AMD is gearing up to launch its next-generation Radeon RX 7000 graphics cards next month, and today the company shared more details on pricing, performance levels and the new RDNA 3 GPU architecture that will power all of its graphics cards over the next couple of years, reports Ars Technica.
The update of the video card line begins with the top solutions: Radeon RX 7900 XTX and RX 7900 XT. AMD will release both models on December 13, with the 7900 XTX starting at $999 and the XT starting at $899 (cards made by AMD’s partners will probably push those prices up a bit). Both price tags provide serious competition for the NVIDIA’s RTX 4000 series, which starts at $1,599 for the top-of-the-line GeForce RTX 4090 and $1,199 for the RTX 4080.
NVIDIA had planned to launch the 12GB RTX 4080 in this price range, but consumer backlash over its potentially confusing branding and lower performance levels forced NVIDIA to change plans. The company has canceled the release and still does not say when this video card can be released, what it will be called or how much it will cost. The 16GB version of the 4080 will hit the market on November 16, about a month ahead of AMD’s new GPUs.
During the presentation, AMD did not compare its new video cards with models from NVIDIA, focusing only on the increase in speed compared to the previous generation. Thus, according to the company, the RX 7900 XTX is 1.5-1.7 times faster than the 6950 XT when playing in 4K.
In terms of raw computing performance, AMD claims the 7900 XTX can deliver up to 61 teraflops, while NVIDIA claims the RTX 4090 delivers up to 83 teraflops and the RTX 4080 up to 49 teraflops. That’s an extremely high performance figure that says little about how the graphics card will perform in real-world games (the RTX 4090 isn’t 70% faster than the RTX 4080, according to NVIDIA’s own benchmarks). But along with the pricing and the lack of direct 4090 comparisons, this vague number suggests that AMD is targeting the GeForce 4080 level with the new cards, rather than trying to outdo the RTX 4090. This is the same relative position that AMD has been in for generations.
As expected, the RDNA 3 GPUs take a lot from Ryzen, using a group of chips rather than one big slab of silicon like NVIDIA does. By making a GPU out of many smaller, interconnected silicon wafers, AMD can potentially lower the cost and improve the performance of its graphics cards by ensuring that less silicon is thrown away or sent to the landfill due to small imperfections or defects.
Unlike Ryzen, where the actual CPU cores are split between multiple chipsets, AMD still packs all of the GPU’s actual computing power into a single silicon die called the Graphics Compute Die, or GCD. Manufactured on TSMC’s 5nm process, the GCD contains all the actual RDNA 3 computing units, as well as the display engine (to control video output) and the multimedia data encoding and decoding engine.
The other chips in each GPU are smaller cache memory chips, or MCDs; each containing a 64-bit memory controller plus Infinity Cache. The 7900 XTX has six such MCDs, while the 7900 XT has five (this can be said because the XTX has a 384-bit memory bus, while the XT has a 320-bit bus). AMD manufactures the MCD chips on a 6nm process, just like the I/O chips in the Ryzen 7000 processors.
In addition to frame rate improvements, RDNA 3 focuses on improving the ray tracing performance of RDNA 2. Compared to GPUs and NVIDIA RTX 3000 series and Intel Arc GPUs, AMD RDNA 2 ray tracing accelerators processed fewer steps of the ray tracing pipeline – in particular, they unloaded the so-called binary tree traversing on general-purpose stream processors, while NVIDIA and Intel ray tracing units handled it independently. The RDNA 3 raytracing accelerators handle this workaround themselves, which, along with other improvements, should deliver “up to 50% more raytracing performance per CPU,” according to AMD.
The GPU’s multimedia encoding capabilities are also getting a boost. Most importantly, they will catch up to the RTX 4000 series and Intel Arc cards by supporting hardware acceleration of encoding and decoding of AV1 video streams (up to 8K at 60 frames per second, similar to what NVIDIA and Intel offer). RDNA 2 GPUs support only hardware-accelerated AV1 video decoding. AMD also noted that the new GPUs support the DisplayPort 2.1 specification, which will enable higher refresh rates on DisplayPort 2.1-enabled monitors as they become more available.
While AMD did not go into performance comparisons in its announcement, the company did not shy away from the size and power requirements of NVIDIA’s RTX 4090, showing that the AMD 7900 XTX reference design can easily replace the RX 6950 XT without requiring an upgraded PSU, adapter for cable or larger housing. The graphics card’s peak power consumption has increased from 330W to 355W, but that’s about 100W lower than the RTX 4090’s peak power consumption and still allows for a pair of 8-pin PCIe power connectors.