In addition to the awaited RTX 4070 Ti, Nvidia presented during CES 2023 the GeForce RTX 4000 family for notebooks, promising massive gains in performance and energy efficiency. Already fully revealed, with 5 models at this early stage, the debut solutions would provide laptops with performance similar to some of the more advanced cards in the RTX 3000 family, such as the RTX 3080, but consuming only 1/3 of the energy, while offering the advanced features of the new Ada Lovelace microarchitecture.
Nvidia’s new mobile GPUs are led by the unprecedented GeForce RTX 4090 — despite the name, the specs of that chip are far from the RTX 4090, but they are still impressive. With the AD103 chip, the same used in the RTX 4080 for desktops, the component has 76 Streaming Multiprocessors (SMs) enabled, composed of 9,728 CUDA cores, 76 RT Cores for Ray Tracing and 304 Tensor Cores for DLSS and other Artificial Intelligence resources.
Its consumption range is established between 80 W and 175 W (150 W + 25 W of Dynamic Boost), the same as the RTX 3080 Ti mobile, while clocks are modest for the architecture, reaching 2,040 MHz. With that, the RTX 4090 mobile could offer up to 39.7 TFLOPs of computational power, which, on paper, would put it on par with the RTX 3090 Ti and the RTX 4070 Ti for desktops. Although not confirmed, leaks already indicated that the flagship mobile of the new family could offer similar performance to the RTX 3090.
The technical sheet completes with 16 GB of GDDR6 VRAM, working on a wide 256-bit interface, but of unknown bandwidth, since the speed of these memories is not confirmed. Just below it is the GeForce RTX 4080 mobile, equipped with the AD104 chip (the same as the RTX 4070 Ti), enabled with 58 SMs (two less than the 4070 Ti) composed of 7,424 CUDA cores, 58 RT cores and 232 Tensor cores.
Consumption ranges from 60 W to 175 W (150 W + 25 W), while the clocks are more aggressive when reaching 2,280 MHz, thus offering up to 33.8 TFLOPs of computational power, a number close to that of the RTX 3080 Ti for desktops. The release ships 12GB of GDDR6 VRAM, on a 192-bit interface and, again, with unknown bandwidth, as speeds were not reported.
Nvidia made a point of separating these two models from the rest of the family, as a way of highlighting the very high level of performance they are capable of providing. According to the company, both would have the strength to play games on up to 3 4K monitors at 60 FPS, and can be up to 4 times faster than their predecessors, as long as using the generation of frames with AI from DLSS 3 — as always, you have to wait for the reviews to know the real capabilities of these GPUs.
Notebooks equipped with the new RTX 4090 and RTX 4080 mobile, such as models from the Alienware or ASUS ROG Strix Scar family, are expected to hit the market on February 8, with prices that should start at a steep US$ 1,999 (~R$ 10,925) . It is not possible to predict whether we will see these solutions arrive in Brazil, as for laptops, availability is linked to the device manufacturers’ plans.
RTX 4070, 4060 and 4050 with 5th Gen Max-Q Suite
Perhaps the most interesting part of the announcement, which should be of interest to most users, is the intermediate solutions, the new RTX 4070, RTX 4060 and RTX 4050 mobile. More balanced and very focused on cost-effectiveness, the trio would offer performance equivalent to the RTX 3080 (probably in its mobile version), but consuming only 1/3 of the energy.
Chip models are still unknown, but leaks suggest that the RTX 4070 will ship the unreleased AD106, while the RTX 4060 and RTX 4050 will use the also debuting AD107. More robust, the 4070 brings 36 SMs, composed of 4,608 CUDA cores, 36 RT Cores and 144 Tensor Cores, operating in a consumption range between 35 W and 140 W (115 W + 25 W of Dynamic Boost). Reaching clocks of 2,175 MHz, the solution would deliver up to 20 TFLOPs of computational power, something close to the RTX 3070 for desktops.
There is also 8 GB of GDDR6 VRAM, in a 128-bit interface, but of unknown speed and bandwidth. The GeForce RTX 4060, in turn, has 24 SMs, featuring 3,072 CUDA cores, 24 RT Cores and 96 Tensor Cores, in a consumption range of 35 W to 140 W (115 W + 25 W of Dynamic Boost). The clocks reach 2,370 MHz, guaranteeing up to 14.6 TFLOPs, a capacity similar to that of the predecessor, which makes room for an alert — as the architectures are different, it is not exactly possible to make a direct comparison.
The memory configurations are the same as the RTX 4070, with 8GB GDDR6 on a 128-bit interface and unknown speeds and bandwidth. Finally, the RTX 4050 debuts with 20 SMs, composed of 2,560 CUDA cores, 20 RT Cores and 80 Tensor Cores, working with consumption between 35 W and 140 W (115 W + 25 W of Dynamic Boost). Frequencies top out at 2370 MHz, providing up to 12 TFLOPs of computational power. There is also 6 GB of VRAM, in a 96-bit interface.
The three mid-range chips don’t draw attention just because of their very low consumption — Nvidia’s performance promises are impressive. In addition to positioning them as solutions prepared for gameplay at 1440P on Ultra with frame rates above 80 FPS, the giant guarantees a reduction in the rendering time of 3D models in Blender from 2.5 hours to just 10 minutes and enough performance to make 14-inch notebooks provide twice the performance of the PlayStation 5 at 1/5 the size.
Statistics suggest that even devices with the RTX 4050, and even low-thickness ultrabooks and much more limited consumption, will be able to play heavy titles with high quality without suffering. This possibility is complemented by the 5th generation Max-Q Technologies suite, which brings three innovations: GDDR6 memories with “the lowest voltage on the market”, a new Tri-speed memory controller that allows GPUs to dynamically increase or reduce RAM voltages, and the Ada Lovelace architecture’s own improvements.
The first notebooks with Nvidia’s intermediate solutions, such as the ASUS TUF Gaming F15 and A15, and the new Dell G15, have already been presented, and are scheduled to hit the market on February 22nd, with prices starting at US$999 ( ~BRL 5,460). Brazil should receive a significant number of laptops with these more basic chips, but there is still no information on price and availability.
Nvidia closed the announcement by highlighting some of the brands that have been confirmed to use the Nvidia GeForce RTX 4000 mobile family, including Acer, Alienware, Gigabyte, HP, Lenovo, MSI, Razer and Samsung (possibly with the Galaxy Book 3 family, leaked recently), in addition to ASUS and Dell.
Among the manufacturers of white-label devices, whose chassis is standardized and used by distributors around the world, CyberPower PC, Eluktronics, Hasee, PC Specialist 3XS by Scan and Schenker are confirmed — the Brazilian Avell is not mentioned, but can be included in the list. package for using the same Eluktronics chassis.
Nvidia GeForce RTX 4090 Mobile: Datasheet
- GPU: AD103
- Streaming Multiprocessors (SMs): 76
- CUDA Cores (Core): 9,728
- Tensor Colors (AI): 304
- RT Colors (Ray Tracing): 76
- Frequencies: 1455 MHz (Base), 2040 MHz (Boost)
- Compute power (FP32): up to 39.7 TFLOPs
- Memory VRAM: 16 GB GDDR6
- Memory interface: 256-bit
- Memory Speed: N/A
- Bandwidth: N/A
- Consumption: 80-150 W (175 W with Dynamic Boost)
Nvidia GeForce RTX 4080 Mobile: Datasheet
- GPU: AD104
- Streaming Multiprocessors (SMs): 58
- CUDA Cores (Core): 7,424
- Tensor Colors (AI): 232
- RT Colors (Ray Tracing): 58
- Frequencies: 1350 MHz (Base), 2280 MHz (Boost)
- Compute power (FP32): up to 33.8 TFLOPs
- Memory VRAM: 12GB GDDR6
- Memory interface: 192-bit
- Memory Speed: N/A
- Bandwidth: N/A
- Consumption: 80-150 W (175 W with Dynamic Boost)
Nvidia GeForce RTX 4070 Mobile: Datasheet
- GPU: N/A (possibly AD106)
- Streaming Multiprocessors (SMs): 36
- CUDA Cores (Core): 4,608
- Tensor Colors (AI): 144
- RT Colors (Ray Tracing): 36
- Frequencies: 1230 MHz (Base), 2175 MHz (Boost)
- Compute power (FP32): up to 20 TFLOPs
- Memory VRAM: 8GB GDDR6
- Memory interface: 128-bit
- Memory Speed: N/A
- Bandwidth: N/A
- Consumption: 35-115 W (140 W with Dynamic Boost)
Nvidia GeForce RTX 4060 Mobile: Datasheet
- GPU: N/A (possibly AD107)
- Streaming Multiprocessors (SMs): 24
- CUDA Cores (Core): 3,072
- Tensor Colors (AI): 96
- RT Colors (Ray Tracing): 24
- Frequencies: 1470 MHz (Base), 2370 MHz (Boost)
- Compute power (FP32): up to 14.6 TFLOPs
- Memory VRAM: 8GB GDDR6
- Memory Speed: N/A
- Bandwidth: N/A
- Consumption: 35-115 W (140 W with Dynamic Boost)
Nvidia GeForce RTX 4050 Mobile: Datasheet
- GPU: N/A (possibly AD107)
- Streaming Multiprocessors (SMs): 20
- CUDA Cores (Core): 2,560
- Color Tensor (AI): 80
- RT Colors (Ray Tracing): 20
- Frequencies: 1605 MHz (Base), 2370 MHz (Boost)
- Compute power (FP32): up to 12 TFLOPs
- Memory VRAM: 6GB GDDR6
- Memory Speed: N/A
- Bandwidth: N/A
- Consumption: 35-115 W (140 W with Dynamic Boost)
Source: WCCFTech
