Because i use afterburner and had windows check box checked, My settings were pulling on reboot. For some reason my ETH settings wont run on LOL. The Core Voltage and Core Clock have a lot more affect on 144,5 via LOLminer
Whereas my ETH based algos rely much more on memory clock, LOL is definitely using more Core than i am used to.
...
*can probably be pushed higher, but i am not comfortable pushing the core voltage or clock any higher.
Hope this helps someone
James
Glad to hear its working now
To add some points to the current technical discussion:
Yes, Equihash profits more from higher core clock then Ethash does. The reason is Ethash almost only does memory operations - loading stuff from memory, do small stuff and store back. So here memory bandwidth is crucial only. For Equihash this is different. The first phase of the algorithm is Blake2B, which is very compute heavy - this is approx 1/5th of the computation time (For 144/5 and 192/7 - for 210/9 it is a little less then this). Also my implementation uses in the later phases the shared memory a lot which works at core clock. So here the bandwidth also scales often with the clock.
So a general tuning advice of me would be to use clock near stock and memory can be upped a bit (lolMiner profits from fast memory, but not so much as Ethereum mining). Instead one can safely lower the power target, because most rounds either use high clock or fast mem, but often not at the same time. Also what helps a lot to optimize performance over watts is undervolting. This helps in two ways: undervolted core use less energy. But this also means if you set a lower power target this means that its less likely the undervolted core has to throttle clock to stay within power target. Overall I would go for 1200 - 1350 clock + undervolt.
Thanks. Obviously there is some space for improvement for VEGA. It's ok for 4years old furi to do 34sol/s but not for vega.
Without a question: Vega cards could do better. I will continue working on this!
But I think you underestimate here what monsters this Fury cards were
Just to put some numbers:
Vega 64:
#Cores: 4096
Core Clock: 1247 / 1548
Memory: 8GB HBM 2, 483 GB/s
Fury X:
#Cores: 4096
Core Clock: ? / 1050
Memory: 4GB HBM, 512 GB/s
So you see ... the Vega clocks higher, but Fury even has more bandwidth. Vega had a lot of improvements regarding power management and so on and also some new compute features. But Equihash just do not need this. The operations done are almost all (64 bit) integer magic, memory operations, atomic operations and vector xor of integers. All the improvements in new features are almost unused. This is why lolMiner uses in fact OpenCL 1.2 and would still run on an ages old HD 5xxx series if it has enough memory. What makes Nvidia strong on Equihash is their cache line size, so some small detail in memory controller that is unchanged within AMD GCN series. By the way you see it here at the 96/5 and 144/5 algorithms. On Equi 144/5 it is hard for me to get an 580 close to a 1060. We are still approx 10 to 15% off. Here the Pascal architecture helps much.
On 96/5 instead this cards are on par. Why? Because 96/5's problem sizes allow to sort memory access better and so the disadvantage of cache line size vanishes.
So the only advantage of Vega over Fury is its higher clock (which is only useful in about 1/5 to 1/4th of the algorithm) and Vega does better sols/w. But overall it does not surprise me that Fury X can get close to Vega 56 at least.
so... can I add 2x6gb 1060s to my 4x8gb 580 rig now? will it run red and green at the same time? not concerned about big hashing, just don't like the 1060s sitting idle. thanks for responding.
If your operation system allows using Nvidia and AMD mixed it will work. So you will need Windows or a custom Linux (most standard Linux do not allow AMD and NV drivers at the same time). But I will not guaranty for stability and speed. But - just to point an option - you can run the AMDs in lolMiner with --devices AMD (so only this cards are used) and open a 2nd command line with EWBF on the Nvidia cards. This constellation will give you most likely best performance.
I just click on it? dunno how to "open" it and look at it? sorry for my ignorance, but thank you for your help.. 