Can confirm that the latest (0.7.21) TRM miner works well with my 2 x RX 570 Sapphire Pulse 4Gb cards,
currently both mining ETH at 28.3 MH/s each.   This is on Xubuntu 16.04 with driver version 17.30.  
Here is my command line

teamredminer --platform=0  -a ethash -o stratum+tcp://europe.ethash-hub.miningpoolhub.com:20535 -u xxxxxx.$WK -p x \
 --eth_4g_max_alloc=4076,4076 --fan_control=:::44,:::26 --pool_no_ensub


A good thing of this miner compared to Phoenixminer is the clean startup and shutdown, as well as lower power consp.   No more hangup requiring box reboot.
Will be keen to see how well the DAG spillover to system mem works into the coming weeks.

Why stop at CUDA 7.5?
I just compiled a version of ccminer for my GT 640 with compute capability 3.0 with CUDA 6.5  and it is now mining Digibyte :-)
Not for profit, just for fun.  The poor thing doesnt even have a fan so will see if I dare to leave it alone for a day..

So I upgraded to PhoenixMiner 5.3b and AMD driver version 19.7.5.  Those changes did not change anything!
Then I added command line options  -rxboost 50 -mt 1  and voila - all cards except 1 are above 30MH/s again :-)
Some further tuning is required and I will monitor the amount of memory errors in HWinfo64.
Since the rig is in a remote location and I log in using Chrome remote desktop I dont want to do any BIOS mods before I get to go on site.
Thanks for the tip to use rxboost, mate!

Guys,
I have a rig with RX 470 8GB cards mining ETH and they were running at around 30MH/s each, but today I noticed a drop in hasrate to around 29-29.5  on all of them.   Surely, they should not be affected by the increased DAG size yet?  Any of you noticed the same?
The current driver version is 18.5.1.   All cards are BIOS modded.   I wonder if a higher driver version will bring hashrate back up to 30?
I already checked ping rates and they are unchanged.
EDIT: I am using Phoenixminer 5.1c

Update:
Had to move to Linux (Xubuntu 16.04) after hashing speed dropped to 27 in Windows.
In Linux I am back at 30MH/s on both GPUs, guys 
Using miner version 5.1c  and AMDGPU-Pro 17.40  and kernel version 4.18 - old but trusty
Cheers,
#Sapphire rules  #Phoenixminer rules

Here is my miner command line

PhoenixMiner.exe -amd -clKernel 1 -mode 1   -pool %POOL1B% -pool2 %POOL1% -wal bobben.%WK% -pass x -proto 1

This is how it looks like (Remoted in to rig via Chrome Remote desktop, so a bit blurry):
https://imgur.com/a/T1juwRi

Can confirm that 5.2c is working great with my RX 570  4Gb cards

Windows 10
Adrenalin 2019-Edition-18.12.3-Dec19 driver

2 x Sapphire Pulse Rx 570 4Gb cards mining ETH at 1100/2000 core/mem

hashrate now 29.9MH/s
hashrate before the DAG limitation: 30.3MH/s

That depends on a couple of things (at least):
1. CPU type.  Generally, Ryzen cpus will be a better choice for mining these days, especially those with higher core counts and larger L3 cache.
2. Your strategy.   You can mine a coin believing it will moon one day in the future.   Then you might not mind it not returning more than a few cents a day now.  You can let the miner do its work and forget about it until the day the coin moons and then rake in the profits selling.
Or, you can spend a lot of time researching the coin of the day and mine it short term to see if that will earn you a quick buck.
I think XMR has potential to go higher in price and am mining it.

It depends what you mean by best.  I am trottling back for best fuel economy.

I have set my MSI 5700 Mech to mine ETH at around 50MH/s.
Settings are as follows:
Core freq  1200
Mem freq   875
Core volt  731 mV    <- used Red Bios Editor to adjust from 800 to 725, but it wont go below 731
Power consumption reported by miner (Phoenixminr 4.9c) is 105Watts
GPU temp is 53 deg
Hot spot temp, which on my GPU is memory temp, is 82.
Fans are at 44%

This is a story on how I managed to boot a headless rig when the mobo had entered into the BIOS.

My one month new MSI RX 5700 Mech GP Edition decided to quit on me this morning  .  I could not get it undead.
I suspect maybe memory got damaged by heat - I found it was impossible to mine with it and keep the hot spot temperature under 100 degrees while mining ETH.  I watched a Youtube vid on that same card where they found the thermal pads did not cover the mem chips.
I took the GPU out of the rig, then faced a new problem:  How to boot the now headless rig back into Windows without a GPU since the Ryzen 3700X CPU in it doesnt have a built in GPU to drive the display.  The mobo I have is an ASUS TUF 450M.

I know the mobo usually enters into the BIOS after a hardware change.   I was not able to RDP into it from my laptop.
After a lot of trials with reboots and new keyboard sequences I found the correct key sequence to get it to exit BIOS and boot into Windows.  Immediately after power-on, I hit the Delete key and held it down for a few seconds until I was confident it was now in the BIOS.  I then pressed F10 and Y to confirm.  And then it booted into Windows.
The rig is now cpu-mining Monero while waiting for a GPU replacement.  I will have to RMA that defective GPU. 

I can confirm the numbers for a 3700X.  Mine is mining Loki at ~7000 H/s  atm.     Miner is XMRig 4.3.1-beta.
For comparison my old clunker PC with an i5 4460 CPU does 1450 H/s on the same algo/miner with 3 threads.

Theoretically, that should give me about half the hashrate.
I tried it and it did not work.  It threw an error and told me to enable large pages.
Interestingly, I was able to cpu-mine Monero at the same hashrate as before - 460 H/s with 7 threads (XMR-Stak 2.10.6).

Ryzen 5 2600 running at 3,4GHz

randomx-benchmark.exe --mine --jit --largePages --threads 6 --init 6 --nonces 300000

Performance: 3630 h/s

test bench:
Asus TUF B450
Memory 2 x 4Gb @ 2133 MHz 

Try using SRB Polaris bios editor and use the one-click memory strap function
To modify the mem strap.

Holy crap
You must be like 500 Watts over where the power should be.
Do yourself a favor and read up on undervolting Vegas.
Heck I am doing 220 MH with 8 x RX570 at 850 Watts total power

Does anybody know how to undervolt the memory of a Vega 64 reference card ? 
HWiNfO reports it as being 1.356V (MVDDC).  I am not talking about the memory voltage you set in OverdriveNTool..
My card is mining XMR with TMR at 2kH/s and consuming around 170 Watts.  I am looking to reducing power a bit more...

I am using RDP and Reminna (Ubuntu)  on my LAN.  I have also tried MobaXterm and Visionapp, but didnt like them too much.  It is for the remote location I currently use Teamviewer, because I have not wanted to open my router at the remote location for direct connections on the port that RDP, for example, is using.   If someone could point me to a good tutorial on how to make port forwarding on routers  acceptably secure then I am all ears.

See if you need to disable UPLS and crossfire.
Do a search for EnableULPS and registry and see if it applies to your card

WinAMDTweak (the command line version) should be available in this thread.   I DLed  the pixel patcher from here.  It seems legit.
https://www.monitortests.com/forum/Thread-AMD-ATI-Pixel-Clock-Patcher

My two test cards both have Elpida mem.  I simply used SRBPolaris bios editor with the one-click method to get good timing straps for my custom bios.
I will use that as a starting point to further tweak my cards, when I get the time.

Use version 1.4.6 of the pixel patcher - then you can sign newer versions of the driver.  I have patched my 19.4.1 install with it.

You could use  WinAMDTweak --current from your custom bios, then store those timings and use them as a starting point after you have reverted the card back to orig.bios.


EDIT:
I had to verify what I just mentioned above:
I upgraded the driver on my test rig with 2 x RX 570 from 18.6.1 to 19.4.1, patched it and restarted the rig. 
I then learned I had to upgrade my OverDrivenTool because I no longer had fan control.  Great, eh
I found version 0.2.8. bet1 and am playing with it as we speak..  Seems like the new driver runs the cards cooler... less power Yay.