should be done (in addition to talking about the thermal setup).

xen2xen1 4y ago

Like hitting the button on my 486sx/16.