i have no idea what it means 
Buying it one Ebay doesn’t make the hardware grade bad. Just remember when a 3 million dollar data center upgrades where do you think a lot of the old older harware goes?
AMD ATI card is not working with CUDA_PLOT. Nvidia CUDA core video card only.
I am sure there will be “all GPU plot” in the near future.
1 Like
AMD still bad drivers…
I see quite a bit lower performance than stated with GigaHorse CUDA - but it’s in part that I have my GPU power limit set lower due to overheating issues, and partly waiting for Disk writes (I’ve never had available M.2 or even non-USB SSDs at all to use on plotting).
The lower memory requirement and the ability to “direct” output plots to multiple outputs and via network automagically however makes up for the handicaps vs Bladebit RAM.
I also VASTLY prefer the “don’t need to access my node just to farm” from FlexFarmer.
Overall a definite plus over other existing software.
I will be looking at Howard’s plotter and farmer when they actually release, though, given the no dev fee aspect.
3 Likes
Note - “older” hardware, which tends to be lower performance.
Doesn’t make it BAD - just not AS good as current stuff tends to be.
goes back to putting XUbuntu on my “new to me, arrived today” SuperMicro 6047 72-drive system
1 Like
That was the second-gen Pentiums - the P75/P90/P100.
ONLY.
And it was rarely seen unless you worked a lot with floating point - I had a P90 for years.
I tried in unbuntu with 1080 and tested for 4.5 minutes
2 Likes
Guys,
Bladebit Cuda ( GPU ) working very good, smooth, but i can’t get a final destination.
sudo ./bladebit_cuda -z 6 -n 1 -t 8 -f b2e7c68f24fb89c94a0d2a84dc224d288254f08aa52fd291f03c6d22db4f266d10 -c xch19wfxgjt96hs8h3ae86ndu7supkz7aha6jn5mpgvq4tkleq cudaplot /mnt/nvme/ -d /mnt/hdd1/
– it is ignoring -d ( final destinatio )
it only plots and save it to nvme directly.
"
Completed Plot 1 in 141.20 seconds ( 2.35 minutes )
/mnt/nvme/plot-k32-c06-2023-03-24-16-04-1343a1dfd2dd6d7f7e3370039bbe5cc9072adb68ffea572e4880ed9e97a8c7ad.plot.tmp → /mnt/nvme/plot-k32-c06-2023-03-24-16-04-1343a1dfd2dd6d7f7e3370039bbe5cc9072adb68ffea572e4880ed9e97a8c7ad.plot
Completed writing plot in 7.37 seconds"
Should I accept it and let it plot on nvme and then make a script to move the plots to hdds ?
What are you using to farm these plots from Bladebit_cuda?
alpha 3.0
https://downloads.chia.net/bladebit/current/
256 GB RAM + GPU : nvidia RTX 3060
- with 3080 TI ( I have 1.35 min/plot - but it takes 250W)
This is kind of in the wrong topic for this 
, but I think you have to put -d before the output directory, or perhaps specify with something like -t1.
so
./bladebit_cuda -z 6 -n 1 -t 8 -f xxx -c xxx cudaplot -d /mnt/hdd1/ /mnt/nvme/
or
./bladebit_cuda -z 6 -n 1 -t 8 -f xxx -c xxx cudaplot -d /mnt/hdd1/ -t1 /mnt/nvme/
OR, this isn’t working yet on the cuda version 
best ask in keybase for a more solid answer
1 Like
That link only has Linux files.
I downloaded that file March 9th, “Beta’s???”
Good afternoon.
I have 1080TI
Please tell me if it is supported for plotting by the GigaHorse and the BladeBit? (Or are there limitations due to the CUDA version or something else?)
The same question about the RTX A2000, can it be carved on it and how much can it process if it is farming?
What volume at what level of compression can be obtained from these cards?
Thank you in advance!!!
P.S. I know that you can test it yourself, but the cards are in different servers / PCs and in different places.
I never used Windows for CHIA.
I ONLY use ubuntu 20.04 and 22.04
Yes. it is OK.
Keep in mind : Plotting : needs 6500 - 7100 MB from GPU ( GigaHorse) and 5700 MB ( BladeBit)
A2000 - is better. – coz it needs 40W insteed of 200 W( 1080 TI)
I already tested.
Not working. It is not like Gigahorse.
Thank you very much.
Thanks, I looked and saw the numbers for 1080, sorry for not doing it right away.
I’m worried about the A2000 because it only has 6GB of memory.