Are we in for the biggest jump in performance ever from one generation to the next with Nvidia’s upcoming flagship graphics card RTX 5090? At least that’s what the leaker AGF wants to know. In a tweet on the X platform (formerly Twitter) he writes:
Link to Twitter content
Translated: People are confused about the memory bus and VRAM capacity of Nvidia Blackwell top die gaming GPU. I can’t say much at the moment, but the increase in performance from generation to generation is the largest it has ever been. Don’t forget that Blackwell was designed to work with a monster 3nm MCM RDN4[Grafikkarte] to compete. More on this soon…
However, AGF does not go into specific technical details or benchmarks. The latter will probably take a while to arrive, as the RTX 5090 is not yet in the testing phase.
How big could the jump in performance be?
If this is indeed the biggest leap in performance of all time, the RTX 5090 would probably have to be at least 70 percent faster than the RTX 4090.
Because the current Nvidia flagship has already made one of the biggest cross-generational leaps we’ve ever seen. In the GameStar test from October 2022, the RTX 4090 outperformed the RTX 3090 in 4K resolution by an incredible 65 percent.
How credible is the leak?
The leaker AGF has been quiet lately, but in the past his predictions have often proven to be correct.
At the same time, his current tweet coincides with leaks and rumors about the RTX 5090, which have been in circulation for a few weeks. The leaker already had at the end of February Moore’s Law Is Dead There was talk of a 70 percent jump in performance compared to the RTX 4090.
The specifications of the RTX 5090 that have been rumored so far also indicate a significant increase in performance:
RTX 4090 | RTX 5090* | |
---|---|---|
production | TSMC 5NM | TSMC 3NM |
Streaming-Multiprozessoren (SM) | 128 | 192 |
computing cores | 16.384 | 24.576 |
Boost clock speed | 2.520 MHz | 2.900 MHz |
Storage | 24,0 GB GDDR6X | 24.0 GB GDDR7 |
The RTX 4090 shows the way
Based on 50 percent more computing cores (24,576 versus 16,384), architectural improvements, faster VRAM (GDDR7 versus GDDR6X) and potentially higher clock speeds, 70 percent more performance seems well within the realm of possibility.
The question remains whether the theoretically higher computing power can also be brought onto the road. Here it is worth taking a look at the RTX 4090 again. Expressed in frames per second, it delivers – as described above – a 65 percent increase in performance compared to the RTX 3090 with 56 percent more computing cores (16,384 versus 10,496).
There are also significantly higher clock speeds (nominally 2,520 MHz versus 1,695 MHz) and improvements to the architecture.
Why should Nvidia make such a big leap again?
This is also evident from AGF’s tweet: According to this, the RTX 5090 was originally intended to compete against an AMD flagship in a multi-chip module design (MCM design for short), which potentially offered an extremely high number of computing cores.
However, AMD now seems to have backed away from these plans. Leaks and rumors suggest that the RX 8000 series will attack the mid-price segment and leave the high-end segment to Nvidia.
But AGF gives another reason why Nvidia is making full use of its resources:
Link to Twitter content
Accordingly, the RTX 5090 will have to deal with both RDNA4 (RX 8000) and RDNA5 (RX 9000) over the course of its lifespan.
When is the RTX 5090 expected and how much should it cost?
The rumor mill puts the launch of the RTX 5090 towards the end of 2024. However, it is said to be the only Blackwell GPU released this year.
Nothing concrete can be said about the price at the moment. However, the upcoming flagship will certainly not be a bargain, which is why we place it at least in the price range of the RTX 4090 (RRP at market launch: 1,949 dollars).
Will other RTX 5000 GPUs also make such a leap?
While the RTX 5090 could actually achieve the biggest jump in performance of all time, the RTX 5080 and Co. only seem to be able to enjoy moderate increases in performance.
The RTX 5080 is said to have 12,228 computing cores (versus 10,240 cores in the RTX 4080), which indicates a performance jump of 20 to 30 percent.
how do you see it? Do you think the RTX 5090 will really make the biggest leap in performance ever? Or do you think she will make a moderate leap forward? Feel free to write it to us in the comments!
Table of Contents