Quote from: boristhebladecs on June 02, 2020, 02:33:16 PM

OpenCL 19.20 AMD Driver
OS: HiveOS 0.6-142@200528
linux Ядpo 5.0.21-hiveos
4Gb AMD Asus Cards STRIX
Phoenix 5.0b

after epoch 350 in Nicehash Ethash I have low hashrates ~ 10-15 mhs from card. after creation ETC DAG
-dagrestart 1, dont solve the problem. ((
What I am doing wrong? Thx a lot

We have tested with the same Linux drivers 19.20-812932 on Ubuntu 18.04 LTS (slightly older kernel but still 5.x) and everything works fine. Please check your other settings (i.e. clocks), and make sure that you are using auto-tune to find the best -gt values (they could be very different than before DAG epoch 350). Also, sometimes just restart of the rig solves such strange problems.

Quote from: pp_miner on June 02, 2020, 02:35:18 PM

NO I DIDNT SEE THAT
AND IAM USUING THE BLOCK CHAIN DRIVER FROM 2017 "22.19.659.0"

The blockchain driver is not tested with the new big DAG kernels and probably is not working properly. Install newer drivers (anything from 18.12.1.1 to 19.7.5) and make sure that you have switched the GPU Workload to Compute mode (by default they are in Graphics) in the AMD Control Center.

Quote from: BradT on June 03, 2020, 12:18:32 AM

Dear Phoenix,

Do you have any plans to add the DAG-split feature to your miner?

For AMD - to allow 8-gig cards to continue mining ETH after the DAG exceeds 4GB
For NVIDIA - to fix the ETH speed reduction because of the raising DAG (on 10xx cards)

We will definitely continue to support AMD 8 GB cards in the future. However this approach probably won't solve the Nvidia slow degradation of hashrate with each new DAG epoch.

Quote from: djzigi on June 03, 2020, 03:50:07 AM

I did some testing and it appears that there is a problem with -clkernel 3 on 8GB cards, the turbo kernel does not work anymore, changing back to -clkernel 1 did the job but it gives a little less hashpower.
PhoenixMiner or someone, could you please verify and confirm the same problem.
PhoenixMiner tnx and keep up the good work. Wink

There are no changes in -clkernel 3 - it is working as it was before. The problem is in the AMD drivers, which start to partially allocate shared memory (i.e. the mainboard RAM) as part of the DAG buffer as the used VRAM approaches 7.4-7.5 GB, which leads to sharp decrease in the hashrate. So, unless they fix it in the future drivers, -clkernel 3 is not viable for epoch 350 and above.

Quote from: djzmei on June 03, 2020, 12:30:57 PM

I tried both versions of the miner 5b-5c and at least 5 types of drivers and the maximum speed I achieved is 10.4m per card. The cards are RX 580 4G. I will be happy if someone shares if this problem can be fixed.

Please check if your cards are switched to Compute mode in AMD control center. It is easy to miss if you are coming from the blockchain drivers, where there wasn't such option.

Quote from: MTL_Borg76 on June 03, 2020, 08:22:27 PM

Hi all,
Fairly new here to mining (as of 2 hours ago),

Can anyone help me with this please

?

Eth: Giving up after 3 retries, switching to next pool
Eth: Reconnecting in 5 seconds...
Eth: Connecting to ethash pool us2.ethermine.org:14444 (proto: EthProxy)
Eth: Connected to ethash pool us2.ethermine.org:14444 (172.65.226.101)
Eth: New job #c02829d6 from us2.ethermine.org:14444; diff: 4000MH
GPU1: Starting up... (0)
GPU1: Generating ethash light cache for epoch #339
Eth: New job #89396b4b from us2.ethermine.org:14444; diff: 4000MH
Eth speed: 0.000 MH/s, shares: 0/0/0, time: 0:01
Light cache generated in 3.8 s (15.6 MB/s)
GPU1: Allocating DAG (3.66) GB; good for epoch up to #341
CUDA error in CudaProgram.cu:373 : out of memory (2)
GPU1: CUDA memory: 4.00 GB total, 3.33 GB free
GPU1 initMiner error: out of memory
Fatal error detected. Restarting.

Error 2 while trying to restart

I have a Nvidia GeForce GTX960 4gb ddr4 on a AMD FX8320 rig with 32gb ddr3 running win10

The problem is that your video card is probably used by Windows and there are only 3.33 GB free VRAM, while the Ethereum DAG buffer requires 3.66 GB VRAM as you can see the the log file. So, you have either mine another coin with lower DAG epoch, or mine on another video card, with more free VRAM.