Phoenixminer 5.6d - CUDA error in CudaProgram.cu:388 : out of memory

[kenstars]

I have Phoenixminer working well for the few newer graphics cards that I can get my hands on, but having issues with an older GTX 970.  I have tried multiple version of drivers (latest 466.27 -> 445.87) with no luck.  I realize the card only has 4GB and obviously running out of memory.  The card benchmarks fine, but when trying to mine it runs out of CUDA.  I have increase virtual memory to 32 GB with no luck and now trying  --daglimit 1000  and even lower settings with no luck.  Phoenixminer DAG always seems to attempt to allocate 4.23GB .  Any suggestions? 



CUDA version: 11.0, CUDA runtime: 8.0
No OpenCL platforms found
Available GPUs for mining:
GPU1: GeForce GTX 970 (pcie 1), CUDA cap. 5.2, 4 GB VRAM, 13 CUs
Nvidia driver version: 445.87
...
...
...
Eth: New job #e7cb3ad4 from us1.ethermine.org:4444; diff: 4295MH
Eth: New job #47cccc97 from us1.ethermine.org:4444; diff: 4295MH
Light cache generated in 3.6 s (19.0 MB/s)
GPU1: Allocating DAG for epoch #414 (4.23) GB
CUDA error in CudaProgram.cu:388 : out of memory (2)
GPU1: CUDA memory: 4.00 GB total, 3.30 GB free
GPU1 initMiner error: out of memory
Fatal error detected. Restarting.

[miner29]

Mine something else the 970 is done on Eth.

You can try lolminer but hoenstly 4gb cards are no longer worth it on eth.

[kenstars]

Thank you.  Earlier today, I found a chart showing that ETH needs a min of 4GB and that must be why the --daglim not working.