Has the memory controller load issue been solved, or a workaround found yet?

Running the latest claymore 9.3 on windows 10 with 16.11.5 drivers using the usual setx parameters. (Issue occurs on every driver I've tried)

I currently have 2 4gb reference 480s acting up with the last few versions of the miner. Troublesome cards are both Hynix memory.

This particular rig is my daily use PC and newly built with 3 RX 480. One card works great, while the other two have difficulties getting to 100% memory controller load. Getting stuck at 70%, 83%, and sometimes 90-94% depending on memory clocks.
It doesn't seem to matter which slot, or driver they are using.

I did spend some time testing various memory straps and timings for on the of the RX 480's.
The only way I was reliably able get 100% memory controller load was to use 1375 or 1425 straps copied over all others and keep the memory clocks no higher than 1750mhz. This seemed to be the only option despite the card being fully stable at much higher clocks. Occasionally a PC reboot would allow 1800mhz at 100% However more often than not, going over 1750mhz will instantly drop memory controller load to around 83% and hashrate down to around 19 mh/s.

Keeping the memory at 1750, and modding the bios to a rather low strap/tight timings it's possible to get 25.8 Eth and 774 DCR with dcri 30 (lowering dcr, only bumps it up to 25.9 give or take.) Compared to my other RX 480 which is working well at 29.3 Eth and 877 Dcr, or even a hynix 470 at 27 Eth these 480 seem stuck at a low hashrate.

I have yet to play around with the newest card, but I expect to hit a similar limitation.


This almost seems like a windows resource management/allocation issue than an problem with the card.

I saw a few similar posts back when 9.0 came out, and I'm hoping someone else with more experience, or luck has figured this out.