Which hardware should handle Archicad 21 better?

[Anonymous]

I have a question regarding 2 possible setups for Archicad 21. I have a computer (lets call it OPT[ION 1) which I'm not sure whether it's utilised in its full potential when rendering and I might want to build another one (lets call it OPTION 2)

OPTION 1:

CPU: i7-9700K
GPU: Nvidia RTX2080 Ti
RAM: 16GB

OPTION 2:
CPU: Ryzen 3950X (or perhaps any threadripper under 600$?)
GPU: Any other Video card (perhaps Vega, 1070? something else?)
RAM: Would upgrade to 32GB

I noticed that rendering on 1000$ machine (build from Ryzen 2700X and 1070 GTX) is 20% faster than rendering with the same settings on computer OPTION 1.

Is there a cheaper option I could fit in OPTION 2, which will give similiar if not better results than OPTION 1? This way I could use OPTION 1 for other purposes.

I've heard Archicad does not utilise GPU, so I might get better CPU instead and a bit slower GPU?

[Erwin Edel]

Cinerender doesn't use GPU for rendering, this is true. However you want a somewhat decent GPU for working in the 3D window and for showing the 2D views at reasonable speed / frame rate.

That said, I'm working with a pretty old Nvidia Quatro K2000 and it is still holding up well enough for AC22.

For cinerender: a good multicore CPU and at least 16 GB of RAM (for say max A4 size 300 dpi renders) or more if you need to render larger / more complex scenes.

Again though, aging Xeon processor here, does the job. HP workstation with 16 GB ECC ram still holds up for AC22 and does our renders quick enough.

[Lingwisyer]

So you are not planning to use any 3rd party renderers?

The Ryzen is probably faster for CPU rendering as it runs 16 threads at 3.7GHz while the i7 is 8 threads at 4.9GHz. For things outside of rendering the i7 will perform better as a lot of things are not multithreaded so the faster single core speed is more advantageous.

Threadrippers are amazing value for high core counts, but I have heard of some people having issues in games caused by the latency between it's two CPUs when it was first released. This is probably not that noticeable under a work application, or may have been solved since. It apparently works amazingly well in Maxwell and I would assume CineRender. Again, for single thread applications, you will lose performance due to the slower core speed.

If you are not doing GPU intensive work, the GTX2080 Ti is way overkill. You will never make full use of the speed, let alone the VRAM.



Ling.

[Anonymous]

Thank you both for your answers, much appreciated. I have few follow up questions

@Erwin Edel, I'm definitely look at building at least 32GB. What do you consider "somewhat decent GPU"? I was thinking about getting something in the range of GTX 1070 performance wise (perhaps cheap quadro/firepro?)

@Lingwisyer, If games are the only issues that threadrippers have issues with, I'm not bothered as computer will definitely not be used for gaming. I wonder how threadrippers compare to new Ryzen 3000 series, when it comes to rendering speeds in Archicad.
3rd party renderers are not planned to be used in near future, so perhaps going with something like 1070 GTX is good enough? If (in theory) 3rd party renderers would be used, how much of a difference does 2080 RTX make compared to 1070 GTX? I know that for gaming, it is far better, just not sure how much of a difference it makes when rendering.

[Erwin Edel]

I've used Nvidia Quatro cards for years, they are very stable and have excellent driver support for windows. However, a gaming card might give you a bit more value for your money if all you are after is raw performance.

I think there are a few more recent threads here where people chime in on experience with cards.

I'd also check some hardware sites for 'best value' type guides. I don't think you need the latest and greatest for ArchiCAD, not sure at which model / make the sweet spot is currently at for performance v.s. price.

[Lingwisyer]

You will have to check out bench scores for the performance difference between different cards. Sometimes there are even renderer / engine specific benchmarks. Performance gains will also vary depending the size of the project / scene.



Ling.