I've been rocking an i7 920 for like 10 years and mi Mobo

>>514215
Looks mostly fine, although I'd have gotten a Noctua D15 over a CLC cooler, especially one with a 120mm rad. Wouldn't want the pump to fail during an overnight render, RIP CPU, or house fire.

The 970 you'll definitely want to change sometime soon, in fact you probably should have gotten a 980 because that 3.5GB RAM might fuck you over if you use a GPU renderer and it decides to use more than that. 3.5GB is easy to burn through if you have more than a basic scene and high-res textures.

I'm not sure how relevant DX12 will be to GPU rendering, since raytracing already uses the GPU to its fullest extent, compared to games which do several unrelated tasks simultaneously that would actually benefit from parallel scheduling. CUDA is just so much more useful that I'd stick with Nvidia for Pascal, even if AMD holds its own in performance benchmarks. OpenCL can eat a dick.

>>514216
ok you just scared me about the cooling,it has to have some kind of security system,and pcs normally shut down when they get too hot.
and i didnt though the 3.5 fail of the 970.

i can always say its broken and make them change it for another card since its marketed as a 4gb card.

btw its 1700€ wich is a lot for me so i expect it to be a monster.

>>514217
Computers should auto shutdown if the CPU gets too hot, but when there's nothing but a bit of water above the CPU you might get a heat spike that's potentially dangerous. At least an air cooler can passively work if the CPU fan stops.

And yes, GPU RAM is important to pay attention to for compute, since game engines can adapt to the amount of RAM you have and continue to work, but stuff like rendering does not.
Of course it all depends on scene complexity, I think that studio or interior arch setups would fit into that much RAM just fine.

>>514219
The Redshift devs claim to have solved the GPU VRAM bottleneck by using an "out of core" feature to stream in relevant data. Scene size is not limited by your VRAM. Other GPU renderer either catch up or Redshift takes the lead.
If this works out, it might be better to have multiple (cheap) GPUs rather than one expensive thing with lots of VRAM.

>>514221
I do hope things like data streaming and memory pooling will make their way into popular renderers soon, but even so, you'll be limited by RAM for the content that's in screen space. there shouldn't be much content outside of your screen space in the first place though.

the 5820k will do jack shit over the i7 920 because - and heres the rub - you're not going to be cpu bound with new apis like vulkan / dx12 which are for the very first time truly multithreaded...you could get a huge speedup if properly programmed on the same hardware you have now

if you really needed to do cpu rendering like vray / prman you'd literally need a renderfarm to get any decent amount of frames done at good quality....

the 970 is real weaksauce

just replace your mobo and get a noctua uh9-be and OC what you already have. if you need a video card get a used 750 ti until pascal

>>514228
On the contrary, because the new APIs will allow CPUs to become even more utilized than they are now, you'll be able to make use of as many cores as you can throw at the task, especially when multiple GPUs are involved. If you look at the DX12 tests of the latest Singularity beta, adding a second GPU already causes the CPU to become the bottleneck at resolutions below 4K - a 4960X of all things.

In the future rendering will be able to make effective use of all the resources you have, so it's never a waste.

>>514229
for all practical purposes the cpu is never the bottleneck, especially now. if you want to have 20 million ships firing at each other you will bottleneck, but even recent 2015 hollywood movies dont have that much