Author

Topic: Claymore's ZCash/BTG AMD GPU Miner v12.6 (Windows/Linux) - page 128. (Read 3839201 times)

newbie
Activity: 61
Merit: 0
To all: you can see low speed at highest possible "-i" value, if so, reduce it. For example, you specify "-i 8" for 280X 3GB card, miner reduces it to "-i 7" automatically (shows "not enough GPU memory" message), but it may work much slower than "-i 6".
So if you see that new version is slower than old version - try to reduce "-i" value a bit.
In some cases you can increase "-i" value, for example, in my tests Nano cards work best with v12.2 using "-i 7", but for v12.3 "-i 8" is faster.

@Claymore Can you make something to improve 380 series? Currently I have 210H. Too low... Thank you in advance

Tonga is really weak for ZEC.

why my nano can't use -i 8?

AMD Cards available: 3
GPU #0: Fiji, 4096 MB available, 56 compute units
GPU #1: Fiji, 4096 MB available, 64 compute units
GPU #1 recognized as Fury X/Nano
GPU #2: Fiji, 4096 MB available, 64 compute units
GPU #2 recognized as Fury X/Nano
POOL version
GPU #0 is going to use too high intensity (Cool, not enough GPU memory, intensity value reduced (7)
GPU #0 algorithm ASM, intensity 7
GPU #1 is going to use too high intensity (Cool, not enough GPU memory, intensity value reduced (7)
GPU #1 algorithm ASM, intensity 7
GPU #2 is going to use too high intensity (Cool, not enough GPU memory, intensity value reduced (7)
GPU #2 algorithm ASM, intensity 7

literally 3 posts earlier:

To all: you can see low speed at highest possible "-i" value, if so, reduce it. For example, you specify "-i 8" for 280X 3GB card, miner reduces it to "-i 7" automatically (shows "not enough GPU memory" message), but it may work much slower than "-i 6".
So if you see that new version is slower than old version - try to reduce "-i" value a bit.
In some cases you can increase "-i" value, for example, in my tests Nano cards work best with v12.2 using "-i 7", but for v12.3 "-i 8" is faster.

Also, why do the compute units differ between your cards?

ZEC: GPU0 455.346 H/s, GPU1 481.857 H/s, GPU2 460.153 H/s
v12.3 -i 7
gpu0 is fury
gpu1 is fury x
gpu2 is nano
for v12.2 can't use -i 7 , for v.12.3 i can use -i 7 but can't -i 8

I think that's pretty normal, my Fury default to i6 and I see the same ballpark speed. Honestly doesn't even change if I force it to i5 or i4

That is right. My R9 Nano run better on -i 7 in 12.3 and -i 6 in 12.2.
donator
Activity: 1610
Merit: 1325
Miners developer
To all: you can see low speed at highest possible "-i" value, if so, reduce it. For example, you specify "-i 8" for 280X 3GB card, miner reduces it to "-i 7" automatically (shows "not enough GPU memory" message), but it may work much slower than "-i 6".
So if you see that new version is slower than old version - try to reduce "-i" value a bit.
In some cases you can increase "-i" value, for example, in my tests Nano cards work best with v12.2 using "-i 7", but for v12.3 "-i 8" is faster.

@Claymore Can you make something to improve 380 series? Currently I have 210H. Too low... Thank you in advance

Tonga is really weak for ZEC.

why my nano can't use -i 8?

AMD Cards available: 3
GPU #0: Fiji, 4096 MB available, 56 compute units
GPU #1: Fiji, 4096 MB available, 64 compute units
GPU #1 recognized as Fury X/Nano
GPU #2: Fiji, 4096 MB available, 64 compute units
GPU #2 recognized as Fury X/Nano
POOL version
GPU #0 is going to use too high intensity (Cool, not enough GPU memory, intensity value reduced (7)
GPU #0 algorithm ASM, intensity 7
GPU #1 is going to use too high intensity (Cool, not enough GPU memory, intensity value reduced (7)
GPU #1 algorithm ASM, intensity 7
GPU #2 is going to use too high intensity (Cool, not enough GPU memory, intensity value reduced (7)
GPU #2 algorithm ASM, intensity 7

Make sure you set these env variables:

GPU_MAX_HEAP_SIZE 100
GPU_MAX_ALLOC_PERCENT 100
GPU_SINGLE_ALLOC_PERCENT 100
sr. member
Activity: 504
Merit: 250
To all: you can see low speed at highest possible "-i" value, if so, reduce it. For example, you specify "-i 8" for 280X 3GB card, miner reduces it to "-i 7" automatically (shows "not enough GPU memory" message), but it may work much slower than "-i 6".
So if you see that new version is slower than old version - try to reduce "-i" value a bit.
In some cases you can increase "-i" value, for example, in my tests Nano cards work best with v12.2 using "-i 7", but for v12.3 "-i 8" is faster.

@Claymore Can you make something to improve 380 series? Currently I have 210H. Too low... Thank you in advance

Tonga is really weak for ZEC.

why my nano can't use -i 8?

AMD Cards available: 3
GPU #0: Fiji, 4096 MB available, 56 compute units
GPU #1: Fiji, 4096 MB available, 64 compute units
GPU #1 recognized as Fury X/Nano
GPU #2: Fiji, 4096 MB available, 64 compute units
GPU #2 recognized as Fury X/Nano
POOL version
GPU #0 is going to use too high intensity (Cool, not enough GPU memory, intensity value reduced (7)
GPU #0 algorithm ASM, intensity 7
GPU #1 is going to use too high intensity (Cool, not enough GPU memory, intensity value reduced (7)
GPU #1 algorithm ASM, intensity 7
GPU #2 is going to use too high intensity (Cool, not enough GPU memory, intensity value reduced (7)
GPU #2 algorithm ASM, intensity 7

literally 3 posts earlier:

To all: you can see low speed at highest possible "-i" value, if so, reduce it. For example, you specify "-i 8" for 280X 3GB card, miner reduces it to "-i 7" automatically (shows "not enough GPU memory" message), but it may work much slower than "-i 6".
So if you see that new version is slower than old version - try to reduce "-i" value a bit.
In some cases you can increase "-i" value, for example, in my tests Nano cards work best with v12.2 using "-i 7", but for v12.3 "-i 8" is faster.

Also, why do the compute units differ between your cards?

ZEC: GPU0 455.346 H/s, GPU1 481.857 H/s, GPU2 460.153 H/s
v12.3 -i 7
gpu0 is fury
gpu1 is fury x
gpu2 is nano
for v12.2 can't use -i 7 , for v.12.3 i can use -i 7 but can't -i 8

I think that's pretty normal, my Fury default to i6 and I see the same ballpark speed. Honestly doesn't even change if I force it to i5 or i4
newbie
Activity: 18
Merit: 0
To all: you can see low speed at highest possible "-i" value, if so, reduce it. For example, you specify "-i 8" for 280X 3GB card, miner reduces it to "-i 7" automatically (shows "not enough GPU memory" message), but it may work much slower than "-i 6".
So if you see that new version is slower than old version - try to reduce "-i" value a bit.
In some cases you can increase "-i" value, for example, in my tests Nano cards work best with v12.2 using "-i 7", but for v12.3 "-i 8" is faster.

@Claymore Can you make something to improve 380 series? Currently I have 210H. Too low... Thank you in advance

Tonga is really weak for ZEC.

why my nano can't use -i 8?

AMD Cards available: 3
GPU #0: Fiji, 4096 MB available, 56 compute units
GPU #1: Fiji, 4096 MB available, 64 compute units
GPU #1 recognized as Fury X/Nano
GPU #2: Fiji, 4096 MB available, 64 compute units
GPU #2 recognized as Fury X/Nano
POOL version
GPU #0 is going to use too high intensity (Cool, not enough GPU memory, intensity value reduced (7)
GPU #0 algorithm ASM, intensity 7
GPU #1 is going to use too high intensity (Cool, not enough GPU memory, intensity value reduced (7)
GPU #1 algorithm ASM, intensity 7
GPU #2 is going to use too high intensity (Cool, not enough GPU memory, intensity value reduced (7)
GPU #2 algorithm ASM, intensity 7

literally 3 posts earlier:

To all: you can see low speed at highest possible "-i" value, if so, reduce it. For example, you specify "-i 8" for 280X 3GB card, miner reduces it to "-i 7" automatically (shows "not enough GPU memory" message), but it may work much slower than "-i 6".
So if you see that new version is slower than old version - try to reduce "-i" value a bit.
In some cases you can increase "-i" value, for example, in my tests Nano cards work best with v12.2 using "-i 7", but for v12.3 "-i 8" is faster.

Also, why do the compute units differ between your cards?

ZEC: GPU0 455.346 H/s, GPU1 481.857 H/s, GPU2 460.153 H/s
v12.3 -i 7
gpu0 is fury
gpu1 is fury x
gpu2 is nano
for v12.2 can't use -i 7 , for v.12.3 i can use -i 7 but can't -i 8
hero member
Activity: 798
Merit: 1000
have you noticed a speed increase of 470/480?
sr. member
Activity: 504
Merit: 250
To all: you can see low speed at highest possible "-i" value, if so, reduce it. For example, you specify "-i 8" for 280X 3GB card, miner reduces it to "-i 7" automatically (shows "not enough GPU memory" message), but it may work much slower than "-i 6".
So if you see that new version is slower than old version - try to reduce "-i" value a bit.
In some cases you can increase "-i" value, for example, in my tests Nano cards work best with v12.2 using "-i 7", but for v12.3 "-i 8" is faster.

@Claymore Can you make something to improve 380 series? Currently I have 210H. Too low... Thank you in advance

Tonga is really weak for ZEC.

why my nano can't use -i 8?

AMD Cards available: 3
GPU #0: Fiji, 4096 MB available, 56 compute units
GPU #1: Fiji, 4096 MB available, 64 compute units
GPU #1 recognized as Fury X/Nano
GPU #2: Fiji, 4096 MB available, 64 compute units
GPU #2 recognized as Fury X/Nano
POOL version
GPU #0 is going to use too high intensity (Cool, not enough GPU memory, intensity value reduced (7)
GPU #0 algorithm ASM, intensity 7
GPU #1 is going to use too high intensity (Cool, not enough GPU memory, intensity value reduced (7)
GPU #1 algorithm ASM, intensity 7
GPU #2 is going to use too high intensity (Cool, not enough GPU memory, intensity value reduced (7)
GPU #2 algorithm ASM, intensity 7

literally 3 posts earlier:

To all: you can see low speed at highest possible "-i" value, if so, reduce it. For example, you specify "-i 8" for 280X 3GB card, miner reduces it to "-i 7" automatically (shows "not enough GPU memory" message), but it may work much slower than "-i 6".
So if you see that new version is slower than old version - try to reduce "-i" value a bit.
In some cases you can increase "-i" value, for example, in my tests Nano cards work best with v12.2 using "-i 7", but for v12.3 "-i 8" is faster.

Also, why do the compute units differ between your cards?
newbie
Activity: 18
Merit: 0
To all: you can see low speed at highest possible "-i" value, if so, reduce it. For example, you specify "-i 8" for 280X 3GB card, miner reduces it to "-i 7" automatically (shows "not enough GPU memory" message), but it may work much slower than "-i 6".
So if you see that new version is slower than old version - try to reduce "-i" value a bit.
In some cases you can increase "-i" value, for example, in my tests Nano cards work best with v12.2 using "-i 7", but for v12.3 "-i 8" is faster.

@Claymore Can you make something to improve 380 series? Currently I have 210H. Too low... Thank you in advance

Tonga is really weak for ZEC.

why my nano can't use -i 8?

AMD Cards available: 3
GPU #0: Fiji, 4096 MB available, 56 compute units
GPU #1: Fiji, 4096 MB available, 64 compute units
GPU #1 recognized as Fury X/Nano
GPU #2: Fiji, 4096 MB available, 64 compute units
GPU #2 recognized as Fury X/Nano
POOL version
GPU #0 is going to use too high intensity (Cool, not enough GPU memory, intensity value reduced (7)
GPU #0 algorithm ASM, intensity 7
GPU #1 is going to use too high intensity (Cool, not enough GPU memory, intensity value reduced (7)
GPU #1 algorithm ASM, intensity 7
GPU #2 is going to use too high intensity (Cool, not enough GPU memory, intensity value reduced (7)
GPU #2 algorithm ASM, intensity 7
sr. member
Activity: 504
Merit: 250
v12.3:

- slightly improved speed, about 410H/s on stock 390X, 465H/s on stock Nano, 300H/s on stock 280X, 305H/s on stock RX480.
- improved stability for high "-i" option values, added "-i 9" option value.
- fixed a few bugs.

Anyone got 16.3.2 drivers on Linux?
Now i've got 2240 sol/S with 1070W - 1095W on wall using 5 x r9 Furys, on latest drivers.
I want check how that work with 16.3.2

I'll see what I can do, in exchange for your settings and BIOS Smiley
I assume you modded the Bios and run voltage, memory and core clock custom?


Stock GPUs and i just Update Motherboard bios. Thats it.
Wink

Impressively low power draw I have to say. You don't use any flags in the Claymore software either? Just standard intensity and Claymore 12.3?

I don't think 16.3.2 exists for Fury/Nano under linux. Official Linux driver packages here: http://support.amd.com/en-us/download/linux
donator
Activity: 1610
Merit: 1325
Miners developer
To all: you can see low speed at highest possible "-i" value, if so, reduce it. For example, you specify "-i 8" for 280X 3GB card, miner reduces it to "-i 7" automatically (shows "not enough GPU memory" message), but it may work much slower than "-i 6".
So if you see that new version is slower than old version - try to reduce "-i" value a bit.
In some cases you can increase "-i" value, for example, in my tests Nano cards work best with v12.2 using "-i 7", but for v12.3 "-i 8" is faster.

@Claymore Can you make something to improve 380 series? Currently I have 210H. Too low... Thank you in advance

Tonga is really weak for ZEC.
YIz
hero member
Activity: 686
Merit: 502
20H/s~ improvement on Hawaii GPUs with v12.3.
newbie
Activity: 12
Merit: 0
v12.3:

- slightly improved speed, about 410H/s on stock 390X, 465H/s on stock Nano, 300H/s on stock 280X, 305H/s on stock RX480.
- improved stability for high "-i" option values, added "-i 9" option value.
- fixed a few bugs.

Anyone got 16.3.2 drivers on Linux?
Now i've got 2240 sol/S with 1070W - 1095W on wall using 5 x r9 Furys, on latest drivers.
I want check how that work with 16.3.2

I'll see what I can do, in exchange for your settings and BIOS Smiley
I assume you modded the Bios and run voltage, memory and core clock custom?


Stock GPUs and i just Update Motherboard bios. Thats it.
Wink
sr. member
Activity: 504
Merit: 250
v12.3:

- slightly improved speed, about 410H/s on stock 390X, 465H/s on stock Nano, 300H/s on stock 280X, 305H/s on stock RX480.
- improved stability for high "-i" option values, added "-i 9" option value.
- fixed a few bugs.

Anyone got 16.3.2 drivers on Linux?
Now i've got 2240 sol/S with 1070W on wall using 5 x r9 Furys, on latest drivers.
I want check how that work with 16.3.2

I'll see what I can do, in exchange for your settings and BIOS Smiley
I assume you modded the Bios and run voltage, memory and core clock custom?
newbie
Activity: 31
Merit: 0
@Claymore Can you make something to improve 380 series? Currently I have 210H. Too low... Thank you in advance
newbie
Activity: 12
Merit: 0
v12.3:

- slightly improved speed, about 410H/s on stock 390X, 465H/s on stock Nano, 300H/s on stock 280X, 305H/s on stock RX480.
- improved stability for high "-i" option values, added "-i 9" option value.
- fixed a few bugs.

Anyone got 16.3.2 drivers on Linux?
Now i've got 2240 sol/S with 1070W - 1095W on wall using 5 x r9 Furys, on latest drivers.
I want check how that work with 16.3.2
newbie
Activity: 16
Merit: 0
For bonaire cards stick with 12.1 went from 102h/s with 12.1 to 73h/s with 12.3

Just increase Volt core

v12.3:

- slightly improved speed, about 410H/s on stock 390X, 465H/s on stock Nano, 300H/s on stock 280X, 305H/s on stock RX480.
- improved stability for high "-i" option values, added "-i 9" option value.
- fixed a few bugs.

there is no speed improvement for the R9 390 or 7970 with the same intensity. Shall I use higher intensity?

Increase about 2%. From 230 to 234-235 sol for HD7950. Saphie faster than HIS 7950 with the same setting
sr. member
Activity: 779
Merit: 250
v12.3:

- slightly improved speed, about 410H/s on stock 390X, 465H/s on stock Nano, 300H/s on stock 280X, 305H/s on stock RX480.
- improved stability for high "-i" option values, added "-i 9" option value.
- fixed a few bugs.

there is no speed improvement for the R9 390 or 7970 with the same intensity. Shall I use higher intensity?
sr. member
Activity: 450
Merit: 255
For bonaire cards stick with 12.1 went from 102h/s with 12.1 to 73h/s with 12.3
sr. member
Activity: 504
Merit: 250
Make sure your AMD Gaming Evolve app didn't automatically update your drivers, I had to use DDU and reinstall 4 times until I figured out it does that EVEN IF you uncheck the box during installation. You have to manually open AMD Gaming Evolved, click on the wheel besides deriver update and uncheck auto updates.

That's not it. And this is a 100% clean and updated Windows 10 x64 install with 4GB ram, 20 GB virtual mem, and NOTHING but AMD 16.3 drivers and claymore installed. Blows my damn mind.

General behavior is:

1. A card (random) will hash at 0% while all others are fine
2. OpenCL call hangs.
3. Crash

Sometimes all cards will be hashing just dandy for a minute or two, then one drops, then OpenCL call fail on 1 card.
newbie
Activity: 1
Merit: 0
Make sure your AMD Gaming Evolve app didn't automatically update your drivers, I had to use DDU and reinstall 4 times until I figured out it does that EVEN IF you uncheck the box during installation. You have to manually open AMD Gaming Evolved, click on the wheel besides deriver update and uncheck auto updates.
sr. member
Activity: 504
Merit: 250
v12.3:

- slightly improved speed, about 410H/s on stock 390X, 465H/s on stock Nano, 300H/s on stock 280X, 305H/s on stock RX480.
- improved stability for high "-i" option values, added "-i 9" option value.
- fixed a few bugs.

Claymore, I have had constant issues with my Rx 480. The latest being 'GPU x hangs in OpenCL call' got any clue? Tried literally all drivers possible, currently on 16.3x all other settings and system variables are correct.
Jump to: