NVIDIA GeForce GTX 1660 Ti OpenCL Benchmarks, 14-Way NVIDIA/AMD GPU Compute Tests

[vegabook]

Wow, this really reinforces that you have to select the right tool for the job, and ignore all the marketing hype. AMD dominates floating point performance with the ROCm OpenCL stack, meanwhile Nvidia wins on integer performance. And of course AMD wins on open source, while the Nouveau open source Nvidia driver remains unusable for any accelerated workload.

Wait till you see Vega II (Radeon VII)'s FP64 performance. It's 7x faster than the 2080ti. Basically if you can live without CUDA, Radeon VII is simply the best compute card on the market by a vast margin.

[theriddick]

Vega64 is pretty good value now, being only $30aud (20usd) more then the Vega56, if I didn't already own a 1080TI MINI it would certainly be the card to get, and some vendors even sell short form PCB AIB cards (HSF are still full sized).

Shame Rad7 didn't go full AIB with custom PCB's, it probably would have made the card allot more interesting, but we all know stock is ultra low on that GPU (plus they cost more then 2080 here).

[torsionbar28]

Vega64 is pretty good value now, being only $30aud (20usd) more then the Vega56, if I didn't already own a 1080TI MINI it would certainly be the card to get, and some vendors even sell short form PCB AIB cards (HSF are still full sized).

Shame Rad7 didn't go full AIB with custom PCB's, it probably would have made the card allot more interesting, but we all know stock is ultra low on that GPU (plus they cost more then 2080 here).

I don't know about Vega 64, but AMD just dropped the price on Vega 56 to US$279 to compete with the new nvidia 1660ti. All the US retailers are sold out of Vega 56's now, lol.

[piotrj3]

Wait till you see Vega II (Radeon VII)'s FP64 performance. It's 7x faster than the 2080ti. Basically if you can live without CUDA, Radeon VII is simply the best compute card on the market by a vast margin.

Depends what you do. Actually you very rarely need FP64 performance in most stuff, while you really more often welcome int performance of RTXes and tensor cores.