If its gaming, Latency and timings make as much of an increase as clk speed, 3000CL15 would likely trump 3200CL16 but you are talking marginal performance, not worth chasing if you already have a decent kit unless you are the type who wants to tweak the crap out of something. Maxing out speed 3866 and fclk to 1900 might see you get ~10% fps bump but you'll have to pay for it, work for it and still need an element of luck, the 3200 ballistix sport LT regularly hits 3733 though allowing you to get part of the way there if you are a tweaker on a budget.
It's not the speed dude. It comes down to the timings. I found some tests a guy had done on the 3000 series and he recommended 3200 @12 or 3600@15 or less. And it is not all about gaming. It comes down to the entire Infinity Fabric process. The faster your RAM is the faster your rig is. It's not really like this on Intel, it's down to AMD (well, Keller's) design. Run your RAM stock as it comes and use Aida 64 memory benchmark. Then start on the timings and keep testing as you go.
TBH playing with the memory was the best bit of fun I had. There is absolutely 0 point in manual overclocking with PBO doing all the work, so I had a whale of time with the RAM.
IF, that's it! I remembered that RAM speed had a bearing on the whole internal system via something, but couldn't remember its name. We'll have to see how it all looks when I finally get to jump into the Ryzen ecosystem.
On Zen 1 it was pretty critical. The IFCLK literally made everything 10-15% faster if you had the correct RAM. On 3000 series it's not quite as important, as they have made huge improvements to it. However, it still likes fast memory
Thanks V-T, they seem to have ironed out the Gen 1 inconsistencies thankfully. I still try to buy used, so would probably test the water initially with a 2600, giving headroom to jump to 3000 series later. Those older system glitches have put me off trying to look for a cheap B350 / X370 board and 1000 series CPU to tinker with. As far as I could see, older boards were still fetching silly money on the 'bay.
don't buy the 2600 unless it is the same price as the 1600AF. There is no point, it's exactly the same CPU.
I installed it and for once it actually booted first time. Whole thing's near-silent, except the low buzz of the Stealth. RAM was 2133MHz but I just changed it manually to the 3200 it should be and now it's 3200. Here's how it compares @ stock (3.6 - 4.2GHz) in AC Origins' benchmark against the 2600K @ 4.2GHz: Stable FPS, staying at the capped 60FPS practically the whole time. Also CPU time is miles better and the GPU doesn't have spikes any more. That's before I get a new cooler and figure out all this overclocking clock. Sounds more complicated than just upping the multiplier and voltage these days. The voltage already fluctuates from 1v to 1.4v so this is already a bit weird.
I just need to wait for my waterblock mount, otherwise I'll be right behind you in a few days! Its my understanding that PBO will do overclocking automatically given the thermal and power excess it detects!
With Ryzen 3rd Gen there isn't much point in manual overclocking unless you're looking for marginal gains. The best thing to do is enable PBO (precision boost overdrive) and give the CPU adequate cooling. It will then automatically boost the CPU cores to their best frequency for the workload.
Wow the stock Stealth cooler really is balls. Playing Origins and its max reached 89C according to Core Temp. Getting a be quiet Pure Rock 120mm tomorrow. Yeah I just read an article where 'stock' v 'PBO + AO' v 'manual OC' had identical results for FPS. Especially as I'm on 60Hz anyway, it doesn't seem worth it. If I'm worrying about FPS I may as well just upgrade GPU. AMD said they don't want to shut out people who don't know about or want to OC so are making sure it's at its best out the box. I'm fine with that! I'm guessing I should turn on PBO after getting a better cooler though.
Only real way is to monitor core speed whilst doing different things. I found Cinebench R20 to be consistent enough between tests to see if any bios changes made a difference. From my experience of a 3900X and a 3600X messing around with any of the "overclocking" options in the bios (not related to just upping the multiplier) resulted in no material change in Windows. Saying that my 3900X also boosted to 4.6Ghz very early on in the X570 bios iterations so I think I was one of the luckier ones.
The best way to know is summed up in one sentence at 23:05 of this video: "Patrick, who wrote most of this script, noted that he hopes everyone can appreciate the amount of time and effort he put into testing a feature that nobody understands and that barely does anything..."
As does HWMON but without all of the bugs and issues. I tried to run RM once on my Threadripper. Never again.