This is a great thread because I went back and forth a lot with this. This is a GPU discussion and I am going to talk a lot about RAM and system cores, but it's all important to the GPU discussion because of the way the systems are sold:
I have both an 8 core 56 base model on my desk (Micro center deal) and now also a 10 core 64 GB Vega 64 (16GB VRAM) model in for testing. (There's a very good $700 off deal at B&H via Apple Insider).
I sang the praises of the base model when I got it, and not without good reason: it IS fast. Way faster than a maxed out 2017 iMac in many cases. WAY faster than a 4 core MBP. And there are only a relatively few use cases when all those CPU cores light up, even when using FCPX. But they DO light up when doing things like noise reduction, Red RAW playback, CPU heavy codecs etc.
The Vega 56 chip is only about 15-30% slower (in FCPX, in general) in the iMP. It's not a big jump at all speed wise, especially if you game. But it DOES have 2X the VRAM.. and I think that, combined with the extra cores and slightly higher clock speed - do edge it out over the 56.
I never, ever thought I would say that. I have to eat my words. (I was thinking of a DESKTOP Vega 56, which is very fast and I tried as an eGPU for a while). The Vega PRO 64 in the iMP is about the speed of a desktop 56. Keep that in mind.
I run an additional 5K monitor at 5K off the iMP. At first all was well, but then I noticed a few times when editing that I was tapping out the both the system RAM and the VRAM... and I often (usually) ran with a swap file going on the base 32GB model. The swap file use was moderate to light - but it was happening.
I am not sure if it is because the GPU VRAM was being saturated, and swapping out, or what. However the result was that this seemed in some specific 4K+ editing scenarios to actually slow down the machine - as seen as some stuttering. Although memory "pressure" was low, the system was still chewing through all the RAM and VRAM it had.
I have since performed some very basic testing on the 10 core model. So far, it is noticeably faster than the base model. Measurably, it's about a 25% bump. I think it's because now I'm not saturating the RAM (main or Video). It's not because of the additional processing power, though of course that helps.
I thought of adding RAM. It probably would have solved my stutter issues (it did on my old 8 core D700 nMP). But to go from 32 to 64GB is over $1000 at OWC, and you have to send it in. It's a PITA to upgrade. That also doesn't upgrade the GPU, or CPU. In the end I paid $2100 more to upgrade the machine to 64GB, 2 more cores of GPU, and a Vega 64 with double the VRAM. That was a better deal.
As a result, I will be selling the base iMP model. Luckily because few people live near a Micro Center, I can easily recoup all of what I paid, even selling it with a $500 discount.
Bottom line - it depends on what you're doing. I suspect without another 5K screen attached, or not performing lots of heavy layered 4K video, that the base model screams. It sure did in some of my tests. But in MY case, for even $4,000 paid - that's a lot of money to have a machine that even occasionally slightly stutters. Seeing as I need this thing to last 3-6 years.. it made sense to pony up now to the 10 core model.
The base model is very very fast. The 10 core / 64 GB / Vega 64 model, at a $2100 (current) upcharge, is definitely worth the extra $. It's almost a dollar for dollar jump in performance for what you pay.