In the present day, AMD takes the wraps off its upcoming Radeon RX 7900 XTX and Radeon RX 7900 XT graphics playing cards. We had a few of the particulars with the November 3 announcement, however AMD wished to carry a number of extra items of knowledge till at the moment — simply forward of the Nvidia RTX 4080 launch. These new AMD GPUs are set to compete towards the finest graphics playing cards, and we anticipate each will land close to the highest of the GPU benchmarks hierarchy come December 13.
We even have a deep dive into the AMD RDNA 3 structure, a have a look at AMD’s Gaming and ISV relations, and a few particulars on the Software program and Platform ecosystem. However for the 7900 sequence, let’s begin with the uncooked specs.
| Graphics Card | RX 7900 XTX | RX 7900 XT | RX 6950 XT |
|---|---|---|---|
| Structure | Navi 31 | Navi 31 | Navi 21 |
| Course of Know-how | TSMC N5 + N6 | TSMC N5 + N6 | TSMC N7 |
| Transistors (Billion) | 58 (45.7 + 6x 2.05) | 56 (45.7 + 5x 2.05) | 26.8 |
| Die measurement (mm^2) | 300 + 222 | 300 + 185 | 519 |
| CUs / SMs | 96 | 84 | 80 |
| SPs / Cores (Shaders) | 6144 (12288) | 5376 (10752) | 5120 |
| Tensor / Matrix Cores | ? | ? | ? |
| Ray Tracing “Cores” | 96 | 84 | 80 |
| Enhance Clock (MHz) | 2500 | 2400 | 2310 |
| VRAM Pace (Gbps) | 20 | 20 | 18 |
| VRAM (GB) | 24 | 20 | 16 |
| VRAM Bus Width | 384 | 320 | 256 |
| L2 / Infinity Cache | 96 | 80 | 128 |
| ROPs | 192 | 192 | 128 |
| TMUs | 384 | 336 | 320 |
| TFLOPS FP32 (Enhance) | 61.4 | 51.6 | 23.7 |
| TFLOPS FP16 (FP8) | 123 | 103 | 47.4 |
| Bandwidth (GBps) | 960 | 800 | 576 |
| TDP (watts) | 355 | 300 | 335 |
| Dimensions | 287x123x49mm | 276x113x49mm | 267x118x49mm |
| Launch Date | Dec 13, 2022 | Dec 13, 2022 | Might 2022 |
| Launch Worth | $999 | $899 | $1,099 |
There are quite a lot of modifications in comparison with RDNA 2, however basically the RX 7900 XT comes packing 20% extra Compute Models (CUs), with double the variety of FP32 (32-bit floating level) computational performance per CU. AMD nonetheless makes use of 64 Streaming Processors (SPs) per CU, although, in quite a lot of methods, it is extra like double that determine.
The die sizes are comparatively similar to the earlier era Navi 21, besides now there is a single GCD chiplet (Graphics Compute Die) and as much as six MCD chiplets (Reminiscence Cache Die). The XTX has the absolutely enabled GCD and 6 MCDs, whereas the XT has {a partially} enabled GCD with 84 CUs and pairs that with 5 MCDs — the sixth MCD slot will nonetheless be occupied, both with a dummy piece of silicon or a non-functional MCD (non-functional as a result of it is fused off, although it might or is probably not faulty).
Transistor counts have shot up so much relative to the earlier era RX 6000-series, with 58 billion whole transistors in comparison with simply 26.8 billion on Navi 21. Due to the doubled FP32 functionality, theoretical compute has jumped by as much as 159% in comparison with the RX 6950 XT. with 61.4 teraflops on the XTX card and 51.6 teraflops on the XT mannequin.
Picture 1 of 5
The core design of the AMD reference playing cards appears to be like much like what we noticed with the RX 6900 XT and 6800 XT. You get three followers and a 2.5-slot card. Beneath the hood, nonetheless, there are many updates. AMD says it targeted its personal designs on delivering increased efficiency per watt and effectivity — that is in distinction to the RTX 4090, which appears to have pushed nicely past the candy spot in pursuit of upper potential efficiency.
AMD’s 7900 XTX mannequin does have a barely increased 355W energy restrict than the 6950 XT’s 335W, whereas the 7900 XT drops right down to the identical 300W TBP because the 6900 XT and 6800 XT fashions. Word that the above slides present the 7900 XTX, and the XT variant could have a barely completely different PCB and cooler. Each will use twin 8-pin energy connectors, and AMD was eager to level out that customers will not must take care of any adapters — melting or in any other case.
AMD will use a 14-layer PCB on the XTX card, with a 20-stage energy supply subsystem. The XTX and XT playing cards can even have two DisplayPort 2.1 outputs, a single HDMI 2.1, and a single USB Kind-C connector that additionally helps DP2.1.
On the 7900 XTX, AMD will use a ten% bigger vapor chamber (in comparison with the 6900 XT reference design), and it appears to be like just like the shroud and different supplies might be barely increased high quality than earlier than. We had some hands-on time with the 2 playing cards, and each felt stable and good.
One fascinating change with the XTX card would be the addition of a fan consumption temperature sensor situated beneath the fan furthest from the IO ports. It isn’t clear if that is only for informational functions or if the consumption temperature might be used to assist modulate fan speeds, however different software program also needs to be capable to use the sensor if desired. Each playing cards even have RGB lighting.
Radeon RX 7900 XTX/XT Efficiency Preview
Picture 1 of 2
AMD supplied the above preview of the efficiency provided by the RX 7900 XTX and XT playing cards in comparison with the present RX 6950 XT. In rasterization efficiency, AMD exhibits 53% increased efficiency in Resident Evil Village, 51% extra efficiency in Name of Obligation, 67% features in Cyberpunk 2077, and 47% increased framerates in Watch Canine Legion.
Ray tracing efficiency probably sees even increased features, although he have each FSR and native efficiency to contemplate. Right here Resident Evil with most RT settings ran 44% sooner at native, and 38% sooner with FSR. Dying Mild 2, utilizing the Excessive RT setting, was twice as quick at native and 85% sooner with FSR. Cyberpunk exhibits features of 48% with FSR, and 62% at native, and lastly, Hitman 3 with Extremely RT was 65% sooner at native and 56% sooner with FSR. Nonetheless, native 4K efficiency fell into the unplayable vary on Dying Mild 2 and Cyberpunk 2077,
It is also value mentioning that Nvidia’s RTX 4090 delivers 44 fps at native in our Cyberpunk testing, whereas the RTX 3090 Ti scored 24.6 fps. In different phrases, not less than in essentially the most complicated ray tracing video games, it appears to be like like AMD’s new 7900 XTX at native 4K will ship comparable efficiency to Nvidia’s quickest earlier era Ampere GPU.
What we do not know (but… quickly!) is how the RTX 4080 performs. We have seen a number of benchmarks from Nvidia, however as with AMD’s checks, we have to think about the supply. Contemplating the huge increase in uncooked compute, features of fifty% on common in rasterization testing aren’t significantly spectacular. That is possible as a result of architectural modifications, the place quite a lot of the theoretical compute efficiency does not at all times immediately translate into increased efficiency.
We have coated quite a lot of the opposite materials in our different AMD articles for this morning, so be sure you examine these out for extra particulars. The complete Radeon 7900 sequence slide deck is beneath for many who wish to view extra content material.
Picture 1 of 34