ETH Mining with NVIDIA - cuda error - cannot allocate big buffer for dag

[Shoxc]

Hey everybody 

I wanted to mine some Ether with my Rig today, but I got this strange error :/
"cuda error - cannot allocate big buffer for dag. check readme.txt for possible solutions claymore"

Well I read the readme.txt, but the suggestions in there didn't lead me to a possible solution.
Therefore I am asking for your help

I am currently mining with:
8x 3GB KFA2 GeForce GTX 1060 OC Aktiv PCIe 3.0 x16 (Retail)
8x ZOTAC GeForce GTX 1060 GF GTX 1060 - 3 GB GDDR5 AMP Edition

Can the DAG already be too big for 3GB cards? I had in mind that this would be the case somewhat around mid 2018 :/

Thank you all very much for your helpful answers in advance!

[Shoxc]

No one? 

[papasmurf1337]

DAG is 2.11 GB
so you can still mine eth.

Set up virtual memory for your PC to 16GB and thats it, it will work.

[Shoxc]

DAG is 2.11 GB
so you can still mine eth.

Set up virtual memory for your PC to 16GB and thats it, it will work.

Hello, thank you for your answer

I followed this guide here : https://www.youtube.com/watch?v=XH4JSt-fQlk

He states that you can push your virtual memory to 2 times the RAM on your board.
This would mean that I need 8 GB RAM on my board to push it, right? (Just want to crosscheck, because I tried to set it to 16GB for my 4 GB of RAM like he does in the video and it doesn't fix the problem)

[Shoxc]

DAG is 2.11 GB
so you can still mine eth.

Set up virtual memory for your PC to 16GB and thats it, it will work.

You have been perfectly right, sorry.
I had to set both, the min and max value to 16GB, because Windows seems to be too stupid to take the max value, if some lower one doesn't work.

Thank you very much for your help!

[knohult]

Same issue here with gtx 1060 3gb. Tried -lidag 1 + 2 + 3 without results.
Virtual memory 16gb to maximum 32gb.

[IITravel01]

When people increase Virtual Memory to 16GB, do they input 16000MB or do you put the actual correct 16384MB?

[mesavoida]

When people increase Virtual Memory to 16GB, do they input 16000MB or do you put the actual correct 16384MB?

It doesn't matter, as long as it's enough.

[amedeusx]

i have same problem today.
mining with 3x 1060 3gb for months, but today i got this cuda error. virtual memory is 32gb. already reinstall the dirvers but still same problem?

edit: -eres 0 command fixed the problem for now

[nitrobg]

Use another card as the primary one. Enable the iGPU on your motherboard.

[r4er]

i have same problem today.
mining with 3x 1060 3gb for months, but today i got this cuda error. virtual memory is 32gb. already reinstall the dirvers but still same problem?

edit: -eres 0 command fixed the problem for now

-eres 0  works! Thanks!

[Vellerephonte]

eres 0 worked for me too. many thanks 

[Martobtc]

Use "-eres 0"

-eres setting is related to Ethereum mining stability. Every next Ethereum epoch requires a bit more GPU memory, miner can crash during reallocating GPU buffer for new DAG. To avoid it, miner reserves a bit larger GPU buffer at startup, so it can process several epochs without buffer reallocation. This setting defines how many epochs miner must foresee when it reserves GPU buffer, i.e. how many epochs will be processed without buffer reallocation. Default value of -eres is 2.

[Berzi]

I have two rigs with 4x 1060 3GB and both stopped working with the same error message. (claymore ETH & DCR)

"-eres 0" helped to go on for a few days, but now they have the same error message since a few days again.

Any recommendations? Shouldn't the 1060 3GB work for ETH about until the beginning of 2019?

Thanks for your help!

[leonix007]

I have two rigs with 4x 1060 3GB and both stopped working with the same error message. (claymore ETH & DCR)

"-eres 0" helped to go on for a few days, but now they have the same error message since a few days again.

Any recommendations? Shouldn't the 1060 3GB work for ETH about until the beginning of 2019?

Thanks for your help!

Use the new version, it could work for a while

v11.6:

- improved hashrate for weak Nvidia cards by about 1%.
- zero devfee for 3GB cards in Windows 10, same as for all 2GB cards.
- applied some tricks to increase available memory for 3GB cards in Windows 10 so you can mine ETH a bit longer.
- now you can press "y" key to turn on "Compute Mode" in AMD drivers for all cards (Windows only).
- Linux version: removed openssl library dependency.
- improved "-logfile" option, now you can use it to specify a folder for log files, use slash at the end to do it, for example, "-logfile logs\".
- added "-epoolsfile" and "-dpoolsfile" options.
- remote management: now "miner_getstat2" command also returns PCI bus index for every GPU.
- a few minor bug fixes and improvements.

Thanks Claymore, for bringing life back to 3GB cards, worked right off the bat!

+1, I can mine Eth on 1060 3GB cards in W10 only with Claymore, all other miners fail now. Also 0% developer fee

[fanatic26]

If you want to mine on your 3gb cards you need to run linux. The way Windows allocates mining GPUs is to blame, it holds half a gig of memory in reserve making your cards effectively 2.5gb and not 3.

[Berzi]

@leonix007:
thanks a lot for this tip, I will give it a try!!

@fanatic26:
thanks also to you. You're right, I am using Windows 10. I have a little linux experience, could you recommend a linux distribution for this kind of work?

[Vellerephonte]

Claymore stopped working last week. I installed windows 7 and the nvidias work great. Altho you need to tweak Windows a bit to get all 6 gpus running. I also had aquite a bit of trouble getting my 7th rx580 to work.
I spent a whole day troubleshooting but it worthed it, rig is also 2mhs faster than in win10.
Ask me if for solutions if you need.

[Berzi]

I tried different Linux systems on Live-Sticks but always the same issue. The OS is booting, but the cards are not found under the system.
I think the problem is that the live linux has not the right drivers, but I don't know how to solve this problem? :/