So im planning to upgrade my PC to AMD ryzen platform, most likely the 1800X CPU. With it im planning to buy GTX 1060 6GB, will use it most likely also for GPU rendering...
Im planning to MAYBE (but probably yes) to buy another card for GPU rendering, somewhere in the future... GPU rendering most likely in Redshift (or any other good GPU renderer). I will also by using Vray (RT also).
I have few questions:
1) Rendering with Multiple GPU (probably 2) requires to have them connected in SLI or not? Because only the highest chipset in AMD ryzen family (the x370 i think) will provide the ability to use 2 GPUs in SLI/crossfire. But the mainboards with these chipsets will be more expensive. So i want to know- does GPU rendering need the cards to be connected via SLI? Or can i just plug two GTX (not even the same chip) in PCIE slots and they will render ok?
2) Does plugging 2 same GPUs ( GTX 1060 for example) equal 2x the speed of rendering? Or only 1.7x or something?
Thats all, thank you :-)!
Replies
1) They don't have to be connected in SLI in all renderers that I know of.
2) It won't be exactly twice the rendering speed, but it should be very close to that in general. The third will probably be closer to 1.9 times the speed. Generally, path-tracing renderers scale pretty well with more GPUs.
Also when using multiple cards in other profesional work (like after effects, rendering videos or whatnot) it also works even if the mainboard doesnt support SLI crossfire?
I need to know this, because there will be two mainchipsets for Ryzen (for mainboards with Ryzen support) X370 and B350. The B350 is quite cheaper and only lacks in the SLI support (which it doesnt have), some other features are missing as well but i dont care about it. So even when B350 is missing SLI support, it doesnt matter. As long as the mainboard has 2 PCIE 3.0 slots where i can put 2 GPU cards, it will run ok in graphic processing apps? I just wont be able to play video game using both cards...?
The SLI bridge is mostly used to quickly send the rendered image from the secondary GPU to the monitor output on the first GPU, to decrease latency and so that it doesn't take up bandwidth on the PCI-E bus. Since most GPU compute applications send their result back to the CPU rather than to a monitor, it's not really needed. For most applications, especially path tracing which can be broken up into tiles, there's no need for communication between the two GPUs, so SLI is useless.
I don't know if this still holds true for video encoding.
citation... :"...Ryzen's support for AVX instructions is 128-bit in nature, meaning two cycles are required to achieve what an Intel core does in one..."
so phew... probably gonna be better to wait for the tests...
http://www.eurogamer.net/articles/digitalfoundry-2017-how-amds-ryzen-will-disrupt-the-cpu-market
Also it shows in some test like CPU mark physics (using Bullet physics engine i think) and CPU mark Prime Numbers... + i believe one more test when the performance of 1800x amd ryzen (or 1700X dont know now) are almost half the performance of 6900K... in majority of other tests they are more or less equal (1700Xor1800X vs intel 6900K)...
They do not know yet why but i believe it is because of the AVX instructions...
So now to find out which graphic aplications we artists might be using that support/need AVX instructions... (and what set)...
Does anybody have a clue?
Just about nothing uses AVX512 instructions at the moment, intel consumer Skylake processors (6700K etc) don't even support AVX512 and their Xeon counterparts don't even support all the AVX512 instructions.
The possible use in 3D would be for ray tracing operations but a lot of this kind of work is being pushed to the GPU now so I don't really see it being much of an issue.