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

[Claymore]

There are bugs in ADL, as always. If you have problems with clocks/voltage settings in 17.7.2 - either don't use these options or use previous drivers. I suspect it will take a lot of time to make it work properly... and then AMD will release another ADL update...

[Branko]

There are bugs in ADL, as always. If you have problems with clocks/voltage settings in 17.7.2 - either don't use these options or use previous drivers. I suspect it will take a lot of time to make it work properly... and then AMD will release another ADL update...

watching wattman, memory speed tick-tock from 625 to max defined in miner command line and back, every 2 seconds

Whole problem probably comes from here (driver release notes):

 Radeon WattMan(6)

        Now supports memory underclocking.
        Now supports power state controls.

[pr0ximus]

Hey Claymore,

It would be great if you could tell us what "few minor bug fixes and improvements" you made.

Not sure if you have seen that comment here but somebody already suggested this. It would be great if you could consider these.

Devfee would be more efficient if you could detect best server, my miners use Asia servers, and i saw that devfee mine on US with 2-3 times higher ping. Though it is the same pool...


Could you try to avoid DAG recreation each devfee? Maybe if you can detect that you can use same pools as miner is using currently for your fees you could just switch Workers/wallets. Would be much faster (for your dev fees as well as us )

Thanks for your work

[kkourmousis]

There are bugs in ADL, as always. If you have problems with clocks/voltage settings in 17.7.2 - either don't use these options or use previous drivers. I suspect it will take a lot of time to make it work properly... and then AMD will release another ADL update...

No pressure man, just do it before the new AMD drivers come out      

Kidding of course, don't hurt yourself trying hard 'cause then we are doomed

[MinerosYV]

Seven days with no new deposits showing in Polonies, wondering what is going on, they are not too good at responding tickets

***deposits showing in Poloniex*** what is it your deposit coin

ETC, already sending the coins to another address, is up to 8 days now.

[BennyT]

Thank you, working great!

7 x 1070's were getting 198 mh/s on eth and 3040 on decred. Now getting 203 mh/s and 3100 each. Not quite 3-5% increase but still an increase.

[justchil]

Thank you, working great!

7 x 1070's were getting 198 mh/s on eth and 3040 on decred. Now getting 203 mh/s and 3100 each. Not quite 3-5% increase but still an increase.

From 9.7 > 9.8 you saw an increase?  Did you see any increase on ETH only? 

[BennyT]

Thank you, working great!

7 x 1070's were getting 198 mh/s on eth and 3040 on decred. Now getting 203 mh/s and 3100 each. Not quite 3-5% increase but still an increase.

From 9.7 > 9.8 you saw an increase?  Did you see any increase on ETH only?  

On 9.7 I was running dcri at 38 to get the numbers above. Now I'm running dcri at 55 and able to keep ETH at 200 but decred is able to go up to 3700 which seems to be the sweet spot for profitability. So while I didn't get the 3-5% increase straight out of the box, with some tweaking, there is a noticeable improvement.

I also just tried single mining ETH right now. On 9.7 I was getting 212 mh/s, and now I'm getting 218 which would be just about under a 3% increase in performance --> which leads me to think the dual mining has increased significantly with this build. I wonder if adjusting each card individually for power/memory will respond differently in 9.8. Each card is pulling just over 31.xx mh/s on single mining. Although 218 mh/s single mining is slightly less profitable than 200 mh/s and dual mining another coin.

[Claymore]

I'd like to say a bit more about ADL (library that is used to manage clocks/voltage/fan). I have been working with ADL API for a couple of years, so I have some experience.
When you use ADL, you must ask what version of API supports your GPU. For example, Tahiti reports v5, Hawaii v6.
Every API version uses own set of functions, sample code can be found in ADL SDK. I implemented v5 and v6 and it worked fine for some time.
Then in some drivers AMD disabled underclocking via ADL for some reason, API calls report "OK" but GPU still uses old clocks.
Then AMD releases Polaris cards which report API v7. And no new ADL SDK was released so no headers for new API or any samples. We had to wait for a couple of months before it was finally released.
I added code that uses v7 and it worked fine, except one thing: if I set fan 50%, API sets it to 41% or 46% and I have no idea why. I had to change fan management logic to make it work properly with this issue.
Then in some drivers AMD changed something again and OC stopped working, partially or completely.
Then AMD releases new drivers and Hawaii cards started to report that they use API v7 instead of v6. But in this mode they support v7 a bit differently than Polaris so I had to change my code again to support this new "feature" of drivers.
Then AMD releases 17.7.2 and API v7 returns "not supported" error code. After some research I found that AMD added new functions for API v7 and blocked old functions. So this time they decided not to create API v8, but changed functions for already released API version. It's awesome. No new ADL SDK is released for these changes, so again no headers or code samples for "new" API v7. Finally I made it work, but sometimes it works with strange issues, currently I try to find the reason.
Vega support in ADL has own problems currently, for example, you can set fan to 90% but then there is no way to set it to any lower value.
For Linux, AMD decided to drop ADL support completely in gpu-pro drivers and use new way that I have to support too.

So ADL is the most weird API I've ever seen, but when you see that some options don't work as expected, probably you will blame the miner
On other hand, when I decided to add fan/OC management I found that Nvidia has no public API for it at all. From this point of view, I think weird API is better than no API at all.

[Branko]

I'd like to say a bit more about ADL (library that is used to manage clocks/voltage/fan). I have been working with ADL API for a couple of years, so I have some experience.
When you use ADL, you must ask what version of API supports your GPU. For example, Tahiti reports v5, Hawaii v6.
Every API version uses own set of functions, sample code can be found in ADL SDK. I implemented v5 and v6 and it worked fine for some time.
Then in some drivers AMD disabled underclocking via ADL for some reason, API calls report "OK" but GPU still uses old clocks.
Then AMD releases Polaris cards which report API v7. And no new ADL SDK was released so no headers for new API or any samples. We had to wait for a couple of months before it was finally released.
I added code that uses v7 and it worked fine, except one thing: if I set fan 50%, API sets it to 41% or 46% and I have no idea why. I had to change fan management logic to make it work properly with this issue.
Then in some drivers AMD changed something again and OC stopped working, partially or completely.
Then AMD releases new drivers and Hawaii cards started to report that they use API v7 instead of v6. But in this mode they support v7 a bit differently than Polaris so I had to change my code again to support this new "feature" of drivers.
Then AMD releases 17.7.2 and API v7 returns "not supported" error code. After some research I found that AMD added new functions for API v7 and blocked old functions. So this time they decided not to create API v8, but changed functions for already released API version. It's awesome. No new ADL SDK is released for these changes, so again no headers or code samples for "new" API v7. Finally I made it work, but sometimes it works with strange issues, currently I try to find the reason.
Vega support in ADL has own problems currently, for example, you can set fan to 90% but then there is no way to set it to any lower value.
For Linux, AMD decided to drop ADL support completely in gpu-pro drivers and use new way that I have to support too.

So ADL is the most weird API I've ever seen, but when you see that some options don't work as expected, probably you will blame the miner
On other hand, when I decided to add fan/OC management I found that Nvidia has no public API for it at all. From this point of view, I think weird API is better than no API at all.

Still, bad way to do things from AMD, breaking compatibility is always bad idea, specially when you can fry the card doing it...who wants API that OCs your
card and stops fan at the same time?

[doktor83]

From this point of view, I think weird API is better than no API at all.

Wrong, no API had been better, as people would just use other tools, they used earlier.
This way they blame the miner software, like you said.

[Luisjc69]

hello guys,

with claymore 9.8 can also mining BCC (Cash!) ?

Or which kind of software we need for mining Bitcoincash??


Thank you

Luisa

[jackbox]

hello guys,

with claymore 9.8 can also mining BCC (Cash!) ?

Or which kind of software we need for mining Bitcoincash??


Thank you

Luisa

No. You need ASIC hardware the same as Bitcoin at this time.

[eddiechi]

v9.8:

- added Vega cards support (ASM mode).
- added fan/OC support for 17.7.2 (and higher) drivers.
- fixed issue with -ttli and -ttdcr option for Nvidia cards.
- improved support for systems with more than 10 GPUs, check readme for "-di" and "-mode" options description for details.
- a few minor bug fixes and improvements.

Thank you for the quick updates, much appreciated here!

[letuandnvn]

Hi Claymore and all,

I'm facing intermittent issue.

The Rig can run from an hour to over 14 hours without any issues but the issue occurs as shown below:

20:37:39:527   1044   NVML: cannot get current temperature, error 15
20:37:39:589   1044   NVML: cannot get fan speed, error 15
20:37:39:636   1044   srv bs: 0
20:37:39:636   c08   NVML: cannot get current temperature, error 15
20:37:39:636   1044   sent: 235
20:37:39:636   c08   NVML: cannot get fan speed, error 15
20:37:39:652   1524   NVML: cannot get current temperature, error 15
20:37:39:652   1524   NVML: cannot get fan speed, error 15

Almost same problem with him. Any body can help if "NVML: cannot get current temperature, error 15" message appear always with 6 VGA rig?

After that it is working stable with 5 VGA only (1 disconnect automatically).

Thanks you.

[ozzy1926]

Thank you, working great!

7 x 1070's were getting 198 mh/s on eth and 3040 on decred. Now getting 203 mh/s and 3100 each. Not quite 3-5% increase but still an increase.

From 9.7 > 9.8 you saw an increase?  Did you see any increase on ETH only?  

On 9.7 I was running dcri at 38 to get the numbers above. Now I'm running dcri at 55 and able to keep ETH at 200 but decred is able to go up to 3700 which seems to be the sweet spot for profitability. So while I didn't get the 3-5% increase straight out of the box, with some tweaking, there is a noticeable improvement.

I also just tried single mining ETH right now. On 9.7 I was getting 212 mh/s, and now I'm getting 218 which would be just about under a 3% increase in performance --> which leads me to think the dual mining has increased significantly with this build. I wonder if adjusting each card individually for power/memory will respond differently in 9.8. Each card is pulling just over 31.xx mh/s on single mining. Although 218 mh/s single mining is slightly less profitable than 200 mh/s and dual mining another coin.

Can you share your powerlimit and oc settings?

[l30z1n]

Is it the 17.7.2 that will solve the DAG problem?

[ffthomas]

@Claymore: Thank you for your work. Can you please share with me/us your instructions about updating our rigs?

I have several RX470 and RX480 rigs, all cards are working with modded bios (only memory timing changed).

First, I updated the AMD driver to 17.7.2 (without using DDU, I just installed the new version), then restarted the rig. Second, I ran atikmdag-patcher-1.4.5, and reboot again. I logged in, started CDM 9.8. The miner works, but I can't use WattTool to set clock and voltage parameters. What should I repair, or did I miss something during the upgrading process?

[Yohow]

Hi! I'm getting the FAILOVER 4 fault, so I'm going to edit epools.txt? For ethermine the command line should be "-allpools 1" and "ESM 0" right? And in the "epools.txt" I must write -allpools 1 and ESM 0 too? My main pool is eu1.ethermine.org:1444, alternative pools will be eu1.ethermine.org:4444, eu2:ethermine.org:14444, eu2.ethermine.org:4444, us1, us2 or asian server? Which is better? I'm from spain. Thanks!

[haxiboy]

@Claymore: Thank you for your work. Can you please share with me/us your instructions about updating our rigs?

I have several RX470 and RX480 rigs, all cards are working with modded bios (only memory timing changed).

First, I updated the AMD driver to 17.7.2 (withoud DDU, I just installed the new version), then restarted the rig. Second, I ran atikmdag-patcher-1.4.5, and reboot again. I logged in, started CDM 9.8. The miner works, but I can't use WattTool to set clock and voltage parameters. What should I repair, or did I miss something during the upgrading process?

Amd wants you to use their WattMan, no 3rd party. Rollback to older driver this will be solved...