Why you can trust Tom's Hardware Our expert reviewers spend hours testing and comparing products and services so you can choose the best for you. Find out more about how we test.
SoC & Chipset Connectivity
As we've discussed, Raven Ridge gets eight PCIe 3.0 lanes for add-in graphics, rather than Zeppelin's 16, along with four general-purpose lanes. This isn't a deal-breaker, though. Modern graphics cards (even the high-end ones) don't fully utilize wide PCIe links. Moreover, these processors include capable on-die graphics.
The processor also has its own USB and SATA controllers, which complement the I/O you get from an X370, B350, or A320 chipset.
| Row 0 - Cell 0 | USB 3.2 Gen2 | USB 3.1 Gen1 | USB 2.0 | PCIe Gen3 (Gfx) | PCIe Gen2 (General Use) | SATA | SATA Express |
| Raven Ridge | 4 | 1 | 1 | 8 Lanes | 4 lanes | 2 | - |
| X370 Chipset Provides | 2 | 6 | 6 | - | 8 | 4 | 2 (or 4 more SATA) |
| B350 Chipset Provides | 2 | 2 | 6 | - | 6 | 2 | 2 (or 4 more SATA) |
| A320 Chipset Provides | 1 | 2 | 6 | - | - | 2 | - |
Display
Raven Ridge supports FreeSync with supported displays and motherboards. It also supports HDCP 1.4/2.2 for streaming 4K+HDR content. AMD plans to have a production-class PlayReady 3.0 DRM graphics driver in Q3, which you'll need to stream 4K content from Netflix. Wireless display via Miracast is also supported.
The processors sport a wide range of hardware-accelerated video encode and decode features that execute on Vega's Video Core Next (VCN) silicon. Naturally, the most important comparison is to Intel's UHD Graphics 630, which has a broader range of hardware-accelerated video encode capabilities, such as MPEG-2, VP8, and VP9 8-bit. AMD does support VP9 10-bit decode, which Intel has yet to offer.
The Vega Graphics Engine
Chris Angelini covered the Vega architecture in-depth in our AMD Radeon Vega RX 64 8GB Review, so check that story out for more detail on Vega itself.
Ryzen 5 2400G features a Vega-based graphics engine with 11 Compute Units, while the lower-end Ryzen 3 2200G includes eight CUs. The 2400G wields 44 texture units (four per CU), 704 Stream processors, and 16 ROPs. That's an impressive list of resources crammed next to a quad-core CPU. But it pales in comparison to Radeon RX Vega 64's 4096 Stream processors.
AMD uses the same Raven Ridge die for its mobile and desktop products. As such, Ryzen 5 2400G looks a lot like the Ryzen Mobile 7 2700U, though the 2400G features an extra CU. It also has a lower maximum graphics frequency of 1250 MHz compared to the 7200U's 1300 MHz.
Ryzen 3 2200G and the Ryzen Mobile 5 2500U both have eight CUs, and they share the same 1100 MHz peak graphics clock rate.
Of course, comparisons to Intel's Kaby Lake-G family are inevitable. Those Intel models come with two flavors of Radeon RX Vega graphics: 100W processors featuring "GH" graphics and 65W models with "GL" graphics.
The GH implementation sports 24 CUs and 1536 Stream processors. It features a base clock of 1063 MHz that accelerates up to 1190 MHz, plus 4GB of HBM2 (4-hi stack) directly attached via Intel's EMIB technology. Single-precision performance tops out at 3.7 TFLOPS, compared to the 2400G's 1.76 TFLOPS.
Taking a step down, the GL graphics engine features 20 CUs. Lower base/boost frequencies of 931 and 1011 MHz, respectively, further differentiate the two configurations. Intel does maintain 4GB of HBM2. But peak compute performance falls to 2.6 TFLOPS compared to the 2200G's 1.126 TFLOPS.
Aside from the brawnier allocation of CUs on Intel's Kaby Lake-G models, they also profit from HBM2 and its massive throughput advantage. Raven Ridge is fed by much slower DDR4 system memory. While overclocking is going to help augment AMD's stock graphics performance, Intel is going to enjoy the leg up in frame rate comparisons.
As an aside, AMD announced the Radeon Vega Mobile at this year's CES. It features HBM2 and the same 1.7mm Z-height as Intel's Kaby Lake-G processors. With Kaby Lake-G going into Intel's NUC form factor, there's a chance we could see AMD take a similar path to the desktop. That would give those CUs a lot more bandwidth to work with, if the company could incorporate the solution into a high-end processor. This also raises questions of whether Intel would make EMIB available to AMD.
MORE: Best CPUs
MORE: Intel & AMD Processor Hierarchy
MORE: All CPUs Content
