AMD Unveils Ryzen 3000 CPUs, 12-Core At $499, IPC Parity With Intel, Navi Ahead Of RTX 2070
Dr. Su didn’t delve too deeply into 2nd Generation EPYC processors. She discussed the recent win with Oakridge National Labs to build a 1.5 Exaflop super-computer. And announced a partnership with Microsoft to put EPYC in the Azure cloud. Perhaps the most exciting bit about EPYC, however, was an update to a molecular dynamics demo that was first shown at CES.
During the new demo a 2P (dual-socket) AMD EPYC system was pitted against a 2P Intel Xeon 8280 28-core Cascade Lake system (56 cores total), and the EPYC system was more than twice as fast (9.68ns vs. 19.60ns per day). This demo was initially shown with a 1P Rome processor versus a previous-gen Xeon, but this new 2P demo is more indicative of the actual competition AMD will face in the data center when 2nd Generation EPYC processors launch next quarter.
In an exciting turn of events, Dr. Su not only made some new disclosures regarding Navi, the company’s next-generation GPU architecture, but also demoed a Navi GPU for the first time publicly. As regular readers probably already know, Navi will be the foundation of a future-generation of AMD Radeon graphics cards. The GPUs will be built on TSMC’s 7nm process technology using a brand-new compute unit design, optimized for higher clocks, higher IPC, and lower power, and focused on gaming. AMD says Navi’s compute units are based on new RDNA – or Radeon DNA – and not the older GCN designs and that the entire graphics pipeline has been optimized for improved responsiveness and smoother gaming.
The initial wave of graphics cards base don Navi will be branded the Radeon RX 5000 series, in honor of AMD’s 50th anniversary. They will be the first PCIe Gen 4 enabled GPUs, and according to AMD, they’re seeing 1.25X performance per clock and 1.5X (or higher, depending on the workload) performance per watt than Vega.
In its first public demo, a Radeon RX 5700 series GPU was put up against a GeForce RTX 2070, running Strange Brigade. The resolution used in the demo wasn’t mentioned, nor was the API (Strange Bridge can use DX12 or Vulkan), but performance showed the Radeon RX 5700 outperforming the RTX 2070, by over 15% -- 117FPS vs. 101 FPS. Dr. Su mentioned, however, than on average, the Radeon RX 5700 is about 10% faster than an RTX 2070, and that cards would be available in July.
The close the keynote, Dr. Su brought out a handful of partners – namely Microsoft, ASUS, and Acer – to lay the foundation for the big 3rd Generation Ryzen processors reveal. The rep from Microsoft talked at a high-level about the collaboration between the two companies, and ASUS and Acer revealed a few new systems based on Ryzen and Radeon. ASUS also showed off a handful of AMD X570 based motherboard and mentioned that it has 30 designs currently in the works. Once AMD’s partners were done speaking though, Dr. Su officially revealed the new 3rd Generation Ryzen processor family, more specifically the Ryzen 7 3700X, Ryzen 7 3800X, and the Ryzen 9 3900X, though in a press release a couple of Ryzen 5 SKUs were announced as well.
Here's a look at the entire line-up that was announced, along with price and availability...
All of the 3rd Generation AMD Ryzen processors are PCIe Gen 4 enabled and leverage the same socket AM4 as existing mainstream Ryzen processors. 3rd Gen Ryzen processors, however, are based on the next-generation Zen 2 microarchitecture, manufactured at 7nm, and have double the cache, double the floating point performance, and significantly better IPC – Dr. Su announced a 15% IPC uplift with 3rd Gen Ryzen, which essentially brings AMD up to parity with Intel’s latest Core architecture. 3rd Gen Ryzen processors are PCIe Gen 4 enabled as well and will hit store shelves July 7.
The $329 AMD Ryzen 7 3700X is an 8-core / 16-thread processors, with 36MB of total cache. It has a base clock of 3.6GHz, a boost clock of 4.4GHz, and only a 65W TDP. In a side-by-side comparison with an Intel Core i9-9700K – which is only an 8-core / 8-thread processor -- running Cinebench R20, the Ryzen 7 3700X was about 30% faster, due to its ability to handle additional threads.
The Ryzen 7 3800X was also shown off as part of what AMD called the World’s First PCIe Gen 4 gaming system. The CPU was paired with a Radeon RX 5700 series GPU and an X570-based motherboard, and it ran an upcoming PCI Express feature test that will be incorporated into 3DMark. During the test, the AMD system was pitted again an Intel-based rig with a GeForce RTX 2080 Ti (which feature PCIe Gen 3), and the AMD PCIe Gen 4 system offers about 69% better performance – 25FPS vs. 14FPS. It is worth noting, that this feature test is specifically designed to heavily tax PCIe bandwidth, and isn’t necessarily indicative of any real-world gaming scenario. All that bandwidth could really do fun things for storage, however.
Finally, in a “one more thing” moment, Dr. Su announced the Ryzen 9 family and showed off the $499 Ryzen 9 3900X. The Ryzen 9 3900X features two Zen 2 core-chiplets, with an IO die (Ryzen 7s feature a single core chiplet), with 12 total cores enabled (24-threads). The processor has a base clock of 3.8GHz, a boost clock of 4.6GHz, and a whopping 70MB of cache – all in a 105W power envelope.
To show off its capabilities, the Ryzen 9 3900X was put up against a 165W Intel HEDT Core i9-9920X processor (also 12-cores / 24-threads), running Blender. When all was said and done, the AMD system churned through the workload in 32 seconds, versus the Intel system’s 38 seconds, which was about an 18% advantage.
All told, Dr. Su’s keynote was impressive. AMD will be putting even more pressure on Intel in the CPU space very shortly, not only in terms of technology and performance, but price pressure as well. All of the 3rd Gen Ryzen processors undercut their Intel-built counterparts in terms of pricing by a wide margin. Seeing Navi in action, outperforming a GeForce RTX 2070, was also a positive development. We look forward to learning more about Navi at E3 in the coming weeks.
