Nvidia and AMD are currently working on the next generation of GeForce RTX and Radeon RX. New rumors appear. They solve various problems including video memory.
The next family of GeForce RTX graphics cards will not be called the GeForce RTX40 Series. The name is now “temporary” due to several inaccuracies. Nvidia is likely to adopt the GeForce RTX50 series name for its generation Ada Lovelace. kopite7kim Specify these solutions as ** 90 / ** 80, not RTX 40 but RTX 50.
GeForce RTX Ada Lovelace
The new speculation about video memory is 21 Gbps, not 24 Gbps. We will suddenly be with a profile like the one newly proposed GeForce RTX 3090 Ti. The heat sink should reach 600 watts. This figure is reliable for the PCIe 5.0 connector currently fitted to the 3090 Ti.
GeForce RTX “Ada Lovelace”.
|Graphics Processing Clusters (GPC)||7||12|
|Texture Processing Clusters (TPC)||42||72|
|Streaming Multiprocessors (SM)||84||144|
|L2 cache||6 MB||96 MB|
|Memory bus||384 bit||384 bit|
|Memory capacity||24 GB (3090)||24 GB (4090 or 5090?)|
|Release date||September 20||2H 2022?|
Based on rumors
GPU Ada Lovelace The AD102 has 144 SMS, which is 71% more than the current GA102 GPU! At its core of the dynamics are 18,432 CUDA cores in operation, as well as the 384-bit bus, which is sufficient to ensure 1004 Gb / s bandwidth.
Radeon RX RDNA 3 (Radeon RX 7950 XT?)
It’s on the AMD side Murmur The next Navi 31 will be equipped with the RDNA 3 family’s primary GPU (Radeon RX 7950 XT?) 7 chiplet. In this case everything is not so clear. Generally AMD should use the MCM compression multi-chip module approach with 2 GCD (5 nm) and 4 MCD (6 nm).
This imaginary Radeon RX 7950 XT with its Navi 31 GPU is expected to deliver 15,360 stream processors running at 2.5 GHz. We do not know whether this frequency represents the maximum possible value or the base value. Meanwhile, this is an 11% improvement over the Radeon RX 6900 XT, with other modifications, sufficient to ensure the theoretical power of 38.4 TFLOPs in FP32, a good + 67% RX6900XT compared to the Radeon. The card should have 500 watts DTP, again 67% higher than its predecessor.
On the equipment side, it is not impossible to offer 32GB of RAM, which is twice as much as the current AMD flagship. The functionality of this GDDR6 is guaranteed by a 256-bit bus at 21 Gbps. Similarly, the Infinity cache should be doubled (512 MB).
Here is the first rumor-based report signed Video Cards.
|GeForce RTX4090||Radeon RX 7950XT|
|Elegance of workmanship||
|TSMC N5 / N6|
|Design||Monolith||Multi-chip module (MCM)|
|Dies||1||2 GCDs + 4 MCDs + 1 IOD|
|Mega clusters||12 GPC (Graphics Processing Cluster)||2 × 3 shader motors|
|Superclusters||72 TPCs (Texture Processing Clusters)||2 × 30 WGP (RDNA)|
|GPU clusters||144 SMS||–|
|FP32 cores||18,432 CUDA cores||15,360 stream processors|
|RAM||24GB GDDR6X||? GB GDTR6|
|Memory frequency and memory bus||21 Gbps – 384-bit||? Gbit / V – 256-bit|
|Hidden||96 MB L2 cache||256 or 512 MB Infinity cache|
|Consumption||600W (TDP)||500W (TBP)|
|Release date||Q3 / Q4 2022||Q3 / Q4 2022|
2.5x NAVI 21
“Hardcore explorer. Extreme communicator. Professional writer. General music practitioner. Prone to fits of apathy.”