In recent weeks, there have been new rumors and leaks about the GeForce RTX 5070, the much-anticipated graphics card as it becomes the model entry into the high-end of the new NVIDIA generationand because the RTX X070 series usually offers good value for money.

All this new information was not limited to talking about the possible specifications of the GeForce RTX 5070, they also touched on such important topics as its possible performance, graphics memory configuration and the possible price of this graphics card. To offer you a more complete vision, I created a compilation all we know so far. This special is an update of an article I published last May.

It’s important to keep in mind that nothing is confirmed yet, so whatever we see next is subject to change. However, it is unlikely because everything points to it NVIDIA already has the GeForce RTX 50 readyand that it is delaying its launch a bit to facilitate the clearing of GeForce RTX 40 stock.

Possible specifications of GeForce RTX 5070

AI generated image.

• Graphics core GB205 made on TSMC 4nm node.
• 50 SM units.
• 6,400 shaders at speeds between 2.8 GHz and 3.2 GHz.
• 200 texture units.
• 64 raster units.
• 50 fourth generation RT cores.
• 200 fifth generation tensor cores.
• 192-bit bus.
• 12 GB of GDDR7 memory at 28 Gbps (672 GB/s bandwidth).
• 40 MB L2 cache.

The new GeForce RTX 5070 specs place it just slightly above the GeForce RTX 4070 shader count, as the former would have 6,400 shaders and the latter has 5,888 shaders. This lowers the gross performance predictions that we had originally and means that its overall performance will be lower than we expected.

This new graphics card will repeat the configuration of the s 12 GB of graphics memory. All the sources I’ve consulted point in that direction and also suggest that there will be an intergenerational refresh that will be configured with more graphics memory. Said refresh could be a GeForce RTX 5070 SUPER and would be configured with 16GB or 18GB of graphics memory.

So there would be no improvement in this regard compared to the GeForce RTX 4070, which It also has 12 GB of graphics memory. This type of configuration is optimal for playing at 1440p resolution with maximum qualities, although the game can also handle 4K resolution without problems. The TGP of this model should be around 200 watts, i.e. at the same level at which the GeForce RTX 4070 works.

Compared to the GeForce RTX 4070, the GeForce RTX 5070 would also offer a larger number of RT cores, tensor cores, texture units, and L2 cache memory. All this, together with the increase in operating frequencies and improvements derived from the new Blackwell architecture, will lead to significant performance improvement both in rasterization and ray tracing.

L2 cache will increase from 36 MB to 40 MB. It’s an interesting upgrade because the extra 4MB of cache will increase the success rate The GPU calls into this cache to load the data and stuff you need, reducing the dependency on the GPU subsystem. The 40 MB is the optimal amount for 1440p resolution, but not for 4K.

The GB205 graphics core that the GeForce RTX 5070 will use will be manufactured in TSMC 4nm nodea minor evolution compared to the 5nm node also from TSMC, which uses the GeForce RTX 4070. This new node will reduce the size of the transistors and allow an improvement in efficiency compared to the previous generation, i.e. more power per watt consumed.

Performance, optimal resolution and comparable performance

AI generated image.

GeForce RTX 5070 will be the graphics card play optimally in 1440p resolution. That will be the ideal level, but as I said before, that doesn’t mean it won’t be able to push games in 4K, nothing could be further from the truth. It will be able to move very well at this resolution, but it will not be an ideal level, since its configuration at the level of L2 cache and graphics memory is designed for 1440p.

In terms of raw performance, if the GeForce RTX 5070 runs at a stable average of 3.2 GHz, we should 40.96 TFLOPs in FP32. For comparison purposes, the GeForce RTX 4070 has 29.15 TFLOPs and the GeForce RTX 4070 Ti SUPER has 44.10 TFLOPsso the GeForce RTX 5070 would be slightly less powerful than the other for the FP32.

Keep in mind that FP32 performance does not directly equate to gaming performance, so this data is only a rough reference to place the GeForce RTX 5070 within a specific range. Personally, I think that this graphics card will meet these specifications almost as powerful as the GeForce RTX 4070 Ti SUPER in 1440p games.

Going a little deeper into this topic, which I know you’re always interested in, the closest thing in the previous generation would be a GeForce RTX 3090 Tiand on the AMD side we should Radeon RX 7900 XT as the closest equivalent in terms of gross performance.

In ray tracing, it is very likely that the GeForce RTX 5070 will ultimately outperform the GeForce RTX 4070 Ti SUPER, thanks to its new fourth generation RT cores and on the improvements that NVIDIA will incorporate as a result of the architecture change. The difference in performance in this case in favor of the former should be between 10% and 20%, depending on each specific game.

Configuration with 12 GB of graphics memory won’t cause any problem in 1440p, not even with ray tracing enabled, but in specific 4K cases it could be a limiting factor. That’s why I told you that the GeForce RTX 5070 has an optimal resolution of 1440p, although this could change if NVIDIA introduces some new technology that will somehow reduce the consumption of graphics memory.

And speaking of new technologies, it should be one of the most important keys to the next generation of NVIDIA graphics, although I don’t have specific details yet. The company may use the launch of the GeForce RTX 50 to introduce optimizations in ray tracing as well as launch news in its NVIDIA DLSS ecosystem. We have to wait and see what they surprise us with.

Ideal processor and resource for GeForce RTX 5070

If the TGP of 200 watts is confirmed, we will need a quality power supply of at least 500 watts, although this will depend on the rest of the components of our device. For example, if we have a Ryzen 5 7600 processor, 32GB DDR5, NVMe M.2 SSD, 240mm AIO liquid cooling and a 6-fan chassis with a GeForce RTX 5070, we would have an average load consumption 372 watts.

Things would change, a lot, if we had a different processor. For example, the same configuration with an Intel Core i9-13900K should load consumption 545 wattsand in this case we would need a 600 watt power supply to leave some margin and not be too tight. If we have more fans or other components that increase consumption, we will have to look for a more powerful power supply.

NVIDIA will repeat 16-pin power connector with a GeForce RTX 5070, so it would be ideal to look for a source that has the mentioned connector so that you don’t have to use adapters. This connector is standardized in most quality sources from brands like Corsair, MSI, ASUS and GIGABYTE, so it won’t be difficult to find an option that fits our budget.

As for the processor needed to move the GeForce RTX 5070 with all guarantees, it’s a very simple question, since we will basically need the same one that we would need to move the GeForce RTX 4070 Ti SUPER without any problem, since both will offer very similar performance. It would be ideal to have a Ryzen 5 7600 or Intel Core i5-12600K.

The GeForce RTX 5070 will be a graphics card focused on 1440p playback, and at this resolution the processor has a significant weight, which is an important detail that we must take into account, since this component can act as a bottleneck and limit the performance of the computer. graphics card. If we enable NVIDIA DLSS Super Resolution, the rendering resolution will decrease and CPU dependency will increaseso the best thing you can do to get the most out of this graphics card would be to not go below the recommendations i gave you.

Possible release date and price of GeForce RTX 5070

The GeForce RTX 5070 will be launched after the arrival of the GeForce RTX 5090 and GeForce RTX 5080. There is no specific date yet, but most likely we will see it when second quarter of 2025a few months after the launch of two more top-of-the-line NVIDIA models, which will arrive in January 2025.

By jumping on the selling price, NVIDIA has two options, they can list the GeForce RTX 5070 at the same price as the GeForce RTX 4070, or they can list it at a slightly higher price. I think the company is betting on placing this graphics card in the $600 range and that its price in Spain after converting taxes and using the currency will be approx. 649 euros.

Keep in mind that it would be recommended base pricei.e. minimum price and that assemblers will market models with premium modifications and overclocking at higher prices. GeForce RTX 5070 will have a Founders Edition versionas has always been the case with this range, which will be available at MSRP.

AI generated cover image.