While we wait to see the much talked about and “complicated” GeForce RTX 3090 Ti at work – we are waiting for March 29th (hopefully) – there is a lot of excitement for what will be the next generation of NVIDIA home GPUs, let’s talk obviously the GeForce RTX 4000 based on Ada Lovelace architecture. Over the past few weeks we have witnessed a considerable leak of news on the new NVIDIA project, not only from classic industry sources, but also due to the attack carried out by the LAPSU $ group against the company; at this point, therefore, all the information that will leak between now and the official launch can be compared with the data available, essentially allowing the less consistent rumors to be discarded a priori.
This is not the case, for example, with the latest report from Igor’s Lab that, while maintaining a certain caution on the possibility that NVIDIA could change plans until the last minute, it publishes a series of new information coming from an unspecified Chinese source, we presume close to the production chain.
At present we know more or less what will be the GPUs that NVIDIA should produce for the GeForce RTX 4000 series, guided remember by the chip AD102, what we will find on the future GeForce RTX 4090, RTX 4080 and – at this point – on the probable RTX 4090 Ti and RTX 4080 Ti. At its most powerful variant, the NVIDIA AD102 should give life to a graphics card with an 600 watt TBP that, consumption and performance aside, in the end it could have many things in common with the RTX 3090 Ti.
This is broadly what emerges from the post in question, accompanied by a unpublished diagram of the NVIDIA AD102 PCB (Founders Edition), apparently received from sources close to Asian producers. Basically, the GeForce RTX 3090 Ti with GA102 GPU and PCI-E 5.0 power connector (12 + 4 pin or 3x 8pin with adapter) it could in effect be a real “test” in view of the most advanced next-gen models such as the hypothetical GeForce RTX 4090/4090 Ti (but also RTX 4080 / RTX 4080 Ti). Going even more in detail, Igor’s Lab goes further by declaring that NVIDIA GA102 and AD102 are expected to be pin compatiblenews of no small importance that – if confirmed – would mean a lot not only for NVIDIA but also for the partners in terms of development / costs.
Returning instead to the PCB for a moment, it seems further confirmed that the GeForce RTX 4000 series will arrive substantially in the same memory configurations as the current RTX 3000 “Ampere”, ie 24GB, 16GB, 12GB and 8GB. The top models could use GDDR7 memories while on the dissipation side there is a top of the range with 3.5 slot design and PCI-E 5.0 or 4x PCI-E 8pin power connector, for current generation PSUs. In short, a lot of meat in the fire that, at least for the moment, does not see NVIDIA comment (obviously), thus suggesting that these new details may not be far from reality. Tell us yours in the comments.