AMD has a competitive processor lineup ; it’s their Radeon (graphics) department that lacks behind. AMD partially unveiled their new GPU architecture called RDNA which can be referred to as a GCN/Navi hybrid during their time at Computex. They were present during the E3 as well, where they announced the flagship of the Ryzen 3000 series and their Graphics cards for the year.
They announced three Graphics card one of which is a limited edition 50th anniversary Graphis card for AMD enthusiasts. For masses, they announced the Radeon RX 5700XT and RX 5700. The Radeon RX 5700XT will also have a 50th anniversary Gold edition. These Graphics cards will be based on the new RDNA technology that AMD introduced. These Graphics cards are mostly referred to as the new Navi Graphics cards, but it should be noted that the ‘original’ Navi architecture is set to be released next year. So, without further ado let’s talk about the new RDNA architecture and the individual graphics cards
As reported earlier, AMD’s new RDNA architecture carries some features over from the historical GCN architecture.
Moreover, it also employs some of the gains from the upcoming Navi architecture. Hence, it is called a hybrid architecture. The architecture is fabricated over TSMC’s 7nm process, and it boasts substantial improvements over the GCN architecture. The most significant features of the architecture are the new compute unit designs, multi-level cache hierarchy, and streamlined graphics pipeline. All of these features and the generational leap adds up to a 1.25x performance leap per clock and 1.5x performance gain per Watt.
One must digest these numbers with a grain of salt as these are provided solely by AMD. Coming to the architectural improvements that AMD has been boasting. The new compute unit design helps in maintaining higher clock speeds consistently while the number of stream processors stays the same at 64 per CU. The streamlined graphics pipeline helps in maintaining hyper-effective clock gating, which results in significant power saving. Lastly, the use of multilevel cache hierarchy allows the new GPUs to access data faster, allowing them to process it efficiently. It stores the data at a different level depending upon their use so that it can access the most used data faster.
AMD is still using their old ‘mid-tier market’ strategy. These graphics cards are for midtier or top midtier market. AMD has no answer to the RTX 2080Ti yet.
AMD Radeon RX 5700
Coming to the individual cards AMD launched last night. The AMD Radeon RX 5700 is marketed against the Nvidia RTX 2060 graphics card. The graphics card is based on the new RDNA architecture and fabbed on TSMC’s 7nm manufacturing process. It has 36 compute units resulting in 2304 stream processors with a base clock speed of 1465MHz, while the ‘near’ boost clock speed is 1725MHz. With the new architecture, AMD has announced a new metric called Game clock speeds which allows these graphics card to stay persistent at slightly higher clock speeds during gaming sessions. It is 1625MHz for the RX 5700.
Like Nvidia, AMD has also switched to the GDDR6 memory module. The did not specify the memory speed or bus size of the modules they are using for their respective graphics cards. A reasonable guess for the RX 5700 could be 14Gbps speed and 192-bit interface. Lastly, the graphics card is rated at 180W TDP and requires a six + eight pin power connector to work. The price of the RX 5700 is only $379, which is almost equal to what Nvidia is asking for its RTX 2060.
AMD also showcased the gaming performance of their new mid-tier graphics card against the RTX 2060. The graphics card performed 2% to 21% better than Nvidia’s offering across the board depending upon the game.
AMD Radeon RX 5700XT
With the arrival of new architecture, AMD is bringing back their old XT naming regime. The better version of the RX 5700 is called RX 5700XT. This graphics card is placed directly against the RTX 2070 graphics card. The specifications include 40 CUs resulting in 2560 stream processors. The base clock speed of the GPU is 1605MHz, while the gaming and boost clock speeds are 1755MHz and 1905MHz, respectively.
They are using an 8GB GDDR6 memory module for this graphics card too. There’s no information regarding the memory specifications, although an educated guess would place it at 14Gbps with a 256-bit memory bandwidth. The graphics card is rated at a power consumption of 225Watt and requires a six + eight pin power connector to work. It will cost around $449, which is significantly less than the cost of its direct rival.
Coming to the gaming benchmarks, AMD displayed during their press conference. Again, these numbers should be taken with a grain of salt. AMD claims that their new RX 5700XT performs 2% to 22% better than the RTX 2070 in the games they tested at 1440p. The RTX 2070 performs slightly better in the shadow of the Tomb Raider and Sid Meier’s Civilization 6 according to the graph AMD provided.
Lastly, they also revealed the 50th anniversary Gold edition of the RX 5700XT. These cards will have better clock speeds which means it will be made by the “cherry-picked” GPUs powering the regular 5700XT. They said that single precision compute performance of the Gold edition is 10.14 TFLOPs, which is higher than the compute performance of the regular RX 5700XT. Lastly, it will be priced at $499 and will only be available for a limited period.
This new lineup will be available for purchase starting July 2019.