AMD promised extra particulars on its 14nm FinFET Vega structure—which debuted earlier this week within the type of the compute-focused Radeon Vega Frontier Edition—at Computex in May. But these eager for a quick launch following the reveal are out of luck. Radeon RX Vega, a shopper model of Vega, will probably be proven off at Computex, however will not be in the stores till later within the yr. Radeon Vega FE, the workstation/cloud-oriented half, is at the moment earmarked for a “late June” launch.
News on the provision of RX Vega comes from Radeon VP Raja Koduri, who took to Reddit yesterday in a AMA.
“We’ll be showing Radeon RX Vega off at Computex, but it won’t be on store shelves that week,” mentioned Koduri. “Some of Vega’s features, like our High Bandwidth Cache Controller, HBM2, Rapid-Packed Math, or the new geometry pipeline, have the potential to really break new ground and fundamentally improve game development. These aren’t things that can be mastered overnight…We believe those experiences are worth waiting for and shouldn’t be rushed out the door.”
Koduri additionally took the chance to have a wee dig at Nvidia, which has to date solely carried out HBM2 reminiscence in extraordinarily costly enterprise graphics playing cards just like the lately introduced Tesla V100. Vega FE “employs two stacks of HBM2,” in line with Koduri, providing speeds of as much as 480GB/s. Weirdly, that is much less than the 512GB/s of the Fury X and the 547.7GB/s of Nvidia’s Titan Xp, though it is hardly gradual.
“We’re effectively putting a technology that’s been limited to super expensive, out-of-reach GPUs into a consumer product,” mentioned Koduri. “Right now only insanely priced graphics cards from our competitors that aren’t within reach of any gamer or consumer make use of it…The good news is that unlike HBM1, HBM2 is offered from multiple memory vendors—including Samsung and Hynix—and production is ramping to meet the level of demand that we believe Radeon Vega products will see in the market.”
Elsewhere within the AMA, Koduri confirmed a number of different Vega titbits, together with that it is potential to run Vega FE with Eight-pin and 6-pin PCIe energy connectors, somewhat than the 2 Eight-pin connectors that made it onto manufacturing boards for “extra headroom.” It’s potential the RX model of Vega opts for extra thrifty use of energy. Vega FE may also assist the consumer-focused RX driver for these with deep pockets that use a Vega FE for gaming, whereas the liquid cooled model of the cardboard will function “a slight difference in clock speed.” That mentioned, RX Vega will in the end be sooner for players.
“Consumer RX will be much better optimised for all the top gaming titles, and flavours of RX Vega will actually be faster than Frontier version!”
Finally, Koduri famous that multi-GPU designs, just like the lately revealed Ryzen Epyc CPU—a 32C/64T server chip that options 4 eight-core CPUs—are “possible with Infinity Fabric.”
“Infinity Fabric allows us to join different engines together on a die much easier than before,” mentioned Koduri. “As well it enables some really low-latency and high-bandwidth interconnects. This is important to tie together our different IPs (and partner IPs) together efficiently and quickly. It forms the basis of all of our future ASIC designs.”
ASIC (application-specific built-in circuit) normally refers to a extremely specialised chip that’s designed to execute a sure kind of workload extra rapidly than a general-purpose chip (reminiscent of a CPU or GPU). In this case, Koduri appears to be referring to the chip packaging that connects collectively a number of discrete parts on a single circuit board.
This publish originated on Ars Technica UK