Launch with everything else on June 7 —
AMD claims performance-per-dollar leadership over Nvidia, but for how long?
– Jun 11, 2019 11:18 pm UTC
The Radeon RX 5700 XT. It has a funny dent in the top.
Specs on the 5700 XT.
Specs on the non-XT version.
AMD also has a special gold “50th-anniversary edition” it will sell on AMD.com for $499.
The XT all pulled apart.
A shot of the GPU.
The new RDNA architecture promises better performance per watt and per clock, and it will show up in all manner of devices.
If you want to believe AMD’s benchmarks, here’s a chart showing the RX 5700 XT versus what AMD considers to be its competition, the RTX 2070.
The 5700 non-XT against the RTX 2060. Again, wait for independent benchmarks.
AMD’s many design wins recently.
AMD took the stage at E3 to announce its “Navi” family of GPUs. The company’s new graphics cards are officially the AMD Radeon RX 5700 XT and Radeon RX 5700. The 5700 series is launching July 7, making the GPUs a one-two punch alongside AMD’s Ryzen 3000 series CPUs.
AMD isn’t tackling the flagship GPU market with the 5700 series. Instead, the company is aiming for more mainstream pricing with mainstream performance: the 5700XT is $449, while the 5700 is $379. AMD is positioning the cards against Nvidia’s GeForce RTX 2070 ($499) and 2060 ($349), respectively, and claims performance wins in each comparison.
The cards introduce AMD’s new “RDNA” architecture, which AMD says has 1.25x performance-per-clock and 1.5x performance-per-watt over the previous generation. The chips are built on TSMC’s 7nm manufacturing process, a significant shrink from the 12nm process used on the Radeon RX 590, and on Nvidia’s GeForce RTX 2080. The Navi die is significantly smaller than the previous-generation Vega design, with a die area of only 251mm2 compared to the 495 mm2 die area for Vega. The smaller die should make the Navi significantly cheaper to produce than Vega.
For specs, the top-end Radeon RX 5700 XT has 40 compute units with 2560 stream processors total. AMD gives three numbers for the clock rate on the XT: a 1605MHz “Base” clock, a 1755MHz “Game” clock, and a 1905MHz “Boost” clock. As usual, the cheaper 5700 disables compute units and lowers the clock rate, so you have 36 compute units for 2304 stream processor total, a base clock of 1465MHz, game clock of 1625MHz, and a boost clock of 1725MHz.
If any of these clock rates are too conservative for you, AMD is promising the 5700 XT is “overclocking ready” thanks to a power solution with room to grown. Both of AMD’s reference designs come with 8GB of GDDR6 memory.
Enlarge / An exploded view of the XT.AMDFor cooling, AMD’s reference design is a blower-style card with an aluminum shroud and back plate. A vapor chamber draws heat into the heatsink, and everything gets blown out the back of the card. Blower cooling designs have the downside of producing a lot of noise, but AMD is promising a “quiet” card thanks to an “acoustically tuned” design. If you’re unhappy with the cooler design, partner-produced cards with alternative cooling solutions should be out eventually. The cards both take a 8-pin + 6-pin power-supply connection, with AMD providing Board Power figures of 225W for the XT and 180W for the cheaper card.
If you get the whole AMD package in July (meaning a Ryzen 3000 CPU, an X570 motherboard, and a Radeon 5700 card), you’ll be all up and running with the new PCI Express 4.0 bus standard. This faster interconnect will be great for next-gen SSDs and 200Gbps (!) Ethernet controllers. But for video cards, we haven’t seen a compelling use case yet. Still, it provides room for upgrades!
All these specs are nice to know, but they don’t really tell us anything about actual performance. As always, it’s best to wait for independent benchmarks, and if you’re in the market for a GPU, it’s probably also best to see if Nvidia does anything in response.
And speaking of Nvidia’s response, the company has been teasing a “Super” GeForce product for a few weeks now. The rumors point to “Super” cards being up-clocked editions of Nvidia’s existing cards and price drops for some non-super cards, making the company more competitive with AMD’s freshly announced lineup. Again though, benchmarks will tell the real story. AMD also doesn’t have an answer this generation for the real-time raytracing technology that Nvidia introduced with the RTX 2080 Ti. But given the huge hit in performance that ray tracing causes for even Nvidia’s fastest cards (and the limited game support for raytracing), it’s not a huge loss for AMD.
Regardless of what happens between AMD and Nvidia in the PC graphics card wars, AMD GPUs promise to be almost everywhere else in the future. At the beginning of the presentation, AMD talked about the company’s design wins, pointing out that AMD is supplying graphics chips to both the next-gen Xbox and Playstation consoles, Apple’s Mac Pro, and Google’s Stadia game-streaming platform. AMD also recently inked an RDNA licensing deal with Samsung, which will build the graphics technology into its Exynos SoCs for smartphones and tablets.