We’re in the home stretch of the year, and both NVIDIA and AMD are already have finalized their pending releasesso it’s a good time to find out which are the ten most powerful graphics cards of 2022, an article that updates the original one we also published at the end of last year.
It was precisely these launches, and also the catalog renewals that NVIDIA and AMD have made, that have forced us to introduce important modifications in this new list of the most powerful graphics cards available today, although in the end quite a few 2021 models have been retained, as they still offer excellent performance.
Generally speaking, we can say that NVIDIA has managed to. keep the raw performance crown with the launch of the GeForce RTX 4080 and RTX 4090, and has improved power efficiency (performance per watt). The green giant has also managed to strengthen its leadership in ray tracing, and with DLSS 3 has shaped a technology that significantly improves performance and is not affected by CPU-level bottlenecks.
AMD has generated contrasting sensations with the Radeon RX 7900 XT and RX 7900 XTX. It is true that they represent a clear leap in performance compared to the previous generation, but they have not met the expectations they had generated, neither in raw power nor in power consumptionand only the latter is capable of coping with the GeForce RTX 4080 in rasterization, as in ray tracing they are still far behind NVIDIA, and this has long ceased to be a technology little used or of minor importance, and has become quite the opposite.
All of this, coupled with the fact that AMD has no real alternative to DLSS 3, means that in the end the price difference that exists between the Radeon RX 7900 and the GeForce RTX 40 is very diluted. The GeForce RTX 4080 right now costs €1,399, and the Radeon RX 7900 XTX has a recommended price of €1,199, but it is impossible to find it available for that money. The difference between the two is 200 euros, but is it really worth that saving and giving up higher ray tracing performance, DLSS 3 value and lower power consumption?
Everyone will give the answer they deem appropriate, but under my point of view I have clear that it is a no, it is not worth saving 200 euros and give up all that, but. especially for the difference that the DLSS 3 makes.. In the absence of that technology my answer would not have been so forceful.
The 10 most powerful graphics cards of 2022 in rasterization.
As I did in the 2022 guide, I’m going to differentiate between the most powerful graphics cards in raster from those that are superior in ray tracing, since limiting the comparison to a single value would be totally unfairand would put graphics cards that are very powerful in ray tracing in a position of inferiority to others that would not accurately reflect reality.
If we focus on rasterization performance, we see that the most powerful models that occupy the top three are the GeForce RTX 4090, Radeon RX 7900 XTX and GeForce RTX 4080.. The Radeon RX 7900 XT drops out of the top three and is almost at the same level as the GeForce RTX 3090 Ti, which is followed by the GeForce RTX 3090. The remaining positions are occupied by the Radeon RX 6950 XT, the GeForce RTX 3080 Ti and the Radeon RX 6900 XT.
All of those graphics cards are capable of moving all current games in 4K resolution with maximum quality while maintaining full fluidity.. They make sense also for playing in 1440p with high frame rates per second, but they lose value in 1080p because they suffer from clear CPU bottlenecks in most cases, a problem that is especially severe from the GeForce RTX 3090 Ti onwards in that resolution.
We have not taken into account second- and third-generation DLSS technologies and FSR because. would greatly complicate the comparisonand because its value is not limited to performance enhancement.since they also affect image quality. In this regard, DLSS maintains higher image quality and offers clearly superior value, so limiting the comparison to a performance-focused approach would not have been fair either.
The 10 most powerful graphics cards of 2022 in ray tracing.
I’ve said it before, but I’ll say it again, ray tracing is here to stay, and. is no longer the future, but the present.. This technology has been supported for years in PS5 and Xbox Series X, despite the power limitations of both consoles, and today many games support it, a reality that has become a trend and that will become more and more.
Cyberpunk 2077 looks spectacular with ray tracing.
Not taking it into account when assessing the raw power of a graphics card no longer makes any sense, and it means forcing a biased and unrealistic view of the value offered by a graphics card.something that is obviously not good for anyone. On this subject I want to add an anecdote, an acquaintance asked me for advice a few weeks ago to buy a new graphics card now that prices had dropped.
He told me he was going to play seriously, and that wanted to get into Cyberpunk 2077 Because he played it on PS4 and decided to return it because of how bad it was running on that console. He had a PC consisting of a Core i7-6700K, 16 GB of RAM and had a GeForce GTX 950. It is clear that at the time he made an unbalanced assembly and chose a very weak CPU for his CPU, and he was also wrong to think that such a game would go well on that console.
I told him that the ideal for his configuration was, if he wanted to spend little money, to opt for a 6GB GeForce RTX 2060 Which, at the time, was available for 249 euros. In the end he decided to follow someone else’s advice and bought a Radeon RX 6600 XT which cost him almost 400 euros, because he was told it had more graphics memory and was newer and more powerful. He was disappointed with the purchase when he saw the performance in Cyberpunk 2077 when he turned on ray tracingand decided to return it.
As we can see in the attached graph many positions have moved. The three most powerful graphics cards in ray tracing in 2022 are the GeForce RTX 4090, the GeForce RTX 4080 and the Radeon RX 7900 XTX, which narrowly outperforms the GeForce RTX 3090 Ti. In fourth and fifth place are the GeForce RTX 3090 and GeForce RTX 3080 Ti, and in sixth place we find the Radeon RX 7900 XT, which performs only slightly better than the 12GB GeForce RTX 3080. Closing out the top ten are the GeForce RTX 3080 and Radeon RX 6900 XT.
That AMD has improved ray tracing performance is unquestionable, but there is also no denying that. is still a generation behind NVIDIAand still has a lot of work ahead of it if it wants to catch up with the green guys.
Note that the figures we’ve given represent relative performance expressed as a percentage, which means that the GeForce RTX 4080 performs 36% better than the Radeon RX 7900 XT in ray tracing, but this does not mean that both offer a smooth FPS rate in 4K with that technology enabled.as the GeForce RTX 4080 hovers between 27 and 29 FPS in Cyberpunk at that setting, and the Radeon RX 7900 XT hovers between 16 and 18 FPS (ray tracing in insane on both).
Specifications of the 10 most powerful graphics cards of 2022.
To compile this selection we have taken into account the combined performance in rasterization and ray tracing.but we have given more weight to the latter because it is a much more demanding technology, so keep it in mind when evaluating it. As always, if you have any questions you can leave them in the comments.
GeForce RTX 4090
- AD102 GPU manufactured in 5 nm process (TSMC).
- 76.3 billion transistors.
- 16,384 shaders at 2,235 MHz-2,520 MHz, normal and turbo mode.
- 512 texturing units.
- 176 raster units.
- 512 fourth-generation tensor cores.
- 128 third-generation RT cores.
- 384-bit bus.
- 24 GB of GDDR6X memory at 21 Gbps.
- L2 cache: 72 MB.
- 1.008 GB/s bandwidth.
- Power at FP32: 82.58 TFLOPs.
- TGP: 450 watts.
GeForce RTX 4080
- AD103 GPU manufactured in 5 nm process (TSMC).
- 45.9 billion transistors.
- 9,728 shaders at 2,205 MHz-2,505 MHz, normal and turbo mode.
- 304 texturing units.
- 112 raster units.
- 304 fourth-generation tensor cores.
- 76 third-generation RT cores.
- 256-bit bus.
- 16 GB of GDDR6X memory at 22.4 Gbps.
- L2 cache: 65 MB.
- 716.8 GB/s bandwidth.
- Power at FP32: 48.74 TFLOPs.
- TGP: 320 watts.
Radeon RX 7900 XTX
- Navi 31 graphics core with MCD (Multi-chiplet Die) design in 5 nm node the GPU block and 6 nm the cache block.
- 57.7 billion transistors.
- GPU area: 300 mm square.
- Area of cache encapsulations: 36.6 square mm per unit.
- 6,144 shaders.
- 384 texturing units.
- 192 raster units.
- 1.85 GHz-2.49 GHz GPU, normal and turbo mode.
- 61.42 TFLOPs of power in FP32.
- 96 computing units.
- 96 units for second generation ray tracing acceleration.
- 192 units for artificial intelligence acceleration.
- 384-bit bus.
- 24 GB of GDDR6 memory at 20 GHz with a bandwidth of 960 GB/s.
- 96 MB of infinite cache with a bandwidth of 5.3 TB/s.
- 355 watts TBP.
GeForce RTX 3090 Ti
- GA102 GPU manufactured in 8 nm process (Samsung).
- 28.3 billion transistors.
- 10,496 shaders at 1,395 MHz-1,695 MHz, normal and turbo mode.
- 328 texturing units.
- 112 raster units.
- 328 third-generation tensor cores.
- 82 second-generation RT cores.
- 384-bit bus.
- 24 GB of GDDR6X memory at 19.5 Gbps.
- L2 cache: 6 MB.
- 936.2 GB/s bandwidth.
- Power at FP32: 35.58 TFLOPs.
- TGP: 350 watts.
GeForce RTX 3090
- GA102 GPU manufactured in 8 nm process (Samsung).
- 28.3 billion transistors.
- 10,752 shaders at 1,560 MHz-1,860 MHz, normal and turbo mode.
- 336 texturing units.
- 112 raster units.
- 336 third generation tensor cores.
- 84 second-generation RT cores.
- 384-bit bus.
- 24 GB of GDDR6X memory at 21 Gbps.
- L2 cache: 6 MB.
- 1.008 GB/s bandwidth.
- Power at FP32: 40 TFLOPs.
- TGP: 450 watts.
GeForce RTX 3080 Ti
- Graphics core: GA102 in 8 nm process (Samsung).
- 28.3 billion transistors.
- 10,240 shaders at 1,365 MHz-1,665 MHz, normal and turbo mode.
- 320 texturing units.
- 112 raster units.
- 320 tensor cores.
- 80 RT cores.
- 384-bit bus.
- 34.10 TFLOPs of power on FP32.
- L2 cache: 6 MB.
- 12 GB of GDDR6X memory at 19 GHz (912 GB/s).
- TGP: 350 watts.
Radeon RX 7900 XT
- Navi 31 graphics core with MCD (Multi-chiplet Die) design in 5 nm node the GPU block and 6 nm the cache block.
- 57.7 billion transistors.
- GPU area: 300 mm square.
- Area of cache encapsulations: 36.6 square mm per unit.
- 5,376 shaders.
- 336 texturing units.
- 192 raster units.
- 2.025 GHz-2.394 GHz GPU, normal and turbo mode.
- 51.48 TFLOPs of power at FP32.
- 84 computing units.
- 84 units for second generation ray tracing acceleration.
- 168 units for artificial intelligence acceleration.
- 320-bit bus.
- 20 GB of GDDR6 memory at 20 GHz with a bandwidth of 800 GB/s.
- 80 MB of infinite cache.
- 300 watt TBP.
GeForce RTX 3080 12 GB
- GA102 GPU manufactured in 8 nm process (Samsung).
- 28.3 billion transistors.
- Package area: 628 square mm.
- 8,960 shaders at 1,260 MHz-1,710 MHz, normal and turbo mode.
- 280 texturing units.
- 96 raster units.
- 280 third-generation tensor cores.
- 70 second-generation RT cores.
- 384-bit bus.
- 10 GB of GDDR6X memory at 19 Gbps.
- L2 cache: 5 MB.
- 912.4 GB/s bandwidth.
- Power at FP32: 30.64 TFLOPs.
- TGP: 350 watts.
GeForce RTX 3080
- GA102 GPU manufactured in 8 nm process (Samsung).
- 28.3 billion transistors.
- Package area: 628 square mm.
- 8,704 shaders at 1,440 MHz-1,710 MHz, normal and turbo mode.
- 272 texturing units.
- 96 raster units.
- 272 third generation tensor cores.
- 68 second-generation RT cores.
- 320-bit bus.
- 10 GB of GDDR6X memory at 19 Gbps.
- L2 cache: 5 MB.
- 760.3 GB/s bandwidth.
- Power at FP32: 28.77 TFLOPs.
- TGP: 320 watts.
Radeon RX 6950 XT
- 7 nm Navi 21 XTX graphics core (TSMC).
- 26.8 billion transistors.
- Package area: 520 square mm.
- 5,120 shaders at 1,860 MHz-2,310 MHz, normal and turbo mode.
- 320 texturing units.
- 80 units for ray tracing.
- 128 raster units.
- 256-bit bus.
- 16 GB of GDDR6 memory at 18 GHz.
- Bandwidth: 576 GB/s.
- L2 cache: 4 MB.
- 128 MB infinite cache.
- 23.65 TFLOPs of power in FP32.
- 335 watt TBP.
