AMD’s advertising for Ryzen appeared to be a lot more on-point as to the processor’s capabilities than how it used to be when they released Bulldozer (the benchmarks didn’t disappoint). I’m not too worried that they’ll suddenly revert to their old ways for Threadripper (the ball is in their court with all the consumer fury towards Intel).
I found some more benchmarks on the web for both Threadripper and Epyc and its blown intel right out of the water. Cannibalizing their xeon market in price/performance in rendering.
http://www.cpu-monkey.com/en/cpu_benchmark-cinebench_r15_multi_core-8
Cinebench R15 (Multi-Core)
Cinebench R15 (Multi-Core)
Sever chips
AMD Epyc 7601
32x 2.20 GHz (3.20 GHz) HT 5424 points.
Intel Xeon E5-2699 v4
22x 2.20 GHz (3.60 GHz) HT 3240 points
High end consumer chips
AMD Ryzen Threadripper 1998X
16x 3.50 GHz (3.90 GHz) HT 3416 points
Intel Core i9-7920X
12x 2.90 GHz (4.00 GHz) HT 2438 points
Cinebench R11.5, 64bit (Multi-Core) is the same story
AMD Ryzen Threadripper 1998X
16x 3.50 GHz (3.90 GHz) HT 38.36 points… And highly disruptive to Intels Xeon market because.
Intel Xeon E5-2698 v4
20x 2.20 GHz (3.60 GHz) HT 36.28 points.
AMD Epyc 7601 AMD Epyc 7601
32x 2.20 GHz (3.20 GHz) HT 54.95 points. No intel chip can touch it.
Other benchmarks here http://www.cpu-monkey.com/en/
Cannibalization occurs when sales of one product hurt the sales of another product from the same company. Both Intel and AMD take steps to avoid this, only Xeons for example support ECC, only EPYC supports multi-socket, etc. There’s no strong technical reason for this, it’s just market segmentation.
The 32 core Epyc in a dual-chip setup would be an excellent choice for small studios that can afford it, it would allow the equivalent of a sizable render-farm in a single PC (and even 5 years ago, you would’ve needed a long row of PC’s to match that with most chip models).
I’m sure such a workstation for rendering tasks would still be out of reach in terms of pricing for most individuals (I wouldn’t be surprised to see the highest end all-AMD machines sell for at least 8K).
Small studios don’t own expensive hardware in the tens of thousands, it doesn’t make any financial sense. Cloud and other services are way better suited for them.
Technically it depends on many factors. generally speaking having at least a small-sized renderfarm is cheaper than cloud rendering, but it depends on how the client could be charged for a project.
in either case a small studio probably can afford one or two headless systems running threadripper or epyc…and if people use a cloud…or a dedicated rendering service…those ladies and gentlemen may purchase threadripper or epyc…or none of them will…I have no idea. 
if I were rendering cg, I would certainly purchase an amd system for the extra pcie lanes…and fill it with many cards(graphics)…and would write a script to split the image tiles up, render them across the cards and stitch them back together…or send them off to another pc that would simple stitch them back together…that would be nice and I assume would work well with some tweaking.
Yea but when you think about actual workflow, how are you going to preview render your work in the view port if you outsourced all your processing to the cloud. Its not like an actual processor you can connect to over the internet, you have to finalize a render file and send it off to them.
You wouldn’t just send the project off blind. You’d do some test frames on your own system first, and possibly a short test on the service farm if really paranoid. Everything would be locked in by the time you commit to the full run. A lot of it comes down to proper communication with the service provider.
As for cost, don’t forget time required for maintenance such as upgrading hardware, OS patches and upgrades, and so on. One of the benefits of using a service is they will handle that in the background, it won’t matter if you are anywhere close to being an IT professional.
Having a cloud rendering service vs. a local render farm can make a lot of sense. But having a fast machine on your desk is almost more important.
I actually spend most of my day in Clarisse iFx or Renderman (not cycles so much) so my experiences are not going to translate to Blender perfectly. That said, there are many similarities. Being able to iterate quickly (far more quickly than sending an entire scene over to a farm and waiting for finished frames) is the equivalent of spending less money for the same output. As we all know, with cycles (and the afore mentioned renderers) you get low quality, but nearly instant feedback as you work. That iterative process is really important and, so far, not one I’ve seen with a cloud rendering solution.
So I agree that outsourcing a renderfarm can make a lot of sense for a small studio, but even there I think having a fast machine on the artists desk will more than pay for its higher initial cost very quickly.
Benchmarks for Intel’s 18 core chips are starting to come out
Overall, it is 30 percent faster in most multithreaded tasks than Threadripper and retakes the performance crown for Intel. The catch though is that you will end up paying double the price (and the chance that Intel will keep the same socket design for next year’s model will probably be slim to none).
If money is no object though and if you value bragging rights more than common sense, then this is for you (and I wouldn’t be surprised if a lot of enthusiasts buy it anyway).
Sure. I’ll be sure to worry about useless 7-Zip, WPrime, CPU-Z as what I concern myself with when it comes to heavy threaded workloads. Not. There is a reason none of the applications shown are Photoshop, Premiere, Blender, Maya, CAD, etc. Intel’s losing.
By all means buy that while I grab a couple more GPGPUs on a 4 way Threadripper board.
I recommend reading that page again.
In CineBench R15, the Intel outperforms the AMD by ~37% in multi-thread mode
Cinebench is actual rendering engine used in Cinema 4D and has direct linkage to Blender, Maya, and other rendering performance.
I love AMD and their Threadripper, but I’m not blind to performance. Intel simply has better IPC and higher clocks. So anyone with basic calculation skills can do the math and see that Intel will be ahead. Just at a mind boggling double the price…
Main thing Intel got scared of Threadripper, and finally offers something interesting to high end users. with out AMD’s TR, Intel fans would still get the 6 core 12 thread at triple the price of the previous generation. So thank you AMD for that.
Price/Performance though, Threadripper rips through the competition And if I was to build a setup, that would be one of the two options (when a 7 PCIE board comes out). Second would be Dual Epyc, when better motherboards come out. Dual Epyc 7281 (16 core 32 thread each) would be a good rendering setup.
Does multithreading work differently in Premiere?
Indeed, there is one big winner when there´s competition in the market - end users 
While the 18 core chips are impressive, I could never justify paying so much for them…I still side with AMD on this one…
It’s actually 18% faster with same core count. It’s nice but not really if you compare the price.
Well if someone is working on big projects, even freelancer, every second does count. time is money, and for some cases, even the extreme price will be an easy pill to swallow.
I’m just eager to see how (if at all) AMD responds. yesterdays news indicates Zen refresh (not Zen2) on 12nm instead of 14nm early next year. That could mean faster clocks at lower power? Still not enough to dethrone the Intel’s king, but get themselves closer…
I just love AMD for getting back to competition. Haven’t been excited by this tech progress since Athlon 64, and later Phenom… but then … FX… 
Giving Intel room to relax and not push tech forward at the paste then could have.
Still Eger to see how this progresses. And to see finally some Blender benchmarks on this 18core monster. Honestly this already puts my Dual Xeon e5-2687w to shame …