One of the most controversial issues regarding NVIDIA’s top performance graphics cards is the fact that they make use of the 12+4-pin connector that can give them up to 600 W of power by itself, not counting the base provided by the own PCI Express. Since this is a new standard, we were wondering if AMD would follow the same path. However, they have made it clear to us that their next graphics cards with architecture RDNA 3 will not use the 12vVHPWR as a power connector.
There is a week left for AMD to present its next-generation graphics cards, based on the RDNA 3 architecture and being the first domestic GPUs to be presented in a disaggregated way, with the chip divided into several that together will fulfill the function of one. However, one question we had was whether Lisa Su’s were going to use the new 12+4-pin connector, at least with their more powerful graphics card, however, they have confirmed that it will not. So the rivals to the RTX 40 are not going to have such a high impact on the electricity bill.
AMD will not use the 12VHPWR connector on its RDNA 3 GPUs
So this means that the maximum consumption that we can expect is that corresponding to 2 8-pin cables together with the 75 W of the PCI Express port itself. So that they would not expand the consumption of their graphics cards compared to the current generation. Which means that if we make a comparison by consumption within the range, the RX 7000 based on RDNA 3 will not have an equivalent to the RTX 4090. And how do we know all this? Well, through a tweet from Scott Herkelman, who is vice president at Radeon, AMD’s graphics card division.
Scott Herkelmann
@sherkelman
@KyleBennett @NVIDIAGeForce @Radeon The Radeon RX 6000 series and upcoming RDNA 3 GPUs will not use this power connector.
October 25, 2022 • 17:34
So the information is completely official and this together with the statement that its future generation of graphics cards will have a power per watt 50% higher than its current graphics cards, as it allows us to get a rough idea of what we can expect in in terms of theoretical computing power of the next generation of AMD.
What can we expect from the Radeon RX 7900 XTX?
Taking into account the consumption of the current generation of AMD and the promise in terms of performance jump, we can extrapolate the following data:
- If AMD maintains consumption in the next generation, 335 W for the top of the range, and extrapolating the greater efficiency of the RX 6000, then the most powerful graphics card would have a 38.6 TFLOPS power compared to 23.8 TFLOPS of the RX 6950 XT. which is approximately a 63% performance per additional watt.
- We will see important changes in the architecture, the jump will not be only in brute force.
- Increased floating point calculation units from 64 to 128 per core.
- Out-of-order execution or reordering of threads on the fly as already implemented by Intel ARC and NVIDIA RTX 40.
- Improved Ray Tracing system, compared to that of the RX 6000.
With the information still unconfirmed, the most powerful RDNA 3 graphics chip would have 12,288 compute units spread over 96 Compute Units in total. A figure above the RTX 4080 with 9728 calculation units spread over 76 SM. In addition, AMD’s unbundled chip will have a 384-bit GDDR6 buswhich will allow you to have up to 24 GB of video memorymatching in this case with the RTX 4090 and the RTX 3090 Ti and bringing with it 192 MB of Infinity Cache.
What translated for the average user, we are talking about that in performance it would be well above the RTX 3090 Ti and RTX 4080, but with much greater efficiency, since we will not see the 12VHPWR connector in RDNA 3. However, in When it comes to raw power, NVIDIA will continue to have the lead.
