Main menu

Pages

NVIDIA GeForce RTX 4090 & RTX 4080 PCB Detailed News

 NVIDIA GeForce RTX 4090 & RTX 4080 PCB Detailed: AD102 ‘Ada Lovelace’ GPU Support, 24 GB GDDR6X Memory & Up To 600W TDP


NVIDIA's cutting-edge GeForce RTX 4090 and GeForce RTX 4080 GPUs in light of the AD102 Ada Lovelace GPUs have been point by point.

NVIDIA GeForce RTX 4090 and RTX 4080 PCB Detailed: AD102 GPU Pin Compatible With GA102, 24 GB GDDR6X Memory, Up To 600W TDP
The most recent data with respect to the PCB of cutting edge NVIDIA GeForce gaming designs cards comes from Igor's Lab. Igor states that the impending GeForce RTX 3090 Ti has been one vital expectation to learn and adapt for NVIDIA and its AIB accomplices to prepared themselves for the cutting edge Ada Lovelace GPU setup included on the GeForce RTX 40 series. That is the explanation the RTX 3090 Ti presents support for higher-recurrence memory, higher TDP, and the main buyer card pressing PCIe Gen 5 connectors.


As indicated by Igor, the Ampere GA102 and Ada Lovelace AD102 GPUs are probably going to be pin-viable which implies they can be highlighted on comparable PCB plans or ones with little changes than to configuration totally fresh out of the box new sheets. This will permit AIBs to mosey down the improvement interaction of their cutting edge GeForce RTX 40 series custom models, assisting them with setting aside creation time and cash.

Yet again alongside that, the PCB outline that Igor has stirred up in light of data from his sources shows that the AD102 GPU sheets for the GeForce RTX 4090 and RTX 4080 will arrive in a reduced format. The format shows a 28 stage power conveyance of which 24 stages have been devoted to the GPU and PLL while the excess 4 stages power the memory. As of now, the RTX 3090 Founders Edition and reference variation for AIBs include 20 stages, and both are reference plans. All things considered, the custom models could include a much higher count assuming the following piece is valid about as far as possible.

NVIDIA Ada Lovelace and Ampere GPU Comparison

Ada Lovelace GPUMCUDA CoresTop SKUMemory BusAmpere GPUMCUDA CoresTop SKUMemory BusSM Increase (% Over Ampere)
AD10214418432RTX 4090?384-bitGA1028410752RTX 3090 Ti384-bit+71%
AD1038410752RTX 4070?256-bitGA103S607680RTX 3080 Ti256-bit+40%
AD104607680RTX 4060?192-bitGA104486144RTX 3070 Ti256-bit+25%
AD106364608RTX 4050 Ti?128-bitGA106303840RTX 3060192-bit+20%
AD107243072RTX 4050?128-bitGA107202560RTX 3050128-bit+20%

The NVIDIA GeForce RTX 40 series illustrations cards furnished with the AD102 GPUs are supposed to offer TDPs of up to 600W. That is basically the thing the current BIOS transportation to board accomplices is appraised at so the bits of hearsay around 450-600W TDPs may be valid however we haven't yet seen the last figures. The power appraisals are as a rule on the high side during the testing stage so those could be enhanced when the cards really send-off. The cards will be furnished with the PCIe Gen 5 connectors and will transport with a 4 x 8-Pin to 1 x 16-pin connector to help the gigantic power draw. The impending GeForce RTX 3090 Ti itself will transport with a 3 x 8-Pin to 1 x 16-Pin connector.

The last piece of the PCB configuration discusses the memory and as may be obvious, there are 12 bind focuses on the PCB and all are viable with Micron's GDDR6X memory. Better quality cards could go with single-sided and double limit memory since that offers the best power/temperature equilibrium and component up to 24 GB limits however at higher rates (21 Gbps+). Concerning the standard fragment, we are probably going to see 20 Gbps+ plans yet in 8 GB and up to 16 GB flavors which can assist with diminishing power since the power guideline will be dropped to 3 VRMs for the memory.


Concerning cooling these beast PCBs, Intel is allegedly going to reuse their triple-space BFGPU plan while board accomplices will use 3.5 and even quad-opening cooling arrangements weighing north of 2 kg. Most AIBs may very well wind up using AIO and Hybrid cooling plans, something that you will be found in the RTX 3090 Ti.

NVIDIA CUDA GPU (RUMORED) Preliminary:

GPUTU102GA102AD102
ArchitectureTuringAmpereAda Lovelace
ProcessTSMC 12nm NFFSamsung 8nmTSMC 5nm
Die Size754mm2628mm2~600mm2
Graphics Processing Clusters (GPC)6712
Texture Processing Clusters (TPC)364272
Streaming Multiprocessors (SM)7284144
CUDA Cores46081075218432
L2 Cache6 MB6 MB96 MB
Theoretical TFLOPs16.137.6~90 TFLOPs?
Memory TypeGDDR6GDDR6XGDDR6X
Memory Bus384-bit384-bit384-bit
Memory Capacity11 GB (2080 Ti)24 GB (3090)24 GB (4090?)
Flagship SKURTX 2080 TiRTX 3090RTX 4090?
TGP250W350W450-850W?
ReleaseSep. 2018Sept. 202H 2022 (TBC)

The NVIDIA Ada Lovelace GPU family is expected to bring a generational jump similar to Maxwell to Pascal. It is expected to launch in the second half of 2022 but expects supply and pricing to be similar to current cards despite NVIDIA spending billions of dollars to acquire those good TSMC 5nm wafers.

SOURCE: wccftech

Comments