Now featuring the advanced AMD chiplet design, AMD has finally brought its RDNA 3 architecture to its professional workstation graphics cards, the Radeon PRO W7800 and W7900. The chiplet design provides higher performance and greater efficiency than the previous generation, thanks to its inclusion of 5nm Graphics Compute Die (GCD) and up to six 6nm Memory Cache Die (MCD). Add features from RDNA 3 architecture, such as dedicated AI acceleration, second-generation raytracing, AV1 encode/decode, and more, and the AMD Radeon PRO W7000 Series workstation graphics cards provide the performance and features necessary to handle demanding creative, production, and visualization workloads quickly and efficiently.
Let’s look at the flagship AMD Radeon PRO W7900 first. The W7900 goes all in when it comes to power and performance. The W7900 is rated for peak single precision (FP32) throughput of 61 TFLOPS with 48GB of GDDR6 VRAM, which is double the memory of AMD’s consumer flagship Radeon RX 7900 XTX. More memory means the trade-off is slightly lower clock speeds, at 18 Gb/s, for 864 Gb/s of memory bandwidth. The W7900 features soft ECC support for the VRAM, which allows the GPU to correct memory errors. AMD states that the W7900 delivers 1.5x the performance and memory of its closest predecessor, the AMD Radeon PRO W6800.
While AMD’s workstation graphics cards typically have dual-slot form factors, due to the power and performance of the W7900, AMD has opted for a triple-slot design. While triple-slot GPUs are nothing new to consumers, such as gamers, it’s a notable change for the workstation market. The W7900 has a TDP of 295W, which means AMD had to increase the size of the card to provide efficient cooling. Fortunately, the W7900 is only wider, so there shouldn’t be any major compatibility issues. However, it does mean that workstation owners will need to make sure they have enough room compared to previous generations.
The AMD Radeon PRO W7800 is more a traditional workstation graphics card compared to the W7900. With a conventional dual-slot design and a 260W TDP, the W7800 makes a compelling replacement to the previous AMD Radeon PRO GPU. The W7800 features 32GB of GDDR6 VRAM and 576 GB/s of memory bandwidth, as well as soft ECC support. Rated at 45 TFLOPs of peak (FP32) throughput, the W7800 still offers powerful performance for heavy workflows.
Both GPUs will feature DisplayPort 2.1 outputs, now with UHBR 20 support. UHBR 20 provides 20 Gb/s of raw bandwidth per lane, for a total of 80 Gb/s (77.4 Gb/s without overhead). Besides being the only workstation graphics cards currently on the market to feature UHBR 20 support, the inclusion is also important because UHBR 20 speeds are required for fully utilizing an 8K display without any compromises. That means both the W7900 and W7800 GPUs can push 7680 x 4320 at 60 Hz with 10-bit color and full (4:4:4) chroma subsampling. For context, the DisplayPort 2.1 UHBR 13.5 outputs found on AMD’s consumer RDNA 3-based GPUs won’t be able to do that without some kind of trade-off. Both GPUs can even go as far as supporting a single 12K display @ 120 Hz with DSC. It is important to note that while each DisplayPort 2.1 output now provides more bandwidth, there are fewer physical outputs. Both GPUs feature three DisplayPort 2.1 outputs and a single Mini DisplayPort 2.1 output. This means each card can only directly output to four displays, as opposed to six displays from previous AMD Radeon PRO GPUs.
AMD works with leading professional software applications on a comprehensive certification program. This ensures that AMD Radeon PRO graphics cards are built for demanding 24/7 environments and tested to meet high standards to deliver performance and stability. Be sure to check the list of certified applications on AMD’s official website for more details.
The AMD Radeon PRO W7900 and W7800 professional workstation graphics cards are expected to be available in Q2 of 2023. Product availability in OEM systems is expected in the second half of 2023.
What does AMD's new workstation graphics cards mean for your workflow? Let us know in the Comments section, below.
0 Comments