Besides RicardoTM's take on this, HA mainly has to do with:
- the quality of the CPU vs the GPU
- the type of operations to perform
- the efficiency of the implementation
The 1st means that if the CPU is much more powerful than your GPU, things will work better with HA disabled, and if the other way around, with HA enabled.
The 2nd means that the CPU is better at serial computing (a single task is performed at a time), while the GPU is better at parallel computing (multiple tasks are performed simultaneously). The CPU is focused on low latency, aka running different tasks as fast as possible, while the GPU is focused on high throughput, aka running as many similar tasks as possible. This is why the type of operations to perform matters, because if the tasks are very different and more complex, the CPU will handle them better, and if they are very similar and less complex, the GPU will be more efficient. It's a consequence of the different purpose of a CPU compared to a GPU. The CPU having multiple cores / threads helps to a degree in making the CPU better at parallelism, but then the GPU has many more such cores, albeit much weaker and more suited for simpler parallel operations (like drawing a bunch of pixels with the same color, or computing 1 + 1 a million times, to give some rudimentar examples).
The 3rd means that the above will also depend on how well the software itself takes advantage of these things, from compatibility, driver quality, how much the operating system facilitates it, and obviously how well written such a software is in that regard.
All these and a few others are the main reason why disabling or enabling HA might produce inconsistent results at times, as both the hardware and the software parts vary from case to case, just like what HA is used for is.
P.S. Yep, my old laptop began to die after roughly 10 years, so I agree that a similar period is a reasonable time to improve things in that field.
- the quality of the CPU vs the GPU
- the type of operations to perform
- the efficiency of the implementation
The 1st means that if the CPU is much more powerful than your GPU, things will work better with HA disabled, and if the other way around, with HA enabled.
The 2nd means that the CPU is better at serial computing (a single task is performed at a time), while the GPU is better at parallel computing (multiple tasks are performed simultaneously). The CPU is focused on low latency, aka running different tasks as fast as possible, while the GPU is focused on high throughput, aka running as many similar tasks as possible. This is why the type of operations to perform matters, because if the tasks are very different and more complex, the CPU will handle them better, and if they are very similar and less complex, the GPU will be more efficient. It's a consequence of the different purpose of a CPU compared to a GPU. The CPU having multiple cores / threads helps to a degree in making the CPU better at parallelism, but then the GPU has many more such cores, albeit much weaker and more suited for simpler parallel operations (like drawing a bunch of pixels with the same color, or computing 1 + 1 a million times, to give some rudimentar examples).
The 3rd means that the above will also depend on how well the software itself takes advantage of these things, from compatibility, driver quality, how much the operating system facilitates it, and obviously how well written such a software is in that regard.
All these and a few others are the main reason why disabling or enabling HA might produce inconsistent results at times, as both the hardware and the software parts vary from case to case, just like what HA is used for is.
P.S. Yep, my old laptop began to die after roughly 10 years, so I agree that a similar period is a reasonable time to improve things in that field.
Statistics: Posted by Yincognito — Yesterday, 2:42 pm