Combine GPU Power

[leonos007]

Heey all,

Maybe I have a silly question, I don't know, so forgive me if so.

I have 4 GPU's, 1060's.

For Ether the Hashrate is about 23Mh/s per GPU with Ethminer (genoil), so all oke.

But if I look to the console screen, the output says things like 'GPU 0: Solution found' and 'GPU 2: Solution found'.
So I think all GPU are doing there own calculations, so in fact they act like they are 4 seperate computers with each having a hashrate of 23Mh/s.

I think the efficiency would be much, much better, if the mining software would combine the 4 GPU power together, having a combined (calculation) hashrate of 92Mh/s
This would be good if the difficulty is high and calculations are running faster.

Maybe I'm mistaken and is the mining-software already doing this, but the console says different.

[1714045889]

1714045889

[1714045889]

1714045889

[1714045889]

1714045889

[1714045889]

1714045889

[1714045889]

1714045889

[giveen1]

Heey all,

Maybe I have a silly question, I don't know, so forgive me if so.

I have 4 GPU's, 1060's.

For Ether the Hashrate is about 23Mh/s per GPU with Ethminer (genoil), so all oke.

But if I look to the console screen, the output says things like 'GPU 0: Solution found' and 'GPU 2: Solution found'.
So I think all GPU are doing there own calculations, so in fact they act like they are 4 seperate computers with each having a hashrate of 23Mh/s.

I think the efficiency would be much, much better, if the mining software would combine the 4 GPU power together, having a combined (calculation) hashrate of 92Mh/s
This would be good if the difficulty is high and calculations are running faster.

Maybe I'm mistaken and is the mining-software already doing this, but the console says different.

Show us a screenshot of the miner program running and we can let you know.

[crypticj]

I feel like we have "blind leading the blind here"

No you cannot combine 4 gpu to run as 1, it doesn't make it faster infact in distributed computing when you have multiple processors/cores/gpus acting as one they run slow [or depends on how fast they can share data between than... thats why supercomputers employ "infini band 10gb/s lanes to share data so they can run all the cores as 1 " but thats expensive proposition. your 4 gpus are on x1 connected with USB so lot of latency issue between them.

so it would be counterproductive.

I can give you 10 other reason why it won't be a good idea, but just do some basic search on how dagger hashimoto algorithm runs by creating DAG file and where it resides in GPU GDR5 ram

hope that helps

[leonos007]

Ok, it's very clear now to me.

Thanks for the reply