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

[mares992]

When I overclock GPU result is same.

do NOT overclock GPU, overclock memory only. my 1060 does almost 25 mh/s with +850 MHz mem

Yes I'm overclock memory, here is my Afterburner
http://imgur.com/a/vsv2x

Don't know why GPU clock goes to Boost it can be 1800MHz but Core clock is 0.

Maybe I need older nVidia drivers and windows 10 with anniversary update?

[Odetas]

Which NVIDIA driver version are you guys finding works best w/ 9.7?

[zenono]

When start mining with this miner hash rate is 18mh/s per GPU.

Check if cards have hynix memory (gpu-z for example)

[TechPark]

Is there any real major change between v9.5 and v9.8  
I still use v9.5 and it is very stable for months. I am scared to change it to v9.8  
Can anybody advise?

well, you should install 9.7 it is about 10-15% faster for 1060 and about 3-5% faster for 1070
there is less downdrop in MHs for dual mode as well

That is not true at all. Do you really think Claymore would release a version that is so much slower?

I have done quite a bit of testing both versions and they are identical on both 1070 and 1060.

However, if you are using DCR as dual coin and you get "Server full" message on Yiimp (if you are using that pool) then the ETH hashrate will be much higher because it is not dual mining at that moment.

Ypsi, I believe you got Makak4R suggestion in the wrong way.
As I can see it he advised to go from 9.5 to 9.7 because its faster.
It might be worth not to update to 9.8 as it does not bring performance increase and people reporting all sorts of issues.

[Tarv]

Maybe I need older nVidia drivers and windows 10 with anniversary update?

You need to update to Win10 Anniversary Update or better.  Try the Creators update build 15063.

Latest nVidia driver will be fine.

[Odetas]

Hey guys, both on 9.7 and 9.8, I'm getting "CUDA error - cannot allocate big buffer for DAG."  "GPU 0, CUDA error 11 - cannot write buffer for DAG"

This is on a GTX 960 which was working fine on 9.6

[AzzAz]

Hey guys, both on 9.7 and 9.8, I'm getting "CUDA error - cannot allocate big buffer for DAG."  "GPU 0, CUDA error 11 - cannot write buffer for DAG"

This is on a GTX 960 which was working fine on 9.6

DAG (for ETH ) is now much over 2GB...

[P00P135]

Hey guys, both on 9.7 and 9.8, I'm getting "CUDA error - cannot allocate big buffer for DAG."  "GPU 0, CUDA error 11 - cannot write buffer for DAG"

This is on a GTX 960 which was working fine on 9.6

DAG (for ETH ) is now much over 2GB...

Not "much" just a little.

[Odetas]

It still works on 9.6, just not 9.7 or 9.8 for mining UBIQ

[Zorky]

Amd chance for better performance on rx VEGA?

[mares992]

Maybe I need older nVidia drivers and windows 10 with anniversary update?

You need to update to Win10 Anniversary Update or better.  Try the Creators update build 15063.

Latest nVidia driver will be fine.

I have all last versions. Have creators update and have last nVidia driver

[Makak4R]

Is there any real major change between v9.5 and v9.8  
I still use v9.5 and it is very stable for months. I am scared to change it to v9.8  
Can anybody advise?

well, you should install 9.7 it is about 10-15% faster for 1060 and about 3-5% faster for 1070
there is less downdrop in MHs for dual mode as well

That is not true at all. Do you really think Claymore would release a version that is so much slower?

I have done quite a bit of testing both versions and they are identical on both 1070 and 1060.

However, if you are using DCR as dual coin and you get "Server full" message on Yiimp (if you are using that pool) then the ETH hashrate will be much higher because it is not dual mining at that moment.

man, you got me wrong! I wasn't saying that 9.7 is faster then 9.8! I was saying that 9.7 is faster then 9.6 and that is why the one has to upgrade to newer version.
as of now , with new drivers, better choose 9.8 I think, it will regain speed which was lost due to DAG bug.

that's my opinion only, of course

[Kumic]

Yeah, I have windows 7 and a rig with 570 cards. It is quite a hassle to setup things to work properly. There are some solutions, but it is an impossible task to do it if you have more than a few rigs. I will wait for Claymore upgrade or something. I hope it will come out soon.

[marimach77]

OK, after some tuning I can confirm below setup is working fine for me with stable 30.4MH/s since 24H:

- clean and fresh Windows 10x64 Home installation + all updates
- latest AMD beta drivers
- AB 4.4.0 beta 16 with applied "voltage control" fix and this one
- no atikmdag-patcher applied
- 4x Sapphire Nitro+ RX580 8GB (11265-01-20G, HYNIX memory) with modded BIOS
- latest Claymore 9.8 with no overclocking/undervolting in BAT file

AB config: Screen shot
mv: 970
Power limit: +0
GPU core: 1250
GPU mem: 2150
Fan: 50%
Power from wall: 787W

I managed to get my MSI RX580 Armor 8GB with modded VBIOS to work properly with the New crypto driver by following the above with a slight twist.

http://imgur.com/a/o7srK
After Installing the driver without the atikmdag-patcher applied, i observed that straight after reboot GPU-Z was not displaying accurate readings (GPU load 16% and mem controller 100% without any load). HwInfo64 however was indicating no load of the GPU. I started Claymore 9.8 without any OC settings in the batch file as above.
as indicated below:
http://imgur.com/a/GBkim
It was getting 30.5MH/s but the 990mV from AB was not being applied properly as the card was not undervolting at 990mV but rather 1.1375V using almost 130W and running hotter at 66 Deg Celcius.

Then I did the following that seemed to work:
http://imgur.com/a/RYZgy
I opened Global Wattman that installs with the driver. I went to Voltage control and set it to manual and inserted 990mV there. This dropped my hashrate to 21MH/s!
After setting Global Wattman back to Automatic and applying it then seemed to default to the 990mV setting in AB. Weird right?
From there the card undervolted to the 990mV going back to 30.5MH/s with a power draw of under 100W and descent Temp at 59C. 0.98V stable

As Claymore has posted everytime AMD brings out a new driver they mess with the ADL API which is what 3rd party software use for Overclocking and Voltage settings.
It seems this time not even their own software can properly control the OC and Voltage Settings as i saw the other day with the RX64 Vega Cards and the new driver.

It seems however that by following these steps you can get it to work properly with the new crypto driver.

Indeed, I can confirm your observation. I just read your post and went to check my rig. mV was 970 in AB but GPU-Z showed around 1.23V and 134W per GPU. Then what I did was only to decrease mV in AB from 970 to 960 and now mV and W dropped in GPU-Z. Power from wall decreased from 787W to 610W.

SCREEN SHOT

[biggee]

Hi Claymore,

I believe the voltage issues with 9.8 and the beta driver, are specific to the RX580 cards, from what I've read the RX480s are just fine with the beta driver.

Not true, 480s here, voltage control doesn't work at all on the new driver.

on Afterburner, latest beta, try to set the voltage control in settings to thirdparty

Legend!

I hadn't checked that as the new AB had reset that option.

Thanks bro!

** EDIT **

Actually this half works.

One of my rigs (the one mentioned in my OP) works fine.

Just tried the same to another 6 x GPU rig (5 x RX470 4GB ref and 1 x RX480 8GB all Sapphires) and the voltage control remains locked.

Maybe it's an issue with the 470 Sapphires?

** Double EDIT **

This is nasty with the 470 refs.. Clay keeps crashing (unable to allocate DAG in the GPU memory) resulting in a total system freeze up.

May have to roll back.

[alelingen]

When I overclock GPU result is same.

do NOT overclock GPU, overclock memory only. my 1060 does almost 25 mh/s with +850 MHz mem

Yes I'm overclock memory, here is my Afterburner
http://imgur.com/a/vsv2x

Don't know why GPU clock goes to Boost it can be 1800MHz but Core clock is 0.

Maybe I need older nVidia drivers and windows 10 with anniversary update?

I have the same issue for my gtx 1060 msi armor 6gb. My gtx 1060's core clock stuck 177 mhz when under load. However, my memory clock can be boosted to 3800 mhz. (If I use previous drivers - before release 2017 june-)

And I don't find any solution except update my windows 10.

I tried to set up lots od nvidia drivers and updted vbios. Unfortunately I can't solve my problem.

I will be very pleased any helps.

Thank you.

[JustCauseBIH]

Hi Claymore,

can you share with us any near update release that will work with this new driver?

Also, can you share your opinion on this driver? Should we install, or wait for AB and other software update? Will there be update from them so we can control undervolt on RX series?

[fatsenmy]

 :)hi...

 my rig is not stable for these few days.... message that always show " hang, opencl.. ?  GPU1 0.000 Mh/s......, restart many time  

i wish who have this experience can help me ^^

thanks...

22:23:19:273   11f0   ETH: 08/21/17-22:23:18 - New job from asia1.ethermine.org:14444
22:23:19:289   11f0   target: 0x0000000112e0be82 (diff: 4000MH), epoch 139(2.09GB)
22:23:19:430   11f0   ETH - Total Speed: 43.129 Mh/s, Total Shares: 1, Rejected: 0, Time: 00:02
22:23:19:445   11f0   ETH: GPU0 21.565 Mh/s, GPU1 0.000 Mh/s, GPU2 21.564 Mh/s
22:23:19:461   11f0   ETH: checking pool connection...
22:23:19:461   11f0   send: {"worker": "", "jsonrpc": "2.0", "params": [], "id": 3, "method": "eth_getWork"}

22:23:19:523   11f0   got 248 bytes
22:23:19:523   11f0   buf: {"id":3,"jsonrpc":"2.0","result":["0xe28918b14428415139639df4dbaba800fb514416015c56136c52ba4ee94f7107","0xc50f0eb9498c019d38f28db0aefec38f0a243f1bcbc67281ce622916ee182e4a","0x0112e0be826d694b2e62d01511f12a6061fbaec8bc02357593e70e52ba","0x3fe04c"]}

22:23:19:523   11f0   parse packet: 247
22:23:19:523   11f0   ETH: job is the same
22:23:19:523   11f0   new buf size: 0
22:23:19:633   db0   GPU0 t=70C fan=68%, GPU1 t=50C fan=0%, GPU2 t=70C fan=63%
22:23:19:648   db0   em hbt: 0, fm hbt: 0,
22:23:19:648   db0   watchdog - thread 0 (gpu0), hb time 250
22:23:19:648   db0   watchdog - thread 1 (gpu0), hb time 47
22:23:19:664   db0   watchdog - thread 2 (gpu1), hb time 66813
22:23:19:664   db0   WATCHDOG: GPU 1 hangs in OpenCL call, exit
22:23:19:664   db0   watchdog - thread 3 (gpu1), hb time 67016
22:23:19:680   db0   WATCHDOG: GPU 1 hangs in OpenCL call, exit
22:23:19:680   db0   watchdog - thread 4 (gpu2), hb time 281
22:23:19:680   db0   watchdog - thread 5 (gpu2), hb time 78
22:23:20:977   db0   Restarting OK, exit...

[lentyna]

New AMD drivers (https://support.amd.com/en-us/kb-articles/Pages/Radeon-Software-Crimson-ReLive-Edition-Beta-for-Blockchain-Compute-Release-Notes.aspx) improved my RX580s performance almost my 2 Mh/s

[Claymore]

Hi Claymore,

can you share with us any near update release that will work with this new driver?
Also, can you share your opinion on this driver? Should we install, or wait for AB and other software update? Will there be update from them so we can control undervolt on RX series?

v9.8 works with new driver. Voltage/clocks management fails in some cases because of bugs in drivers.
Right now I'm working on new version with x2 speed for LBC (for AMD cards).