A good amount of information about the GeForce RTX 5080 has been leaked from several sources, and this led me to do a full review of the special article I published about this graphics card at the time, so that you have all updated information.

The positioning of the GeForce RTX 5080 has not changed, it will be a high-end graphics card that will sit below the GeForce RTX 5090, but we have an interesting new detail that comes from Chinese sources, that NVIDIA will launch two versions of which one with 16 GB of graphics memory and the other with 24 GB of graphics memory.

The 16GB model will be the original model and the first to hit the market, and the 24GB model will arrive later, presumably as an intergenerational refresh accompanied by a tagline. “You” or “SUPER”.

Image: Chiphell.

GeForce RTX 5080 specifications

Latest information have decreased drastically the total number of shaders of this graphics card and pushed it to a level that would place it just slightly above the GeForce RTX 4080. However, thanks to the improvements that the Blackell architecture will introduce and the significant increase that we expect in the clock frequency year, the difference between the two in terms of performance will be relatively large.

• Graphics core GB203 made on TSMC’s 4nm node.
• 84 SM units.
• 10,752 shaders at speeds between 2.8 GHz and 3.1 GHz.
• 336 texturing units.
• 112 raster units.
• 84 cores of the fourth generation RT.
• 336 fifth generation tensor kernels.
• 256-bit bus.
• 16 GB of GDDR7 memory at 28 Gbps (896 GB/s).
• 64 MB of L2 cache.
• 400W TGP.

GeForce RTX 5080 vs GeForce RTX 4080

• AD103 GPU manufactured by TSMC’s 5nm process.
• 76 SM units.
• 9728 shaders at 2205MHz-2505MHz, normal and turbo mode.
• 304 texturing units.
• 112 raster units.
• 304 fourth generation tensor kernels.
• 76 cores of the third generation RT.
• 256-bit bus.
• 16 GB of GDDR6X memory at 22.4 Gb/s (716.8 GB/s).
• TGP 320W.

With these specs, the GeForce RTX 5080 would only have a small increase in total shaders compared to the GeForce RTX 4080, as it has 9,728 shaders. The difference between the two would be 1024 shaders, and as we can see when comparing the specs of the two, there would also be an increase in the rest of the key GPU components such as texture and raster units, tensor cores and RT cores.

GeForce RTX 5080
PG144/147-SKU45
GB203-400-A1
10752FP32
256-bit GDDR7 16G
400W

— kopite7kimi (@kopite7kimi) September 26, 2024

An increase in working frequencies would be spectacular as we speak up to 600 MHz more in turbo modeand the bandwidth would also increase thanks to the use of GDDR7 memory, although this amount will remain the same as the previous generation, 16 GB, a number that is still more than enough to play with all guarantees in 4K.

The GeForce RTX 5080 is much more powerful, but also will have higher consumption. The latest rumors are talking about 400 watts of TGP, if confirmed, the difference with the current model would be 80 watts, which is a reasonable increase considering the differences that will be between the two graphics cards in terms of performance.

New model with 24 GB GDDR7

As I said at the beginning of the article, Chinese sources assure that NVIDIA is going to launch a second GeForce RTX 5080 model, which will have 24 GB of graphics memory. It will be an intergenerational restoration as well as the GeForce RTX 4080 SUPERand therefore it should also come with a more powerful GPU.

This means that it should have a larger number of active SM units, which would translate into more shaders, texture and raster units, tensor cores, and RT cores. No concrete details have been leaked yet, but https://www.muycomputer.com/2024/09/04/geforce-rtx-5080-mas-potente-que-la-geforce-rtx-4090-y-con-menor-consumption maybe this second model exceeds 12,000 shaders and has a 384-bit memory bus.

The difference in performance compared to the 16 GB model It will depend on the improvement it brings at the GPU levelmore than an increase in the amount of graphics memory, because as I said, having more than 16 GB of graphics memory currently does not make a difference in gaming performance, and will not in the short or medium term.

Possible performance and raw power

If the GeForce RTX 5080 arrives with 10,752 shaders at a maximum of 3 GHz, its performance will 64.51 TFLOPs in FP32, a number that would exceed 48.74 TFLOPs which reaches the GeForce RTX 4080, but would be behind the GeForce RTX 4090, which has a raw performance of 82.58 TFLOPs in FP32.

Although this comparison in numbers suggests the opposite, A GeForce RTX 5080 would be about 10% more powerful than the GeForce RTX 4090 in games, thanks to the various improvements it will bring at the architectural level. It could also have support for new technologies within the DLSS ecosystem, but this is not yet confirmed.

Performance difference in ray tracing could have been even higher thanks to the possible improvements that the Blackwell architecture will introduce in the fourth generation of RT cores. We will have to wait for NVIDIA to officially introduce this architecture to clarify any doubts.

If we are talking about a model with 24 GB of graphics memory, the performance jump would be greater. We don’t have an exact number of shaders, but we’ll use a realistic estimate to arrive at an approximate performance value that isn’t crazy, like 12,288 shaders, which is the equivalent of 96 active SM units.

A GeForce RTX 5080 SUPER or Ti with 12,288 shaders at 3GHz would have a raw performance of 73.72 TFLOPs at FP32. It still wouldn’t be enough to beat the GeForce RTX 4090 in raw performance, but the former would be more powerful in games. In this case, I think a hypothetical GeForce RTX 5080 SUPER or Ti it would beat the GeForce RTX 4090 in games by an average of 15% or 18%.

Power supply, consumption and electrical installation

If the TGP of 400 watts is confirmed, it will be normalized consumption GeForce RTX 5080 working at 100% load. In situations where the load on the GPU is lower, because the game is not very demanding or because we have NVIDIA DLSS activated and these technologies reduce the load on the GPU, the consumption will be lower.

To be able to use the GeForce RTX 5080 with all the guarantees we will need 750W power supply, and ideally it will have a 16 pin power connector so you don’t have to use adapters. This will also make the installation much more aesthetic.

In most cases, this power supply will be enough, so we will not have any type of problem in terms of stability and performance. However, it is very important that you keep this in mind everything will depend on the consumption of the rest of the components your PC.

If you are using processors like Ryzen 7 and below and Core i5 and below You won’t have a problem, but starting with Ryzen 9 and Core i7 the CPU consumption may increase and in these cases it might be advisable to have a slightly more powerful resource. In these cases, we will have no problems with the 850 watt source.

Recommended configuration for GeForce RTX 5080

For the GeForce RTX 5080 to develop its full potential, we will need powerful PC, although we won’t necessarily have to achieve a top configuration in every sense, which would be recommended with the GeForce RTX 5090.

Starting with the processor, it would be ideal not to go lower Ryzen 5 7600, Intel Core i5-12600K or Ryzen 7 57000X3D. Having 32 GB of RAM and a PCIe NVMe SSD capable of at least 3.5 GB/s sequential read speed are two other important recommendations for balancing a PC with this graphics card.

We’ve already seen what kind of power supply we’d need, and now we need to talk about it.in the recommended resolution. The GeForce RTX 5080 will be a graphics card designed to achieve a perfect gaming experience below 1440p with ray tracing enabled and maximum qualities, as well as to play in 4K without any problems, so these would be the two ideal target resolutions for this card. chart.

That’s not to say that we won’t be able to use it in 1080p, but at that resolution we might find cases where it will be so sufficient that eventually may suffer from a CPU bottleneck. This is something we’ve already seen with the GeForce RTX 4090, and the GeForce RTX 5080 will be more powerful than it is, so the story speaks for itself.

Release date and price

It seems that a certain consensus is gradually emerging between various sources and that the introduction of the GeForce RTX 5080 could happen in CES 2025. A preview is not the same as a release date, so there may be a few weeks between this preview and the actual availability date of said graphics card.

Rumors that NVIDIA could introduce the GeForce RTX 5080 before the end of the year have not resurfaced, so right now has more credibility that a possible presentation in 2025. However, be clear that nothing is confirmed yet.

The price of the GeForce RTX 5080 could be similar to that of the GeForce RTX 4080 at launch, but it could also be slightly lower or slightly higher. NVIDIA fine-tuned the price of its new graphics cards it increases significantly after each new generationa strategy that could be repeated with the GeForce RTX 50 series.

I think at best the GeForce RTX 5080 could stand 999.99 euros, and that could end up being the worst case scenario 1,299.99 euros. As a reference, I remind you that the GeForce RTX 4080 came to the market with a price of 1,469 euros, but the GeForce RTX 4080 SUPER, which was launched with a price of 1,119 euros, was much more affordable.