NVIDIA AD102 GPUs Powering Next-Gen GeForce RTX 40 Series Expected to Stick to PCIe Gen 4.0 Protocol

NVIDIA’s GeForce RTX 40 series graphics cards based on the Ada Lovelace GPU architecture are expected to retain their existing PCIe Gen 4.0 compliance, as reported by Kopite7kimi.

NVIDIA GeForce RTX 40 ‘AD102 GPU’ graphics cards to maintain PCIe Gen 4.0 compliance

NVIDIA will launch its GeForce RTX 40 series graphics cards based on the all-new Ada Lovelace GPU architecture later this year. Specs and specific configurations for the line of graphics cards have already been leaked, but the design of the card itself is a more interesting aspect.

NVIDIA AD102 GPU flagships consume insane amounts of power: Gaming GeForce RTX 4090 at 600W TBP and Workstation RTX L6000 at 375W TBP

PCIe Gen4

– kopite7kimi (@kopite7kimi) April 24, 2022

So far we know that NVIDIA GeForce RTX 40 series graphics cards will adopt the new ATX 3.0 compliant 16-pin 12PVHPWR connector which allows up to 600W of power consumption through a new power connector interface PCIe Gen 5. This power connector was previously featured on the GeForce RTX 3090 Ti graphics card and currently allows up to 450W of power consumption through a triple 8-pin adapter. But there’s another aspect to enabling full PCIe Gen 5.0 compliance and that’s the interface slot itself.

Currently, modern graphics cards communicate with the CPU via the PCIe Gen 4.0 protocol. The PCIe Gen 4.0 protocol allows 64 GB/s of total bandwidth and 32 GB/s of bidirectional bandwidth. But the latest platforms from Intel and AMD support the brand new PCIe Gen 5.0 interface protocol. This new standard allows up to 128 GB/s of total bandwidth and 64 GB/s of bidirectional bandwidth. This will essentially double the bandwidth, but it seems that upcoming graphics cards or at least the high-end GeForce RTX 40 graphics cards based on the AD102 GPU will not yet feature a PCIe Gen 5.0 interface.

Based on a tweet from Kopite7kimi, the upcoming GeForce RTX 40 lineup will retain the PCIe Gen 4.0 protocol, which is a bold move by NVIDIA not to jump to the next-gen standard, even though they do in the HPC segment where their GPU Hopper will be among the first to use the new protocol. It now makes sense for the HPC line to have it because servers are bandwidth intensive and the Gen 5.0 protocol will help those environments. As far as consumers are concerned, the PCIe Gen 5.0 interface is simply too wide, and current GPUs have yet to fully define the PCIe Gen 4.0 interface.

NVIDIA reportedly begins testing its fastest next-gen GPU, the AD102, for the GeForce RTX 4090 graphics card, featuring 24Gbps GDDR6X memory

Now having PCIe Gen 4.0 bodes well for the entry level range which doesn’t have to worry about bottlenecks if equipped with lower lanes as was the case with the Radeon RX 6500 series and RX 6400 which when upgrading to Gen 3 end up with less than required graphics bandwidth resulting in poor performance compared to the PCIe Gen 4.0 compliant standard. If the high-end range does not starve the Gen 4.0 standard, the low-end range is far from reaching the maximum threshold. So far, we can’t say for sure if NVIDIA will really keep PCIe Gen 4.0 on its upcoming RTX 40 series cards, but that might change as marketing likes to have the PCIe Gen 5.0 logo for new cards.

Along with PCIe Gen 5.0 and PCIe Gen 4.0 support, NVIDIA is apparently going to make major changes to the way its CUDA cores are laid out in the Ada Lovelace architecture. The GeForce RTX 40-series GPUs won’t just be a simple CUDA core bump from Ampere, but could include a range of new mixed-precision cores that aren’t detailed yet. The lineup is still a few months away from introduction, so a lot could change, but we’ll be sure to keep you posted.

NVIDIA CUDA (RUMORD) GPU Preview:

GPUs	TU102	GA102	AD102
Architecture	Turing	Ampere	Ada Lovelace
Process	TSMC 12nm NFF	Samsung 8nm	TSMC 5nm
Die size	754mm2	628mm2	~600mm2
Graphics Processing Clusters (GPC)	6	7	1 2
Texture Processing Clusters (TPC)	36	42	72
Streaming Multiprocessors (SM)	72	84	144
CUDA Cores	4608	10752	18432
L2 cache	6 MB	6 MB	96 MB
Theoretical TFLOPs	16.1	37.6	~90 TFLOPs?
Memory type	GDDR6	GDDR6X	GDDR6X
Memory bus	384 bit	384 bit	384 bit
Memory capacity	11 GB (2080 Ti)	24 GB (3090)	24 GB (4090?)
Flagship SKU	RTX-2080Ti	RTX-3090	RTX4090?
TGP	250W	350W	450-850W?
Release	Sep. 2018	Sep 20	2H 2022 (to be confirmed)

News Source: Videocardz