In the coming months, new graphics cards from NVIDIA and AMD will appear. The first companies will offer the Ada generation in the form of GeForce RTX 4000 systems. On the other hand, AMD is working on the RDNA 3 architecture, which will debut on the Radeon RX 7000 generation chips codenamed Battlemage. Another handful of information about upcoming graphics chips has been released on the web, which will lead to a significant increase in performance over current generations. In the case of NVIDIA, we’ll have a flagship graphics card that’s very power-hungry. AMD is planning an extensive GPU in a chipset architecture, while Intel may also offer a much larger system, consisting of four computing squares.
A batch of information has surfaced on the web about upcoming NVIDIA GeForce RTX 4000, AMD Radeon RX 7000, and 2nd generation Intel ARC graphics systems codenamed Battlemage.
NVIDIA GeForce RTX 4000 – Lots of information about possible specifications and performance of Ada graphics cards
The best NVIDIA graphics card (probably a GeForce RTX 4090 model, unless the manufacturer decides to change the label, as pointed out by Twitter user – kopite7kimi) is to use a board with the designation PG137 / PG139 and an Ada AD102-300 graphics core in a 144 SM block configuration and 18432 CUDA FP32 cores . Interestingly, according to information provided by kopite7kimi, NVIDIA decided to lower the speed of GDDR6X memory from 24 Gbps to 21 Gbps. This change also applies to lower-end systems, such as the GeForce RTX 4080 or GeForce RTX 4070. However, Kopite7kimi clearly indicates that the main card should have a TDP of 600W, indicating highly efficient but also highly polished graphics. The top card will again receive 24GB of VRAM on the 384-bit bus.
** 90 PG137/139-SKU330, AD102-300, 24Gbps 21Gbps, 600W
– kopite7kimi (@kopite7kimi) April 11, 2022
Two 5nm GCD
Four 6nm MCD
Single threaded controller
7 slices in total
– Greymon55 (@greymon55) April 11, 2022
Information about the competing AMD NAVI 31 graphics chip, in turn, comes from another Twitter user – @greymon55, who often shares information about the latest AMD components. According to the published content, NAVI 31 should have a fairly complex architecture, based on two main computing blocks GCD (graphics compound template) and four MCD (memory compound template) blocks – the latter likely containing memory controllers and/or storage memory Extended infiniti timer. The GCD computation matrices will be manufactured in TSMC’s 5nm lithography, while the MCD blocks will benefit from the company’s 6nm process. Therefore, NAVI 31 should consist of a total of seven chipsets, which will certainly be unique compared to other consumer GPUs.
In each configuration below, it will have the corresponding amount of FP32 ALUs.
1T: 5120 (+25% ACM SoC1/G10)
2T: 10240 (x2.5 ACM SoC1/G10)
4T: 20480 (×5 ACM SoC1/G10)
The tile 2 configuration is most likely as it currently stands.
– Redfire (@Redfire75369) March 30, 2022
Intel ARC A770M, ARC A730M, ARC A550M, ARC A370M, ARC A350M – Announcing Alchemist Portable Systems for Laptops
Something was revealed in the context of future Intel solutions. In this case, the leading representative of the Battlemage generation (2nd generation Intel ARC) can use a tiled architecture, similar to what Intel has already presented on the occasion of the Ponte Vecchio graphics accelerator. Battlemage’s top layout will use four arithmetic tiles. Each block will contain a 5120 FP32 processor, which is a 25% increase in the number of compute modules compared to the current ARC ACM-G10 chip (4096 SP FP32). Using four squares in the main graphics chip, the number of FP32 processors will be 20,480. However, we do not know right now when the Intel ARC Battlemage generation will appear on the market – in this case, we can wait until the end of 2023.
Source: VideoCardz, Twitter @Greymon55, @kopite7kimi, @Redfire
“Prone to fits of apathy. Introvert. Award-winning internet evangelist. Extreme beer expert.”