More than one article with this sort of headline:

Quote
AMD Radeon VII Won’t Offer Full-Speed Double Precision [Compute function]


These sorts of headline make it seem like a "fault". Thing is, double precision compute is not used in games -- and Radeon VII is for games.

Games that use compute will use single precision Compute -- Radion VII is blazingly fast at single precision:

Quote
Radeon VII — ... 13.8TFLOPs [single precision]..

the recent focus on AI and ML, .. tend to prioritize lower types of mathematical precision [i.e. single precision compute]


Radeon VII "gaming relevant" FP32 (float) performance is 13.8TFLOPs.

Nvidia RTX 2080 Ti is 13.5TFLOPs.

Nvidia RTX 2080 (what Radeon VII competes with pricewise) is only 10.0TFLOPs.

Then, the article above continues:

Quote
the Radeon VII will offer 862GFLOPS [of FP64 double precision]


RTX 2080 Ti only offers 420GFLOPs and RTX 2080 only offers 315GFLOPs. But the article doesn't mention that.

Compute is not yet a big deal in games. However, AMD has traditionally lead in "compute" (why it was favored for cryptocurrency mining) and it continues to lead gaming GPUs in that function.


Sapphire Pulse RX7900XTX, 3 monitors = 23P (1080p) + SAMSUNG 32" Odyssey Neo G7 1000R curve (4K/2160p) + 23P (1080p), AMD R9-7950X (ARCTIC Liquid Freezer II 420), 64GB RAM@6.0GHz, Gigabyte X670E AORUS MASTER MB, (4x M.2 SSD + 2xSSD + 2xHD) = ~52TB storage, EVGA 1600W PSU, Phanteks Enthoo Pro Full Tower, ASUS RT-AX89X 6000Mbps WiFi router, VKB Gladiator WW2 Stick, Pedals, G.Skill RGB KB, AORUS Thunder M7 Mouse, W11 Pro