The new beta version is ready. You can download PhoenixMiner 5.1b from here:

https://mega.nz/folder/XYkTQZiI#HT_dcKdLtsU26nwbE8pMBQ  (MEGA)

If you want to check the integrity of the downloaded file, please use the following hashes (you need the last file PhoenixMiner_NVRTC_Windows.zip only if you want to mine BCI with Nvdia cards under Windows):

The new features in this release are:

• Added support for VRAM timing adjustments for Nvidia cards of 10x0 series (see the new command-line parameters -straps, -vmt1, -vmt2, -vmt3, and -vmr for more information)
• Added -ttli option to automatically decrease the mining speed to avoid overheating the GPUs over the target temperature (useful when -tmax option is not supported by the drivers)
• Added support for latest AMD drivers 20.7.1 under Windows and 20.20-1089974 under Linux
• Fixed long-standing problems with -gpow option, which now works properly
• Many other small improvements and fixes

Here are some additional notes about the VRAM timing adjustments of Nvidia 10x0 cards:

• Most recent Nvidia drivers require running as administrator (or as root under Linux) to allow hardware control, so you need to run PhoenixMiner as administrator for the VRAM timing options to work.
• The four independently adjustable VRAM timings are controlled by four different parameters: -vmt1, -vmt2, -vmt3, and -vmr. The last one controls the memory refresh rate, and the first three are other timings. The possible values for each of them are from 0 to 100, where 0 are the default timings, and 100 is the most aggressive timing (most GPUs probably wont work with 100).
• The -straps option gives an easy way to set one of the predefined levels of VRAM timing adjustment. The possible values are 0 to 6. 0 is the default value and uses the default timings from the VBIOS. Each strap level corresponds to a predefined set of memory timings ("-vmt1", "-vmt2", "-vmt3", "-vmr"). Strap level 3 is the fastest predefined level and may not work on most cards, 1 is the slowest (but still faster than the default timings). Strap levels 4 to 6 are the same as 1 to 3 but with less aggressive refresh rates (i.e. lower "-vmr" values).
• When using the VRAM timing options, start with lower values and make sure that the cards are stable before trying higher and more aggressive settings. You can use -straps along with the other options. For example -straps 1 -vmt1 60 will use the timings from 1st strap level but -vmt1 will be set to 60 instead of whatever value is specified by the 1st strap level. In such case the -straps option must be specified first.
• The VRAM timing options can be quite different between the GPUs, even when the GPUs are the same model. Therefore, you can (and probably should) specify the VRAM timing options separately for each GPU.

Please let us know if you have any problems or questions related to PhoenixMiner 5.1b.

You should use AMD Memory Tweak (https://github.com/Eliovp/amdmemorytweak binaries are here compiled by me: https://drive.google.com/open?id=1sgYugFgFq5Ja89DexDvwgvP1dtRO2d7A) and you will 100% sure get 1.0%-1.5% more hashrate than Claymore.

Unzip WinAMDTweak.zip to some folder XXX. Run Claymore with your best settings. After that run a command prompt window with Administrative privileges, navigate to WinAMDTweak.zip unzipped folder XXX using 'cd XXX' command and after that run:

WinAMDTweak.exe --current



You will see the current setting of Claymore. Something like this:
https://www.cryptoprofi.info/wp-content/uploads/2019/04/word-image-49.png

Copy/Paste them down in text file.
After that stop Claymore. Run WinAMDTweak.exe --...  with these settings and run Phoenixminer.exe after that with the  same -cclock -mclock -cvddc -mvddc parameters like Claymore.

If you have NVidia GPUs using -strap option leave them with Claymore using Claymore's -di option, and run Phoenix miner with all other AMD GPUs using PhoenixMiner's -gpus option. For example: 'PhoenixMiner.exe -gpus 135' will run PM on your first, third and fifth GPU. The GPU index starts from 1 not from 0 in PM and from zero not from 1 in Claymore.


Another more simple way (that works on Nvidia too) without WinAMDTweak.exe is to run Claymore (don't stop during mining) with your best settings, pause all the GPUs (by pressing the number keys from '0' to '9' or 'p' for all) and run Phoenixminer.

I have 1070 ti titanium, best i get is on claymore with +150 core +650 memory,  65% power limit, strap 1 getting 34.5 mhs 120 watts power draw.

I have 4 x 1070 TI's and I get between 33 and 34 with them. I have +0 core, 70% power and +650 Mem. I am using CM on those ones. I haven't tried PM yet but I am looking at moving all my rigs there soon.

With Phoneix you can get a Little bit more...
- 1st change P2 State off
- With that you should change your memory -200...

Now I'm in Phoneix 32.7 with 150 Core (that depends on model, my model can come with more core) and Memory +525 (but with P2 State off, that means +200 of your set) and Power Limit 66%

If you wanto arribe to 34.7 with the same Power Limit you have to change to the competence...  This values are with ETC.


I tried to change P2 State Off with nvidia profile inspector...
In afterburner, i change to 70% power limit, +0 core, +500 memory
I can get hashrate 32mh/s but after 1minute there is an error showing Cuda error CudaProgram.cu 456

8 GB cards not work Linux, sharing is close to zero.

i'm doing a rig maintenance before i check this... full apt-get update and -upgrade before i install v1.28 but prior to this i thought i'd give the miner a check.. It immediately ramped a 1070ti to 24.5sol/s on Beam, previous version I had was v1.25 and it got at best 23sol/s so after my maintenance finishes i'm hopeful of a few sols more per rig..

edit:

after full apt-get upgrade and update on the rig.. the 1070ti ramped in and runs stable at 24.4sol/s (previously 22sol/s on gminer 1.25)
all other cards seeing similar increases.. however.. this might just be due to updated drivers and not necessarily the miner, but as stated.. from 1.25 to 1.28 the 1070ti ramped in at 24sol/s so i think there is very slight performance improvements in the architecture deployed in 1.28.