my experience (xmr only here)
rx480 - 4gb Sapphire nitro+ oc:
Samsung Memory non-flashed: 650 h/s
Elpida/Hynx Memory non-flashed: 580 h/s
with rom flashed (just the simple copy/paste of straps and overclocking to around 2040, nothing fancy)
Samsung: 730-750 range
Elpida/hynx memory: up to 710 to 730 range

Note: I'm not sure exactly what the other memory type is, but the 'other' polaris bios editor works with it (check in comments of the github page). 
Also, using Ubuntu 16.04, not windows..

Most popular do seem to be ethereum, monero, and zerocash (at the moment).  Why not research those three coins and see which one interests you?  If you are in it for profit, you sadly will find out that unless your hardware and electricity is free, it is most often (for hobbyists like yourself) much smarter to buy the coin outright and wait for the price to appreciate.   
In terms of electricity useage, ETH uses the most, ZEC the second-most, and XMR demands the least amount of electricity (if that matters to you). 
Have a great time mining!   
Websites that I find interesting..   
cryptowat.ch 
coingecko.com   
https://anorak.tech/   
https://monerotalk.org/   
https://forum.ethereum.org/categories/mining

Maybe this is the wrong thread to jump in, but I was hoping someone could help me clarify something with ram timings.. My understanding is that they're the time allotted for the completion of a certain task done by the gpu.  If they are too large, your card will be sitting around doing nothing after it is done with the task.  Too little, and that obviously makes problems.   
I recently copy-pasted the straps from the 1500 to 1625 on my 480 Elpida's.  (Above that had all been copied from 1625).  These were my worst performers, going at 650h/s on xmr, but with the strap copy, they got to 710 with far fewer errors.  If the GPU is overclocked to 2000 Mhz anyway, why would this simple strap-copy have such an effect? 
More to the point: if you are clocked at 2000 Mhz (BIOS, in Polaris editor), why does fine-tuning the timing on all (or at least down to 1500) Mhz levels have an effect?  My simple mind is thrown off. 

did you check out the github issue that wolf0 rose on the github page?  that helped me get through compiling and working correctly with amd

Using sgminer, I see a number of different ways that the mining is measured, and even after reading through a few posts, I am not entirely sure what the difference between these calculations is.  Namely:
[from the sgminer-gm Readme](https://github.com/sgminer-dev/sgminer/blob/master/README.md)
Each column is as follows:
5s: A 5 second exponentially decaying average hash rate   
avg: An all time average hash rate   
A: The total difficulty of Accepted shares   
R: The total difficulty of Rejected shares   
HW: The number of HardWare errors   
WU: The Work Utility defined as the number of diff1 shares work / minute (accepted or rejected)   

It seems that miner output can be measured in number of accepted shares / minute (work utility) or in hash rate.  Which one of these should be tuned for optimal pool mining?  Are pools measuring accepted shares or hash rate or some variation of the two?   

As a concrete example, using sgminer, I am getting the following from my cards:   

GPU1                | (5s):652.4 (avg):652.2h/s | A:26316032 R:144000 HW:0 WU:26914.2 rI:1016   
GPU3                | (5s):652.3 (avg):652.3h/s | A:25902032 R:90000 HW:0 WU:27196.71 rI:1016   

GPU0                | (5s):741.5 (avg):741.0h/s | A:21024026 R:36000 HW:480 WU:22027. rI:1016     
GPU2                | (5s):744.7 (avg):744.8h/s | A:22500028 R:54000 HW:436 WU:22573. rI:1016   

GPU4                | (5s):723.0 (avg):723.1h/s | A:28782035 R:108000 HW:0 WU:29469.1 rI:1016   
GPU5                | (5s):723.0 (avg):723.0h/s | A:29196036 R:90000 HW:0 WU:30270.72 rI:1016   

uptime about 12 hours... 
I understand that we have to take electricity consumption into account to get a true metric, but ignoring that for now, what is the strongest card? 
GPU1 and 3 have noticeably low hash rates, but their work utility outperforms that of GPU0,2. 

GPU0,2 have the highest hashrate, but they have hardware errors (is that a lot for 12 hours?) and their work utility is the lowest of all cards. 

GPU4,5 have the highest work utility, however their hash rates are lower than GPU0,2.   

If anyone has insight, it would be appreciated. 

For the curious, I am using sgminer-gm, all Sapphire rx480 4gb Nitro+OC gpu's with custom roms on Ubuntu 16.04.  All ROM's are custom, but GPU1,3 had different ROM's to start with. 

How are you modifying your bios?  You just going into UEFI boot and changing the options?  I'm new to this obviously. 

Core 1160 = gpu core clock rate?
Mem = memclock rate?
What is straps? 
Also, have you tried undervolting the gpu's or is it possible on ubuntu? 

I have 2 unmodded sapphire rx480's running at 577H/s
"threads": 1,
"rawintensity": 1024,
"worksize": 8