AMD Announces Radeon RX 7900 XT & XTX Based on RDNA3 Architecture – Computer – News

AMD has introduced its new technology of video playing cards primarily based on the RDNA 3 structure. The Radeon RX 7900 XT and Radeon RX 7900 XTX are the brand new prime fashions which can be the primary gaming GPUs to make use of chiplets. The playing cards will probably be launched on December 13.

With the Radeon RX 7000 sequence video playing cards, AMD brings the RDNA 3 structure with vital enhancements over the earlier technology. The producer says that in growth it has emphasised good efficiency per watt with out sacrificing higher gaming efficiency. In accordance with AMD, the efficiency per watt is as much as 54 p.c greater at RDNA 3 than at RDNA 2.

Chiplets on a GPU

By far the most important change we see with the brand new Radeon playing cards is the transfer from a single, monolithic chip to a number of chiplets. Similar to AMD has been doing with Ryzen processors for a number of generations, the producer makes use of a number of chips and combines them into one cooperating complete. The Radeon RX 7900 XT and XTX are primarily based on the Navi 31 chip. A most of seven chiplets are mixed on this, of which the Graphics Compute Die is the most important chip and accommodates all stream processors. This single GCD is surrounded by smaller chiplets referred to as Reminiscence Cache Dies, which comprise the Infinity Cache, in addition to a 64bit reminiscence controller to the GDDR6 reminiscence.

As on the RDNA 2 video playing cards from the RX 6000 sequence, AMD once more makes use of Infinity Cache as a reminiscence buffer between the graphics chip and the GDDR6 video reminiscence, however this connection is of the second technology and due to this fact significantly sooner. AMD states that the utmost bandwidth at 5.3TB/s under the road is 170 p.c greater than with the reminiscence association on the earlier technology. The quantity of Infinity Cache has decreased from 128MB on Navi 21 to 96MB on the brand new Navi 31 chip, which, due to smarter reminiscence use, doesn’t end in a lack of efficiency, however does make a distinction in vitality consumption.

New compute unit, sooner ray tracing and AI

The compute unit in RDNA 3 is supplied with a twin difficulty SIMD unit, with which AMD makes roughly the identical adjustment as Nvidia has achieved with its Ampere structure. Any stream processor can now work concurrently on a floating level, integer or ai job. It’s nonetheless unclear how flexibly the brand new compute unit can arrange and distribute this work, however the theoretical computing energy is at the very least doubled with this step. With this adjustment AMD doesn’t specify the variety of calculation cores twice as excessive, as Nvidia has began to do with its CUDA cores. With the position of two AI accelerators per Compute Unit and the addition of latest directions, the efficiency on this space ought to be 2.7 instances greater than on RDNA 2.

Disconnected Clock Speeds

To get the efficiency per watt as excessive as doable, AMD has decoupled the clock speeds of the shaders and the entrance finish within the design of Navi 31. In accordance with the producer, excessive clock speeds of the entrance finish are necessary for efficiency, however the shaders themselves can run barely slower with out lowering efficiency. On the RX 7900 XTX it was determined to have the shaders work at a most of 2300MHz, whereas the entrance finish of this chip works at 2500MHz. The producer states that it chooses one of the best steadiness as a result of the clock speeds are successfully 15 p.c greater than on the earlier technology, whereas on the similar time saving as much as 25 p.c in vitality consumption.

Radiance Show engine

With the RX 7000 sequence, the show output additionally will get a serious replace, which AMD has even renamed the Radiance Show Engine. The brand new technology Radeons function the model new DisplayPort 2.1 normal and may output 12bit per channel. With DP 2.1, a large 54Gbps bandwidth makes it doable to drive an 8K show as much as 165Hz with out compression, or a 4K show as much as 480Hz. That is an enormous step up from DisplayPort 1.4, as used on the RX 6000 sequence in addition to on Nvidia’s new RTX 40 sequence playing cards. With DP 1.4, the restrict is for 8K at 60Hz and for 4K at 240Hz with out utilizing compression. Throughout the presentation, the producer favored to emphasise this benefit in connectivity in comparison with its competitor.

Media Engine, FSR 3 en Hypr-RX

With the RX 7000 sequence, AMD is bringing AV1 encoding help to its graphics playing cards for the primary time. The brand new Media Engine must also allow higher video encoding via the usage of AI, though the producer doesn’t but present particulars how that is achieved precisely and whether or not the AI ​​accelerators are used for this. Nevertheless, SmartAccess Video is launched, the place a Ryzen 7000 processor and Radeon RX 7000 video card mixed can carry out a heavy encoding or decoding job at an accelerated price.

Throughout the presentation, AMD briefly talked about {that a} new model of FSR is coming, with an anticipated launch someday in 2023. Below the identify FSR 3, AMD will launch a counterpart to DLSS 3 for RDNA 3 graphics playing cards, during which the AI accelerators on the gpu will probably be used to generate frames and show between rendered frames.

Within the space of ​​software program, AMD additionally plans to introduce Hypr-RX within the first half of 2023. This perform ought to make it simpler for avid gamers to decide on an optimum recreation profile within the Adrenalin software program, with out having to manually choose options akin to Enhance, FSR and SuperResolution.

Anticipated Efficiency

Within the presentation, AMD makes quite a lot of efficiency claims for the RX 7900 XTX, however the producer doesn’t make a direct comparability with Nvidia’s GeForce RTX 4090. Total, as much as 50 p.c higher ray tracing efficiency is promised than on the RX 6950 XT, and in rasterization this might be between 4K. the 50 and 70 p.c ought to be. If AMD could make this occur, will probably be a matter of recreation whether or not the RTX 4090 or the RX 7900 XTX is quicker. It’s also hanging that AMD appears to counsel in its presentation that FSR on RDNA 3 gives a higher efficiency enhance than is the case proportionally on the RX 6950 XT.

Costs and availability

The AMD Radeon RX 7900 XT and RX 7900 XTX will probably be accessible from December 13, in line with the producer in each reference variations and customized playing cards that the board companions will include. The instructed retail value of the RX 7900 XT is USD 899, the RX 7900 XTX may have a instructed retail value of USD 999. Transformed to euros, with the present trade price together with VAT, that’s at present 1110 euros for the RX 7900 XT and 1233 euros for the RX 7900 XTX.