AMD Vega 56 and Vega 64 GPUs New Information Available
The GPUs that will be found built-in the iMac Prom Vega 56 and Vega 64 have been particularized by an AMD-provided driver update for Linux, the cards being able to use twice as much data in both registers as the last cards, 32 bit of precision not being needed.
The collected data was gathered from the flagship graphic chipset’s drivers for Linux’s Direct Rendering Manager and brought together by Wccftech and others.
Vega Pro 56 is made from 56 compute units with 3584 stream processors and 8 gigabyte of HBM2 RAM thrusting over 400GB per second of data. While single-precision (FP32) calculations are about 11 tflop analogous to the Nvidia GTX 1080ti, half-precision 16-bit calculations (FP16) such as that used for image and graphic processing, artificial intelligence, ray tracing the game rendering hit the top of 22 tflops.
Vega Pro 64 has 64 compute units with 4096 stream processors and 16 gigabyte of HBM2 RAM but the bandwidth has not been announced yet. It is known to have FP32 single precision calculations at 13 tflops and with FP16 at 25 tflops.
The prices have not been yet disclosed and can’t be determined from the drivers either.
A new set of Vega powered cards were made widely known in affiliation with Vega Pro 56 and Pro 64 chipsets. The Radeon RX Vega gaming card has little specification upon itself and it is still not made clear if macOS will have drivers for either card.
The pre-orders for RX Vega Frontier card have begun at $1200 for the air-cooled version and $1800 for the water-cooled one, but the authenticity of these prices is yet to be known. The card is scheduled for 27th June and is intended for workstations with the same market segment as the iMac Pro.
More details related to the price, the official launch and about the marketing information is expected to be made at SIGGRAPH trade show in July.
The Nvidia GTX 1080ti has about 11 tflop of FP32 performance, but comes short on the FP16 cores with one per streaming multiprocessor while FP32 cores can’t do FP16 calculations. The latest eGPUs tests have shown cards performances when doing FP32 calculations for simplicity.
The FP16 advantages are not yet clear for the consumer and it can’t be compared either to any Nvidia cards. The new architecture of the Vega card increases FP16 performances twice as much, offering calculations that lean on the big boost over other versions of Radeon and Nvidia cards.
Some of the newest versions of Nvidia cards have the same FP16 results to FP32, some considerably lower FP16 performances than FP32. The new Nvidia 1080ti card, often used for builds trying to match the iMac Pro by the admirers, offer less than a tflop of FP16 results.
With specifications leaked from vendor supplied drivers from Linux for Nvidia 10 series and Radeon 500 series cards still may not be accurate enough, but with the announcement of Vega RX Frontier cards, they might be slightly correct.