SRBMiner Cryptonight AMD GPU Miner V1.9.3 - native algo switching

[Kgonla]

V1.6.0
+ As some have issues with miner 'just waiting some time' on startup, i added more logging option so we can catch where it hangs, and i can possibly fix that

Reproduced
[2018-06-10 02:38:44] Heating up system, please wait...

[2018-06-10 02:38:43] CryptonightV7 mode enabled
[2018-06-10 02:38:44] DevFee pools loaded
[2018-06-10 02:38:44] DevFee pool SET
[2018-06-10 02:38:44] DevFee address SET

ok, now i know its not the new devfee method that is problematic.
But i would suggest to try now a previous version that you know was working good.
Nothing was changed in the new version in the GPU's initialisation process (this is where it hangs), so my guess is something at your side is causing this.

Comparing his log with mine:

[2018-06-09 13:52:22] CryptonightV7 mode enabled
[2018-06-09 13:53:46] DevFee pool SET
[2018-06-09 13:53:46] DevFee address SET

I don't have the "DevFee pools loaded", maybe is this what is slowing down my startup.

well then your firewall is probably blocking srbminer.com domain, and it cant get the devfee pools from online. add an exception to your firewall for the srbminer.com domain.

I think this too
I set firewall to allow all (Firewall says is disabled)
Added an entry to srbminer expressly allowing port 80
Also added srbminer.com to trusted sites
But no changes

I'm preparing a usb stick with W10 to test this & the self close after the 4th retry when using max_difficulty

[1713283480]

1713283480

[1713283480]

1713283480

[abg00]

test a lot of configs

18.5.1 drivers is best. stable high rate

Anyone have noticed the inability of 18.5.1 driver to recognize RX 570's???
In my system it sees all my RX 570 as 580!!!
So I couldn't run miner.
I reverted back to 18.3.4.
win update 1803.
Even the 18.5.2 driver does the same.

Thanks in advance.

win1803 should be avoided at all cost.

Gentlemen,

it's not 1803 update definitely.
I installed 18.5.1 on my previous win 1709 and there was the problem!
So I decided to update to 1803.

what trouble?

[livada]

What is this?
https://image.prntscr.com/image/EomGSxQKREKg0w97UYvkYQ.jpg

0HR 5 sec and go back and mining work ok.

[popow_sergei]

[2018-06-08 23:07:33] Error CL_OUT_OF_HOST_MEMORY when creating clCreateCommandQueue for DeviceID 0 (Thread 0)
[2018-06-08 23:07:33] Error initing GPU's. Stopping miner process

what caused the error ?

You can read? Miner all wrote to you. Not enogh memory on card. Lower intensity.

1.4.9 works, and 1.5.9 gives the following error
with the same settings

From 1.4.9 to 1.5.9 was 10 versions. I don't want to find in wich version what was changed. But you must understand that through 10 versions same settings can used by miner different ways. Just try to lower intensity and increase pagefile size to 8Gb * number of videocards.

that's obviously not the point. installed 8 GB of RAM and 80 GB of virtual 6 cards .

What algo? What intensity, worksize and threads?

normalv7
"gpu_conf" :
[
   { "id" : 0, "intensity" : 112, "worksize" : 16, "threads" : 2},
   { "id" : 1, "intensity" : 112, "worksize" : 16, "threads" : 2},
   { "id" : 2, "intensity" : 112, "worksize" : 16, "threads" : 2},
   { "id" : 3, "intensity" : 112, "worksize" : 16, "threads" : 2},
   { "id" : 4, "intensity" : 112, "worksize" : 16, "threads" : 2},
   { "id" : 5, "intensity" : 112, "worksize" : 16, "threads" : 2}
]

[UnclWish]

[2018-06-08 23:07:33] Error CL_OUT_OF_HOST_MEMORY when creating clCreateCommandQueue for DeviceID 0 (Thread 0)
[2018-06-08 23:07:33] Error initing GPU's. Stopping miner process

what caused the error ?

You can read? Miner all wrote to you. Not enogh memory on card. Lower intensity.

1.4.9 works, and 1.5.9 gives the following error
with the same settings

From 1.4.9 to 1.5.9 was 10 versions. I don't want to find in wich version what was changed. But you must understand that through 10 versions same settings can used by miner different ways. Just try to lower intensity and increase pagefile size to 8Gb * number of videocards.

that's obviously not the point. installed 8 GB of RAM and 80 GB of virtual 6 cards .

What algo? What intensity, worksize and threads?

normalv7
"gpu_conf" :
[
   { "id" : 0, "intensity" : 112, "worksize" : 16, "threads" : 2},
   { "id" : 1, "intensity" : 112, "worksize" : 16, "threads" : 2},
   { "id" : 2, "intensity" : 112, "worksize" : 16, "threads" : 2},
   { "id" : 3, "intensity" : 112, "worksize" : 16, "threads" : 2},
   { "id" : 4, "intensity" : 112, "worksize" : 16, "threads" : 2},
   { "id" : 5, "intensity" : 112, "worksize" : 16, "threads" : 2}
]

Cards is 8Gb vmemory? Display connected to card id 0? If so, just try to id 0 card set intensity 108 or 110. And why worksize 16? 8 is works perfect. Try it. I noticed that 1.6.0 with worksize 16 and high intensity on 580 8Gb significant lower speed on the same intensity as worksize 8.

[marcos5872]

what memory and what clocks?

Samsung 1250/2050.

thanks

not sure whats up but sometimes i get 1100h/s on Heavy on my samsung and sometimes 1000h/s

when samsung does 1100h/s my micron does less and sometimes other way around
cant get both to do 1100h/s at the same time

GPU0: RX 570 4GB Samsung Memory
GPU1: RX 580 8GB Samsung Memory
GPU2: RX 580 8GB Micron Memory

Hello, what's worksize and threads do you use? (on your RX 580 8GB mircron)

[MaxMidnite]

Some of these results are not making sense at all.

I have tweaked my Asus RX580 (8GB) according to best guide and not getting over 1000 hash on each.
I dont think its possible unless you having clocked it to have hundreds of memory errors and card then fuck up..

[dingdongtobias]

Some of these results are not making sense at all.

I have tweaked my Asus RX580 (8GB) according to best guide and not getting over 1000 hash on each.
I dont think its possible unless you having clocked it to have hundreds of memory errors and card then fuck up..

So! If you don't know how to do it then it is impossible?

[MaxMidnite]

Some of these results are not making sense at all.

I have tweaked my Asus RX580 (8GB) according to best guide and not getting over 1000 hash on each.
I dont think its possible unless you having clocked it to have hundreds of memory errors and card then fuck up..

So! If you don't know how to do it then it is impossible?

Didnt say its impossible, just saying with the dodgy overlocks your not considering power draw.
Considering the safety of your cards, then good luck with your power bill vs profit of mining.
Good luck with the life of your cards..

[heavyarms1912]

Some of these results are not making sense at all.

I have tweaked my Asus RX580 (8GB) according to best guide and not getting over 1000 hash on each.
I dont think its possible unless you having clocked it to have hundreds of memory errors and card then fuck up..

So! If you don't know how to do it then it is impossible?

Didnt say its impossible, just saying with the dodgy overlocks your not considering power draw.
Considering the safety of your cards, then good luck with your power bill vs profit of mining.
Good luck with the life of your cards..

You need good timing straps to reach 1k+ on CN-V7.  8gb cards can do 1k+ on CN-Heavy easily. 
I have also got 1260 H/s on an RX 570 8gb card (1466/2300 clocks).  They do 1140 H/s @ 1250/2240 clocks and 900mV core (105w from wall).  So it's pretty much possible.

Also, All 4gb cards that I have Elpida, Hynix do 1k+ on CN-V7.

[Steak12]

Some of these results are not making sense at all.

I have tweaked my Asus RX580 (8GB) according to best guide and not getting over 1000 hash on each.
I dont think its possible unless you having clocked it to have hundreds of memory errors and card then fuck up..

So! If you don't know how to do it then it is impossible?

Didnt say its impossible, just saying with the dodgy overlocks your not considering power draw.
Considering the safety of your cards, then good luck with your power bill vs profit of mining.
Good luck with the life of your cards..

You need good timing straps to reach 1k+ on CN-V7.  8gb cards can do 1k+ on CN-Heavy easily. 
I have also got 1260 H/s on an RX 570 8gb card (1466/2300 clocks).  They do 1140 H/s @ 1250/2240 clocks and 900mV core (105w from wall).  So it's pretty much possible.

Also, All 4gb cards that I have Elpida, Hynix do 1k+ on CN-V7.

I want to see that...
Can't get more that 700h/s with Elpida 570 pulse 4gb (with stak and SRB)

[dragonmike]

Some of these results are not making sense at all.

I have tweaked my Asus RX580 (8GB) according to best guide and not getting over 1000 hash on each.
I dont think its possible unless you having clocked it to have hundreds of memory errors and card then fuck up..

So! If you don't know how to do it then it is impossible?

Didnt say its impossible, just saying with the dodgy overlocks your not considering power draw.
Considering the safety of your cards, then good luck with your power bill vs profit of mining.
Good luck with the life of your cards..

You need good timing straps to reach 1k+ on CN-V7.  8gb cards can do 1k+ on CN-Heavy easily. 
I have also got 1260 H/s on an RX 570 8gb card (1466/2300 clocks).  They do 1140 H/s @ 1250/2240 clocks and 900mV core (105w from wall).  So it's pretty much possible.

Also, All 4gb cards that I have Elpida, Hynix do 1k+ on CN-V7.

I want to see that...
Can't get more that 700h/s with Elpida 570 pulse 4gb (with stak and SRB)

You need tweaked memory straps to achieve these speeds... And you need to make sure compute mode is active in the registry.

I've got 14 RX 570's with Elpida mem hashing at 960 on average.

[nordmann666]

you are in SRBMiner thread!

try here https://bitcointalk.org/index.php?topic=1882656.0

[altsay]

I'm getting ~500 H/s on R9 280x @ default settings, v1.6.0. Is that what should i expect? I was hoping more though

[UnclWish]

I'm getting ~500 H/s on R9 280x @ default settings, v1.6.0. Is that what should i expect? I was hoping more though

Point algo. Double threads enabled? Try to manual increase intensity.

[darkneorus]

getting around 980-1030 H/s on CN-Heavy from 8GB RX470s with Samsung memory and 8GB RX570s with Micron. great!

[Matkurb]

getting around 980-1030 H/s on CN-Heavy from 8GB RX470s with Samsung memory and 8GB RX570s with Micron. great!

That is good speed.

[popow_sergei]

[2018-06-08 23:07:33] Error CL_OUT_OF_HOST_MEMORY when creating clCreateCommandQueue for DeviceID 0 (Thread 0)
[2018-06-08 23:07:33] Error initing GPU's. Stopping miner process

what caused the error ?

You can read? Miner all wrote to you. Not enogh memory on card. Lower intensity.

1.4.9 works, and 1.5.9 gives the following error
with the same settings

From 1.4.9 to 1.5.9 was 10 versions. I don't want to find in wich version what was changed. But you must understand that through 10 versions same settings can used by miner different ways. Just try to lower intensity and increase pagefile size to 8Gb * number of videocards.

that's obviously not the point. installed 8 GB of RAM and 80 GB of virtual 6 cards .

What algo? What intensity, worksize and threads?

normalv7
"gpu_conf" :
[
   { "id" : 0, "intensity" : 112, "worksize" : 16, "threads" : 2},
   { "id" : 1, "intensity" : 112, "worksize" : 16, "threads" : 2},
   { "id" : 2, "intensity" : 112, "worksize" : 16, "threads" : 2},
   { "id" : 3, "intensity" : 112, "worksize" : 16, "threads" : 2},
   { "id" : 4, "intensity" : 112, "worksize" : 16, "threads" : 2},
   { "id" : 5, "intensity" : 112, "worksize" : 16, "threads" : 2}
]

Cards is 8Gb vmemory? Display connected to card id 0? If so, just try to id 0 card set intensity 108 or 110. And why worksize 16? 8 is works perfect. Try it. I noticed that 1.6.0 with worksize 16 and high intensity on 580 8Gb significant lower speed on the same intensity as worksize 8.

vega 56  8 gb
the monitor is connected to an integrated integrated card.
at worksize 16 more gives

[livada]

vega 56  8 gb
the monitor is connected to an integrated integrated card.
at worksize 16 more gives

ok. what intensyti and worksize u use for HEAVY algo ?

[Larvitar]

Some of these results are not making sense at all.

I have tweaked my Asus RX580 (8GB) according to best guide and not getting over 1000 hash on each.
I dont think its possible unless you having clocked it to have hundreds of memory errors and card then fuck up..

So! If you don't know how to do it then it is impossible?

Didnt say its impossible, just saying with the dodgy overlocks your not considering power draw.
Considering the safety of your cards, then good luck with your power bill vs profit of mining.
Good luck with the life of your cards..

Impossible? Nope.
Crazy overclock? Nope.
Insane power draw? Nope.

I can reach average 1100H/s in heavy with 1310/2020 (Samsung, 940mv cvddc) with no pain. Well, I did some tests with intensity but was not hard to find the perfect balance (between 56 and 58).

For Micron @heavyarms shared his own timings some pages ago and they are great. Ubermix timings are good for Samsung. So, there are many knowledge around this thread to find best performance.

I'm trying to find a way to break 1000H/s with stability with my Elpida. I had good results with some mixing, but when I reach 1000H/s with low memory clock, the ETHash performance drops even with high clock (it doesn't makes sense).