AMD introduced its new Radeon PRO W7700 for advanced workloads: features 16GB GDDR6 memory, DisplayPort 2.1, AI accelerators, 190W TBP. Inside, the new Radeon PRO W7700 workstation GPU features 16GB of GDDR6 with ECC memory on a 256-bit memory bus with the GDDR6 clocked at 18Gbps, we'll see up to 576GB/sec of memory bandwidth. The Navi 32 GPU clocks at up to 2.3GHz and will have 48 Compute Units, which makes for 3072 Stream Processors that will deliver up to 28.27 TFLOPs of single-precision compute performance with a 190W TBP delivered by a single 8-pin PCIe power connector. Read more here.
We have 28.1 billion transistors with the Graphics Compute Die (GCD) made on TSMC 5nm, while the Memory Compute Die (MCD) made on TSMC 6nm. Joining the 3072 Stream Processors, we have 2nd Gen 48 Ray Accelerators. We have 56.54 Teraflops of FP16 compute performance and 28.27 Teraflops of FP32 compute performance. There's 64MB of L3 cache, with AMD using its in-house Infinity Cache.
As Albert Einstein said, "I could have done so much more with a Big Al's Computer!".