Nvidia Ada RTX 40 Laptop GPU analysis: 175 W RTX 4090 Laptop GPU surpasses a 350 W RTX 3090 in performance and power efficiency

The RTX 4090 Laptop GPU can be configured in a TDP range between 80 W and 150 W, which results in boost clocks ranging from 1,455 MHz to 2,040 MHz.

Learning from the Ampere fiasco, OEMs are now proactively indicating the GPU wattage in the specifications which should give buyers a better perspective of performance relative to the laptop's configuration and chassis.

However, do note that not all OEMs publish Dynamic Boost values. 

The RTX 4090 Laptop GPU uses the same 256-bit memory bus as the desktop RTX 4080 but suffers from a ~20% reduction in memory bandwidth albeit being slightly faster than the RTX 3080 Ti Laptop GPU.

The power-hungry nature of GDDR6X means that it cannot be used in a laptop form factor. Nvidia said that the GDDR6 VRAM in Ada mobile is designed to be highly efficient with improved clock gating, low voltage, and added lower power memory states.

The RTX 4080 Laptop GPU uses the desktop RTX 4070 Ti's AD104 die with a slight reduction in CUDA core counts that matches with what we had seen before in the RTX 3080 Ti Laptop GPU.

Although the RTX 3080 Ti Laptop GPU seems better on paper, at least in terms of VRAM characteristics, the RTX 4080 mobile makes up for this deficiency with architectural improvements in Ada.

The Ada generation also brings with it some notable changes to Nvidia's Max-Q implementation. When Max-Q first debuted with Turing, the focus was more towards the TDP rating alone. The cards themselves were distinctly labeled as Max-Q, which made selection easier.

However, from Ampere, Nvidia chose to put both the press and end-users in a conundrum by removing the Max-Q branding and instead using it as an umbrella term for technologies such as Whisper Mode, Resizeable BAR, Dynamic Boost, Advanced Optimus, and the likes.

OEMs now have the choice of using whichever subset of Max-Q technologies they deem fit, and also fit in a large GPU in a thin chassis for bragging rights whilst actually configuring its TDP near base values.

With Ada, Nvidia is introducing what it calls 5th gen Max-Q in which DLSS 3 and low-power GDDR6 memory play an integral part.

We've touched briefly upon DLSS 3 and new ray tracing functions in Ada in our review of the RTX 4090 Founders Edition. Essentially, DLSS 3 uses Ada's new standalone Optical Flow Accelerator (OFA) to insert frames on the fly. The frame generation happens on the GPU and not in the game.

More precisely, DLSS 3 is a combination of existing DLSS 2 Super Resolution and frame generation. According to Nvidia, DLSS 3 reconstructs 7/8th of the image — 3/4th of the first frame is reconstructed with DLSS 2 Super Resolution while the complete second frame is created with frame generation.

Since frame generation happens at the GPU, DLSS 3 can help boost frames in CPU-bound or physics-heavy titles like Microsoft Flight Simulator 2020.

Note that DLSS 3 requires an RTX 40 GPU, supported title, and enabling Hardware accelerated GPU scheduling and Nvidia Reflex.

Note: We are the first to receive an early variant of the Schenker XMG Neo 16. Since this is a pre-production model, the scores may not be reflective of the device's actual performance in our official review with a retail sample.

Cumulative 3DMark performance sees the RTX 4090 Laptop GPU in both the Titan GT77 HX and the XMG Neo 16 performing on similar lines, indicating no potential bottlenecking elsewhere in the pipeline.

The Razer Blade 16's RTX 4080 Laptop GPU trails by 19% compared to the RTX 4090 Laptop, which is similar to the percentage delta between the desktop variants of these GPUs.

That being said, the RTX 4090 Laptop GPU is up to 35% slower than the desktop RTX 4090 Founders Edition card while trailing by up to 26% compared to the Radeon RX 7900 XTX. However, its performance is comparable to that of the desktop RTX 4070 Ti.

Despite featuring a similar number of CUDA cores, TDP limitations mean that the 320 W RTX 4080 Founders Edition is 20% faster compared to a 175 W RTX 4090 Laptop GPU in overall 3DMark performance.

Compared to previous generation mobile GPUs, the Ada generation seems to offer significant performance uplifts. The RTX 4090 Laptop GPU is a whopping 76% faster in all 3DMark tests taken together than the average RTX 3080 Ti Laptop GPU. 

Good 43% gains are also seen with the RTX 4080 Laptop GPU compared to the average RTX 3080 Ti mobile variant. 

A similar trend is seen in gaming as well, wherein the RTX 4090 Laptop GPU in the XMG Neo 16 is within striking distance of the RTX 4070 Ti desktop in cumulative gaming performance.

We see that the RTX 4080 Laptop GPU trails the RTX 4090 mobile variant by up to 17%, which is similar to the deltas we see between the desktop versions of the RTX 4090 and the RTX 4080.

Architecture improvements seem to confer the RTX 4080 Laptop a 33% advantage over the RTX 3080 Ti Laptop GPU in overall gaming performance, but the RTX 4090 offers a whopping 61% gain over Ampere's mobile flagship.

Just like their desktop counterparts, RTX 40 mobile GPUs also seem to get easily CPU-bottlenecked at lower resolutions. Metro Exodus (non-Enhanced), for instance, sees the RTX 4090 Laptop GPU trailing the desktop RTX 4070 Ti by up to 19% at 1080p Ultra. However, at 4K Ultra, the flagship Ada mobile GPU performs on par with the RTX 4070 Ti.

We also see widening deltas between the RTX 4090 Laptop GPU and the RTX 4090 desktop card at 4K Ultra compared to 1080p or QHD Ultra for this reason — Ada simply loves to be taxed at 4K.

Interestingly, the gap between the RTX 4080 Laptop GPU and the RTX 4090 Laptop GPU also increases at 4K.

Exceedingly CPU-bound games like Mafia: Definitive Edition show perceptible differences between the RTX 4070 Ti and RTX 4090 Laptop GPU only at 1080p. As the resolution increases, the RTX 4090 Laptop GPU is able to close the gap or even beat the RTX 4070 Ti by a whisker in this game.

Despite sporting the same number of RT cores as the RTX 4080 desktop, the RTX 4090 Laptop GPU can be up to 25% slower in overall ray tracing performance. However, we see excellent gains up to 58% compared to the RTX 3080 Ti Laptop GPU.

While we see very good fps numbers at native 1080p, 4K native ray tracing at playable frame rates is still not within the realm of possibility as seen in the Cyberpunk 2077 results. Nvidia is pushing DLSS 3 as the only viable means of playing at higher resolutions with RT enabled.