Topic: Gateless Gate Sharp 1.3.8: 30Mh/s (Ethash) on RX 480! - page 85. (Read 214432 times)

I don't sure that memory voltage control works... Core voltage works - I checked it with HWInfo... But memory voltage... I didn't know how to check it... But temperatures of cards is a bit higher than if set voltages via Claymore's miner...

I was able to access PCI configuration space from C#.
It's pretty amazing that you can do such a low-level stuff with such a high-level language like C#.
This is going to be a real fun...

Yeah, that makes more sense... I don't know what I was thinking when I picked the current default values.
Thanks!

zawawa
Thanks for the hard work) Tried you miner today for the first time - neoscrypt AMD and so far so good! The hashrate is much better than nsgminer 750Khs vs 540Khs on my modded RX570|580 cards - http://prntscr.com/i08mrt
Also, the values are kinda way off, please put Intensity at 16 and work size at 64 (the same worked for nsgminer). All other values for all my rigs are either unstable or cause performance drops. This will be a good way to start for neoscrypt miners.

It will also be nice to add skunkhash algo in the future as well with default -i 20



Thanks

Hello

GPU: Gigabyte Windforce 2 NVIDIA GeForce GTX 1060 3GB (Samsung)
OS: Windows 7 Service Pack 1
Video Driver:  390.65
MSI Afterburn: 4.4.2
Gateless_Gate_Sharp_1.1.16_alpha = Last test https://bitcointalksearch.org/topic/m.27601822
Gateless_Gate_Sharp_1.2.2_alpha = New Test done to check progression

Let's begin:

1. Application is very lagy old test > new test same = no improvement
- Pressing Start button action is take after 10-15 sec
- Pressing Stop it take between 10-25 sec program to fully stop from mining & stop using video card (i was watching MSI Afb)
- Closing program it take near 30-40 sec action to be done, all this time GSS windows is not responsive

2. Dashboard = not functional (custom pool used for testing)

Elapsed Time: not working, noo time pass
Speed: 0 all the time, algo used is correctly reported
Device Speed not working, Share not working,  Core not working, Mem Not working
Single place were it can see any activity is "Log Tab" = speed, share are seen

3. Eror
2018-01-14 12:58:46.0756 [1] Exception in InitializeDevices(): Object reference not set to an instance of an object.   at GatelessGateSharp.OpenCLDevice.get_DefaultCoreClock()
   at GatelessGateSharp.MainForm.ResetDeviceOverclockingSettings(Device device)
   at GatelessGateSharp.MainForm.InitializeDevices()
   at GatelessGateSharp.MainForm.MainForm_Load(Object sender, EventArgs e)

4. Second GPU is incorrectly detect, is not a problem coz is not meant for mining but still is a bug, a 650 GTX for personal necessitate

5. GSS still have warning about want windows swap to be set at 24 G & noo disable / ignore option implement with is bad request

6. Msi Afterburner is not happy at all & somehow i think GSS it overight or try to gain control, ignoring setting already done, with is BAD, special if OC is not enable at all, GPU clock, Memory clock out of chart or of limit what suppose to be = BAD

7. Mining speed i will not say anything because i undestend from your previous replay GSS is not tune for Nvidia soo i will ignore this part

cheers

Nah, two threads should be enough. Let me finish experiments with PhyMem first, though...

Maybe 3? Just for future...

Let's just add another parameter for intensity for CryptoNight, then.

for cryptonight 2 are enough

I can't be sure that it will help to gain some extra speed. But in log we allways see that each thread on one device gives different speed. I just think that it can be usefull in some cases. For Vega's GPU's pepole says that it helps. On others maybe not.
Maybe make a separate tab for this or another way to don't make inteface too complicated... For advanced users...

That's not easy, though... Do you need more than two threads?
If not, I would just add another NumericUpDown for the intensity for the second thread.

thanks, i would suggest make it optional so it wont make interface complicated

I still don't quite get the rationale behind this trick, but as you wish...

i dont know if it increase speed for rx 470/480 /570/580, but for vega it does for sure with xmr stak.

you set there intensity like this
"gpu_threads_conf" : [
    { "index" : 0, "intensity" : 2016, "worksize" : 8, "affine_to_cpu" : false },
    { "index" : 0, "intensity" : 1600, "worksize" : 8, "affine_to_cpu" : false },


i ordered some vegas so yeah do it please

Does this make any meaningful differences at all? It is pretty easy to implement this, but I just don't want to make the user interface unnecessarily complicated. If you think we can gain, say, a 3% increase in hashrate, I would be happy to make this feature available to-day.

Yes, in log is speed for each thread.
It's need to find right settings for each card individually. If you find that 96 for your cards better, than it's so.

And it's still no possibility to config each thread separately. As it's in Stak-AMD...

2 threads, in log I get
On dashboard the speed is 400+, so maybe in the log each speed line is per thread?!

1 thread is still way faster..
But intensity 96 increased my speed to 800-810 h/s

hi

good work :)it keep setting now

an other question:
on rig 1 ,monero,nicehash,my wallet ---->server usa nicehash
on rig 2 ,monero,nicehash,my wallet ---->server eu nicehash

how to change server eu to usa?

on my rig its 100watts difference from the wall(running gpus@940mv or 860mv @cryptonight), so for me the uv stuff is importand...but i´ll leave it with trixx for now

Ah, that makes sense. I venture to guess that this limitation is hard-coded somewhere in the ATOMBIOS.
Technically, I can even rewrite the driver on the fly to lift this limitation, but I don't know if it's worth the effort...
I will come back to this after I am done with Equihash and optimizations.