AMD is gearing up to launch its next-generation Radeon RX 7000-series GPUs next month, and today the company shared more details on card pricing, performance tiers, and the new RDNA 3 GPU architecture that will power all of its graphics cards. for the next couple of years.
The launch starts at the high end, with the Radeon RX 7900 XTX and RX 7900 XT. AMD will release both of these GPUs on December 13, with the 7900 XTX starting at $999 and the XT starting at $899 (cards made by AMD partners will no doubt push those prices up a bit). Both of those price tags are lower than Nvidia’s RTX 4000 series, which starts at $1,599 for the top-tier GeForce RTX 4090 and $1,199 for the RTX 4080.
| CPU | Launch of MSRP | KS | game clock | RAM/bus width | TVP |
|---|---|---|---|---|---|
| Radeon RH 7900 HTX | $999 | 96 (RDNA 3) | 2300 MHz | 24 GB GDDR6 (384-bit) | 355 W. |
| Radeon RH 6950 HT | $1,099 | 80 (RDNA 2) | 2100 MHz | 16GB GDDR6 (256bit) | 335 W. |
| Radeon RH 7900 HT | $899 | 84 (RDNA 3) | 2000 MHz | 20GB GDDR6 (320bit) | 300 W. |
| Radeon RH 6800 HT | $649 | 72 (RDNA 2) | 2015 MHz | 16GB GDDR6 (256bit) | 300 W. |
Nvidia planned to release a 12GB version of the RTX 4080 in this price range, but consumer backlash against its potentially confusing branding and lower performance tier prompted Nvidia to “launch”the card without saying when it would be reintroduced, as it would be called. or how much it will cost. The 16GB version of the 4080 will go on sale on November 16th, beating new AMD GPUs on the market by about a month.
When it came to comparing their cards to Nvidia’s, AMD played modestly during the presentation, carefully comparing the RX 7900 XTX series to the RX 6950 XT without mentioning either the RTX 3000 series or the RTX 4000 series. The 7900 XTX is said to be “up to 1. 7x faster than the 6950 XT in 4K games – 1.5x improvement was more common among games tested by AMD.
In terms of raw compute performance, AMD says the 7900 XTX is capable of up to 61 teraflops, while Nvidia says the RTX 4090 is capable of 83 teraflops and the RTX 4080 is 49 teraflops. That’s a very high performance figure that doesn’t tell us much about how the cards will compare in real-world games (according to Nvidia’s own tests, the RTX 4090 isn’t 70 percent faster than the RTX 4080). But along with the price and the lack of direct comparisons of the 4090, this vague number really suggests that AMD is aiming at the level of the Nvidia 4080 with these cards, rather than trying to outdo the RTX 4090. It’s the same relative position as AMD. for several generations.
A bag of chips
As expected, RDNA 3 GPUs will take a page from the Ryzen book by using a bunch of chiplets rather than one big wafer of silicon like Nvidia uses. By building GPUs from many interconnected smaller silicon dies, AMD can potentially lower the cost and improve the performance of its GPUs by ensuring that less silicon is thrown away or scrapped due to small flaws or defects.
Unlike Ryzen, where the actual CPU cores are split across multiple chiplets, AMD still puts all of the GPU’s actual processing power into a single piece of silicon called a graphics compute die, or GCD. Manufactured using TSMC’s 5nm process, the GCD contains all of the actual RDNA 3 compute blocks, as well as a display engine (to control video output) and a media encoding and decoding engine.
The other chiplets in each GPU are smaller cache dies or MCDs; each contains a 64-bit memory controller plus Infinity Cache memory. The 7900 XTX has six of these MCDs, while the 7900 XT has five (you can tell because the XTX has a 384-bit memory bus, while the XT has a 320-bit memory bus). AMD is building the MCD on a 6nm process, similar to the I/O die in the Ryzen 7000 processors.
In addition to boosting frame rates, RDNA 3 focuses on improving the mediocre performance of RDNA 2 ray tracing. Compared to Nvidia’s RTX 3000 series GPUs and Intel’s Arc GPUs, AMD’s RDNA 2 ray tracing accelerators handled fewer ray tracing pipeline steps—in particular, they offloaded what is called “tree traversal”to general purpose stream processors, while Nvidia accelerators and Intel ray tracing units handled the traversal themselves. According to AMD, RDNA 3 ray tracing accelerators handle this bypass, which, along with other improvements, should be useful for “up to 50% performance improvement [of ray tracing] per CU.”
Media graphics encoding capabilities are also getting a boost. In particular, they will catch up with the RTX 4000 series and Intel Arc cards by supporting hardware accelerated encoding and decoding of AV1 video streams (up to 8K at 60fps, similar to what Nvidia and Intel offer). RDNA 2 GPUs only support 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 widely available.
While there was little performance comparison in its announcement, AMD highlighted the size and power requirements of the Nvidia RTX 4090, demonstrating that the AMD 7900 XTX reference design can easily replace the RX 6950 XT without requiring an upgraded PSU, power cable adapter, or a large body. Peak power rises from 330W to 355W – still an increase but about 100W lower than the RTX 4090’s peak power and can still be powered by a pair of 8-pin PCIe power connectors.
On the software side, AMD has teased a next-gen version of its FidelityFX Super Resolution (FSR) upscaling technology that it says will “come out in 2023″and provide “2x the FPS compared to FSR 2″at 4K resolution. FSR 2 helped close the gap in visual quality between the first version of FSR and Nvidia’s similar DLSS scaling technology, as well as running on more GPUs, including older Radeon GPUs and many Intel and Nvidia GPUs. We don’t yet know if FSR 3 will also be GPU dependent or if AMD will limit it to the latest generation of Radeon GPUs like Nvidia is doing with DLSS 3 and the RTX 4000 series.
We’ll take a closer look at the AMD Radeon RX 7000-series cards when they’re a little closer to launch.