3 it is

PS5 itself is a disgrace

I know you mainly talk about Intel products. But by any chance, do you know when consumer AM5 Zen 6 is coming?

Does it hypothetically mean that RZL-S ES/QS will boot on consumer NVL boards with consumer BIOS too? Or will those be specifically hardware locked?

Makes sense. But similar pattern is observed in consumer vs professional GPUs with different dies. It doesn't need a whole new generation for more VRAM. It may need minor die revisions but that is not new generation as OP said. Reason is purely marketing; not hardware.

And so? QC has both and both are useless.

It is you who is flaunting your ignorance.

Then explain the existence of 96GB on RTX6000 Blackwell while the 5090 is limited to just 5090 outside of marketing reasons. You don't need a new generation for more VRAM.

Exynos over Apple A series anyday. Obviously.

The SoundWave chip is basically a placeholder for the ultra low end of AMD. It'd barely have enough horsepower for basic tasks. No way that's going against SDXE/X2E of N1X/GB10.

Just because the release of larger VRAM chips generally coincide with the time period of the next generation. If a specific generation of hardware were to last many years, it is possible to have 2x or even 4x VRAM because of availability of higher capacity chips.

I can say this is pure BS. If whatever you said were true, how does Nvidia make 32GB and 96GB variants with the same GB202 die? Sure, you're limited by the size of the VRAM chips available but that's it. If you were to have 12GB VRAM chips, you could go to 192GB total.

Tried on Blender 5.1.2. Getting 24.66 seconds. So, within the margin of error.

24.89 seconds

ThE hUmAn EyE cAnNoT sEe AbOvE 24 fPs

While the Open Data scores can be compared between the same generation, a lot of deviations in subscores arise when you compare different generations or vendors of hardware. It gets even worse when you compare different individual/custom scenes. (2/2)

Will share screenshot of PartyTug render on 4060 as soon as I can. Blender Open Data isn't very representative of a lot of scenes. I did lot of tests on my (now sold) M4 Mac mini. Usual difference of M4 vs 4060 was 2x. But varied between 1.5-3x depending on the scene. (1/2)

More than enough to have all the other PC GPU vendors, be it integrated or discrete, namely AMD, Apple, Intel, Nvidia dedicate resources to make a GPU rendering codepath for their hardware.

1) 4060 matches a 3070 in Blender. 2) Make sure you take the second or third run for this specific scene because sometimes there's the kernel compilation delay. 3) Make sure you're running OPTIX and not CUDA. I may provide a screenshot later in the day.

What if QC realized that their CPUs do better render times than their GPUs and left it at that without actual public release to stop embarrassing themselves?