OK, I took delivery of Core i9-7900X, Asus TUF Mark 2 mobo, and 16Gb of 3000Mhz DDR4. I saw the price drop into the mid £800's for an OEM part in the last 72 hours (Threadripper money) and so I've upgraded from an 4.2Ghz long-term overclocked Core i7-3930K with a custom loop of 1 x 360mm rad and 1 x 240mm double-thickness rad, which also included a 980Ti in the loop. I run a very mixed range of workloads and only do casual gaming. Most of what I do will make use of strong IPC first and foremost, but also I can use every core I can lay my hands on. I've had a long look at Ryzen and the IPC/core doesn't quite work for me, but it was close. Intel IPC with 10 cores essentially means that this is another 5 to 6 year investment in this rig for my purposes and look at some of now known architectural design details in Threadripper I've decided I'm not waiting it.
So far, I've rebuilt my rig and after reading various reviews I've been left under the impression that my custom loop would be seriously busy potentially trying to shift a possible 240W TDP of heat I've removed my 980Ti from the loop and put it's air-cooler back on.
I've not properly benched my new rig yet and have simply run a bunch of commonly used workloads to get a direct comparison. The configuration of the new rig was simply to enter the BIOS, select the XMP profile offered by my DDR4 modules and then go and load my OS. Some of my workloads are heavily threaded and so I expected a significant uplift of performance but maybe at the cost of quite a bit of heat, and boy was I surprised.
Suffice to say I would take what has been written on Toms and others with a massive sack of salt. Based on their article I've been worrying myself silly about shifting heat and it has been exactly the opposite. My XMP based overclock has set a 4.2Ghz base and my typical have this thing running about 20c cooler than my 3930K for a similar overclock. I need to do more benching but this thing just isn't getting hot at all, and the TDP hasn't been anything close to the fabled 240w that they claimed they were seeing. I thought it didn't stack up at the time that I read that, and my actual experience is indicating that my gut instinct was correct. Either that article is a crock, or they were reviewing an engineering sample and perhaps my OEM lower tray priced unit is from a more recent batch - because it's freakin amazing! It's early days and maybe I've just not hit a workload that is really going to push it. I've got some AVX based workloads I can throw at it (which is another reason why I didn't wait for threadripper) and it might be that I'll start seeing something different then.