Nvidia getting into the ARM game will certainly spice things up a bit. It's certainly going to raise the ante quite a bit, especially when nVidia's custom ARM CPU lands, but it's also going to be facing stiff competition from AMD.

WinterLightDP
Nvidia getting into the ARM game will certainly spice things up a bit. It's certainly going to raise the ante quite a bit, especially when nVidia's custom ARM CPU lands, but it's also going to be facing stiff competition from AMD.

Everyone is gradually realizing that the x86 architecture is outdated. It's fascinating to watch an iPhone 16 Pro (MacBook NEO) processor capable of running Houdini

On the other hand, it’s fascinating to watch Apple processors handle legacy x86 code through the Rosetta 2 translator. I recently saw firsthand how the outdated, physically-based spectral renderer Indigo—which only has an x86 build for Mac—flies like crazy through the translator. After this, I believe in Apple more and more.

alexeyvanzhula1984
Everyone is gradually realizing that the x86 architecture is outdated. It's fascinating to watch an iPhone 16 Pro (MacBook NEO) processor capable of running Houdini

The modern x86 no longer bears much resemblance to the original, which is a good thing.

That said, I was rather surprised that AMD's SoundWave does not appear to be intended to replace Zen any time soon. But then Zen5 exceeded expectations, though it didn't get the adoption in mobile devices that I'd hoped for -- apparently the laptop vendors in particular are still loathe or contractually (or both) still stuck on pleasing Intel. Most of the Strix Halo laptops (all two or three) are crippled, apparently to keep them from steamrolling Intel.

It's frustrating because Zen5 is so far ahead of Intel efficiency wise, yet still able to scale up to a sizeable performance lead... but not if the system it's in keeps it throttled to 45 watts, which most of them do so the Halo can't run at its maximum clock speed.

Having 128GB is nice, but not having Karma XPU support for my GPU isn't.

alexeyvanzhula1984
The modern x86 no longer bears much resemblance to the original, which is a good thing.

I think they still drag a ton of useless x86 legacy crap anyway.

alexeyvanzhula1984
I think they still drag a ton of useless x86 legacy crap anyway.

There's less and less of that nowadays, as the legacy crap is from so long ago that even a crappy and inefficient emulation is faster than anything compiled for legacy x86 had access to.

I still hope that it will fade away entirely, and hopefully be replaced with ARM or MIPS V or something much more efficient and probably also more scalable.

alexeyvanzhula1984
About the “unsuitability” of Apple Silicon for Houdini and 3D: myths vs. real-world experience

1. Viewport (OpenGL vs Vulkan)
Right after the release of H21, there were indeed graphical bugs in the Vulkan viewport, and the good old OpenGL seemed more stable and faster, so I used it myself and advised others to do the same. But recently I tested Vulkan on one of the latest builds — all display bugs (like highlighting selected edges) have completely disappeared, and the viewport itself became maximally responsive and smooth.
As for the limitations in the SOP context, perhaps only the Texture Mask Paint tool really doesn't work on macOS. I haven't encountered any other limitations during my entire time working with it.

2. How a $700 box “bent over” a $4000+ setup
And now the most interesting part — why I completely moved my development to Apple.
Initially, I got a base Mac mini M4 just as a TV box. I had no thoughts of running heavy software there. But just out of curiosity, I installed Houdini on it. My main workstation at that time was a desktop PC with an AMD Ryzen 9 9950X (128 GB RAM) + RTX 4080 Super. A setup costing over 4 thousand dollars.
I couldn't believe my eyes and re-checked the tests several times. This little box performed incredibly cool in CPU-oriented tasks. The “phone” ARM architecture digests algorithms and OpenCL computations much more efficiently than heavy x86 hardware. The only area where the Mac mini predictably felt weak was viewport.

3. Bottom line
After this test, my life was never the same. I went out and bought a top-tier MacBook Pro with M4 Max with 128Gb unified memory — and I'm happy to this day. The same goes for working in modern Blender: it feels like you're sitting not at a laptop, but at an expensive Threadripper computer with 512 GB RAM.
It's great that Houdini 22 brought its own improvements for Apple. But personally, no one will ever be able to tell me again that Houdini or 3D runs poorly on Apple Silicon. The fairy tales that Mac is “not suitable for graphics” are a thing of the past.

4. Everything I wrote does not apply to the Karma renderer since I don't use it. But of course, I would dream of seeing a Metal backend for Apple. However, I am a realist and understand why this won't happen. Apple is a closed ecosystem, and it is not in their interest to work on Vulkan support. And support from SideFX for both Vulkan and Metal is a very difficult task, since the architectures are very different, and making some kind of universal layer for both Metal and Vulkan is impractical.

Can you not running your reply through LLMs? You've been in this community for quite a while. Everyone has been okay with your text. No need AI to write for you.

raincole

alexeyvanzhula1984
About the “unsuitability” of Apple Silicon for Houdini and 3D: myths vs. real-world experience

1. Viewport (OpenGL vs Vulkan)
Right after the release of H21, there were indeed graphical bugs in the Vulkan viewport, and the good old OpenGL seemed more stable and faster, so I used it myself and advised others to do the same. But recently I tested Vulkan on one of the latest builds — all display bugs (like highlighting selected edges) have completely disappeared, and the viewport itself became maximally responsive and smooth.
As for the limitations in the SOP context, perhaps only the Texture Mask Paint tool really doesn't work on macOS. I haven't encountered any other limitations during my entire time working with it.

2. How a $700 box “bent over” a $4000+ setup
And now the most interesting part — why I completely moved my development to Apple.
Initially, I got a base Mac mini M4 just as a TV box. I had no thoughts of running heavy software there. But just out of curiosity, I installed Houdini on it. My main workstation at that time was a desktop PC with an AMD Ryzen 9 9950X (128 GB RAM) + RTX 4080 Super. A setup costing over 4 thousand dollars.
I couldn't believe my eyes and re-checked the tests several times. This little box performed incredibly cool in CPU-oriented tasks. The “phone” ARM architecture digests algorithms and OpenCL computations much more efficiently than heavy x86 hardware. The only area where the Mac mini predictably felt weak was viewport.

3. Bottom line
After this test, my life was never the same. I went out and bought a top-tier MacBook Pro with M4 Max with 128Gb unified memory — and I'm happy to this day. The same goes for working in modern Blender: it feels like you're sitting not at a laptop, but at an expensive Threadripper computer with 512 GB RAM.
It's great that Houdini 22 brought its own improvements for Apple. But personally, no one will ever be able to tell me again that Houdini or 3D runs poorly on Apple Silicon. The fairy tales that Mac is “not suitable for graphics” are a thing of the past.

4. Everything I wrote does not apply to the Karma renderer since I don't use it. But of course, I would dream of seeing a Metal backend for Apple. However, I am a realist and understand why this won't happen. Apple is a closed ecosystem, and it is not in their interest to work on Vulkan support. And support from SideFX for both Vulkan and Metal is a very difficult task, since the architectures are very different, and making some kind of universal layer for both Metal and Vulkan is impractical.

Can you not running your reply through LLMs? You've been in this community for quite a while. Everyone has been okay with your text. No need AI to write for you.

I constantly use Gemini for translation since I am not a native speaker. The fact that it automatically breaks text into groups is its own feature, but I am not asking it to improve the text.

I think if you wanted to get your point across to a Russian speaker, you would have used an LLM instead of Google Translate.

So I don't understand what the problem is at all. Is it that I translate even though I can write in English myself? Well, it's just faster; why should I spend three hours writing a long message and wasting time on its translation when I can spend it on something more useful. And I always number the replies in order.


I don't improve the text, and I don't do vibe coding.

But I partially agree, so in the future, I will ask not to change the text at all when translating.

This is a second, newer account. Actually, I have been on this forum since 2006.