Possible specifications and performance GeForce RTX 4090: Up to twice as powerful as GeForce RTX 3090
May 16, 2022
0
The GeForce RTX 4090 will become the next NVIDIA graphics card and will be based on the recent leak up to twice as powerful as the GeForce RTX
The GeForce RTX 4090 will become the next NVIDIA graphics card and will be based on the recent leak up to twice as powerful as the GeForce RTX 3090. The same leak allowed us to see what the final specifications of that graphics card might be, and the truth is that it’s impressive.
To put it in context, we have to remind you that the GeForce RTX 3090 came on the market 10,496 shadersand that it is manufactured on an 8nm Samsung node. The GeForce RTX 4090 will use the AD102 graphics core, which means it will be built on a 5nm TSMC node and will contain a huge 16,128 shaders. This graphics core has a maximum of 18,432 shaders in full configuration, but this will theoretically be reserved for the GeForce RTX 4090 Ti.
The rest of its specifications are not clear, though at the time I made an estimate what we could expect from the AD102 kernel in its full version. I enclose it below:
Made by 5nm process.
18,432 shaders.
576 texturing units.
576 tensor cores.
144 RT cores.
128 raster units.
It is important to note that these specifications could be met if NVIDIA retains the structure we saw in Ampere in its GeForce RTX 40. However, in the latest leaks about Ada Lovelace, we have seen this architecture can bring profound changesand one of them would suggest double the number of rasterization units, also known as ROPs for its acronym in English. If this is confirmed, the AD102 core would have a huge amount of ROP. It’s not confirmed, so let’s take it easy.
More details about GeForce RTX 4090: Lower consumption than expected
Those 16,128 shaders would be supported by a 384-bit bus and a GDDR6 memory system running at 21 GHz, leaving us with a bandwidth of 1 TB / s. We would need a powerful power supply to power this graphics card, but not as much as the first rumors suggested, because in theory your TGP will be around 450 watts. With this in mind, an 850-watt power supply could suffice if it is high quality and depending on the consumption of the rest of the components.
Another important detail to keep in mind is that the GeForce RTX 4090 will take advantage 4th generation tensor cores and 3rd generation RT cores, which means it will provide much higher performance under AI load and beam tracking. NVIDIA has already made a significant leap forward with the 2nd generation RT cores in Ampere, so I’m looking forward to seeing what they’ve accomplished with the 3rd generation.
In the attached table, courtesy of WCCFTech, we can see the keys to the rest of the NVIDIA high-end models. GeForce RTX 4070 will be the entry model of this series, and your TGP will be quite reasonable (300 watts). In terms of performance, it will be more powerful than the RTX 3090, although the difference may be small in the end. The GeForce RTX 4090 Ti would be the most powerful of NVIDIA, and 48 GB of 24 GHz GDDR6X memory could be added.
The new NVIDIA GeForce RTX 40 series could launch earlier than expectedand it is rumored that the recommended prices will be a little higher than those who had the GeForce RTX 30 line at the time.
The performance improvement achieved by the second generation RT cores was very large.
RDNA3 disappointing? It’s hard to believe
The source of this information, the well-known leaker “kopite7kimi”, made a comment that some overlooked, but which I find really interesting. He says he is “disappointed” with RDNA3.AMD’s next-generation architecture that will compete with NVIDIA’s Ada Lovelace, but she didn’t say why, so I don’t know if it’s a question of rasterization performance, beam tracking performance, energy efficiency, or anything at all. .
After seeing what happened to the Radeon RX 6000 that in the end they exceeded all expectations performance in rasterization, I think it is best to take that comment very carefully and not anticipate anything.
We will wait for the Radeon RX 7000 and GeForce RTX 40 to perform, and then draw the necessary conclusions, but always with the right foundation. But I personally think so The Radeon RX 7000 may result in insufficient beam trackingat least compared to GeForce RTX 40.
Alice Smith is a seasoned journalist and writer for Div Bracket. She has a keen sense of what’s important and is always on top of the latest trends. Alice provides in-depth coverage of the most talked-about news stories, delivering insightful and thought-provoking articles that keep her readers informed and engaged.