Claymore's Dual Ethereum AMD+NVIDIA GPU Miner v15.0 (Windows/Linux)

[lionelnglk]

Hi Everyone,

I can't mine ETC anymore but i still can mine ETH. It keep showing below errors,  

CUDA error - cannot allocate big buffer for DAG. Check readme.txt for possible solutions.

GPU 0 failed

GPU 0, CUDA error 11 - cannot write buffer for DAG

WATCHDOG: GPU error, you need to restart miner

Setting DAG epoch #296(3.31GB)...

i'm using GTX 1050 4gb ram
windows 10
Pagefile intial and max 50gb.
Also tried using these parameters -eres 0, -lidag 0

But error still persist. Anyone knows the solution to this problem? Appreciate all helps. Thks.

Windows 10 for some reason uses a lot of GPU memory which results in these errors. The way the operating system works is that it keeps some system data loaded on the GPU and you can't actually use the entire 4GB.

The easiest way to fix this was just to use Linux and you should be good for many many more months.

I had this issue before when my 2GB stopped working and one thing that worked was to put it in a system with a 4GB GPU and make that GPU device 0 and it worked for a while that way. So if you got any other GPUs with 6GB memory of higher try that or just switch to Linux.

This GTX 1050 is my spare GPU on my laptop..My main system are using intel GPU. So this GTX GPU should be free from any memory taken up by windows 10.

No I am not talking about your system memory RAM. I am talking about your GDDR which is the memory chips directly on the GPU. When you are using Windows 10 it reserves around 500-700MB of your GDDR for some system data.

For gaming this isn't an issue because the game just stores data on the system RAM and the game still works but with ETH mining the entire DAG needs to fit on the GDDR and if 700MB is taken up by Windows then it will fail.

Try Linux

It shows i have 4gb available for my GPU. Is there any way i can check how much Windows 10 is using for the reserve for this GPU? Does switching to Linux meaning replacing the entire windows 10 OS?

NVIDIA Cards available: 1
CUDA Driver Version/Runtime Version: 10.1/8.0
GPU #0: GeForce GTX 1050, 4096 MB available, 5 compute units, capability: 6.1  (pci bus 1:0:0)
Setting DAG epoch #296(3.31GB)...

[1714886389]

1714886389

[1714886389]

1714886389

[1714886389]

1714886389

[adaseb]

Hi Everyone,

I can't mine ETC anymore but i still can mine ETH. It keep showing below errors,  

CUDA error - cannot allocate big buffer for DAG. Check readme.txt for possible solutions.

GPU 0 failed

GPU 0, CUDA error 11 - cannot write buffer for DAG

WATCHDOG: GPU error, you need to restart miner

Setting DAG epoch #296(3.31GB)...

i'm using GTX 1050 4gb ram
windows 10
Pagefile intial and max 50gb.
Also tried using these parameters -eres 0, -lidag 0

But error still persist. Anyone knows the solution to this problem? Appreciate all helps. Thks.

Windows 10 for some reason uses a lot of GPU memory which results in these errors. The way the operating system works is that it keeps some system data loaded on the GPU and you can't actually use the entire 4GB.

The easiest way to fix this was just to use Linux and you should be good for many many more months.

I had this issue before when my 2GB stopped working and one thing that worked was to put it in a system with a 4GB GPU and make that GPU device 0 and it worked for a while that way. So if you got any other GPUs with 6GB memory of higher try that or just switch to Linux.

This GTX 1050 is my spare GPU on my laptop..My main system are using intel GPU. So this GTX GPU should be free from any memory taken up by windows 10.

No I am not talking about your system memory RAM. I am talking about your GDDR which is the memory chips directly on the GPU. When you are using Windows 10 it reserves around 500-700MB of your GDDR for some system data.

For gaming this isn't an issue because the game just stores data on the system RAM and the game still works but with ETH mining the entire DAG needs to fit on the GDDR and if 700MB is taken up by Windows then it will fail.

Try Linux

It shows i have 4gb available for my GPU. Is there any way i can check how much Windows 10 is using for the reserve for this GPU? Does switching to Linux meaning replacing the entire windows 10 OS?

NVIDIA Cards available: 1
CUDA Driver Version/Runtime Version: 10.1/8.0
GPU #0: GeForce GTX 1050, 4096 MB available, 5 compute units, capability: 6.1  (pci bus 1:0:0)
Setting DAG epoch #296(3.31GB)...

If you go to your system resources it will show you a graph of how much of your GPU memory is in use. You can also use a 3rd party tool like MSI Afterburner to show you this info. You need to open the tool and then start the miner program and see how much memory it uses.

You don't need to replace your Win 10, just dual boot it with Linux. So when your computer boots up you pick which OS you want to run. You can't run 2 at once unfortuntely.

[dortortz]

Hi all,

So today my 1050ti 4gb is now giving me "cannot allocate..." Ive had this issue with that card recently but fixed it with an older driver version that someone here suggested. Today it suddenly started happening again. I have updated to latest driver but that didn't work. I tried -eres 0 and that didnt work either. I wont increase page file size because that didnt work last time and I already have it up very high. Any other fix?

[peteycamey]

nvidia DCH drivers doesnt show temperature

[user0071]

Installed new Adrenaline drivers. Trying to install the pixel clock patcher but it's saying that the driver file is too large? WHAT?!

[abg00]

Installed new Adrenaline drivers. Trying to install the pixel clock patcher but it's saying that the driver file is too large? WHAT?!

use 1.4.7

[mu_enrico]

For those who could not start the miner after the October 3 update https://support.microsoft.com/en-us/help/4524147/windows-10-update-kb4524147
Remove this line on your batch file:

Code:

setx GPU_FORCE_64BIT_PTR 0

Well, that's how I solved my problem.

[riverdip]

MINER.HOST Supports Claymore Directly !

https://miner.host/

[Alexgtr]

-strap stopped working. I have 570 and vega, both are work on basic straps (even if writes strap applied successfully), but 1080ti still works fine.
What could be wrong?

[Ursul0]

I know you can take a log file and if you have a long enough mining duration say 1 month, then you can easily add all the accepted+rejected+stale and get an accurate measurement and being within +- 1% is fairly close.

I think the 1% loss you are talking about is not related to the Claymore software but most likely due to the GPU drivers. I know with my AMD RX series GPUs there is a time when some GPUs hash a little slower for a few seconds and then resume the optimal speed. Been trying to figure out why it happens but couldn't and just left it the way it was because the speed loss was minimal. So find in your logs if your speed was always constant or did it slow down from time to time. You can probably find this easily with MS Excel when you load your log file.

Nobody is talking about speed fluctuations here. The averaged stated speed incorporates them. If miner does not work at some moment it prints 0.000 Mh/s in console (and log) and this zero is heaped when calculating the averaged speed and the number of samples is increased by one for this zero. Come on man use a little the last floor of your body. I talk that this averaged stated speed in console and in the log is 1.5% higher than the real one which can be calculated from the found shares (locally, not at the pools side), difficulty and the time that the miner has worked.

I'm trying to follow but it's a bit confusing. So now you are claiming that the speed that the miner displays in console and reports to the pool is 1.5% higher than the actual one? and is that somehow related to your previous claim of Claymore taking higher than advertised devfee?

I was wrong that Claymore is taking higher than the advertised devfee. But I am 1000% sure that Claymore shows and reports to pools (if he reports what he shows in console) about 1.5% higher hash rate than the real one.

Ohh I see.
If you are into this kind of action can you maybe do all of us a favor and perform the same analysis on this one https://bitcointalk.org/index.php?topic=2647654.0
And maybe post here with numbers and such?
I know for myself I'd be very much interested to see the actual numbers. Thx.

Partial results for PM 4.6C after two days mining with 10 RX570 + 1 RX580 and 14889 shares found (14799 shares + 90 devfee shares):
https://drive.google.com/open?id=1luXj5zlyeFdXeQ4QL9GcS1t7bSUPzAQd

Compile AverageSpeed.cpp with visual studio

Or if you use GCC compiler:
comment row 4:
//#include "stdafx.h"

and compile it with:
g++ -o AverageSpeed.exe AverageSpeed.cpp

and run it with the following parameters:
AverageSpeed.exe 3 main Eth speed: log20190929_174105.txt

It will produce the averaged stated speed: 316.689 MH/s
The real effective speed is: 14889 (total shares) * 4,000,000,000 (ethermine difficulty) / 188886 (the seconds that the miner has worked) = 315.301 MH/s
So PM shows (316.689 / 315,301 - 1) * 100 = 0.44% higher hashrate than the real effective one. In comparision Claymore shows about 1.5% higher than the real one.
Will update the results when one week ends but 14889 found shares seems to me statistically representative test that will probably produce error under 0.1%. So I expect this 0.44% to finish between 0.34%-0.54% after 5 days.

this is interesting.

[batsonxl]

i want share my results for rx470 4gb
i have rx470 4gb msi armors and xfx 470 4gb.
Armors performing best in its history giving 30.5-30.9Mhs samsung memory. im amazed now.with modded bios i never saw this speed.
 showing around 77w-85w depends on different cards.

rx 480 8gb giving 32.5Mhs

so sad 4gb will die soon.

Thanks claymore it is good job

[big_daddy]

so sad 4gb will die soon.

why will die soon? what are you writing about
Can you please write few more words
Tnx

[sonerbo]

is any one mining quarkchain ? If yes can you please share your hashrates. Epoch is 46 and this is very lite. So I am wondering if your rates are higher then ETH mining rates. Thanks in Advance !

[ronnieb]

i want share my results for rx470 4gb
i have rx470 4gb msi armors and xfx 470 4gb.
Armors performing best in its history giving 30.5-30.9Mhs samsung memory. im amazed now.with modded bios i never saw this speed.
 showing around 77w-85w depends on different cards.

rx 480 8gb giving 32.5Mhs

so sad 4gb will die soon.

Thanks claymore it is good job

I've heard something in a reddit mining post that if you have a rig full of 4gb cards and at least 1 8gb card, the dag would read everything as 8gb and allow those cards to mine, same would be true for 2gb cards.

I never tried it so don't take my word. something interesting to research into though.

[ronnieb]

i want share my results for rx470 4gb
i have rx470 4gb msi armors and xfx 470 4gb.
Armors performing best in its history giving 30.5-30.9Mhs samsung memory. im amazed now.with modded bios i never saw this speed.
 showing around 77w-85w depends on different cards.

rx 480 8gb giving 32.5Mhs

so sad 4gb will die soon.

Thanks claymore it is good job

I've heard something in a reddit mining post that if you have a rig full of 4gb cards and at least 1 8gb card, the dag would read everything as 8gb and allow those cards to mine, same would be true for 2gb cards.

I never tried it so don't take my word. something interesting to research into though.

It was in this thread

https://www.reddit.com/r/gpumining/comments/d9csi9/2gb4gb_dag_fix_need_atleast_1_gpu_large_enough_to/

[adaseb]

i want share my results for rx470 4gb
i have rx470 4gb msi armors and xfx 470 4gb.
Armors performing best in its history giving 30.5-30.9Mhs samsung memory. im amazed now.with modded bios i never saw this speed.
 showing around 77w-85w depends on different cards.

rx 480 8gb giving 32.5Mhs

so sad 4gb will die soon.

Thanks claymore it is good job

I've heard something in a reddit mining post that if you have a rig full of 4gb cards and at least 1 8gb card, the dag would read everything as 8gb and allow those cards to mine, same would be true for 2gb cards.

I never tried it so don't take my word. something interesting to research into though.

It was in this thread

https://www.reddit.com/r/gpumining/comments/d9csi9/2gb4gb_dag_fix_need_atleast_1_gpu_large_enough_to/

Yes back in the 2016 days when my 2GB  GPUS stopped working, I noticed that in some rigs they mined while in others they didn't. Then I discovered that in the rigs where I had mixed GPUs like the 3GBs with the 2GB GPUs then mining would work, while in rigs purely with 2GB GPUs they didn't.

I discovered that you need to put a 3GB GPU in the device 0 PCIe slot, usually this is the 16x 3.0 PCIe slot. And you can put the 2GB in the other slots.

Right now I can't test this out since all my cards are still working. But if you are getting issues, try and buy 1 GPU which is 8GB like the RX 470 8GB which you can probably buy for like $75 these days and try this out and it should work. Just make sure to put it in the 0 PCIe slot so that GPU loads the most of the DAG plus the Windows data resources and the rest should fit on the other 4GB GPUs.

Either this or just try Linux. There are distros available these days which are plug n play pretty much.

[Bittap]

That was me on Reddit. Cant get my old Bitcointalk account so made one with my business name but yea this absolutely is true now as mentioned by you. Im getting ready to film a youtube sometime this week and put it up on this topic, demo it. Here is a link to my website and youtube.

https://www.bittapfinancial.com/

https://www.youtube.com/watch?v=qnOmZzE8MsU

[MATHReX]

That was me on Reddit. Cant get my old Bitcointalk account so made one with my business name but yea this absolutely is true now as mentioned by you. Im getting ready to film a youtube sometime this week and put it up on this topic, demo it. Here is a link to my website and youtube.

https://www.bittapfinancial.com/

https://www.youtube.com/watch?v=qnOmZzE8MsU

I sold most of my 1060 3gb because of this.
Anyway, I got two remaining. I'll try this with GTX 1660ti connected in the first PCIe slot and the other two cards in the subsequent one.
Can we use risers or just connect it directly to the PCIe slot?

[Bittap]

Really doesn't matter. You just need to have at least 1 card that is large enough and able to load the dag alone. I have 6 rigs this is functioning on, all were 6x Rx470 4GB rigs, but then I took out a card in each and replaced each slot with an Rx580 8GB once I discovered the fix. Distributed the set ups so every 4GB rig has at least one 8GB card to load the dag. I have not confirmed if an Nvidia card can assist another Nvidia card in this way, I don't any under 8GB however the even stranger thing about all of this is that a p104-100 8GB will assist my Rx470 4GBs and they will mine.

[adaseb]

Really doesn't matter. You just need to have at least 1 card that is large enough and able to load the dag alone. I have 6 rigs this is functioning on, all were 6x Rx470 4GB rigs, but then I took out a card in each and replaced each slot with an Rx580 8GB once I discovered the fix. Distributed the set ups so every 4GB rig has at least one 8GB card to load the dag. I have not confirmed if an Nvidia card can assist another Nvidia card in this way, I don't any under 8GB however the even stranger thing about all of this is that a p104-100 8GB will assist my Rx470 4GBs and they will mine.

You will need at least 1 GPU to load the DAG plus the Windows data system files on the GDDR. But you will still need to have at least 4GB for the other GPUs. What I mean is that you can't have 1 GPU that has 8GB GDDR and have the rest being 3GB GDDR GPUs. It won't work I tried.

The way the Claymore software works is that it needs to load the entire DAG onto each and every GPU. It doesn't just load it only on the first GPU. You can look at the resource history monitor and you will see that the memory for each GPU is completely used up. So this will still only work with any GPUs greater than 4GB but anything 3GB or less is obsolete in mining ETH right now.