1080TI VS Tesla VS 2080TI

Kyan001

Hi guys, going to choose the best videocard for render. I've seen Daz3D works thanks to the number of NVIDIA's CUDA Cores. 
What is the fastest model between 1080TI, Tesla series (?), and the new 2080TI?

ant2505

I guess, 2080 Ti was developped as faster solution than 1080 Ti from the very begining. Moreover, perhaps one day the new ray tracing possibilities of 2080Ti will significantly boost up render speed with iRay. But if you are going to buy 2080Ti right now, you should understand that usually adapting new hardware for Daz Studio software should take some time to Daz Studio developers (maybe quite a long time), and during that time you may not have an oppotunity to render in iRay. 

Looks like Tesla is more about double precision calculations (need for sciense researches etc), so byuing it only for rendering images - it’s an overkill and a huge waste of money).

chris-2599934

There are some timings in this thread: https://www.daz3d.com/forums/discussion/272071/iray-titan-x-titan-v-1080ti-vs-2080-1180-rtx-turing-discussion

Bottom line is: the 2080 is faster, but only if you're running the (currently) public beta version of Daz. The release version isn't using a new enough version of Iray to make full use of the new card.

DrNewcenstein

For GeForce RTX cards to really shine for Iray, they will have to have VRAM pooling over SLI. 4300 cores is nice and all, but really only about 800 more than a 1080ti, or 500 more than a Titan_xp.
Two 1070tis will give 4800 cores at a cost of $900 direct from Nvidia. You are still limited to 8GB total of VRAM, however.

Tesla cards, for the most part, are Quadros without output ports on them. They are intended to be helper monkeys that perform calculations which are fed to the primary video output device. As they are part of the Quadro format, they're more expensive than a GeForce card with the same number of CUDA cores. VRAM matters, but 4000 Cores on a 4GB card means your scenes will dump to the CPU if they exceed 4GB. If your scenes exceed 12GB, then get a higher-level Quadro, as that's where they blow past the 1080ti and Titan_xp. However, for the cost of a P6000, you could have several 1080tis in a mining cluster where their CUDA cores stack (but not VRAM).