Today, AMD unveiled the first 7nm gaming graphics card at the company's CES 2019 keynote. Available February 7th for $699, the AMD Radeon VII promises big performance improvements over the Radeon RX Vega 64, the company's current flagship gaming GPU. The newly announced graphics card's job also includes competing against Nvidia's GeForce RTX 2080 in the high-end segment.
With twice the memory and memory bandwidth, the AMD Radeon VII claims to be up to 29 percent and 39 percent faster in gaming and demanding content creation scenarios, respectively, when compared to the Radeon RX Vega 64.
AMD Radeon VII Specifications
| AMD Radeon VII | AMD Radeon RX Vega 64 | |
|---|---|---|
| Architecture (GPU) | Graphics Core Next 5.0 (Vega 20) | Graphics Core Next 5.0 (Vega 10) |
| Shaders | 3840 | 4096 |
| Peak FP32 Compute | 13.8 TFLOPS | 12.5 TFLOPS |
| Texture Units | 240 | 256 |
| Base Clock Rate | 1450 MHz | 1200 MHz |
| GPU Boost Rate | 1800 MHz | 1536 MHz |
| Memory Capacity | 16GB HBM2 | 8GB HBM2 |
| Memory Clock | 2000 Mbps* | 1890 Mbps |
| Memory Bus | 4096-bit | 2048-bit |
| Memory Bandwidth | 1024 GB/s | 483.8 GB/s |
| ROPs | 64 | 64 |
| TDP | 300W* | 295W |
| Transistor Count | 13.2 billion | 12.5 billion |
| Die Size | 331 mm² | 487 mm² |
*=unconfirmed
The AMD Radeon VII is based on AMD's second-generation Vega graphics card architecture and manufactured with TSMC's 7nm FinFET process. On paper, the Radeon VII resembles a lot like the Radeon Instinct MI50, which was one of the first graphics card to come out of the 7nm baking oven. Like the Radeon Instinct MI50, the Radeon VII sports 60 compute units resulting in 3,840 stream processors. However, the Radeon VII boasts a 1,450MHz base clock and an impressive 1.8GHz boost clock.
AMD has equipped the Radeon VII with nothing less than 16GB of high-performance HBM2 memory (second-generation High-Bandwidth Memory) across a 4096-bit memory interface. AMD stated that the Radeon VII is capable of pumping out 1 TB/s of memory bandwidth, which is twice of the Radeon RX Vega 64. The memory is probably clocked around 2,000MHz to achieve that throughput.
The Radeon VII shown by Dr. Lisa Su features a dual-slot design and triple-fan cooling solution. It draws power from a pair of 8-pin PCIe connectors, so the graphics card could possibly feature a 295W TDP (thermal design power) rating. In regards to video ouputs, the Radeon VII is equipped with three DisplayPort outputs and one HDMI port.
According to AMD's own benchmarks, the Radeon VII excels in real-time 3D and compute applications and gaming. AMD's is touting improvements up to 27 percent in Blender and DaVinci Resolve 15, 29 percent in Adobe Premiere, and 62 percent in LuxMark OpenCL in comparison to the aging Radeon RX Vega 64 graphics card. In terms of gaming, the Radeon VII is 35 percent faster in Battlefield V, 42 percent in Strange Brigade, and 25 percent in Fortnite.
With the purchase of an AMD Radeon VII graphics card or a pre-built system powered by it, customers receive AMD's latest 'Raise the Game Fully Loaded' game bundle that consists of Resident Evil 2, Devil May Cry 5, and Tom Clancy's The Division.
When it launches February 7th, the Radeon VII will be available directly from AMD as a first-party card, something team red hasn't done before. The graphics card is going to cost around $699. As is the custom, several of AMD's add-in-board (AIB) partners will also offer the Radeon VII graphics card.
In any case, given the FE RTX 2080 is $799, the $100 price undercut works. Board partners have a few models of the RTX 2080 (as of today) in the $670-700 range. It's not clear to me if board partners are going to charge less than $699 for the Vega VII, but hopefully so.
Curious as well to see if there are soon to be lower models (less than 16GB memory, slower clock speeds, or fewer ROPs/Shaders/Texture Units)... gotta do something with the ones that aren't 100% flawless, I imagine.
It doesn't make me jump for joy, but it's not bad, either, unless the power consumption is at Vega 64 levels. THAT, to me, would be disappointing.
Still... I wonder if Nvidia knew something ahead of time, prompting them to start supporting Adaptive Refresh in their drivers for non-GSync monitors with the upcoming Jan 15 drivers.