Those tables are so much out of date and irelevant, that I just can not believe it, that they recommend it to even take a look.
I will just point out some things at first glance:
1. Drivers used are old and buggy
2. Graphic cards are waaaay out of date
3. Only low-range gaming gpus are compared
As of upgrading to Archicad 19, GPU tables:
1. Again old Cards with old drives
2. Comparing GTX 750 (150$) to 500$ - 1500$ Pro cards
3. The 750 is the slowest card on the market with only 2GB Vram
4. It says: "some popular gaming cards"
- I don't see them. I see old 2010 and the slowest cards on the market.
The 780Ti was popular, yes. Especially for it's amazing rendering and drawing speed. Its available for 300$ again.
1. Where are the 300$-400$ gaming cards (970, 980, 280, 290, 285, 390) ?
- Example: the 390 offers 8GB VRAM for 300$ and is 3x faster in Maya or OpenGL.
2. Where is the Fury which has blazing fast HBM memory and 4096bus and 500GB/s bandwidth, costs only 600$ (as the K2200) ?
3. Where are the Titan cards, Titan with double precision, Titan X with 12GB Vram, the 980Ti, which scores amazing 150fps in Maya 2013 in specviewperf.
- Recently checked other OpenGL software and software using RT render engines, they all recommend gaming cards from 300$- 1000$ and Tesla cards. Actually none of them recommeng quadro or firepro cards.
The only thing where they recommend those cards are Catia, Creo and Siemens NX. Even solidworks is faster on a 290 (250$) with realhack than a firepro card for 800$.
4. Also we are moving to the FREE Windows 10 (in two weeks) upgrade and there isn't a single test for that either.
Those tables are just wrong on so many levels. I've come here to ask for real relevant comparisons and all you get is this.
This post is no offence in any way, its more of a positive criticism.
Vuego Thank you for explanation...I have other questions, i am sure many will find this interesting ? Lets take two workstation cards which both have aprox 4 TFLOPS SP - AMD W8100 2.1 TFLOPS DP and -M5000 0,1344 TFLOPS Double precision capabilities of the graphic card, how much and where is DP important when working with archicad? And if we use another GPU or accelerator card as Tesla K40 along M5000 (K40 have cca 1,7TFLOPS Double precision as does Titan X black(kepler)), is Archicad going to utilise second card?
No point of utilizing two cards. ArchiCAD is using the video card only in 2d drawing and 3D viewport and ArchiCAD is not well video optimized - we discussed that in other topics.
The most usable and cost effective video card now is quadro k620.