We have just published information on the specifications of the NVIDIA GeForce RTX 4000 Laptop GPU mobile graphics systems, using the Ada Lovelace architecture. We prepared this material before the start of the GeForce Beyond conference, as we received access to information about the parameters of individual GPUs a little earlier. In that material, however, we did not write anything about performance. However, we are already after the presentation of the company as part of CES 2023 and we report what has been announced in terms of performance in games and software, as well as new 5th generation Max-Q options.
Today, NVIDIA introduces the next generation of GeForce RTX 4000 graphics for notebooks. What should we expect from Ada Lovelace systems in mobile versions?
NVIDIA GeForce RTX 4090, RTX 4080, RTX 4070, RTX 4060, RTX 4050 – specification of Ada Lovelace graphics systems for laptops
NVIDIA boasts that the new Ada Lovelace architecture combined with the TSMC 4N technological process will make the energy efficiency of the GeForce RTX 4000 Laptop GPU generation in notebooks up to three times higher compared to the previous generation. During the GeForce Beyond presentation, it was shown in one example that the performance previously offered by the 120 W version of the GeForce RTX 3070 Laptop GPU will now be achieved on a much more energy-efficient Ada Lovelace system with 40 W. We will wait for the actual performance results, especially since NVIDIA in its materials, it focuses primarily on comparing games where the DLSS 3 function is available – as a reminder, the entire GeForce RTX 4000 Laptop GPU generation will support a new version of the upscaling technique, now seasoned with a functionality called Frame Generation.
GeForce RTX 4090 Laptop GPU | GeForce RTX 4080 Laptop GPU | GeForce RTX 4070 Laptop GPU | GeForce RTX 4060 Laptop GPU | GeForce RTX 4050 Laptop GPU | |
Architecture | Ada Lovelace | Ada Lovelace | Ada Lovelace | Ada Lovelace | Ada Lovelace |
Lithography | TSMC 4N | TSMC 4N | TSMC 4N | TSMC 4N | TSMC 4N |
Layout | AD103 | AD104 | AD106 | AD107 | AD107 |
Surface | 378.6mm² | 294.5mm² | ? | ? | ? |
Transistors | 45.9 billion | 35.8 billion | ? | ? | ? |
SM blocks | 76 | 58 | 36 | 24 | twenty |
CUDA cores | 9728 | 7424 | 4608 | 3072 | 2560 |
Gen 3 RT cores | 76 | 58 | 36 | 24 | twenty |
Generation 4 Tensor cores | 304 | 232 | 144 | 96 | 80 |
TMUs | 304 | 232 | 144 | 96 | 80 |
ROP units | 112 | 80 | 48 | 48 | 48 |
GPU boost clock | 1455MHz (80W) 2040MHz (150W) | 1350MHz (60W) 2280MHz (150W) | 1230MHz (35W) 2175MHz (115W) | 1470MHz (35W) 2370MHz (115W) | 1605MHz (35W) 2370MHz (115W) |
Power FP32 (max) | 38.9 TFLOPS | 33.8 TFLOPS | 20 TFLOPS | 14.6 TFLOPS | 12.1 TFLOPS |
VRAM | 16GB DDR6 | 12GB DDR6 | 8GB DDR6 | 8GB DDR6 | 6GB DDR6 |
Main line | 256 bit | 192 bit | 128 bit | 128 bit | 96 bit |
PCIe bus | PCIe 4.0 | PCIe 4.0 | PCIe 4.0 | PCIe 4.0 | PCIe 4.0 |
DLSS support 3 | Yes | Yes | Yes | Yes | Yes |
HDMI 2.1 | Yes | Yes | Yes | Yes | Yes |
NVIDIA codec | 8th Gen. Dual AV1 | 8th Gen. Dual AV1 | 8th Gen. AV1 | 8th Gen. AV1 | 8th Gen. AV1 |
TGP power | 80 – 150W +25W Dynamic Boost | 80 – 150W +25W Dynamic Boost | 35 – 115W + 25W Dynamic Boost | 35 – 115W + 25W Dynamic Boost | 35 – 115W + 25W Dynamic Boost |
Premiere on laptops | February 8, 2023 | February 8, 2023 | February 22, 2023 | February 22, 2023 | February 22, 2023 |
NVIDIA GeForce RTX 4070 Ti – official presentation of the graphics card with the AD104 core. We know its Polish price
For several generations, NVIDIA has strongly promoted a set of certain solutions included in the so-called Max-Q Design. In the previous generation, Max-Q ceased to be a separate entity, instead individual functionalities were integrated with all graphics systems, regardless of TGP. The Ada Lovelace architecture brings the 5th generation of Max-Q, where in addition to the already known functions, i.e. DLSS 2, Advanced Optimus, Whisper Mode 2.0, Dynamic Boost or Resizable BAR, new techniques will be added: DLSS 3, Ultra Low Voltage GDDR6, Tri-Speed Memory Control and Ada High Efficiency On Chip Memory. The introduction of lower-voltage GDDR6 memory and new dynamic power management mechanisms are to properly optimize the operation of graphics systems both on mains power and on batteries. NVIDIA additionally boasts a 20-fold increase in energy efficiency since the introduction of the first generation Max-Q in laptops. A noticeable increase in perf / watt is also expected to bring more notebooks with a screen diagonal of no more than 14 inches – the Lenovo Yoga Pro 14 model was mentioned at the show
Ultra Low Voltage GDDR6 is introduced closely with laptop manufacturers, which is why there are no specific data on the working parameters of GDDR6 cubes in individual devices. However, as the name suggests, laptops based on Ada Lovelace systems will offer the lowest voltages in the history of portable computers for GDDR6 bones to reduce their power consumption both at rest and under load. Tri-Speed memory control is designed to allow the GPU to dynamically transition to lower-power memory states. In turn, the cache memory built into the Ada Lovelace chips has been optimized for Max-Q, increasing e.g. throughput by up to 16 times the size of the Ampere architecture and improved GPU clock gating. A total of over 170 portable computers will be released this year, taking advantage of the capabilities of the GeForce RTX 4000 Laptop GPU. The flagship models with GeForce RTX 4080 and GeForce RTX 4090 Laptop GPU will be introduced first – their market debut is scheduled for February 8 with prices starting at $ 1999. Laptops with other Ada Lovelace systems will go on sale from February 22 at a price starting at $999 (GeForce RTX 4050 Laptop GPU).
Source: NVIDIA, PurePC