I’ll give Infiltrator a go as well but it’ll probably have to wait till tomorrow.
Would be pretty keen to see anyone with a Ryzen system and faster RAM post results as well, otherwise maybe I’ll play with it sometime, I’ve just never OC’d RAM before and don’t want to break anything.
You should be able to go up to 2666 with just mobo OC profiles (if it has them) or by hand. But right now Ryzen RAM OC beyond 2666 is really difficult due to BIOS issues. Some people managed 3200MHz but it involves really specific mobos, single-rank RAM etc.
But with proper CPU cooling you should be able to pull 4GHz 
Here is a guide how to OC it 
Pretty sure I have Samsung B-die single rank memory (G.Skill Trident Z RGB), the only DOCP (XMP) profile they have is 2400Mhz, it wouldn’t post on 2800Mhz Trident Z RAM on the initial BIOS so I just got something I knew would work. So I should be able to get some good speeds if I put in some reading time.
I have the Asus Prime X370 which might not be the C6H but I’m sure I can do better than 2400Mhz.
I’ve been playing around with changing the ProcessorCountMultiplier like in the link you sent me.
Interestingly it’s compiling ~1.5 mins faster than yesterday at the same settings (bone stock with PCM set to 1).
Not sure if this is AMD’s SenseMI doing it’s thing?
Or it’s just that I’ve had my 40+ Chrome tabs closed today.
Edit for today’s results:
vebski pointed me to this: A new, community-hosted Unreal Engine Wiki - Announcements and Releases - Unreal Engine Forums
And I’ve had a play around with “ProcessorCountMultiplier” to get more tasks going in parallel.
The weird thing is every day the times seem to change -_-
So a recap of previous compile times…
Original BIOS, stock R7 1700, windows high performance power plan: 21.78 mins | 20 mins (Not 100% sure this was at stock settings)
Original BIOS, 3.5Ghz, high performance power plan: 21.25 mins | 19.54 mins
Original BIOS, 3.8Ghz, high performance power plan: 20.61 mins | 18.98 mins
Updated BIOS, stock R7 1700, balanced power plan: 24.85 mins | 22.96 mins
Updated BIOS, 3.5Ghz, balanced power plan: 22.78 mins | 20.97 mins
Updated BIOS, 3.77Ghz, balanced power plan: 21.25 mins | 19.54 mins
On to todays stuff, all of these first ones are on the balanced power plan as I didn’t realize it was on that setting -_-
These first ones are also at stock clocks.
8 tasks in parallel: 24 mins | 22.25 mins
12 tasks in parallel: 23.39 mins | 21.51 mins (Why is this over 1 minute faster than yesterday?)
14 tasks in parallel: 23.32 mins | 21.45 mins
16 tasks in parallel: 23.30 mins | 21.49 mins
So we can see that SMT/HT is not giving a whole lot more performance =
Then I went for 3.5Ghz at 1.2v.
16 tasks in parallel: 22.45 mins | 20.64 mins
Tried giving her more juice, 1.3v.
16 tasks in parallel: 22.35 mins | 20.55 mins
Then I realised high performance plan wasn’t on, so I turned it on.
16 tasks in parallel: 22.04 mins | 20.4 mins
Just about to do another Lightroom bake to see how that goes today, not enough time to get Infiltrator today as it’s 10PM and I haven’t actually done anything yet today -__-
I’d love to know why I’ve lost 1+ minutes of build time after swapping to the new BIOS…
Then I would guess due to dependencies in compilation of UE4, more then 8cores/threads will not speed up the process. Unless somebody with Intel (hexa/octa core) will get faster times with 12/16 parallel workers.
Well the other thing is, when doing 8 in parallel task manager shows all 16 threads going at 100%.
At least for the periods where it uses more than ~2/4 threads.
The lightmass test was basically the same time as yesterday, 1m44s.
I will try at home with my 6700k, I wonder what times I will get. Btw. your UE4 code is on HDD or SSD?
It’s on a Samsung 960 Evo M.2 SSD.
Task manager shows it’s utilization almost always between 0% & < 10%, low latency though.
Removed the peak that I saw, can’t possibly watch it all the time.
EDITZ:
Something I didn’t post earlier, was my 7 threads in parallel compilation result…
Back with the balanced power plan at stock settings.
7t :1502.76 seconds (Local executor: 1391.11 seconds)
8t :1444.86 seconds (Local executor: 1335.17 seconds)
12t :1403.50 seconds (Local executor: 1290.93 seconds)
14t :1399.52 seconds (Local executor: 1287.20 seconds)
16t :1398.59 seconds (Local executor: 1289.72 seconds)
So your m.2 is definitely not a limiting factor Mine is on SM951 so at least we can compare without worrying about differences between hard drive read/writes.
I’ve had nothing but an old Seagate Barracuda for the last 10 years.
Feels good man.
EDIT:
GEHHH, Ryzen Master seems to take up a bit of CPU resources, so I’m building again to see if that can explain the discrepancy between my first runs and the newer runs.
Using an SSD provides a noticeable boost in compilation time on my 6700k. Using a RAM drive helps to shave off some extra 5-10% :D. I’m looking forward to 16-core Ryzen if it comes to be real.
Ahhk.
I should be able to compare SSD vs HDD next week.
Oooookay
Closing Ryzen master = 21.49 mins | ??
Closing Ryzen master & Asus Aura = 21.11 mins | 19.5 mins
Noticed them using 6-10% EACH.
Lightroom builds in: 1:35 min total, 138 ms importing, 31 ms setup, 3.75 sec photons, 1:31 min processing, 0 ms extra exporting
Since I’m not getting anything else done tonight… Getting Infiltrator now.
First result with the Infiltrator Demo on production settings
i7 5930k @ 4.2 GHz
51:17 min total, 6.27 sec importing, 13.2 sec setup, 2:10 min photons, 48:47 min processing, 0 ms extra exporting [11951/11951 mappings]
I’m going to run it one more time, was using the computer for web browsing and videos at the same time.
I just knocked a quick preview quality Infiltrator light build out, will do production tomorrow gotta go to bed.
9:52 min total, 4.96 sec importing, 15.6 sec setup, 1:02 min photons, 8:29 min processing, 0 ms extra exporting
EDIT: @3.5Ghz
Preview i7 5930k @ 4.2 GHz
13:18 min total, 6.94 sec importing, 13.2 sec setup, 1:34 min photons, 11:23 min processing, 0 ms extra exporting [11951/11951 mappings].
Production #2 i7 5930k @ 4.2 GHz
48:30 min total, 6.41 sec importing, 13.0 sec setup, 2:09 min photons, 46:02 min processing, 0 ms extra exporting [11951/11951 mappings]
You are likely just RAM limited, so it doesn’t matter how much CPU power your throw at it.
If the CPU is waiting for the RAM it still shows up as 100% used, even though it might only wait for the RAM to deliver new data or write data.
Having a play around with things:
1.25v Infiltrator preview light build: 9:55 min total, 4.98 sec importing, 15.5 sec setup, 1:05 min photons, 8:29 min processing, 0 ms extra exporting
1.2875v Infiltrator preview light build: 9:44 min total, 5.23 sec importing, 15.3 sec setup, 1:02 min photons, 8:21 min processing, 0 ms extra exporting
1.3v Infiltrator preview light build: 10:19 min total, 5.25 sec importing, 15.3 sec setup, 1:04 min photons, 8:53 min processing, 0 ms extra exporting
1.25v Infiltrator production light build: 50:08 min total, 5.26 sec importing, 15.1 sec setup, 1:47 min photons, 48:00 min processing, 0 ms extra exporting
1.2875v Infiltrator production light build: 49:09 min total, 4.81 sec importing, 15.2 sec setup, 1:44 min photons, 47:05 min processing, 0 ms extra exporting
Out of interest, how many cores/threads does lightmass report in the log?
thread 0, 4 & 5 aren’t showing max utilization in task manager.
You could well be right, I’ll get to play with it sometime but I’ve done nothing but this for the last 3 days hahaha.
EDIT: Should have said the above is 3.5Ghz.
Infiltrator production build after doing a Cache Clean on a 1700x @3.8GHz (max stable speed I could reach with a AIO water cooler :/) with 4x8GB DDR4 @2133MHz (the memory is rated at 2666MHz, but as I am running 4 sticks and even 2400MHz is still unstable in Lightmass and the builds fail(Seems Lightmass really stresses Ryzen as this does not seem to happen in synthetic benchmarks) 
) memory, using 16 threads (Lightmass defaults to 14, you can set it manually in dev options), while doing some light web browsing in the background:
46:35 min total, 4.13 sec importing, 14.1 sec setup, 1:35 min photons, 44:41 min processing, 0 ms extra exporting [11951/11951 mappings]. Threads: 12:39:46 hours total, 11:19:36 hours processing.
Did you touch the voltage?
Yup, this is at 1.3125V even cranking it way up to 1.45V (and a bunch of values inbetween) at higher clock speeds I got the mysterious screens turns black crashes and invalid memory read errors(especially in Lightmass under high load)/bsods related to it.
Took me a while but I did a test on mine and friend’s PC.
UE4.15.1 Development Editor
i7 6700k@4.5GHz with 16Gb 2666MHz
21min 5sec
i7 5820k@4.0GHz with 16GB 2666MHz
18min 30sec
both on Samsung m.2 with 2.8GBs reads.
From perspective of a programmer I am bit disappointed with R7. It seems 5820k beats it for same price (and now you can get even faster 6800k) in compilation, even normal i7 gets same time.
But for people with more usage for R7 power like lightmass or other heavy mt it seems like a good choice.