 Show Posts
|
|
Pages: [1] 2 3 »
|
Windows 10 strikes again. It uses quite a bit of GPU VRAM for its own internal needs. Here is the list of possible solutions in order of increasing dificullty:
* You have to add the following command-line options in order to minimize VRAM usage: -rvram -1 -eres 0
* Usually Windows 10 grabs memory only from the GPU that is connected to a display, so connect your display (or the display plug) to the iGPU of the motherboard (if you are using Intel motherboard). On multi-GPU rigs, you can also put an old (unusable for mining) GPU and connect the display to it.
* The heavy-duty solution is to switch to Linux but it has its own problems. This will buy you about 5-9 additional months of operation until the DAG grows too big to fit in 4 GB VRAM.
Note for Linux users: the latest AMD Linux driver (19.30-855429) is not supported by PhoenixMiner 4.5c. For now, stick with 19.30-838629. We will release 4.6 shortly (in a week or two), and it will contain some goodies for Linux, including advanced hardware stats, and hardware control for AMD cards.
-rvram -1 -eres 0 ==> only last about a week not 5-9 additional months and is no longer working. Please HELP! Yes, these options won't buy you much time. We never said that - if you read the original message, the best solution is to use Linux, which doesn't hog GPU memory like Windows 10 does. However, we recognize that this may be quite impractical, especially if the machine is not a dedicated mining rig. So, your best bet is to either use the integrated GPU (if you are using an Intel CPU) as primary GPU (i.e. connected to a display), or "sacrifice" one of the GPUs and use it as primary GPU, and run the miner on the rest, which shouldn't be affected by the problem. comes to the rescue each time Phoenix miner having problems with 4GB cards. The rest...still mining with Phoenix (8GB cards). |
|
|
|
Same here as follows: GeForce GTX 1050 Ti (pcie 8), CUDA cap. 6.1, 4 GB VRAM, 6 CUs
==>GeForce GTX 1050 Ti: CUDA error in CudaProgram.cu:407 : out of memory Windows 10 strikes again. It uses quite a bit of GPU VRAM for its own internal needs. Here is the list of possible solutions in order of increasing dificullty: * You have to add the following command-line options in order to minimize VRAM usage: -rvram -1 -eres 0 * Usually Windows 10 grabs memory only from the GPU that is connected to a display, so connect your display (or the display plug) to the iGPU of the motherboard (if you are using Intel motherboard). On multi-GPU rigs, you can also put an old (unusable for mining) GPU and connect the display to it. * The heavy-duty solution is to switch to Linux but it has its own problems. This will buy you about 5-9 additional months of operation until the DAG grows too big to fit in 4 GB VRAM. Note for Linux users: the latest AMD Linux driver (19.30-855429) is not supported by PhoenixMiner 4.5c. For now, stick with 19.30-838629. We will release 4.6 shortly (in a week or two), and it will contain some goodies for Linux, including advanced hardware stats, and hardware control for AMD cards. ==> only last about a week not 5-9 additional months and is no longer working. Please HELP! |
|
|
|
Having issue with 4gb Nvidia cards in windows not having enough memory to create the DAG. Its not a virtual memory error
Same here as follows: GeForce GTX 1050 Ti (pcie  , CUDA cap. 6.1, 4 GB VRAM, 6 CUs ==>GeForce GTX 1050 Ti: CUDA error in CudaProgram.cu:407 : out of memory |
|
|
|
I wanna bask in what remains of my stay here =)
I guess he's still basking longer than usual.  yep, been non-stop busy handling all the things that fell apart while away from home. I'm finally taking a day for myself, hopefully Ill be up to getting it compiled tonight.... but I wont have it compiled on the updated VS, or CMake, ill be compiling it on my already configured remote machine; I need to get myself to a good internet connection to be able to download and install the newer apps on my laptop. Give it a day or two longer. Ill get it done. I guess he is still non-stop busy handling all the things that fell apart while he's away from home!!!  |
|
|
|
I wanna bask in what remains of my stay here =)
I guess he's still basking longer than usual. |
|
|
|
Why don't you compare with real winners: t-rex and ccminer?
T-Rex/CCMiner does not support ETHASH algo. Pay attention - the comparison started with MTP {RTX 2070 Zotac Gaming AMP (core +100, memory +900) MTP - TTMiner 2.2.6 - 1,623MH/s MTP - T-Rex 0.12.1 - 3,292 MH/s} Look, I do not mine MTP but ETH/ETC. |
|
|
|
Why don't you compare with real winners: t-rex and ccminer?
T-Rex/CCMiner does not support ETHASH algo. |
|
|
|
|
TT-Miner 2.2.5 vs Phoenix Miner 4.2c
1. Hash rate (winner=TT-Miner 2.2.5)
2. Accepted shares (winner=Phoenix Miner 4.2c)
3. Rejected shares (winner=Phoenix Miner 4.2c=NONE)
4. Dev fee (winner=Phoenix Miner 4.2c=0.65%)
TT-Miner 2.2.6 vs Phoenix Miner 4.2c
1. Hash rate (winner=Phoenix Miner 4.2c)
2. Accepted shares (winner=both)
3. Rejected shares (winner=both=NONE)
4. Dev fee (winner=Phoenix Miner 4.2c=0.65%)
|
|
|
|
|
TT-Miner 2.2.5 vs Phoenix Miner 4.2c
1. Hash rate (winner=TT-Miner 2.2.5)
2. Accepted shares (winner=Phoenix Miner 4.2c)
3. Rejected shares (winner=Phoenix Miner 4.2c=NONE)
4. Dev fee (winner=Phoenix Miner 4.2c=0.65%)
|
|
|
|
Hey Trailing,
on MTP your miner atm is the fastest on GTX cards, but have some problems on mixed rigs with 1070/1080 cards, dunno why.
But on RTX cards your miner is much slower than your competitor. Are there any more improvements possible?
Hi, did you try to 'play' around with intensity or grid size. You can set these values for each gpu separate. Please try this for the RTX as well - if you haven't tried it yet. Unfortunately I do not have a RTX available and cannot run my own test, but maybe someone from the community can help - or give me some more information what the problem could be??? Thanks. >>>You can set these values for each gpu separate Either but not both, this case will not work: -d 0,1,2 -i 15,16,17 You can't have both -d & -i at the command line. Can you fix that? Hi, do you use the last version? I just tried it here and it seems to work. Can you send program log please (-log)? Thanks. Assume 10 GPUs (0-9) and I wanted to test first 2 GPUs (0,1). It works but I have to do this way: -d 0,1 -i 16,16,0,0,0,0,0,0,0,0 If you do this it will not work due to TT-Miner knows that you have 10 GPUs not 2 GPUs, -d 0,1 -i 16,16 You will get this error message, >>>>>Number of selected devices and intensity command line parameter didn't match. >>>>>Please check your command line (parameter -d, -i), cannot continue! |
|
|
|
Hey Trailing,
on MTP your miner atm is the fastest on GTX cards, but have some problems on mixed rigs with 1070/1080 cards, dunno why.
But on RTX cards your miner is much slower than your competitor. Are there any more improvements possible?
Hi, did you try to 'play' around with intensity or grid size. You can set these values for each gpu separate. Please try this for the RTX as well - if you haven't tried it yet. Unfortunately I do not have a RTX available and cannot run my own test, but maybe someone from the community can help - or give me some more information what the problem could be??? Thanks. >>>You can set these values for each gpu separate Either but not both, this case will not work: -d 0,1,2 -i 15,16,17 You can't have both -d & -i at the command line. Can you fix that? |
|
|
|
default settings fails on pure 1070 rigs , mixed , 1060 6gb , 1070ti works fine with default settings
Check your OC setting. Mine have mixed of 1070 & 1080 and default (i-15) or i-16 works as expected. @TrailingStop Can you add a new option to NOT submit stale shares? |
|
|
|
New release 2.2.0 is available.
I see here on my system also some hashrate improvements for ProgPoW and MTP, but it may be my 'overoptimized' system.
So I do not know if you will see this improvements as well. I see them at default settings ( default intensity and default grid size).
Please let me know if you run into any issues.
Happy mining.
1. -log option =>  2. hashrate improvements => (-i 16/17 for both 1070 & 1080 GPUs) |
|
|
|
1. OK - so no new file for each day?
2. what can I see if api-bind works? What program do I need to use this? What is the benefit/function of api-bind? Why and how do you want to use it?
1. Could you please just implement that anything passed afer -log is used for the log file name? No need for the fancy logic here. If log exists - append to the end. 2. Honestly - just ignore it. If dude has problems explaining what he needs it's not your fault neither is your obligation to do some research. 1. I have a rig configured of 14 GPUs, a set of 1070x8 and another 1080x6. Can't run all at once because it will crash. So configure to have two processes run per set. With the default log set, I ended up having just 1 log file for both processes which ever run second. In this case, I would have log file for second process but lost the first. 2. Yeah! As a developer, he likes to be challenges so I did just that so he can do better each time. So, just shut your mouth  if you do not know nothing about what others are saying. |
|
|
|
New version 2.1.19 ready for download.
Release notes:
- Headline-color changed from blue to yellow.
- Reduced memory-requirements for MTP (host-side only - not GPU)
Please report all issues and bugs you find. Happy mining.
Thanks
1. option: -log -> work suggestion for v2.1.20: -log rig1@MintPond.log2. hash rate: miner always looking for more per release. 3. api-bind: still does not work Thanks for your suggestion. I have some questions: 1. makes sense - no date? Today TT-Miner creates a new logfile each day - so the date makes sense to have it in the filename 2. 3. I'm still not sure what you are looking for??? Can you please let me know step by step what you do if it works? Thanks! 1. Want to name my log file not auto-generate by TT-Miner. 3. I wish I knew more to explain to you. You can research with TRex-Miner or CryptoDredge. |
|
|
|
|
@dev
How does api-bind work? Thanks
|
|
|
|
New version 2.1.19 ready for download.
Release notes:
- Headline-color changed from blue to yellow.
- Reduced memory-requirements for MTP (host-side only - not GPU)
Please report all issues and bugs you find. Happy mining.
Thanks
1. option: -log -> work suggestion for v2.1.20: -log rig1@MintPond.log2. hash rate: miner always looking for more per release. 3. api-bind: still does not work |
|
|
|
New version 2.1.18 ready for download.
Release notes:
- New option: -log creates a logfile that contains the screen output and
additional information that could be helpful to find bugs.
- New option: -luck shows a second statistic line - additional to the rig
speed. It shows the time it may take to find a new solution(share)
at the current difficulty. It also calculates a 'luck' value that
show the luck in percent for the next share.
- New option: --api-type does nothing so far, TT-Miner support TCP connections only
- Remove the three stars in the 'Accepted share' message.
- The meaning of the -A parameter changed slightly. If you define an algo without
- cuda version, TT-Miner will load the algo that fits best to your installed
- driver version - it will no longer load the algo for the last cuda version!
Please report all issues and bugs you find.
Thanks.
Will test --api-type and report back to you. Thank you. You rock! I hope that is what you expect? I'm not sure that I really got your point with this option. Anyway - let me know if I should change something. Thanks - any happy mining. Report: 1. --api-type (parameter ignored) OR --api-type TCP : does not work 2. CTRL+C still does not work. Have to press 'q' then 'ctrl+c' Thank so much for taking the time to fix and release v2.1.18. Hi, what do you expect with --api-type TCP? ? Please let me know step by step what you are doing please? API uses TCP - so that should work. Here are command-line arguments from CryptoDredge, =========================================================== --api-type Specify API type to use (default: ccminer-tcp) ccminer-tcp (TCP) ccminer-ws (WebSocket) off -b, --api-bind IP:port for the miner API, 0 disabled (default: 127.0.0.1:4068) =========================================================== Can you implement like that? I think if you start TT-Mminer and listen on a simple TCP/IP socket for single string API requests from the same machine running ccminer and reply with a string and then close the socket each time. If you add the "--api-type" option, it will accept API requests from any network attached computer. It works for me using this setup as an example: -b 123.456.789.012:99999 --api-type ccminer-tcp Last but not least, TT-Miner start up with few lines has dark-blue color, please change to a diff. color. Hard to see the text in dark-blue with black in background. |
|
|
|
New version 2.1.18 ready for download.
Release notes:
- New option: -log creates a logfile that contains the screen output and
additional information that could be helpful to find bugs.
- New option: -luck shows a second statistic line - additional to the rig
speed. It shows the time it may take to find a new solution(share)
at the current difficulty. It also calculates a 'luck' value that
show the luck in percent for the next share.
- New option: --api-type does nothing so far, TT-Miner support TCP connections only
- Remove the three stars in the 'Accepted share' message.
- The meaning of the -A parameter changed slightly. If you define an algo without
- cuda version, TT-Miner will load the algo that fits best to your installed
- driver version - it will no longer load the algo for the last cuda version!
Please report all issues and bugs you find.
Thanks.
Will test --api-type and report back to you. Thank you. You rock! I hope that is what you expect? I'm not sure that I really got your point with this option. Anyway - let me know if I should change something. Thanks - any happy mining. Report: 1. --api-type (parameter ignored) OR --api-type TCP : does not work 2. CTRL+C still does not work. Have to press 'q' then 'ctrl+c' 3. Miner start up with few lines has dark-blue color, please change to a diff. collor. Hard to see the text in dark-blue with black in background. Thank so much for taking the time to fix and release v2.1.18. |
|
|
|
New version 2.1.18 ready for download.
Release notes:
- New option: -log creates a logfile that contains the screen output and
additional information that could be helpful to find bugs.
- New option: -luck shows a second statistic line - additional to the rig
speed. It shows the time it may take to find a new solution(share)
at the current difficulty. It also calculates a 'luck' value that
show the luck in percent for the next share.
- New option: --api-type does nothing so far, TT-Miner support TCP connections only
- Remove the three stars in the 'Accepted share' message.
- The meaning of the -A parameter changed slightly. If you define an algo without
- cuda version, TT-Miner will load the algo that fits best to your installed
- driver version - it will no longer load the algo for the last cuda version!
Please report all issues and bugs you find.
Thanks.
Will test --api-type and report back to you. Thank you. You rock! |
|
|
|
|
