Good point, didn't realise it was off as default in the new drivers. I only updated the driver as I couldn't get Claymore's software to run after the latest Windows 10 update (fall creators update). It worked after install the new driver but slower as stated (didn't check the compute mode). After downgrading the driver to the blockchain beta all is running fine again so will leave it.

Just tried the new AMD drivers (Crimson ReLive Edition 17.11.1) and saw a 20% drop off in speed. Went back to the blockchain drivers (23rd Aug) and speed back up. Not sure if this is isolated to my rig of RX480s but noticed drop was on both 8GB and 4GB models.

I get nothing but HW errors when trying to mine DaggerHashimoto using this miner, XMR works fine so not sure. Have messed with the intensity setting with no joy. Running 2x R9 480s 4GB at stock, have switched back to ZEC for now.

I did run a couple of tests before. +400 memory for the -25% TDP I get 355Sols/s. +100 to the core, +400 memory and -25% TDP gives 367Sols/s which is back to the starting number.

Guess plenty of room for adjusting to get much higher numbers/power ratio but I was just after something simple that kept the temps down.

Just in case anybody's interested I had a play with TDP reduction to see what the best performance would be with reduce power usage, heat, wear and tear etc... on my 1070.

At stock I get around 367 Sols/s with the new release. Reducing the TDP gives the following:

TDP Dec.	Sols/s
0%	367
10%	363
20%	355
25%	350
30%	343
40%	317

I've gone with a 25% reduction in power usage which only reduces Sols/s by ~5%.

Also seeing a lot of rejected shares using NiceHash. It mined fine for a few minutes and then seemed to get stuck in a loop with all shares being rejected. Was getting 250H/s with a 1070 GTX at stock

Back down to 1.2Gh now, hope the big miners return or we won't be solving many blocks today!

It does work with 1GB cards I've tried it, however 5850 maybe too old. Try updating your driver, reboot and try again.

Just tried on my HD 7800 1GB and it works fine. Message on start-up states that there is not enough memory to run at intensity 4 so it automatically reduces it to 2. Gets around 72H/s

Very nice 164H/s from 144H/s for my stock RX480 4GB (14% increase) and 111H/s from 89H/s for my old R9 380 2GB (25% increase) both on default intensity. Very low CPU usage even on my older AMD A10-5800K

They're ok reckon around 120-140H/s with latest Claymore release. ROI at this speed based on current NiceHash BTC rate is around 180 days or half a year after electricity costs. Obviously very dependent on the ZEC/BTC ratio holding up for that long.

Maybe of interest to somebody. Here are my speeds, not peak but averaged over 1 day of running v2.1 with -I 2 setting. Interestingly no difference in the two RX 480s running on two different rigs.

Rig 1 - i7-6700k CPU with 1 GPU also mining with 7 threads, CPU usage is around 4% for the GPU miner exe.

• Sapphire Nitro+ RX 480 OC 4GB @ Stock settings (1306,1750) - 40.16 (Hynix memory)

Rig 2 - A10-5800K CPU with 2 GPUs also mining with 3 threads/cores, CPU usage is around 17% for the GPU miner exe.

• Sapphire Nitro+ RX 480 OC 4GB @ Stock settings (1306,1750) - 40.16 (Elpida memory)
• ASUS R9 380 2GB @ Stock settings (990,1375) - 32.64

Try this, I posted it on the zcash forum earlier. It restarts the miner if an error is found.

There are two batch files, you need to run the first one which launches the second


Second batch file named nh.bat

I've investigated this a little more and it may not be memory but something to do with the driver and how it turns on the GPU boost. As a test I opened a browser (Edge) and the clock was at 1531MHz and the Memory Load 0.57GB. Load up this site to type this reply and the clock was still at 1531MHz. I then opened a new tab and simply doing this pushes the core clock to 1873MHz, exactly the same Memory Load. Closing the new tab the Clock drops to 1531MHz??? This is repeatable. I've a ton of things running the background but none of these make any difference. I tried a few other programs, playing a video with the built in Windows 10 player pushes it up to 1873MHz again but playing a video with VLC doesn't? Anyway to keep the core clock high I have to leave Edge open with two tabs, which isn't a problem but am interested why this is happening.

Run for a while now and it is stable with 69-72C temp range and the clock fluctuates between 1847.5Mhz and 1873MHz.

TomDD

Apologies if I've missed this somewhere in the thread but I've noticed an interesting behaviour of my 1070 GTX. Basically I normally get around 272MH/s with 1.8.2 but have noticed sometimes it has been dipping to around 225MH/s and I couldn't figure out why. Anyway I did a bit a digging and noticed if the Memory Usage on the GPU (monitored with GPU-Z) drops below around 512MB the GPU Core Clock drops down to a default speed of around 1531MHz. If I open a web browser or something which increases the GPU memory usage it pushes up the GPU core clock up to max and I get 272MH/s again.

Anybody else seen this? I'm using the latest NVIDIA drivers 372.90

TomDD