I feel like there's such a disconnect between all the talk about Ryzen and the reality of quality of life with a given piece of hardware. There's this assumption that all that matters, at the end of the day, is the raw FPS number on screen, and that FPS number (on a few existing titles) divided by price is the only metric of consequence.
We've become used to ignoring AMD and our only scale of relative comparisons from Intel, generation to generation, has become 'Where is the CPU bottleneck FPS?', on the automatic assumption that virtually all threads are fully utilized. Intel absolutely has had teething issues in the past on new architecture (and avoidance of this, and the backlash similar to the Ryzen reception may be why they haven't bothered to introduce any meaningful architectural differences in 5 years)
I've been running on a 1700X system for about a week now, and while my comparison point is to my old 4690k, instead of a 7700k like most reviewers are, the differences between those two aren't really that dramatic (~15% more instructions through a combination of IPC and higher clocks, 30% more total throughput on the Kabylake due to HT). But the 7700k is 100%, to the balls maxed out in all the comparison circumstances. When someone sees Ryzen producing fewer frames, the automatic assumption is that it, too, is either maxed out, or has a single threaded bottleneck. It's even worse with these nonsense 300-500 FPS tests, that are only testing the draw call capacity of the CPU, not the actual computational bottlenecks that exist in a real gaming situation.
But that isn't what's happening, in any of those circumstances. On most of the games I've tested, I'm getting one of two results:
- The CPU is the 'bottleneck', but no single core is over 75%, and the total use is under 30-40% - the Windows scheduler is moving data across the CCX, and not knowing how to manage that subsystem properly, is keeping any threads from even maxing out. Resultingly, my FPS is about the same than on my 4690k, and lower than the 7700k's in reviews - but at very low utilization, power use, fan use, noise, minimal thermals - while the slightly better results are going up against max fan use, 100% threads maxed out, 80C thermals, etc etc. With a few exceptions, this usually puts me at 120 FPS instead of 140 FPS. Overclocked Kaby Lake's evidently run about as hot as the surface of the sun, by comparison.
- The GPU is the bottleneck, the threads are properly distributed, and I'm getting the same frames as the 4690k/7700k, but again with massively lessened power use, heat, temperatures.
Once there are some optimizations to scheduling, new games that are programmed to use more threads (it wasn't worth anyone's time in the past, when 99% of the population used 4 cores or less), and the memory issues are sorted out, this thing is going to be insane. AMD rushed the launch, no question, but the reception that Ryzen isn't worth looking at because, hurr durr, the frames are lower! - total idiocy. The single thread performance is awesome, the threaded performance is awesome, but the co-ordination of the two needs some work.
As an aside, it's unbelievable how efficient this thing is when you clock it down and drop the voltage. I was pulling a 40 watt delta over idle while running stress tests at 3 GHZ/0.88 volts, and temperatures of 35 F. And it keeps scaling very efficiently up until you get to about 3.8 GHZ. That's going to make for a monster server processor.
Now, Vega, I have zero faith in, but you know Intel is feeling the pressure behind the scenes.