AMD is gearing as much as launch its next-generation Radeon RX 7000-series GPUs subsequent month, and right this moment the corporate shared extra particulars in regards to the playing cards’ pricing, efficiency ranges, and the brand new RDNA 3 GPU structure that may energy all of its graphics playing cards for the following couple of years.
The launch begins on the excessive finish, with the Radeon RX 7900 XTX and RX 7900 XT. AMD will launch each of those GPUs on December 13, with the 7900 XTX beginning at $999 and the XT beginning at $899 (playing cards made by AMD’s companions will certainly push these costs upward a bit). Each of those value tags undercut Nvidia’s RTX 4000 collection, which begins at $1,599 for the top-tier GeForce RTX 4090 and $1,199 for the RTX 4080.
|CPU||Launch MSRP||CUs||Sport clock||RAM/bus width||TBP|
|Radeon RX 7900 XTX||$999||96 (RDNA 3)||2300 MHz||24GB GDDR6 (384-bit)||355 W|
|Radeon RX 6950 XT||$1,099||80 (RDNA 2)||2100 MHz||16GB GDDR6 (256-bit)||335 W|
|Radeon RX 7900 XT||$899||84 (RDNA 3)||2000 MHz||20GB GDDR6 (320-bit)||300 W|
|Radeon RX 6800 XT||$649||72 (RDNA 2)||2015 MHz||16GB GDDR6 (256-bit)||300 W|
Nvidia had deliberate to discipline a 12GB model of the RTX 4080 on this value vary, however a shopper backlash to its doubtlessly complicated branding and decrease efficiency stage prompted Nvidia to “unlaunch” the cardboard, with no phrase on when it will likely be re-introduced, what it will likely be referred to as, or what it’ll value. The 16GB model of the 4080 launches November 16, beating AMD’s new GPUs to market by round a month.
When it got here to evaluating its playing cards towards Nvidia’s, AMD performed coy throughout its presentation, rigorously evaluating the RX 7900 XTX collection to the RX 6950 XT with out mentioning both the RTX 3000 collection or the RTX 4000 collection. The 7900 XTX is claimed to be “as much as 1.7x” quicker than the 6950 XT whereas enjoying video games in 4K—a 1.5x enchancment was extra frequent among the many video games AMD examined.
By way of uncooked compute efficiency, AMD says the 7900 XTX is able to as much as 61 teraflops of efficiency, whereas Nvidia says the RTX 4090 is able to 83 teraflops and the RTX 4080 of 49 teraflops. It is a tremendous high-level efficiency quantity that does not inform us loads about how the playing cards will examine in actual video games (the RTX 4090 will not be 70 % quicker than the RTX 4080, in response to Nvidia’s personal benchmarks). However together with the pricing and lack of head-to-head 4090 comparisons, this obscure quantity does recommend that AMD is aiming at Nvidia’s 4080 tier with these playing cards somewhat than attempting to outperform the RTX 4090. This is identical relative place that AMD has been in for just a few generations now.
A bag of chiplets
As anticipated, the RDNA 3 GPUs will take a web page out of Ryzen’s e-book, utilizing a gaggle of chiplets somewhat than a single giant slab of silicon like Nvidia is utilizing. By making a GPU out of many smaller interconnected silicon dies, AMD can doubtlessly cut back the fee and enhance the yields of its GPUs by guaranteeing that much less silicon is thrown out or binned due to small imperfections or defects.
Not like in Ryzen, the place the precise CPU cores are cut up amongst a number of chiplets, AMD continues to be placing the entire GPU’s precise processing energy right into a single piece of silicon referred to as the Graphics Compute Die, or GCD. Manufactured utilizing TSMC’s 5 nm course of, the GCD incorporates the entire precise RDNA 3 compute items, plus the show engine (for controlling video output) and the media encoding and decoding engine.
The opposite chiplets in every GPU are smaller Reminiscence Cache Dies, or MCDs; these every comprise a 64-bit reminiscence controller plus Infinity Cache reminiscence. The 7900 XTX has six of those MCDs lit up, whereas the 7900 XT has 5 of them (you possibly can inform as a result of the XTX has a 384-bit reminiscence bus, whereas the XT has a 320-bit bus). AMD is constructing the MCDs on a 6 nm manufacturing course of, just like the I/O die within the Ryzen 7000 CPUs.
Other than bettering framerates, RDNA 3 focuses on growing RDNA 2’s mediocre ray-tracing efficiency. In comparison with Nvidia’s RTX 3000-series and Intel’s Arc GPUs, AMD’s RDNA 2 ray-tracing accelerators dealt with fewer steps of the ray-tracing pipeline—particularly, they offloaded one thing referred to as “tree traversal” to the general-purpose stream processors, whereas Nvidia’s and Intel’s ray-tracing items dealt with traversal themselves. RDNA 3’s ray-tracing accelerators do deal with this traversal, which, together with different enhancements, ought to be good for “as much as 50% extra [ray-tracing] efficiency per CU,” in response to AMD.
The GPU’s media encoding capabilities are additionally getting a lift. Most notably, they will catch as much as the RTX 4000-series and Intel’s Arc playing cards by supporting hardware-accelerated encoding and decoding of AV1 video streams (as much as 8K at 60 FPS, just like what Nvidia and Intel supply). RDNA 2 GPUs solely help hardware-accelerated decoding of AV1 movies. And AMD famous that the brand new GPUs help the DisplayPort 2.1 spec, which is able to allow greater refresh charges on DisplayPort 2.1-capable displays once they’re extra extensively out there.
Though its announcement was gentle on efficiency comparisons, AMD took potshots on the sheer dimension and energy necessities of Nvidia’s RTX 4090, demonstrating that AMD’s 7900 XTX reference design may simply substitute an RX 6950 XT with out requiring an upgraded energy provide, an influence cable adapter, or a bigger case. Peak energy consumption goes up from 330 W to 355 W—nonetheless a rise however round 100 W decrease than the RTX 4090’s peak energy consumption and nonetheless able to being powered by a pair of 8-pin PCIe energy connectors.
On the software program aspect, AMD teased a next-generation model of its FidelityFX Tremendous Decision (FSR) upscaling know-how, which it says is “coming in 2023” and can ship “as much as 2x extra FPS vs FSR 2” at 4K resolutions. FSR 2 helped to shut the visible high quality hole between the primary model of FSR and Nvidia’s comparable DLSS upscaling tech whereas additionally operating on extra GPUs, together with older Radeon GPUs and lots of Intel and Nvidia GPUs. We do not but know whether or not FSR 3 can even be GPU-agnostic or if AMD will prohibit it to the latest-generation Radeon GPUs as Nvidia does with DLSS 3 and the RTX 4000-series.
We’ll be taking a better have a look at AMD’s Radeon RX 7000-series playing cards once they’re a bit nearer to launch.