CPU Upgrade, which one?

Hello everyone, I want to upgrade my PC. I have currently installed a Ryzen 1800X and am considering buying the Ryzen 5950X. He is praised everywhere and seems to be a real beast. However, this CPU is not available and I therefore looked for alternatives. I found the Threadripper 3960X which has twice as many cores, but a slightly lower clock rate. In terms of price, it is also slightly higher than the Ryzen 5950X. I mainly render with my 2080ti graphics card (I would like the option to add a second one at some point). I would need the CPU for animating water, flip fluids or mantaflow and clothing. Which CPU would do better in this context? I ask because I’m not sure whether the application benefits from the many CPUs or whether I should rather look at the single core clock.
The thread rib would still have quad-channel (would I benefit from it?) - What do you think? Would an ECC memory be worthwhile? both CPUs could do that. And what I don’t understand is the difference between the PCIe lanes - what are the advantages or disadvantages?
Lots of questions, but hopefully you can help me :slight_smile:

Assuming you’re referring to Blender only, CPU is not that important when dealing with the things you enumerated. They are done by running vertex shaders on the GPU, so your main concern should be the video card.

1 Like

if i bake a simulation with flipfluids i thought it would use only the cpu?

I guess it depends on what you use. I was thinking more about real-time rendering. Yeah, FLIP Fluids uses a lot of CPU. In this case, you should look for whatever CPU has better multi-threading support.

1 Like

CPU is very important in Blender (Sculpt, Edit mode, modifiers, copy attributes, simulations, etc). Processes are very CPU dependent, and many of them are single threaded processes, so when choosing a CPU, don’t neglect single thread performance.
Single Thread Performance

So for blender in general (Rendering and for working), when choosing a CPU you get something balanced both in Single thread and in multi process.

You can check for yourself with a good CPU usage monitor, preferably one that graphically shows the work of each thread. So while you are running a heavy task in Blender, you keep the CPU monitor in the foreground.


mainly I want everything to go faster: D. on my last project i worked with the ocean modifier. if you increase the result and the repetition, my computer slows down. the ram is also full quite quickly. I currently have 32GB. but would like to upgrade to at least 64GB, better still 128 - this would then require the thread ripper. if I also have other programs open, such as photoshop or aftereffects, things get tight very quickly.
would I have any disadvantages (apart from the price, of course) with the thread ripper? I have so many questions - I’m sorry … I still haven’t understood whether you benefit from the ecc ram or quad channel in blender? or what the pcie lanes are all about, which are probably different with the threadripper platform.

As I said, I render with the 2080ti - if the cpu were fast enough, I would perhaps also render with both, I would have to test. But apart from that, I noticed that my pc reaches its limits, especially in more complex scenes. if i want to simulate smoke, liquids or clothing, it takes a long time.
i would like a solid and stable box.

I think for your use case you want a nicely balanced cpu and it kind of depends on budget but out of those two I’d probably go with the 5950x unless you are rendering with cpu also. I think flip fluids is multithreaded but I don’t know about mantaflow. How does your current rig handle these tasks? YAFU has very good answers and you should check out the CPU monitor as suggested.

You can divide the CPU mark and single thread ratings by the price you can actually by them for to get overall and single thread price performance numbers.

Bear in mind if you aren’t rendering unless I missed something the only benefits you’ll be getting from tons of cores are the third party simulators and you should make sure these can handle such an enormous amount of threads. From flip fluids:

• Some calculations in the simulator are not able to be multithreaded efficiently and must be run on a single core. These sections of calculations create a bottleneck which lowers average CPU usage. • The simulator may be running with too many threads enabled. It is possible that simulations could slow down from the overhead of running more threads than the simulator can handle efficiently.

1 Like

On the site that I have posted above you can do side-by-side CPU comparisons.
Regarding RAM, I have always read that for AMD CPUs it is important that you get the fastest RAM you can buy.

Since you say that you mainly render on GPUs, my personal choice would be to wait for the AMD Ryzen 9 5950X.

Regarding Flip fluid addon, I recommend that you ask this to its developer.

But an important question would be, in which area of Blender are you currently having problems with your Ryzen 1800X?
That CPU is not bad. And you should be aware that Blender has its own limitations in some areas and a better CPU will not make Blender significantly better.

1 Like

To beat render time of 3080 you would need more than one 3990X and to mortgage your house. However you would be free from vram limitations on a threadripper which would speed things up by a lot if you use a ton of high res textures or something.

1 Like

thank you for all your answers!

if I look at the comparison, the 3960x does a little better.

i am trying to describe my needs and my problems a little more precisely here.

  1. if i want to simulate smoke and bake the simulation, the process takes a long time.
    the same applies to the baking of the simulation for water and clothing. So here I want more performance.

  2. the computer begins to lag when i have a lot of geometry in the scene. e.g. with the ocean modifier or a lot of vegatation with the grasswald addon. so i want the computer to run more stable and fluid here too.

  3. occasionally my scenes are too big for the graphics card’s vram. so i don’t exclude the possibility of including the cpu when rendering. with the 1800x that doesn’t make sense at the moment because it is too slow compared to the 2080ti.

  4. the next hardware / cpu upgrade i am doing should keep the possibility open for me to use a second graphics card in the future.

Questions I have:

  • is ECC RAM useful to me?
  • What are the benefits / disadvantages of the different pcie lane connections?

5950x has 16 cores and the threatripper has 24 cores. should you really notice it clearly, or am I getting something wrong?

You need to get a motherboard with the best PCIe for your video card configuration (especially if you want 2 video cards). Basically, you want x16 and x16/8 depending on your video card dual compatibility. It’s difficult to build a system and you should carefully check out all hardware features before making a buy.

Here’s more details about PCIe:

Regarding Cycles render, threatripper is probably faster. Regarding simulations, there are not many simulations benchmark out there, so it is difficult to know exactly. You could create a scene with simulation, contact a user with the CPU you want to buy and ask him to take the simulation baking time to compare with your current CPU.
You see this 2.7x particles simulation benchmark what is a multi-thread job (user results at the endd of the link):

  • AMD Ryzen 7 1800X (16 threads - Single Thread score: 2.178) = 03:31.92
  • Intel i5-8600K (6 threads - Single Thread score: 2.642) = 03:32.00
  • AMD Ryzen 7 3700X (16 threads - Single Thread score: 2.690)= 02:07.00

So at least up to Blender 2.7x my experience with that benchmark is that even though it is a multi-thread job, single-threaded performance influences more than number of threaded in this simulation.


thanks for the link.
to render the graphics card does not have to sit in a 16x slot because it is only transmitted once?

well, I still don’t quite understand it. I’m not sure what advantages and disadvantages I have with a threatripper regarding the pcie lanes compared to a ryzen. But well, the only important thing is that there are enough slots that provide at least 8x performance. the problem with current cards is the slot height … So even if there are enough slots, that doesn’t mean that the cards fit in.

what would you say is important for stable and fluid work? especially when it comes to large and complex scenes?

For fluid work in Blender you should get the optimum system requirements:

Then you can look for some workstation builds like this:

I guess i will wait for the 5950x.
Does anyone know when it will be available?

I’m also waiting on the 5950x to be available, been running a 1700x for a couple years now. From what I’ve read, either 5xxx series or Threadrippers will be awesome, but what tipped the scale for me personally is that the 5xxx will fit in my current AM4 motherboard whereas the TR chips require a new mobo as well.

Yes. I also need a new motherboard. I think I need a completely new PC besides the GPU. but I’ve read that the motherboard manufacturers may release a new BIOS version that will allow you to use an x370 board for the 5000 Ryzen. but I do not know for sure