Both AMD and Intel vaunt their boost core clocks. This article talks about the boost clock for single core performance on Intel’s latest:
And this video talks about AMD’s single core boost clock in it’s latest:
But what no one talks about is what happens to these clock speeds when 2 or 4 threads/cores are used equally. Because for argument’s sake, if you have a game that uses 4 threads, and the boost clock is 60 million GHz, but multicore is only 2 GHz, then potentially, for anything but an app that uses just 1 core, that’s a really crappy processor. A game like X4: Foundations is extremely cpu bound, and is written for 4 threads. So I’m wondering if boost speeds get progressively less with each core used, or if it’s binary. For intel’s example, 5.8 HGz for single core, and 3.0 for 2-8 cores.
Heat caps all the speeds. You won’t even get 1 core to published “max boost” long with poor cooling. Extreme overclockers use extreme cryo cooling solutions to get max boost on all cores, and then push all their custom speed and power limits higher.
But in a perfectly chill environment with no overclocking. What do these chips bump up 2 or 4 thread games to, if at all? We know what default speeds they run at when all cores are used, and we know what the default boost speed is under good cooling, but what would the default be at 2 or 4 cores (what most games utilize today)?
Well, nobody might talk about speeds themselves, but there are both single and multi-core benchmarks being used out there. I don’t think there are many 2-4 specific, but I’d expect a fairly linear regression to be close enough wrt throttling.
Reality may vary, I don’t have the money to care :P
I think your question is “if there is workload on multiple cores, can they all get the same max boost as a single core would”. Generally yes, some or all cores can have the same boost, limited by thermals. If the CPU and motherboard support overclocking then you can deviate from the standard behavior in a lot of ways. Maybe someone with more recent examples could help, I haven’t overclocked CPUs since the '90s :)
I’ve been managing the thermals on my new 5800X for the past week, and when benchmarking (mostly for heat and undervolt stability) in ACC, which uses 8 threads, all 8 cores have been hovering around 4500-4600 Mhz, versus a max boost of ~4800 Mhz. Before I figured out both airflow and undervolting, they were throttling down to about 3500 Mhz at the 90C throttle point.
So for AMD, if I understand the question, the answer is all cores will behave similarly and be extremely sensitive to thermal limitations. Since heat is generated in direct proportion to the amount of power used, which is directly affected by how many cores are trying to run fast, your thermal limitations will get worse the more cores you use.
However good your cooling is power limits also come into play.
It gradually degrades.
Undervolting will consume less power and generate less heat at a given clock. Therefore if the processor is stable at those voltages you can drive clocks higher while staying inside your thermal and power limits.
The lousy prebuilt case they’re living in had terrible airflow, so they were both hitting the temperature limit well below max clock speeds. Undervolting decreases the heat output relative to a given performance level, so higher clock speeds reached that same thermal equilibrium. Popping off the front of the case improved airflow enough that the CPU doesn’t thermally throttle at all, and the GPU reaches equilibrium above Nvidia’s stock speed.
I actually wanted a 5700X for the lower power consumption & heat output, but the 5800X was on sale for less, so I’ve essentially turned it into a 5700X+.
