Thats a good question. Since it has basically two (or more) dies inside the CPU itself you would want the best coverageon the entire IHS. Dot might be best or even the X method. However it will also depend on what TIM you are using for it as some spread better than others in certain methods.What Thermal Compound application method would you recommend for the Ryzen 9 3900X (Vertical Line, Middle Dot, etc)?
When testing CPUs you want to put the limit on the CPU not the GPU. This was a CPU review.Okay, I know I'm late to the party, but why did you test a high end system with VRMark Orange Room? This is not a true test for a CPU and GPU! It's just to determine if the system is VR Ready!
From VRMark Developer UL_Jarnis over on the Steam Forums:
"Orange Room, quite true, is not a normal benchmark. It is intended only to figure out if you meet the minimum requirement of the HMDs. It gets CPU limited real fast. It is NOT a graphics card performance test. It is not intended to be that. So comparing two PCs to each other with it is not what you should be doing. The documentation does tell you this, but after almost 20 years of 3DMarks, I can see how it is easy to miss this fact.
Blue Room and Cyan Room are just as much "proper benchmarks" as anything, they are purely GPU-limited so will scale well and specifically test performance in rendering VR. This specific thing does mean that they only scale to two GPUs in multi-GPU, but this is the limitation of VR APIs as they do not support 3+ card configs - and in all honesty vendors are phasing out 3+ GPU multi-GPU configurations. 2080 for example can only be set up in 1 or 2 card configuration at all. "
What "home user" software do you find speed lacking and experience can be improved by and only by faster cores?A 12 core CPU will not be any faster than an 8 core for home users. The software used does not scale with that many cores. If it was so easy to make typical software scale with more cores, programmers would have already done it. Most mainstream software is not capable of being highly parallelized like a graphics workload, so this problem isn't going to get fixed without a completely different type of computing platform. Adding more cores doesn't make a PC faster when what you really need is faster cores.