Claymore's Dual Ethereum AMD+NVIDIA GPU Miner v15.0 (Windows/Linux)

[ajolly87]

Hi everyone, at this moment what are the best cards in terms of price / performance based on RX 5700 for mining eth?
I plan to build a 4-card rig soon and I am undecided between the purchase of reference cards or some 2-fan variants (for example Sapphire Pulse RX 5700 or MSI RX 5700 MECH)...thank you!

+1 vote from me for Saphire

Interesting the lottery, I have put my issues down to thee Sapphire non Pulse RX 5700XT

Tried everything and then went and picked up a Gigabyte OC and no issues at all.

Weirdly the Sapphire would only work with my B250 Mining Expert if plugged into the PCI X16 port... The gigabyte doesn't seem to be that fussy.

[1715198612]

1715198612

[1715198612]

1715198612

[1715198612]

1715198612

[1715198612]

1715198612

[Binary100100]

Well... as expected.
20 minutes in with these settings:
-tt 58 -cvddc 750 -cclock 1300 -fanmin 0 -mclock 904 and I get this:
WATCHDOG: GPU 0 hangs in OpenCL call, exit
WATCHDOG: GPU 0 hangs in OpenCL call, exit
Quit, please wait...

then the computer freezes and I have to reset it. Going to dial back to -mclock 900 and let it sit for a few hours.

EDIT:
51 minutes in and still holding but doesn't seem stable.
58C / 57C / 56C giving ~50.5MH/s each but it's bouncing between ~45MH/s to ~50.5MH/s at times.

DDU, fresh install of the drivers (19.9.1, I had issues with .2) - just install drivers from Device Manager.

Set all clocks/power in miner as you've been trying.

My card is  Saphire (without PULSE) Radeon RX 5700xt. Maybe I had luck with the vendor and the model. Running already 22 hours with stable hashrate (no bouncing), no crashes, mem temp 76 degree C, power draw reported 94W, fans about 44%, room temperature 22 degree C. Using  PM with -tt 58   -cclock 1300 -mclock 910 -cvddc 750

I'm using XFX. Unfortunately, I'm not sure if there's any way that we can determine if there are various types of memory yet. (Micron, Hynix, Samsung, etc)
It could be that XFX might use Micron where Sapphire might use Samsung. Something of that nature.
This would explain why I cannot get stable memory clock beyond 900 where some can even go beyond 910. Of course, every card is different. Question is, how much?

I'll look into DDU some other time. I'm pretty busy this week and that can be time consuming should there be a complication of any kind. Thus far my 5700 XT rig (of three XFX cards) has been running without issue for almost 11 hours with -tt 58 -cvddc 750 -cclock 1300 -fanmin 0 -mclock 900
If you recall when I tried 904 I quickly had issues. Unfortunately, they're not immediately obvious. It could take up to an hour for the OpenCL hang and system freeze. That's a long time to slightly (slowly) make minor adjustments to clock speeds in hopes for 2 or 3 MH/s extra.

I'm also keeping in mind that everything about these cards are still very new. As stated, I don't even know what kind of memory (other than 8GB of DDR6) I have. I don't think that -rxboost or -straps would make any difference yet either. So there's plenty of room for improvement and I see no reason why we can't get at least another 10% from these cards without sacrificing efficiency. There will be driver updates, software updates, bios firmware modifications, etc. I think that we just need to simply settle on whatever we each find to be most stable and wait until these improvements release.

[adaseb]

Hey, ive a question.

I ve RX 570 (Ellesmere) and i use a little OC in memory (like 1600 to 2000). and i use this way always. Now reading the author's notes, i used the Driver install, Strap and RXBoost, and i can see a good changre (since 24 to 31 hasrate), and i dont see changes in the MSI Afterburner (same clock/mermory/tdp/temp that without these options). The question is, if my cards will be afected really with this options. Thanks.

If you have set the OC settings in the Claymore config file and you actually are getting 31 MH/s then you have done it correctly. Depending what AMD drivers you are running, you might get issues with MSI Afterburner. So if you want to verify your clocks you can try using the AMD Driver software called WattMan which is in the AMD display settings, or you can just use GPU-Z. Launch it and you will see your current clocks.

You should also undervolt if you haven't to save some watts and lower the temperature of your GPUs. You can also verify your hash speed in your pool after a few hours since the difference between 24 to 31 mh/s should be noticeable.

[oskoa]

Hey, ive a question.

I ve RX 570 (Ellesmere) and i use a little OC in memory (like 1600 to 2000). and i use this way always. Now reading the author's notes, i used the Driver install, Strap and RXBoost, and i can see a good changre (since 24 to 31 hasrate), and i dont see changes in the MSI Afterburner (same clock/mermory/tdp/temp that without these options). The question is, if my cards will be afected really with this options. Thanks.

If you have set the OC settings in the Claymore config file and you actually are getting 31 MH/s then you have done it correctly. Depending what AMD drivers you are running, you might get issues with MSI Afterburner. So if you want to verify your clocks you can try using the AMD Driver software called WattMan which is in the AMD display settings, or you can just use GPU-Z. Launch it and you will see your current clocks.

You should also undervolt if you haven't to save some watts and lower the temperature of your GPUs. You can also verify your hash speed in your pool after a few hours since the difference between 24 to 31 mh/s should be noticeable.

Hello and thanks. What i win exactly undervolting the Cards?

[batsonxl]

Hi claymore and everyone else.
I need help with this timing straps and OC.
Im using nicehash.there is claymore 15 on it.

1st problem: i use -strap 1 into settings (extra lunch parameters) it works if i OC gpus from after burner.gets 29Mhz for rx 470.but after while every time after new DAG settings happens my cards starts use more power(wattman same result) so every time i need from after burner move slider low or high and hot apply.then again anda again happens.im tired of this, what is the problem? when it drains more power more hotter my gpu are get.

2nd problem: i cant lunch rxboost. on nicehash in extran lunch parametrs i write -strap1 -rxboost, strap works but rxboost not.

Help guys in this difficult days
Thanks

[zenstrive]

Anyone mining with RX 5700,
has any of you experiencing strange phenomena where only the one of the GPU is actually mining/or having its memory loaded with mining job?

[WhackOBill]

Hello and thanks. What i win exactly undervolting the Cards?

Lower power usage and cooler GPU temps.

[Binary100100]

Well... I hate to say this... but I'm getting 60MH/s using PhoenixMiner 4.6c now.

-tt 58 -cvddc 750 -cclock 1300 -fanmin 0 -mclock 900 -amd -acm -mi 12 -gt 126 -mode 1 -clKernel 1 -clNew 1 -clf 0 -lidag 1 -openclLocalWork 128 -openclGlobalMultiplier 4096

~95 watts per card and 59C.

I like it a lot!

(Credit to PedPandaMining)

[Bathmat]

Well... I hate to say this... but I'm getting 60MH/s using PhoenixMiner 4.6c now.

-tt 58 -cvddc 750 -cclock 1300 -fanmin 0 -mclock 900 -amd -acm -mi 12 -gt 126 -mode 1 -clKernel 1 -clNew 1 -clf 0 -lidag 1 -openclLocalWork 128 -openclGlobalMultiplier 4096

~95 watts per card and 59C.

I like it a lot!

I'm getting incorrect shares with those settings... and not just a couple.

[Binary100100]

Well... I hate to say this... but I'm getting 60MH/s using PhoenixMiner 4.6c now.

-tt 58 -cvddc 750 -cclock 1300 -fanmin 0 -mclock 900 -amd -acm -mi 12 -gt 126 -mode 1 -clKernel 1 -clNew 1 -clf 0 -lidag 1 -openclLocalWork 128 -openclGlobalMultiplier 4096

~95 watts per card and 59C.

I like it a lot!

I'm getting incorrect shares with those settings... and not just a couple.

Yeah, I got a lot too. About 30%.
I'm trying this now:
-tt 58 -cvddc 750 -cclock 1300 -fanmin 0 -mclock 900 -amd -acm -mi 12 -mode 1 -clKernel 3 -clNew 1 -clf 0 -lidag 1 -openclLocalWork 128 -openclGlobalMultiplier 4096

I removed the -gt and hope that clears up a bit. Seems like it needs some tuning still but it's progress!


-tt 58 -cvddc 750 -cclock 1300 -fanmin 0 -mclock 900 -amd -mode 1 -clKernel 3 -clNew 1 -openclLocalWork 64
Seems much better. It's been running for about 4 minutes and no incorrect shares (yet).

[bategojko74]

Well... I hate to say this... but I'm getting 60MH/s using PhoenixMiner 4.6c now.

-tt 58 -cvddc 750 -cclock 1300 -fanmin 0 -mclock 900 -amd -acm -mi 12 -gt 126 -mode 1 -clKernel 1 -clNew 1 -clf 0 -lidag 1 -openclLocalWork 128 -openclGlobalMultiplier 4096

~95 watts per card and 59C.

I like it a lot!

It seems to me that PM is optimized especially for 32 threads per work group and doesn't work correctly for more. So I think that this 60 MH/s is fake and that your -openclLocalWork 64/128 causes incorrect shares.

[Binary100100]

Well... I hate to say this... but I'm getting 60MH/s using PhoenixMiner 4.6c now.

-tt 58 -cvddc 750 -cclock 1300 -fanmin 0 -mclock 900 -amd -acm -mi 12 -gt 126 -mode 1 -clKernel 1 -clNew 1 -clf 0 -lidag 1 -openclLocalWork 128 -openclGlobalMultiplier 4096

~95 watts per card and 59C.

I like it a lot!

It seems to me that PM is optimized especially for 32 threads per work group and doesn't work correctly for more. So I think that this 60 MH/s is fake and that your -openclLocalWork 128 causes incorrect shares.

I'm trying this:
-tt 58 -cvddc 750 -cclock 1300 -fanmin 0 -mclock 900 -amd -mode 1 -gt 100 -clKernel 3 -clNew 1 -openclLocalWork 64

And it's pretty solid. I'm only getting 56.6 though. No incorrect shares.

[bategojko74]

Well... I hate to say this... but I'm getting 60MH/s using PhoenixMiner 4.6c now.

-tt 58 -cvddc 750 -cclock 1300 -fanmin 0 -mclock 900 -amd -acm -mi 12 -gt 126 -mode 1 -clKernel 1 -clNew 1 -clf 0 -lidag 1 -openclLocalWork 128 -openclGlobalMultiplier 4096

~95 watts per card and 59C.

I like it a lot!

It seems to me that PM is optimized especially for 32 threads per work group and doesn't work correctly for more. So I think that this 60 MH/s is fake and that your -openclLocalWork 128 causes incorrect shares.

I'm trying this:
-tt 58 -cvddc 750 -cclock 1300 -fanmin 0 -mclock 900 -amd -mode 1 -gt 100 -clKernel 3 -clNew 1 -openclLocalWork 64

And it's pretty solid. I'm only getting 56.6 though. No incorrect shares.

..for now. You are a great optimist.

[Binary100100]

Well... I hate to say this... but I'm getting 60MH/s using PhoenixMiner 4.6c now.

-tt 58 -cvddc 750 -cclock 1300 -fanmin 0 -mclock 900 -amd -acm -mi 12 -gt 126 -mode 1 -clKernel 1 -clNew 1 -clf 0 -lidag 1 -openclLocalWork 128 -openclGlobalMultiplier 4096

~95 watts per card and 59C.

I like it a lot!

It seems to me that PM is optimized especially for 32 threads per work group and doesn't work correctly for more. So I think that this 60 MH/s is fake and that your -openclLocalWork 128 causes incorrect shares.

I'm trying this:
-tt 58 -cvddc 750 -cclock 1300 -fanmin 0 -mclock 900 -amd -mode 1 -gt 100 -clKernel 3 -clNew 1 -openclLocalWork 64

And it's pretty solid. I'm only getting 56.6 though. No incorrect shares.

..for now. You are a great optimist.

As stated, it still needs some tweaking but it's progress!
I'm keeping it with these settings for now. Only 1 incorrect share out of 22 so far.

[Bathmat]

Well... I hate to say this... but I'm getting 60MH/s using PhoenixMiner 4.6c now.

-tt 58 -cvddc 750 -cclock 1300 -fanmin 0 -mclock 900 -amd -acm -mi 12 -gt 126 -mode 1 -clKernel 1 -clNew 1 -clf 0 -lidag 1 -openclLocalWork 128 -openclGlobalMultiplier 4096

~95 watts per card and 59C.

I like it a lot!

It seems to me that PM is optimized especially for 32 threads per work group and doesn't work correctly for more. So I think that this 60 MH/s is fake and that your -openclLocalWork 128 causes incorrect shares.

I'm trying this:
-tt 58 -cvddc 750 -cclock 1300 -fanmin 0 -mclock 900 -amd -mode 1 -gt 100 -clKernel 3 -clNew 1 -openclLocalWork 64

And it's pretty solid. I'm only getting 56.6 though. No incorrect shares.

..for now. You are a great optimist.

As stated, it still needs some tweaking but it's progress!
I'm keeping it with these settings for now. Only 1 incorrect share out of 22 so far.

So you have a 5% speed gain, but losing 5% to incorrect shares.... seems like a wash to me. Until the kernels are written for that worksize (if it's even possible), seems like using the default settings is the best way to go.

[Binary100100]

Well... I hate to say this... but I'm getting 60MH/s using PhoenixMiner 4.6c now.

-tt 58 -cvddc 750 -cclock 1300 -fanmin 0 -mclock 900 -amd -acm -mi 12 -gt 126 -mode 1 -clKernel 1 -clNew 1 -clf 0 -lidag 1 -openclLocalWork 128 -openclGlobalMultiplier 4096

~95 watts per card and 59C.

I like it a lot!

It seems to me that PM is optimized especially for 32 threads per work group and doesn't work correctly for more. So I think that this 60 MH/s is fake and that your -openclLocalWork 128 causes incorrect shares.

I'm trying this:
-tt 58 -cvddc 750 -cclock 1300 -fanmin 0 -mclock 900 -amd -mode 1 -gt 100 -clKernel 3 -clNew 1 -openclLocalWork 64

And it's pretty solid. I'm only getting 56.6 though. No incorrect shares.

..for now. You are a great optimist.

As stated, it still needs some tweaking but it's progress!
I'm keeping it with these settings for now. Only 1 incorrect share out of 22 so far.

So you have a 5% speed gain, but losing 5% to incorrect shares.... seems like a wash to me. Until the kernels are written for that worksize (if it's even possible), seems like using the default settings is the best way to go.

You mean a 10% gain?
50MH/s increased to 55MH/s is 10%. But I get your point.

[Bathmat]

So you have a 5% speed gain, but losing 5% to incorrect shares.... seems like a wash to me. Until the kernels are written for that worksize (if it's even possible), seems like using the default settings is the best way to go.

You mean a 10% gain?
50MH/s increased to 55MH/s is 10%. But I get your point.

Yeah, sure. I tried all the settings you just did, and every time, the % gain was almost exactly offset by % invalid shares. However, I did not run very long tests, so it could be possible that pool-side effective hashrate might go up even with the incorrect shares, but I doubt it.

[Binary100100]

So you have a 5% speed gain, but losing 5% to incorrect shares.... seems like a wash to me. Until the kernels are written for that worksize (if it's even possible), seems like using the default settings is the best way to go.

You mean a 10% gain?
50MH/s increased to 55MH/s is 10%. But I get your point.

Yeah, sure. I tried all the settings you just did, and every time, the % gain was almost exactly offset by % invalid shares. However, I did not run very long tests, so it could be possible that pool-side effective hashrate might go up even with the incorrect shares, but I doubt it.

All of my incorrect shares were from the same GPU (I have three) so it's possible that it's just my XFX GPU since it's also the GPU that won't let me clock past 900Mhz without the watchdog error. Thinking of returning it.

But from 49.5 to 57 seems like an improvement to me. Once Claymore releases some Navi kernels for his software then we'll be in business for sure.

[bategojko74]

So you have a 5% speed gain, but losing 5% to incorrect shares.... seems like a wash to me. Until the kernels are written for that worksize (if it's even possible), seems like using the default settings is the best way to go.

You mean a 10% gain?
50MH/s increased to 55MH/s is 10%. But I get your point.

Yeah, sure. I tried all the settings you just did, and every time, the % gain was almost exactly offset by % invalid shares. However, I did not run very long tests, so it could be possible that pool-side effective hashrate might go up even with the incorrect shares, but I doubt it.

All of my incorrect shares were from the same GPU (I have three) so it's possible that it's just my XFX GPU since it's also the GPU that won't let me clock past 900Mhz without the watchdog error. Thinking of returning it.

But from 49.5 to 57 seems like an improvement to me. Once Claymore releases some Navi kernels for his software then we'll be in business for sure.

Claymore has already released his version with Navi support. Quoting Claymore:
-----------------------------------------------------
v15.0:

- now miner supports up to #384 epoch (4GB DAG size). Note that previous versions support up to #299 epoch, you will not be able to use old versions after #299 epoch.
- added support for Navi cards (ETH-only mode).
- now miner sets environment variables automatically (required for 4GB AMD cards).
- a few minor bug fixes and improvements.

PS. AMD still has no public Navi drivers with DAG-fix, so currently 5700XT shows a very bad hashrate. But AMD already solved this issue and they promise to release public drivers soon...
-----------------------------------------------------

What else do you expect from him, to start walking on water like Jesus did?

[Binary100100]

So you have a 5% speed gain, but losing 5% to incorrect shares.... seems like a wash to me. Until the kernels are written for that worksize (if it's even possible), seems like using the default settings is the best way to go.

You mean a 10% gain?
50MH/s increased to 55MH/s is 10%. But I get your point.

Yeah, sure. I tried all the settings you just did, and every time, the % gain was almost exactly offset by % invalid shares. However, I did not run very long tests, so it could be possible that pool-side effective hashrate might go up even with the incorrect shares, but I doubt it.

All of my incorrect shares were from the same GPU (I have three) so it's possible that it's just my XFX GPU since it's also the GPU that won't let me clock past 900Mhz without the watchdog error. Thinking of returning it.

But from 49.5 to 57 seems like an improvement to me. Once Claymore releases some Navi kernels for his software then we'll be in business for sure.

Claymore has already released his version with Navi support. Quoting Claymore:
-----------------------------------------------------
v15.0:

- now miner supports up to #384 epoch (4GB DAG size). Note that previous versions support up to #299 epoch, you will not be able to use old versions after #299 epoch.
- added support for Navi cards (ETH-only mode).
- now miner sets environment variables automatically (required for 4GB AMD cards).
- a few minor bug fixes and improvements.

PS. AMD still has no public Navi drivers with DAG-fix, so currently 5700XT shows a very bad hashrate. But AMD already solved this issue and they promise to release public drivers soon...
-----------------------------------------------------

What else do you expect from him, to start walking on water like Jesus did?

I've removed the "-openclLocalWork 64" completely and I'm averaging 52 instead of 49.5 with ZERO errors or incorrect shares. PhoenixMiner says that the new kernel is included in the latest version. IF the developers coordinate and assist each other instead of competing against each other more can be accomplished. Granted I don't know if they do share tips with each other. Anyway, I still prefer Claymore... but so far Phoenix yields better results after this update. Not sure how he did it but it is an improvement nonetheless. I hope he can figure something out (and I'm sure he will) because I prefer Claymore.