AMD’s CEO Lisa Su laid out a very aggressive vision for the future of the company at its first keynote ever at Computex 2019. AMD picked a good year to participate in the opening Computex keynote—it had a lot of new products to announce, including its latest CPU and GPU built on the bleeding edge 7nm process node from TSMC. Let’s dive in and take a closer look at what AMD had to say.
Introducing new 7nm architectures
Numerous announcements were made during the press conference, but arguably the biggest were the new CPUs and GPUs, both based on new architectures and the 7nm process node. AMD is leading both in CPUs and GPUs with 7nm which should theoretically help the company gain some share from Intel and NVIDIA. As a prelude to these product announcements, AMD’s CEO Lisa Su emphasized that the company’s 7nm CPU and GPU architectures would power a custom chip inside the next PlayStation.
The new the Ryzen 3000 series of CPUs features a full lineup, with the new 7nm Zen 2 cores, all of which AMD claims are significantly faster than the previous generation. They continue AMD’s tradition of compatibility with AM4 sockets, though it is not a full compatibility because older AM4 motherboards won’t be able to support PCIe Gen 4, one of the Ryzen 3000 series’ key features. PCIe Gen 4 requires a different hardware design and BIOS, which means that its safe to assume that no boards other than X570 boards will support it (the X570 chipset is effectively a copy of the I/O chiplet inside of the Ryzen 3000 series desktop chip, with more I/O—more on this later). AMD is the first company to implement and support PCIe 4.0, which should translate to some significant performance increases for storage devices and potentially improved bandwidth for compatible GPUs.
AMD claims a 15% IPC uplift with the new Zen 2 cores, as well as doubled L3 cache sizes and double floating-point performance over the previous generation of Zen cores. One of the new chips, the Ryzen 7 3700X, features 8 cores, 16 threads, a 3.6 GHz base clock, and 4.4 GHz boost clock. It also features 36MB of cache with an overall TDP of 65W, which should theoretically leave some room for overclocking. AMD put its 3700X head to head against Intel’s i7 9700K in Cinebench, a popular CPU benchmark, in both single core and multi-core tests, with the 3700X coming out on top in both (by as much as 28% in the multi-core and at a lower TDP). Another new CPU, the Ryzen 7 3800X, has all the same specifications as the 3700X but higher base and boost clocks of 3.9 GHz and 4.5 GHz respectively. These clock boosts also mean it has a higher 105W TDP, which is a pretty big jump from 65W for a few hundred megahertz. In addition to those two products, AMD also surprise-announced a new tier for the Ryzen family of processors: the 12-Core Ryzen 9 3900X. The Ryzen 9 3900X also has nearly double the cache of the 8-core processors and utilizes a chiplet design to achieve a 12-Core processor with 3.8 GHz base and 4.6 GHz boost at 105W TDP.
AMD showed off Cinebench performance of its new Ryzen 7 3700X against last year’s 2700X, demonstrating double-digit or better improvements across some of the most popular games (including GTA V, Overwatch, DOTA 2, PUBG, LOL, and CSGO). AMD also compared the 3700X and 3800X against Intel’s 9700K and 9900K in Cinebench, which showed the 3800X narrowly beating both processors in single thread and multi-thread performance. The company also showed the 3900X competing against Intel’s 9920X high-end desktop CPU, which is where AMD originally came after Intel with the Ryzen series. Because Intel generally utilizes an older CPU core in this category, it is ripe for AMD to attack. This really showed in single thread performance, where it beat Intel by 14% in Cinebench. AMD also beat Intel in multi-thread Cinebench performance by 6%, all while being 60W lower in TDP. AMD rounded this all out with the announcement that the Ryzen 7 3700X will be available for a paltry $329, with the Ryzen 3800X selling for $399. Meanwhile, the Ryzen 9 3900X will be available for a more substantial $499. All of these prices are significantly lower than Intel’s competitive chips, which should translate to significant price cuts and or market share gains for AMD.
AMD also showed off the previously unannounced Ryzen 9 3950X, the company’s 16-core Ryzen chip. This chip features a boost clock of up to 4.7 GHz a 3.5 GHz base frequency and a meager 105W TDP for $749. I find it interesting that AMD was somehow able to keep the TDP between the Ryzen 7 3800X, Ryzen 9 3900X, and Ryzen 9 3950X the same. This could be because the base frequencies are lower on the higher core count chips.
Introducing the X570 AM4 chipset
Not to be sold short, AMD also announced a new X570 AM4 chipset, paired with the 3rd Generation Ryzen processors, that is a strong complement to the Ryzen CPU’s core capabilities. For example, the Ryzen 3000 series supports up to 4 USB 10 Gbps ports, 4 Hi-Speed USB ports, 2 SATA 6 Gbps and 24 PCIe 4.0 lanes. The AMD X570 chipset adds up to 8 more USB 10 Gbps ports, up to 12 more SATA 6 Gbps ports, and up to 16 more PCIe 4.0 lanes. While the platform is capable of an impressive amount of connectivity, one must remember that it is ultimately up to the motherboard vendors to implement these port numbers and connectivity on the physical board for this to matter. With AMD finally getting top-tier slots with all 6 major board vendors, though, I believe there will be plenty of models that will implement the full 12 USB 10 Gbps, 4 hi-speed USB, 14 SATA 6Gbps, and 40 PCIe lanes.
“Navi” GPUs are here
Equally important to AMD’s future is the newly announced Radeon RX 5000 family of GPUs, codenamed “Navi.” These GPUs represent a complete rearchitecting of AMD’s GPU architecture, starting with the graphics cores themselves (which are now referred to as RDNA, or Radeon DNA). This architecture leverages TSMC’s 7nm process node and is the second discrete GPU architecture from AMD to utilize 7nm. The new RDNA will have a compute unit design for each core, multi-level cache hierarchy, and a streamlined graphics pipeline. These changes allow AMD to claim an improvement of 25% on performance per clock and a 50% improvement in performance per watt versus the old AMD GCN architecture.
The family will start with the Radeon RX 5700 series, which will be available in July, at the same time as the 3rd Gen Ryzen CPUs and X570 motherboards. At E3, AMD gave more details on the new Radeon RX 5700 series. This included some preliminary performance numbers that show it performing comparatively with the NVIDIA RTX 2070, which should prove interesting for the competitive landscape. AMD says that the RX 5700 will sell for $379 and perform around or above an NVIDIA RTX 2060, while the RX 5700 XT will sell for $449 and perform around the RTX 2070. AMD was very clear at E3 about where it expects these new GPUs to perform, as to not create any confusion amongst gamers. AMD is also planning a limited edition $499 version of the 5700 XT, in commemoration of the company’s 50th anniversary, that will deliver slightly higher performance numbers. What’s interesting about these new GPUs is that they will have three different clock speeds: a base clock, a game clock, and a boost clock. The most relevant clock, I believe, will be the ‘game clock’ since that’s the average clock speed that the GPU will operate at during gameplay. NVIDIA claims that this is akin to its own boost clock, clearly in an effort to stop people from comparing the two companies’ boost clocks side by side.
AMD also announced some interesting software features, geared towards gamers, that will ship with the new RX 5700 series of GPUs. AMD’s anti-lag software may prove to be the most compelling as it claims to reduce input latency from the gamer’s actions to when they are displayed on the monitor. This could prove invaluable to the eSports community, driving adoption of AMD GPUs by pro teams and gamers alike. I have yet to test the software myself, but the claims sound very promising.
AMD is also shipping a new software suite called FidelityFX, with a multitude of features for developers to leverage to improve image quality in games. Certain developers like Gearbox have already committed to utilizing FidelityFX in their games, including the much-anticipated Borderlands 3.
AMD’s approach appears to be to attack NVIDIA where it has the best chance. This makes sense, especially since NVIDIA most likely cannot cut prices as aggressively as it has in the past to counter AMD. We’re also seeing AMD challenge Intel in more than one CPU category, which may trigger Intel to introduce new products like the i9-9900KS or rumored price cuts. With AMD suggesting both strong CPU and GPU offerings this year, we could truly see the company make a comeback of impressive proportions. Gamers could potentially build an incredibly fast and affordable all-AMD gaming PC—something only the fanboys would do before. I look forward to seeing how things shake out in the coming year in this new competitive landscape.