[Cast XMR] high speed XMR/CryptoNight miner for RX Vega GPUs (2 KHash/s)

[dragonmike]

Well, stopped again after a few hours. So core clock now back up to 1408 and lowered mem clocks to 1040. Let's see how that goes. Annoying. I really thought I'd be able to run these cards closer to 1100... Now it seems I'm tweaking the other way

EDIT: I suppose playing with power limit wouldn't do anything for the memory clocks anyway... so pointless playing with that.

[bostonvex]

Well, stopped again after a few hours. So core clock now back up to 1408 and lowered mem clocks to 1040. Let's see how that goes. Annoying. I really thought I'd be able to run these cards closer to 1100... Now it seems I'm tweaking the other way

EDIT: I suppose playing with power limit wouldn't do anything for the memory clocks anyway... so pointless playing with that.

Dude, are you sure it is memclock related? I think it is a bug in cast_xmr.

[dragonmike]

Well, stopped again after a few hours. So core clock now back up to 1408 and lowered mem clocks to 1040. Let's see how that goes. Annoying. I really thought I'd be able to run these cards closer to 1100... Now it seems I'm tweaking the other way

EDIT: I suppose playing with power limit wouldn't do anything for the memory clocks anyway... so pointless playing with that.

Dude, are you sure it is memclock related? I think it is a bug in cast_xmr.

Well, when I was mining on unmodded 56 bios @ 925MHz memory freq the rig was mining several days without a single hickup. So I can only assume stability is memclock related. I'd be very happy to  be proven wrong!

[bostonvex]

Well, stopped again after a few hours. So core clock now back up to 1408 and lowered mem clocks to 1040. Let's see how that goes. Annoying. I really thought I'd be able to run these cards closer to 1100... Now it seems I'm tweaking the other way

EDIT: I suppose playing with power limit wouldn't do anything for the memory clocks anyway... so pointless playing with that.

Dude, are you sure it is memclock related? I think it is a bug in cast_xmr.

Well, when I was mining on unmodded 56 bios @ 925MHz memory freq the rig was mining several days without a single hickup. So I can only assume stability is memclock related. I'd be very happy to  be proven wrong!

Fair enough. I just switched back to version 0.7 and will let you know how it goes.

What version are you on?

[dragonmike]

Well, stopped again after a few hours. So core clock now back up to 1408 and lowered mem clocks to 1040. Let's see how that goes. Annoying. I really thought I'd be able to run these cards closer to 1100... Now it seems I'm tweaking the other way

EDIT: I suppose playing with power limit wouldn't do anything for the memory clocks anyway... so pointless playing with that.

Dude, are you sure it is memclock related? I think it is a bug in cast_xmr.

Well, when I was mining on unmodded 56 bios @ 925MHz memory freq the rig was mining several days without a single hickup. So I can only assume stability is memclock related. I'd be very happy to  be proven wrong!

Fair enough. I just switched back to version 0.7 and will let you know how it goes.

What version are you on?

I'm on the latest, aka v0.8.
Not sure if it makes any difference in stability... How does the intensity affect hashrate? I've set it to the max (I think that's 9), but no idea if lowering it would allow clocking mem higher (or prevent the miner from stopping)?

[h311m4n]

Well, stopped again after a few hours. So core clock now back up to 1408 and lowered mem clocks to 1040. Let's see how that goes. Annoying. I really thought I'd be able to run these cards closer to 1100... Now it seems I'm tweaking the other way

EDIT: I suppose playing with power limit wouldn't do anything for the memory clocks anyway... so pointless playing with that.

Dude, are you sure it is memclock related? I think it is a bug in cast_xmr.

Well, when I was mining on unmodded 56 bios @ 925MHz memory freq the rig was mining several days without a single hickup. So I can only assume stability is memclock related. I'd be very happy to  be proven wrong!

In my experience, stability issues are almost always related to memory frequencies being too high. Some cards are just worse than others. I have some RX470s that hash non stop at 2050Mhz while others have to be dropped to 1900, otherwise my miners end up crashing. Once they have been all fine tuned, they will mine for days on end. The same goes for Vegas. For polaris this could be troubleshooted with the number of memory errors in hwmonitor.

Just remember that what counts is not to have the fastest hashrate, but to have a stable rig.

[bostonvex]

Well, stopped again after a few hours. So core clock now back up to 1408 and lowered mem clocks to 1040. Let's see how that goes. Annoying. I really thought I'd be able to run these cards closer to 1100... Now it seems I'm tweaking the other way

EDIT: I suppose playing with power limit wouldn't do anything for the memory clocks anyway... so pointless playing with that.

Dude, are you sure it is memclock related? I think it is a bug in cast_xmr.

Well, when I was mining on unmodded 56 bios @ 925MHz memory freq the rig was mining several days without a single hickup. So I can only assume stability is memclock related. I'd be very happy to  be proven wrong!

Fair enough. I just switched back to version 0.7 and will let you know how it goes.

What version are you on?

I'm on the latest, aka v0.8.
Not sure if it makes any difference in stability... How does the intensity affect hashrate? I've set it to the max (I think that's 9), but no idea if lowering it would allow clocking mem higher (or prevent the miner from stopping)?

I've used -I 9 on version 0.6 and up and never had an issue until recently. Only that changed on my end was upgrading to v0.8 and swapping out my mobo. "only thing" = swapping out mobo...big change I know but I have the exact same mobo before and no issues till v0.8. I'm not pressing the cards hard either, running 920MHz memclock.

Are you using JJs modified monitoring script? If you are, then next time it hangs, browse to 127.0.0.1:7777 and try to refresh the page...only after you do that will JJs pick up that the process is hung and restart it.

[dragonmike]

Not using any monitoring script for now. Just trying to make it work reliably, basically.

Failing mobo is like the thing you don't want to happen because it's the one variable you tend to always rule out when you have issues. On my two rigs of RX570's I've got some GPUs chugging at 1950MHz mem speed, while other keep crashing at 1850 and lower -h values (in Claymore). I can't think there's anything wrong with the GPUs, so it's gotta be risers... or the damn mobo.

Let's see how the Vega rig behaves...

[kggcel]

weird... I'm getting only 1.2KH/s with a Vega 64. All drivers are set.

Could it be due to different mining pool?

I'm using dark-mine.su

[pyr8]

I see here that many people claim to get 1950 Hs with memory speed of 945 or so. I'm wondering what i'm doing wrong - i'm also getting 1950, but with memory at 1100 (and gpu at 1407). If i set memory to 945, hash rate goes down to ~1700-1750!?

I'm using Vega 56 cards, with 64 bios and registry files for HBM1100. And of course, blockchain drivers and latest version of Cast XMR.

P.S. i have 6 cards, tried with single one and same result. I'm also using only 2GB of RAM, not sure could that be an issue (but i doubt, as all calculations are done inside GPUs).

thanks

One you upgrade to the 64 BIOS, you have to use 64 settings. 1950 H/s @ 935/945 is for 56 un-moded cards!

Try these settings (for your appropriate card and BIOS type) NOTE that it's a starting point, your results may vary slightly:

Vega 64 Hash Rates

1900 Hash/s with 200 watts power consumption

    Core Frequency: -10%
    Memory Clock: 1025 MHz
    Power Limit: -23%

2000 Hash/s with 250 watts power consumption

    Core Frequency: -10%
    Memory Clock: 1095 MHz
    Power Limit: 0

2050 Hash/s with 280 watts power consumption

    Core Frequency: -10%
    Memory Clock: 1105 MHz
    Power Limit: +20%

Vega 56 Hash Rates

1900 Hash/s with 150 watts power consumption

    Core Frequency: -10%
    Memory Clock: 945 MHz
    Power Limit: -22%

2000 Hash/s with 200 watts power consumption

    Core Frequency: -10%
    Memory Clock: 965 MHz
    Power Limit: 0

2050 Hash/s with 225 watts power consumption

    Core Frequency: -10%
    Memory Clock: 990 MHz
    Power Limit: +12%

Hm, sounds strange to me, as i though vega64 bios is used only because of better undervolting and overclocking possibilities. Instead in relative percentages, do you know exact frequencies of GPU at those settings? I'm most interested in lower power consumption (150W) and getting about 1900 Hash/s is just fine to me. Even if cards doesn't complaint about 1100mhz, i believe they will be more relaxed on 945MHz.

And just to be clear, you didn't even change registry for voltage adjusting, just using installed drivers/cards with factory settings?

Btw, is there any explanation of intensity argument, and how to find the best one for the card (except trying all 10)?

Right now I am getting the following using Stock Vega 56 BIOS on "Blockchain Drivers" from AMD and it's been stable for more than 24 hours:

Average Hashrate = 1923 H/s
Fan set to = 2000 RPM
Temp = 60C (Stable)
GPU Freq = 0%
Memory Freq = 935 MHz
Power Limit % = -10
Power draw at the wall (Measured) = 150W to 165W

That sounds quite nice - mine fans are about 3000-3500 in order to keep HBM temp around 60-62C. But i'm using 6 gpu which of course increase total heat and cards are harder to cool - how much cards you have?
However, i'll give it a try (stock bios, unmodified registry).

Thanks.

My HBM is hovering around 63/65C...do you know what the safe maximum is for these temperatures. I read that HBM will usually run hotter than GPU temps as well, but I'm not sure what a safe maximum is?!

[pyr8]

I see here that many people claim to get 1950 Hs with memory speed of 945 or so. I'm wondering what i'm doing wrong - i'm also getting 1950, but with memory at 1100 (and gpu at 1407). If i set memory to 945, hash rate goes down to ~1700-1750!?

I'm using Vega 56 cards, with 64 bios and registry files for HBM1100. And of course, blockchain drivers and latest version of Cast XMR.

P.S. i have 6 cards, tried with single one and same result. I'm also using only 2GB of RAM, not sure could that be an issue (but i doubt, as all calculations are done inside GPUs).

thanks

One you upgrade to the 64 BIOS, you have to use 64 settings. 1950 H/s @ 935/945 is for 56 un-moded cards!

Try these settings (for your appropriate card and BIOS type) NOTE that it's a starting point, your results may vary slightly:

Vega 64 Hash Rates

1900 Hash/s with 200 watts power consumption

    Core Frequency: -10%
    Memory Clock: 1025 MHz
    Power Limit: -23%

2000 Hash/s with 250 watts power consumption

    Core Frequency: -10%
    Memory Clock: 1095 MHz
    Power Limit: 0

2050 Hash/s with 280 watts power consumption

    Core Frequency: -10%
    Memory Clock: 1105 MHz
    Power Limit: +20%

Vega 56 Hash Rates

1900 Hash/s with 150 watts power consumption

    Core Frequency: -10%
    Memory Clock: 945 MHz
    Power Limit: -22%

2000 Hash/s with 200 watts power consumption

    Core Frequency: -10%
    Memory Clock: 965 MHz
    Power Limit: 0

2050 Hash/s with 225 watts power consumption

    Core Frequency: -10%
    Memory Clock: 990 MHz
    Power Limit: +12%

Hm, sounds strange to me, as i though vega64 bios is used only because of better undervolting and overclocking possibilities. Instead in relative percentages, do you know exact frequencies of GPU at those settings? I'm most interested in lower power consumption (150W) and getting about 1900 Hash/s is just fine to me. Even if cards doesn't complaint about 1100mhz, i believe they will be more relaxed on 945MHz.

And just to be clear, you didn't even change registry for voltage adjusting, just using installed drivers/cards with factory settings?

Btw, is there any explanation of intensity argument, and how to find the best one for the card (except trying all 10)?

Right now I am getting the following using Stock Vega 56 BIOS on "Blockchain Drivers" from AMD and it's been stable for more than 24 hours:

Average Hashrate = 1923 H/s
Fan set to = 2000 RPM
Temp = 60C (Stable)
GPU Freq = 0%
Memory Freq = 935 MHz
Power Limit % = -10
Power draw at the wall (Measured) = 150W to 165W

That sounds quite nice - mine fans are about 3000-3500 in order to keep HBM temp around 60-62C. But i'm using 6 gpu which of course increase total heat and cards are harder to cool - how much cards you have?
However, i'll give it a try (stock bios, unmodified registry).

Thanks.

My HBM is hovering around 63/65C...do you know what the safe maximum is for these temperatures. I read that HBM will usually run hotter than GPU temps as well, but I'm not sure what a safe maximum is?!

I answered my own question:  https://www.reddit.com/r/Amd/comments/72f101/vega_56_safe_hbm_temps/

[bostonvex]

weird... I'm getting only 1.2KH/s with a Vega 64. All drivers are set.

Could it be due to different mining pool?

I'm using dark-mine.su

What are you're settings? Are you using OverdriveNtool? Sounds like a HBM issue. Did you disable and enable the devices with devcon.exe? check this page http://vega.miningguides.com/

[bostonvex]

Not using any monitoring script for now. Just trying to make it work reliably, basically.

Failing mobo is like the thing you don't want to happen because it's the one variable you tend to always rule out when you have issues. On my two rigs of RX570's I've got some GPUs chugging at 1950MHz mem speed, while other keep crashing at 1850 and lower -h values (in Claymore). I can't think there's anything wrong with the GPUs, so it's gotta be risers... or the damn mobo.

Let's see how the Vega rig behaves...

No issues so far with version 0.7 and it has been 4 hours.

[vladaepro]

I see here that many people claim to get 1950 Hs with memory speed of 945 or so. I'm wondering what i'm doing wrong - i'm also getting 1950, but with memory at 1100 (and gpu at 1407). If i set memory to 945, hash rate goes down to ~1700-1750!?

I'm using Vega 56 cards, with 64 bios and registry files for HBM1100. And of course, blockchain drivers and latest version of Cast XMR.

P.S. i have 6 cards, tried with single one and same result. I'm also using only 2GB of RAM, not sure could that be an issue (but i doubt, as all calculations are done inside GPUs).

thanks

One you upgrade to the 64 BIOS, you have to use 64 settings. 1950 H/s @ 935/945 is for 56 un-moded cards!

Try these settings (for your appropriate card and BIOS type) NOTE that it's a starting point, your results may vary slightly:

Vega 64 Hash Rates

1900 Hash/s with 200 watts power consumption

    Core Frequency: -10%
    Memory Clock: 1025 MHz
    Power Limit: -23%

2000 Hash/s with 250 watts power consumption

    Core Frequency: -10%
    Memory Clock: 1095 MHz
    Power Limit: 0

2050 Hash/s with 280 watts power consumption

    Core Frequency: -10%
    Memory Clock: 1105 MHz
    Power Limit: +20%

Vega 56 Hash Rates

1900 Hash/s with 150 watts power consumption

    Core Frequency: -10%
    Memory Clock: 945 MHz
    Power Limit: -22%

2000 Hash/s with 200 watts power consumption

    Core Frequency: -10%
    Memory Clock: 965 MHz
    Power Limit: 0

2050 Hash/s with 225 watts power consumption

    Core Frequency: -10%
    Memory Clock: 990 MHz
    Power Limit: +12%

Hm, sounds strange to me, as i though vega64 bios is used only because of better undervolting and overclocking possibilities. Instead in relative percentages, do you know exact frequencies of GPU at those settings? I'm most interested in lower power consumption (150W) and getting about 1900 Hash/s is just fine to me. Even if cards doesn't complaint about 1100mhz, i believe they will be more relaxed on 945MHz.

And just to be clear, you didn't even change registry for voltage adjusting, just using installed drivers/cards with factory settings?

Btw, is there any explanation of intensity argument, and how to find the best one for the card (except trying all 10)?

Right now I am getting the following using Stock Vega 56 BIOS on "Blockchain Drivers" from AMD and it's been stable for more than 24 hours:

Average Hashrate = 1923 H/s
Fan set to = 2000 RPM
Temp = 60C (Stable)
GPU Freq = 0%
Memory Freq = 935 MHz
Power Limit % = -10
Power draw at the wall (Measured) = 150W to 165W

That sounds quite nice - mine fans are about 3000-3500 in order to keep HBM temp around 60-62C. But i'm using 6 gpu which of course increase total heat and cards are harder to cool - how much cards you have?
However, i'll give it a try (stock bios, unmodified registry).

Thanks.

My HBM is hovering around 63/65C...do you know what the safe maximum is for these temperatures. I read that HBM will usually run hotter than GPU temps as well, but I'm not sure what a safe maximum is?!

I think (but can't claim) that it slows down automatically somewhere about 70C, so you shouldn't be able to burn it out. But if you want to avoid slowdowns keep it under 65.

[Crypto_Future]

Really AMAZED at 470 performance.

One question is it better to have tight timmings or higher VRAM mhz for this miner?

[Cyan_Antares]

Can we get support for sumokoin added?

Pretty Please?

It already has it...

how did you achieve that? It doesnt work for me... I had to use other programs to mine on my desktop computer.

[commonorx]

Can we get support for sumokoin added?

Pretty Please?

It already has it...

how did you achieve that? It doesnt work for me... I had to use other programs to mine on my desktop computer.

You know what, I bet he simply put his sumkoin wallet in and then put in the correct server address and then you know what I bet he started the miner up.

Please use this exact setup for Vega cards.. ****DO NOT CHANGE ANYTHING*****

VEGA CARDS USE THIS *****DO NOT CHANGE***** create a bat file with this below
cast_xmr-vega -G 0 -S sumopool.sonofatech.com:3333 -u Sumoo6sncGt3h3r5dkNKkkHnpAfxGbvomDo14cDJAxsGgVZKPRsEMvQ4tL5hVQj2HihL8JoN93qzn8U 3ckjyLFfdcUomFAnkbV5 -p x --nonicehash --fastjobswitch -I 9

[maleone41]

I'm using devcon to disable and enable my Vegas after a reboot, but no matter what I still have to manually go in and toggle the HBCC Memory Segment switch to get the maximum performance. Anyone have a better workaround or is this still the only way to get it to work to its full potential?

[bostonvex]

I'm using devcon to disable and enable my Vegas after a reboot, but no matter what I still have to manually go in and toggle the HBCC Memory Segment switch to get the maximum performance. Anyone have a better workaround or is this still the only way to get it to work to its full potential?

You have to run it every time before you execute the mining executable just to be safe. I recommend either writing your own bat file or leveraging some of the suggestions on this forum. There is also a mining hash monitoring script that I use that can be found here http://vega.miningguides.com/.

[Morgon]

New --forcecompute switch does not work for Vega yet?

According to AMD, all of the features that benefit compute workloads are already enabled by default in Vega's architecture, so the switch isn't necessary. It's basically for other RX GPUs.

Except AMD is lying and exactly zero people have been able to get equivalent hashrates on Vega with the newer drivers using Cryptonight (for Eth, it's fine). This is pretty well-known at this point.