Supported Gpu List
cyber cyb

Supported GPU list for cyberd validators

In our cyber protocol implementation on GO proof of relevance root hash is computed on Cuda GPUs every round as the best way to calculate merkle tree faster. We need to load the whole graph in memory for calculating that’s why memory volume is important. GPU with 6Gb memory can calculate graph with ~200 M links.

GPU Supported Tested CUDA cores Memory Year of production
GEFORCE RTX 2080 Ti ✅ ❌ 4352 11GB GDDR 6 2018
GEFORCE RTX 2080 ✅ ❌ 4352 11GB GDDR 6 2018
GEFORCE RTX 2070 ✅ ❌ 2304 8 GB GDDR6 2019
GeForce RTX 2060 ✅ ❌ 1920 6 GB GDDR6 2019
GEFORCE GTX 1660 Ti ✅ ❌ 1536 6GB GDDR6 2019
GEFORCE GTX 1660 ✅ ❌ 1408 6GB GDDR5 2019
GEFORCE GTX 1650 ✅ ✅ 896 4GB GDDR5 2019
GeForce GTX 1080 ✅ ✅ 2560 8 GB GDDR5X 2016
GeForce GTX 980 ✅ ❌ 2048 4 GB GDDR5 2014
TITAN Xp ✅ ❌ 3840 12 GB GDDR5 2017
GeForce GTX 1080 Ti ✅ ❌ 3584 11 GB GDDR5X 2017
GeForce GTX 980 Ti ✅ ❌ 2816 6 GB GDDR5 2015
GeForce GTX 1070 Ti ✅ ✅ 2432 8 GB GDDR5 2017
GeForce GTX 1070 ✅ ✅ 1920 8 GB GDDR5 2016
GeForce GTX 970 ✅ ❌ 1664 4 GB GDDR5 2015
GEFORCE GTX 1060 6GB ✅ ✅ 1280 6 GB GDDR5 2016
GeForce GTX 1050 Ti 4GB ✅ ❌ 768 4 GB GDDR5 2016
GeForce GTX 745 (OEM) 4GB ✅ ❌ 768 4 GB GDDR3 2014
GeForce GTX TITAN X ✅ ❌ 3072 12 GB GDDR5 2016
GeForce GTX TITAN Z ✅ ❌ 5760 12 GB GDDR5 2014
GeForce GTX TITAN Black ✅ ❌ 2880 6 GB GDDR5 2014
GeForce GTX 770 ✅ ❌ 1536 4 GB GDDR5 2013

If you have used some GPU from column supported but without ✅ at tested column please submit a pull request with corrections. If you have tested GPU and it’s not contained in that list submit PR too.

Note If you using some old cards (like GTX 770, or older) make sure your card will be supported by al least v.410 of NVIDIA diver for Linux.