Graphics Card (GPU) based render engines have overtaken CPU-based render engines quite a bit. We have companies like Intel, NVIDIA and AMD to blame for their success, though IBM is trying its best with talk (only talk) of quantum computing. Though, they are probably trying to sort out a few cryptocurrency markets and bump up their stock price before admitting they can’t do it. In reality, in this day and age, Taiwan, Intel, AMD, and even the smaller producers are making GPUs faster and more efficient.
Why not make GPUs more powerful? We have a cryptocurrency market that is crying out for more and more GPU power, and modern-gen gaming from PCs, Sony and Microsoft are desperate for better and faster graphics engines. Even the car industry is soaking up its fair share of chips for its advanced cars (that police forces will be able to shut down from miles away).
We Live in an Exciting Time
Still, with all these advancements and market demands, which is going to give you the best value for money when conducting your GPU rendering? Oddly, the best bang-for-your-buck solution is to not conduct your own rendering.
If you use a company like https://rebusfarm.net/, then you get the best overall value for money. You don’t have to deal with things like hardware depreciation, breakdowns, and energy supply, and you don’t have to dedicate hardware work toward rendering. You are not tying up your own devices and using data centers and cloud systems is far more eco-friendly and efficient. Still, if you want to conduct your own rendering, here are a few thoughts on your hardware.
AMD and Intel
GPU-Render Engines will use your GPU(s) to do their rendering, so your first thought is probably to go for a max-core-clock CPU like the Intel i9 12900K or the AMD Ryzen 9 5950X that clocks at 3,4Ghz, with added extra for turbo settings. This isn’t a bad train of thought because GPUs are connected on the motherboard to your CPU via PCIe-Lanes. Plus, you want a good CPU to ensure your preparations/designs are as efficient as possible. Your AMD and Intel models allow 16x PCIe 3.0 Lanes to run at full performance without bandwidth throttling, and that is typically what your top-tier GPUs need.
The trouble is that these CPUs will only allow you to use one top-tier GPU at top speed with its best performance. This means having several CPUs or trying something like the AMD threadripper. The AMD Threadripper 3960X or the AMD Threadripper 2950X have 64 PCIe-Lanes, and they can run in lower bandwidth modes too.
Why Is The Topic So Complicated?
It is tough to say which processor is the best for you because you need to match your hardware with your situation. The question of “Best bang for your buck” also depends mostly on how you are set up and what they want. A company that runs the processors all the time, switching between rendering and cryptocurrency mining, will need raw power and plenty of processing units. On the other hand, for a company that spends most of its time designing and animating and just a bit of time rendering, then the setup will be very different.
When you get down to things like Octane and Redshift, you are going to need a GPU that has CUDA-Cores, so NVIDIA is the best place to go. Before picking your hardware setup, you have a lot of questions to answer. Then comes the problem of optimizations. Even if we try our best to optimize the software as best as possible, you still need to match your hardware with other hardware elements like VRAM and RAM. Knowing which setup will suit you, along with which is the best value and which setup you can afford, is very difficult.