Igor’s Lab has the most up-to-date information about the PCB of next-generation NVIDIA GeForce gaming graphics cards. According to Igor, the upcoming GeForce RTX 3090 Ti has served as a critical learning curve for NVIDIA and its AIB partners as they prepare for the next-generation Ada Lovelace GPU lineup featured on the GeForce RTX 40 series. That’s why the RTX 3090 Ti is the first consumer card to include PCIe Gen 5 connectors, as well as support for higher-frequency memory and a higher TDP.
The Ampere GA102 and Ada Lovelace AD102 GPUs, according to Igor, are likely to be pin-compatible, which implies they can be used on similar PCB designs or ones with minor adjustments rather than having to develop entirely new boards. This will allow AIBs to speed up the creation of their next-generation GeForce RTX 40 series bespoke models, saving time and money in the process.
Aside from that, Igor’s PCB blueprint, based on information from his sources, suggests that the AD102 GPU boards for the GeForce RTX 4090 and RTX 4080 will come in a compact configuration once again. The arrangement depicts a 28-phase power supply, with 24 phases allocated to the GPU and PLL and the remaining four phases powering the memory. The RTX 3090 Founders Edition and reference variation for AIBs both have 20 phases and are reference designs at the moment. If the following piece about the power restriction is true, the custom models are likely to have an even greater count.
TDPs of up to 600W are expected for the RTX 40 series graphics cards coupled with the AD102 GPUs. The current BIOS shipping to board partners is rated at that, thus the speculations of 450-600W TDPs could be genuine, but we haven’t seen the final statistics yet. During the testing phase, power ratings are normally on the high side, so they can be tweaked once the cards are released. To support the massive power draw, the cards will be equipped with PCIe Gen 5 connectors and a 4 x 8-Pin to 1 x 16-Pin converter. A 3 x 8-Pin to 1 x 16-Pin adaptor will be included with the future GeForce RTX 3090 Ti.
The memory is discussed in the final section of the PCB design, and as you can see, there are 12 solder locations on the PCB, all of which are compatible with Micron’s GDDR6X memory. Single-sided and dual-capacity memory may be used on higher-end cards since it provides the best power/temperature balance and can hold up to 24 GB of data at greater speeds (21 Gbps+). In the mainstream sector, we’re likely to see 20 Gbps+ architectures in 8 GB and up to 16 GB flavours, which will assist cut power consumption because the memory’s power regulation will be reduced to three VRMs.
Intel is apparently intending to employ their triple-slot BFGPU architecture to cool these massive PCBs, while board partners will use 3.5 and possibly quad-slot cooling solutions weighing over 2 kg. Most AIBs are likely to use AIO and hybrid cooling systems, as shown in the RTX 3090 Ti.
The NVIDIA Ada Lovelace GPU family is intended to be a generational leap in the same way that the Maxwell to Pascal transition was. Despite paying billions of dollars to purchase those good TSMC 5nm wafers, NVIDIA anticipates supply and pricing to be equal to existing cards when it launches in the second half of 2022.