CUDA error 11 - cannot write buffer for DAG

[Ablazed1]

C:\Users\ilhami\Desktop\Ethereum>setx GPU_FORCE_64BIT_PTR 0

SUCCESS: Specified value was saved.

C:\Users\ilhami\Desktop\Ethereum>setx GPU_MAX_HEAP_SIZE 100

SUCCESS: Specified value was saved.

C:\Users\ilhami\Desktop\Ethereum>setx GPU_USE_SYNC_OBJECTS 1

SUCCESS: Specified value was saved.

C:\Users\ilhami\Desktop\Ethereum>setx GPU_MAX_ALLOC_PERCENT 100

SUCCESS: Specified value was saved.

C:\Users\ilhami\Desktop\Ethereum>setx GPU_SINGLE_ALLOC_PERCENT 100

SUCCESS: Specified value was saved.

C:\Users\ilhami\Desktop\Ethereum>EthDcrMiner64.exe -epool eu1.nanopool.org:9999 -ewal "I AM PARANOID, I KNOW NOTHING WILL HAPPEN BUT..." -epsw x -mode 1 -ftime 10

╔════════════════════════════════════════════════════════════════╗ ║ Claymore's Dual ETH + DCR/SC/LBC/PASC GPU Miner v10.0 ║ ╚════════════════════════════════════════════════════════════════╝

This pool (eth-eu2.nanopool.org) does not support Ethereum addresses as login (or requires worker name in "Login.Worker" format) and cannot be used for devfee mining, therefore it is not supported. However, you can mine on this pool if you specify "-allpools 1" option, default pools (different from this pool) will be used for devfee. Please read "Readme" file for details.This pool (eth-eu2.nanopool.org) does not support Ethereum addresses as login (or requires worker name in "Login.Worker" format) and cannot be used for devfee mining, therefore it is not supported. However, you can mine on this pool if you specify "-allpools 1" option, default pools (different from this pool) will be used for devfee. Please read "Readme" file for details.Pool eth-eu2.nanopool.org removed from the list This pool (eth-us-east1.nanopool.org) does not support Ethereum addresses as login (or requires worker name in "Login.Worker" format) and cannot be used for devfee mining, therefore it is not supported. However, you can mine on this pool if you specify "-allpools 1" option, default pools (different from this pool) will be used for devfee. Please read "Readme" file for details.This pool (eth-us-east1.nanopool.org) does not support Ethereum addresses as login (or requires worker name in "Login.Worker" format) and cannot be used for devfee mining, therefore it is not supported. However, you can mine on this pool if you specify "-allpools 1" option, default pools (different from this pool) will be used for devfee. Please read "Readme" file for details.Pool eth-us-east1.nanopool.org removed from the list This pool (eth-us-west1.nanopool.org) does not support Ethereum addresses as login (or requires worker name in "Login.Worker" format) and cannot be used for devfee mining, therefore it is not supported. However, you can mine on this pool if you specify "-allpools 1" option, default pools (different from this pool) will be used for devfee. Please read "Readme" file for details.This pool (eth-us-west1.nanopool.org) does not support Ethereum addresses as login (or requires worker name in "Login.Worker" format) and cannot be used for devfee mining, therefore it is not supported. However, you can mine on this pool if you specify "-allpools 1" option, default pools (different from this pool) will be used for devfee. Please read "Readme" file for details.Pool eth-us-west1.nanopool.org removed from the list This pool (eth-asia1.nanopool.org) does not support Ethereum addresses as login (or requires worker name in "Login.Worker" format) and cannot be used for devfee mining, therefore it is not supported. However, you can mine on this pool if you specify "-allpools 1" option, default pools (different from this pool) will be used for devfee. Please read "Readme" file for details.This pool (eth-asia1.nanopool.org) does not support Ethereum addresses as login (or requires worker name in "Login.Worker" format) and cannot be used for devfee mining, therefore it is not supported. However, you can mine on this pool if you specify "-allpools 1" option, default pools (different from this pool) will be used for devfee. Please read "Readme" file for details.Pool eth-asia1.nanopool.org removed from the list ETH: 1 pool is specified Main Ethereum pool is eu1.nanopool.org:9999 AMD OpenCL platform not found Be careful with overclocking, use default clocks for first tests Press "s" for current statistics, "0".."9" to turn on/off cards, "r" to reload pools, "e" or "d" to select current pool CUDA initializing...

NVIDIA Cards available: 1 CUDA Driver Version/Runtime Version: 9.1/8.0 GPU #0: GeForce GTX 1050, 2048 MB available, 5 compute units, capability: 6.1

Total cards: 1 ETH: Stratum - connecting to 'eu1.nanopool.org' <198.251.88.37> port 9999 ETHEREUM-ONLY MINING MODE ENABLED (-mode 1) ETH: eth-proxy stratum mode Watchdog enabled Remote management (READ-ONLY MODE) is enabled on port 3333

ETH: Stratum - Connected (eu1.nanopool.org:9999) ETH: Authorized Setting DAG epoch #166... Setting DAG epoch #166 for GPU0 Create GPU buffer for GPU0 CUDA error - cannot allocate big buffer for DAG. Check readme.txt for possible solutions. GPU 0 failed Setting DAG epoch #166 for GPU0 GPU 0, CUDA error 11 - cannot write buffer for DAG GPU 0 failed

is the log. Can you guys help me?

Rig: GPU: GTX 1050 CPU: Core i3-7100 Motherboard: Gigabyte H110M-S2V HDD: WD Caviar Blue 1 TB RAM: GSkill RipjawsV 8GB(1x8)

[jeremyisme]

You need 4gb cards, 2gb is too small. The dag is about 2.3gb now.

[sizeg]

@jeremyisme sometimes has the same error on 1070.

[pioniere]

I don't think this is the issue. I have a 1060 with 3GB and I am also getting this error. This just started happening, I was able to mine without an issue a few days ago.

[rafi213]

I have exactly the same problem, it started suddenly on February 1. My cards are 1060 3gb just. What I noticed is when ETH is mining , this problem doesnt occur. What so little memory 3 gb allready for the DAG file for ETC?

[m5wdev]

same problem with DAG file for ETC

[sizeg]

@ilham123836 @pioniere @rafi213 @m5studio make sure u have a pagefile size 16000-20000.

[pioniere]

Still not working.... Set page file to 16GB, still getting the error. Odd that this is suddenly occurring after months of perfect operation, and the DAG file size is < 3GB.

[m5wdev]

same issue, after a mounts of perfect work rig just refuse to mine i've checked every GPU in rig, see no troubles

Other miners works same way with very low hashrate

[drahlf]

Same here with Zotac 1060 3GB card. Since last epoch change in ETC it refuses to work with "cannot allocate big buffer for DAG" error. I tried different pagefile sizes, no fix. Any chance to fix this with a command line parameter? 3GB seems to be an the edge by now...

[rdavis852]

I had the same issue going from Version 10 to 11. I change 2 or 3 options in the batch file and I think the LIDAG was the important one. It makes the DAG creation go slower....I think. I entered into the batch file: "-lidag 1" (without quotes) Please post if it works for you.

[drahlf]

Thanks for the advice, rdavis852, but still no go for ETC with -lidag1 and 0. I tried ETH now and it loads epoch 169 just fine, I think ETC just got too big for 3GB memory :-(

[Infamous540]

Also got the same issue, tried lidag and gser 2 & 5, dident have this issue before could write dag for 8 gpus same time, but after i´ve connected 3 more gpus it wont work, but it goes further with gser 5 but still not good enought, any solution?

[Muaadib]

I did some looking around and it seems this issue is caused by Windows 10 not allocating all VRAM to CUDA apps. Many posts say it reserves about 20% memory even if the gpu is a secondary (not used for output).

3GB - 0.6 (20%) = 2.4GB.

Which is about the size of the current ETC DAG 186 (2.45GB) while ETH is still at DAG 176 (2.37GB)

If my analysis is right then ETH mining itself will give the same error when it's DAG size start using more than 2.4GB

Unfortunately, the issue is Windows based and Microsoft knew about it for almost 3 years now and still there is no solution.

I guess people should try Win 7 or 8.1 and see if the issue remains...

[kulbabskyy]

-eres 0

1060 3gb works with this param

[Muaadib]

already tried: -eres 0 -lidag 0/1 same result, can mine ETH but not ETC.

[danielehrhardt]

Same Problem here also have 1060 3gb Cards. @albalirus your solution worked for me

[savvasd]

-eres 0 fixed mine, thank you.

[FrankBurmo]

can also confirm that the '-eres 0'-setting fixed the issue for my 1060 3gb-cards

[surfman007]

Can someone post their bat file to show how -eres 0 worked? My 1060 3g rig just started doing the "CUDA error - cannot allocate big buffer for DAG" today and I've tried everything. -eres 0 - 16g vram and nadda

[Muaadib]

-eres 0 worked up until today's DAG (178)

If you want to continue mining ETH with 1060 3GB, you need to downgrade to Win 7/8.1 or Linux

[netkingtw]

Set your driver back to V.388.71. It does resolve the problem today. See below for the Nvidia's link for 388.71 driver - http://www.nvidia.com/content/DriverDownload-March2009/confirmation.php?url=%2FWindows%2F388.71%2F388.71-desktop-win10-64bit-international-whql.exe&lang=tw&type=TITAN

[stefanwendelmann]

@Muaadib same here with zotac 1060 3GB -eres 0 worked until today

@netkingtw will try drivers and report, i got 390.65 atm

[danielehrhardt]

@stefanwendelmann yes same here

[stefanwendelmann]

worked for me

[pieiac]

thank you

[Kiranvarkey]

Zotac gtx 1060 3 gb. Same error guys -CUDA error 11-cannot write buffer for DAG gpu failed. I m using Windows 10. Somebody please help.

[remon-nashid]

I confirm that nvidia driver 388.71 resolves the issue. Thank you @netkingtw. However, out of eleven nvidia cards, only nine are detected with 388 driver. That's not an issue with 390.xx or 391.xx tho.

[Aljook]

I have tried all the things mentioned here until now, just installed the 388.71 drivers as well (and checked if it was applied to everything). Also eres 0, lidag combos, starting two instances of miner to force rebuild dag. Nothing works. Have 10 3gb 1060-s and one 6gb 1060 in the rig.

Any help would be great

[netkingtw]

@Aljook @remon-georgy

Driver 388.71 works for most miners that had the same cuda error for 1060 3GB GPU yesterday but does not work for a very small part of miners. Taiwan mining pool forum released a solutions to their miners as below:

Step 1, Add the following 5 line commands in your Claymore's miner's .bat file（ -eres 0 should still be added as it was before yesterday's problem occured)：

setx GPU_FORCE_64BIT_PTR 0 setx GPU_MAX_HEAP_SIZE 100 setx GPU_USE_SYNC_OBJECTS 1 setx GPU_MAX_ALLOC_PERCENT 100 setx GPU_SINGLE_ALLOC_PERCENT 100

Step 2， Enlarge your rig's virtual memory to at least 32GB.

Step 3, If driver 388.71 does not work for you, use the Windows 10's default Nvidia driver or driver version prior to 388.71 such as 388.13 and make sure that your auto Windows update is disabled .

If all above-mentioned steps still do not resolve your problem, use Claymore 10.6 and try to decrease both core and memory clock a bit until it works.

If all steps fail, use Ethminer instead because Ethminer does not have this issue.

[Muaadib]

Claymore quote

"It seems 3GB cards won't be able to mine ETH/ETC in Windows10 soon because this OS reserves about 20% of GPU memory size and there is no way to fix it (at least easily). In next update I will set zero devfee for these cards (same as I did for 2GB) so you can mine ETH forks. If you still want to mine ETH/ETC on 3GB cards you should use Windows7 or Linux."

https://bitcointalk.org/index.php?topic=1433925.msg33471755#msg33471755

Seems my theory is correct

[Kiranvarkey]

Ethminer 14 solved all my problems in windows 10 and disable updates

[Kiranvarkey]

Make sure 388.13 installed

[stefanwendelmann]

Since yesterday downgrading GForce Drivers to 388.13 on Win10 wont work anymore.

[heroborys]

GForce Driver 388.13 ,claymore miner 11.6 ,virtual memory 35GB and commands -setx GPU_FORCE_64BIT_PTR 0 -setx GPU_MAX_HEAP_SIZE 100 -setx GPU_USE_SYNC_OBJECTS 1 -setx GPU_MAX_ALLOC_PERCENT 100 -setx GPU_SINGLE_ALLOC_PERCENT 100 -eres 0

helped for me on win 10 and 7x gtx 1060 3gb

[stefanwendelmann]

@heroborys works just fine. Thx! But is a 35GB Pagefile for Win10 necessary ?

[heroborys]

@stefanwendelmann i think it can be lower but I have not tried . I set 35gb to be safe

[nikusworld]

Hey Fellow miners,

Just wantedto let you @all know I overcome this CUDA errors with my 1060 3GB's. I have been having issues and resolved for a while by adding -eres 0 for few days, after a while it came back and didn't go away.

Thought of installing Win7 as recommended by 'Claymore': https://bitcointalk.org/index.php?topic=1433925.msg33471755#msg33471755 but couldn’t due to some other issues like my GIGABYTE mobo wouldn’t detect 3.0 keyboard and mouse.

Finally installed Claymore latest version 11.6 and OMG worked right off the bat! and as per Claymore there is no mining fees on 3GB's same as 2GB's now. Happy mining!

[Aljook]

Can someone confirm that nikusworld's solution works? Im currently mining zclassic on miningpoolhub and set it to autoexchange to ETH.

[josamito]

@nikusworld thanks! Works just fine (Windows 10)

[nikusworld]

No worries!

[carljune64]

6x GTX1060 3GB have same issue with 391.35 CPU driver. Downgrade GPU driver version to 388.31 work fine. Thank you.

[stefanwendelmann]

Today it stoped woking again. WIN10 GeForce Driver 388.13 Pagefile 35 GB

[Muaadib]

My recommendation is still the same, switch to win 7/8.1 or linux.

[nikusworld]

Same here stopped since AM.

One thing though, my one small rig 1060 3GB’s running on ETHOS, they are just fine!

Is it worth trying Win8.1, someone has success with 8.1?

[savvasd]

Any other ideas? Isn't there a way to virtualize multiple cards to appear as one or more more powerful cards? Or to allocate virtual VRAM

On Sat, Apr 28, 2018 at 7:23 PM, Muaadib notifications@github.com wrote:

My recommendation is still the same, switch to win 7/8.1 or linux.

— You are receiving this because you commented. Reply to this email directly, view it on GitHub https://github.com/nanopool/Claymore-Dual-Miner/issues/207#issuecomment-385192185, or mute the thread https://github.com/notifications/unsubscribe-auth/AMDabt5l40eMN1lMVKOzCd4l1opwvr7wks5ttKWYgaJpZM4RvQBg .

[princegary]

I had the problem for over a day. took me ages but finally changed to windows 7 and it's up back and running.

[nikusworld]

Thanks for the update! Windows 7 and Ethos is as expected. Did someone tried Windows 8.1?

[dzidass]

Hello, I can confirm that on 8.1 Enterprise OS 5xGTX 1060 3GB work fine. It was tested from yesterday.

[nikusworld]

Thanks for the update @dzidass . I am going to give it a shot Win8.1 when I am back from work. I'll update the community if that works.

Off topic but if someone doesn't have access to Win8.1 or don't wana install etc... just switch to UBIQ almost same hash rate and returns, not too bad IMO vs ETH.

[stefanwendelmann]

Time for a second hdd or bootmanager for Linux/Ubuntu any good guide how to set up linux for claymore eth ?