LeStringMan

At launch yea the 970 was probably just a hair faster than the RX 470, with the ReLive update there was about a 4 or 5% increase of performance on the Polaris cards, which pretty much evened the 470 and 970 out. Also i would say it depends on which version of the 470 you are talking about, some have a tdp target of 120w (which is what AMD rates them at), some have a higher target like the RX 570.

You are correct thermal margin in overdrive is how many degrees you are from the thermal throttle temp which is 70c. AMD recommends a max of 61c on the FX series.

Prime B350-Plus, got the bios beta on June 7th, the exact same day as the X370 boards, the B350M has been late for all the updates. For me i could get 2666 previously, now i can get 3066 CL16, and 3200 but at CL18, im running a B350 Prime Plus, and Team Dark DDR4 3000 CL16 with SpecTek (Micron) memory.

Beta bios for the microcode have been out for a little over a week on B350 boards, ASRock and MSI were a little late to the party though.

Slow boot times have been a problem on MSI boards in the past, as for having to reset the bios before booting, that could be a ram speed issue with Ryzen, make sure you update to the newest bios with the AGESA 1.0.0.6 micro code, which should help with memory speeds and stability. On a B350 it will probably be a beta bios, and i don't know if MSI has it out for the PC-MATE, but the board is literally identical to the Tomahawk so i assume it is out.

His main channel is Actually Hardcore Overclocking, he also has a guide with a brief summary of all the B350 boards.

That explains the RAM speeds, the Prime Plus, at this point, has frankly terrible memory speed support. I thought dual channel would help quite a bit as well, like i said i didn't test going from single to dual really on Ryzen , but i did test with Fallout 4 and a few others on my 8300 (4.1ghz @ 1.25 volts.), and dual channel did almost nothing (slightly better mins) (could also have to do with Piledriver's terrible IMC)

just the cheapo stuff Aegis or whatever it is, i bought it because it was almost the same price as 2133, what board are you using? also going from single to dual channel did almost nothing when i tested on my FX-8300 ( was running slow stuff though,JEDEC standard 1600 mhz CL11.), but idk about Ryzen.

So first a little backstory, i upgraded from an FX-8300 to Ryzen a few weeks ago and went with a R5 1600, Asus Prime B350 Plus, and a single 8gb stick of G.Skill rated for DDR4 3000 CL 16-18-18-38 (SK Hynix memory chips.).So that ram is not on the QVL list, and i couldn't even get the stick to run at 2400 mhz CL16, it was kind of sad really, anyways i get an email from Newegg with a promo code for 16 gb of Team Dark rated for DDR4 3000 CL 16-18-18-38 (SpecTek (Micron) memory chips.)($99.99) So i check the Asus QVL list and they are supported, but only at 2666 mhz, i got them today and decided to run some benches, so here are the results for anyone interested. (The games tested are games that i happened to have installed and have an in game benchmark tool.) Test Setup: Ryzen 5 1600 stock. Asus Prime B350 Plus 8GB single channel DDR4 @ 2133 CL15 vs 16 GB Dual Channel @ 2666 CL16 MSI RX 470 1230 core clock, 1750 memory clock +25% power limit. Results: Cinebench R15: ( i only ran this once, and with priority set to realtime.) SC 8gb 2133 Score: 1136 DC 16 gb 2666 Score 1152 Bioshock Infinite 1080p Max Settings SC 8gb 2133 Min 12.41 Max 171.41 Avg. 96.56 DC 16gb 2666 Min 30.37 Max 158.23 Avg. 98.43 Bioshock Infinite 720p Low SC 8gb 2133 Min 29.49 Max 499.38 Avg. 194.38 DC 16gb 2666 Min 29.65 Max 518.89 Avg. 208.39 Deus Ex: Mankind Divided (DX12) High Preset 1080p SC 8gb 2133 Min 48.3 Max 75.5 Avg. 60.2 DC 16gb 2666 Min 49.5 Max 76.2 Avg. 61.0 Deus Ex: Mankind Divided (DX12) 720p low SC 8gb 2133 Min 68.0 Max 112.7 Avg. 86.6 DC 16gb 2666 Min 86.5 Max 146.6 Avg. 111.5 Hitman(2016)(DX12) 1080p ultra SC 8gb 2133 Min 0.87 Max 184.88 Avg. 64.86 DC 16gb 2666 Min 0.94 Max 207.45 Avg. 74.31 Hitman(2016)(DX12) 720p low SC 8gb 2133 Min 1.02 Max 207.18 Avg. 74.31 DC 16gb 2666 Min 0.97 Max 289.44 Avg. 102.07 Total War Rome II 1080p Extreme Preset Unlimited Video Memory On SC 8gb 2133 Min 34 Max 76 Avg. 53.5 DC 16gb 2666 Min 34 Max 78 Avg. 58.2 Total War Rome II 1080p Ultra Preset Unlimited Video Memory On SC 8gb 2133 Min 25 Max 123 Avg. 76.9 DC 16gb 2666 Min 58 Max 136 Avg. 83.6 Total War Rome II 720p Very Low Preset Unlimited Video Memory On SC 8gb 2133 Min 179 Max 285 Avg. 225.2 DC 16gb 2666 Min 174 Max 274 Avg. 231.8 Total War Warhammer (DX11 because DX12 is broken in this game) 1080p Ultra ( should have benched high preset instead) SC 8gb 2133 Min 49 Max 64 Avg 58.02 DC 16gb 2666 Min 34 Max 66 Avg. 59 Total War Warhammer (DX11 because DX12 is broken in this game) 720p Low SC 8gb 2133 Min 129 Max 198 Avg 180.8 DC 16gb 2666 Min 128 Max 214 Avg. 198.5 War Thunder Tank Battle CPU Benchmark SC 8gb 2133 Min 46.2 Avg 70.0 DC 16gb 2666 Min 56.5 Avg 75.5 Total War Shogun II CPU Benchmark SC 8gb 2133 Avg:19.687 DC 16gb 2666 Avg: 22.1739 TL;DR Ram speed helps quite a bit with Ryzen even with something like an RX470. (Hitman 2016 gained nearly 10 fps average at 1080p ultra.)

Did you install an older build of Windows 10 off a disc by any chance?

Lul FX-8300 at 4.1ghz and MSI Armor RX 470, which seemed to stay in the 1280 core clock range for most of the bench.

There are also benches where the 2550k and 3570k barely win, my point is simply that every one was cherry picking benches in this thread, and that the FX series is not nearly as slow as some make it out to be, in fact for most task it is indeed faster than a 2550k, obviously gaming is not one of those, but even that is closer than people give Piledriver credit for. As for the price of the 3570K. http://ark.intel.com/products/65520/Intel-Core-i5-3570K-Processor-6M-Cache-up-to-3_80-GHz

This is normal, 970, 990X, and 990FX chipsets run pretty warm.

@LAwLz, @Strike105X, Here is Tomshardware review of Piledriver when it launched. If you are too lazy to click the link and look for yourselves and just you know.... read.... then i have already tallied it for you, Tom's ran 35 cpu bound benches (i excluded GPU bound things such as 3dmark gpu score, gaming results, and also excluded power consumption.) the ("horrible") FX-8350 was faster than the 2550K in 23 of 35 benches, the 2550k won the other 12 obviously, they tied in none. The 8350 also beat the 3570k in 21 of 35 benches and the 3770k in 1 http://www.tomshardware.com/reviews/fx-8350-vishera-review,3328.html. I would do the same for Anandtech's review of the 8350 but they pretty much exclusively tested synthetic benches in that review, and it is very short.(http://www.anandtech.com/show/6396/the-vishera-review-amd-fx8350-fx8320-fx6300-and-fx4300-tested). Also the 8350's original launch price was $195 and it was marketed against the 3570k which at the time retailed for $235.

Infinity Fabric is based upon, and is an improvement on AMD's Hypertransport Link, HT Link was first introduced in 2003 in AMD cpus, and replaced the traditional Front Side Bus, Intel's QuickPath Interconnect does the same thing and was introduced in 2008.