X370 Gaming K4 - new BIOS with AGESA 1.0.0.6&a

My point is, in game titles where CPU cycles and/or IPC really counts, @3200 CAS 18 has better average and minimum FPS than @2666 CAS 14, just because the infinit fabric runs at half the memory rate. That's the reason why a lot of users try loosening timmings to bump memory speed up: To gain more frames and/or better minimum frames.

AIDA 64 is not the best tool to conclude gaming workloads in regard of Ryzen memory bus speed and latency. It shows what you said, and choosing the best performance between latency vs speed is the best way if you will use the system as a thread ripper :-)) , but not always these results scales in gaming workloads.

I'm sorry, but I don't get your point about 3DMark results. If I don't see wrong, there are different hardware tested there, only the CPU is the same across results, and with different overclocks too. Also, It doesn't show memory configuration tested on each run (take it with a grain of salt, my eyes are bad at this day's time) :-)

Also, 3dMark favours GPU horsepower, so using different GPU models and core frequencies bring different results. Heck, a lot of times I saw even different results with two distinct runs on the same hardware in 3DMark :-( 

Thanks for your time on this.

Cheers.

EDIT: If you have the time, please take a look at this test benching different dram speeds: http://www.legitreviews.com/ddr4-memory-scaling-amd-am4-platform-best-memory-kit-amd-ryzen-cpus_192259

AND this graph: http://imgur.com/tvtkbtb

I will never purchase another POS ASRock board again, and you are a pathetic representative for the company.

Lastly... for people trying to overclock on this POS board... higher memory voltage doesn't help. It is finding and getting lucky at some random voltage which does the most... for example. I found that 1.395V on my DDR4 was stable to 2966mHz. I forget what 3200 was since things went sour even on blender at that frequency. It was between 1.405 - 1.425 volts which netted a "stable" frequency, anything higher made things worse, but I also found that the system was just not "smooth" for lack of a better word, even though tasks would launch faster. Even at 2966, the system suffered some odd behavior over time such as closing windows using the X at the top right of the window wouldn't respond immediately. All that even with the benchmarks passing, I dropped it to 2800 and things cleared up. 

@Xaltar

First and foremost, PC gaming is not only about AAA console ported games, developed with 1080p @60Hz console's maximum target. And even that, these games also will run smoother with an 5 FPS minumum framerate increase. These games runs smoother with at first barely 5 FPS increase on minimum framerates. But these games yes, need more GPU than CPU as you stated if you want to get the most out them.

But, there are a myriad of games on PC which are a lot more CPU bound than AAA ported titles. Personally, I play racing and flying sims, where CPU calculations are more important on frametime delivery than any eye candy AAA game.

In simracing, a 10 or 15 FPS makes a LOT of difference because the reduced input lag. You can be much more precise on your inputs, which in turn allow you to get better lap times and commit less mistakes.

Also, consistent minimum FPS plays a big role. The user experience will be a lot better, even in 60 FPS locked games, if you go from spikes of 30 minimum to 40 minimun. Much less stutter, more smoother gameplay.

Take Codemaster's F1 2016 for example. Going from 2133 memory to 2800, the minimum framerate went from low 60 to steady 70 on my setup at 1440p, making the game experience a lot better, and reduced the input lag in the process. I'm using a GTX 1070 here, so not the high end.

I take F1 2016 as an example because it's one of the titles which shows more variation between AMD and Intel, and it has a benchmark included so it's easy to reproduce different configs and run several tests with each config.

If you take Assetto Corsa (a more robust sim than F1 2016, which is a simcade title) as an example, it needs even more CPU power in order to do the sim calculations, so if I can get 5 more max FPS and 10 more minimum FPS, the experience, smoothness and input lag translated are all well received.

All of this at 1440p, so mind you what the gains would be at 1080p where the GPU is not the bottleneck.

Also, first person shooters needs more FPS because the reduced input lag it represents, and in competitive e-sports titles this is only what matters, so a gain of 15/20 FPS at 1080p is important to these folks.

Last, you have VR. In VR you are constrained at 90 FPS to get a good experience and to avoid nausea induced by the tech. So, it's all about frametimes. You need <13ms frametimes to get that golden 90 FPS in VR. Less than that, and the headset tech makes its magic an lower the FPS to 45.

If you have a CPU-GPU combination which delivers borderline 13ms, you are much on the fence to a good experience in VR. If you, by any means (i.e. overclocking cpu AND memory bus) can reduce these frametimes to 12ms, your experience will be much better because you will be at 90 FPS more of the time.

I saw situations where my GPU stayed at 70% utilization because the CPU can't deliver the data fast enough to meet that <13ms frametime time, so making the experience a stutter fest mess, going from 90 to 45, to 90 FPS again, even if its constantly pulling 85 FPS, the headset introduced in between frames and reduced the delivery to 45 FPS. You get this, or 80-85 FPS with nausea, it's your choice. This, in a racing or flying sim, is unnaceptable because not only the stutter, but because the massive input lag it intruduce.

So, to finish my wall of text, as I stated in my earlier post, in CPU bound gaming titles, a merely 5/10 FPS increment make a huge difference for some titles. I can assure you is not the same to drive in Assetto Corsa at 60 FPS than 75 FPS, and it's a hot subject on AC forums too, so I'm not alone.

I'm sorry, but I can't agree with what you posted, and this is why.

Cheers.