Nvidia's RTX 40-Series Laptops Don't Bode Well for RTX 4060, 4050 Desktop GPUs

Nvidia RTX 40-series laptops
(Image credit: Nvidia)

Nvidia's Ada Lovelace architecture ushers in a new level of performance at the top of the stack, with the RTX 4090 besting the previous generation RTX 3090 Ti by 52% on average in our rasterization benchmarks, and 70% in ray tracing benchmarks — both at 4K, naturally. The 4090 now sits comfortably atop our GPU benchmarks hierarchy and ranks as one of the best graphics cards around, at least if you have deep pockets.

Unfortunately, the step down from the 4090 to the RTX 4080 is rather precipitous, dropping performance by 23% for rasterization and 30% for ray tracing. Dropping down another level to the new RTX 4070 Ti knocks an additional 22% off the performance relative to the 4080. If you're keeping track — and we definitely like to keep score — that means the third-string Ada card sporting the AD104 GPU is slower than the previous generation 3090 Ti, nevermind Nvidia's claims to the contrary that rely on benchmarks using DLSS 3's Frame Generation.

Perhaps more alarming with the RTX 4070 Ti is that it only has a 192-bit memory interface. It still has 12GB of GDDR6X memory, and the large L2 cache in general means that the narrower bus isn't a deal killer, but things don't look so good as we eye future lower-tier RTX 40-series parts like the 4060 and 4050.

Nvidia recently announced the full line of RTX 40-series laptop GPUs, ranging from the RTX 4090 mobile that uses the AD103 GPU (basically a mobile 4080) down to the anemic-sounding RTX 4050. Here's the full list of specs for the mobile parts.

Swipe to scroll horizontally
Nvidia Ada Laptop GPU Specifications
Graphics Card RTX 4090 for Laptops RTX 4080 for Laptops RTX 4070 for Laptops RTX 4060 for Laptops RTX 4050 for Laptops
Architecture AD103 AD104 AD106? AD106? AD107?
Process Technology TSMC 4N TSMC 4N TSMC 4N TSMC 4N TSMC 4N
Transistors (Billion) 45.9 35.8 ? ? ?
Die size (mm^2) 378.6 294.5 ? ? ?
SMs 76 58 36 24 20
GPU Shaders 9728 7424 4608 3072 2560
Tensor Cores 304 232 144 96 80
Ray Tracing "Cores" 76 58 36 24 20
Boost Clock (MHz) 1455-2040 1350-2280 1230-2175 1470-2370 1605-2370
VRAM Speed (Gbps) 18? 18? 18? 18? 18?
VRAM (GB) 16 12 8 8 6
VRAM Bus Width 256 192 128 128 96
L2 Cache 64 48 32 32 24
ROPs 112 80 48 32 32
TMUs 304 232 144 96 80
TFLOPS FP32 (Boost) 28.3-39.7 20.0-33.9 11.3-20.0 9.0-14.6 8.2-12.1
TFLOPS FP16 (FP8) 226-318 (453-635) 160-271 (321-542) 91-160 (181-321) 72-116 (145-233) 66-97 (131-194)
Bandwidth (GBps) 576 432 288 288 216
TDP (watts) 80-150 60-150 35-115 35-115 35-115
Latest Videos FromTom's Hardware
Jarred Walton
Senior Editor

Jarred Walton is a senior editor at Tom's Hardware focusing on everything GPU. He has been working as a tech journalist since 2004, writing for AnandTech, Maximum PC, and PC Gamer. From the first S3 Virge '3D decelerators' to today's GPUs, Jarred keeps up with all the latest graphics trends and is the one to ask about game performance.

  • bigdragon
    The RTX 40-series continues to be a disappointment. I'm surprised how close together the laptop 4070, 4060, and 4050 are. The specs make it look like there won't be a meaningful difference in performance unless you step up to the 4080 -- a GPU that likely will only appear in the oversized, gaudy, overpriced gamer laptops.

    Hopefully some manufacturer puts the 4070 in a 2-in-1 device, sells it in the USA, and prices it like other gaming laptops instead of putting it in some weird premium category. However, given what Nvidia has done on pricing, I'm not sure I want to know how expensive this year's systems are going to be.
    Reply
  • PlaneInTheSky
    There are plenty of games now that can exceed 8GB of VRAM use.

    Isn't this just because the memory is available. Many games just try to saturate the graphical memory with textures and shading code, because it's there so why no use it. A 12GB GPU might show 12GB in use, but that doesn't mean it can't run with 8GB.

    I haven't seen any games that need more than 8GB of VRAM at 1080p, and 1080p would be kind of the target resolution for a 4050 no.

    Steam shows 64% of Steam users are playing in 1080p, 11% plays in 1440p and only 2% at 4k.

    To be honest, I could be wrong though, I don't follow every AAA game release. Needing more than 8GB VRAM at 1080p, just seems like...a lot.
    Reply
  • PlaneInTheSky
    I looked up some of the VRAM requirements of 2022 PC games with "demanding" graphics (subjective).

    Slim pickings in 2022, I really had to dig to find any "AAA" 3D PC games from 2022.

    But still, none of them required anything more than 8GB VRAM, even for recommended specs.

    The 4050 having "just" 8GB VRAM seems like a non issue?

    VRAM requirements

    A Plague Tale: RequiemMinimum: 4GB VRAM
    Recommended: 8GB VRAM

    Elden RingMinimum: 3GB VRAM
    Recommended: 8GB VRAM

    Need for Speed UnboundMinimum: 4GB
    Recommended: 8GB

    StrayMinimum: 1GB
    Recommended: 3GB
    Reply
  • thestryker
    The only game that leaps to mind which can use over 8GB (without mods etc) at 1080p is Doom Eternal. 8GB really needs to be the starting point on desktop though and the bus width that nvidia is playing with doesn't really leave them any options. Given that their charts still show the 30 series covering lower end I'm not sure nvidia is even planning lower desktop parts any time soon.
    Reply
  • ravewulf
    Alternatively, the laptop 4060 and 4050 might both be using AD107 but the laptop 4050 gets its bus cut down to 96-bit from the full 128-bit. This would mean the desktop versions may have higher specs, as NVIDIA is already doing with the other models.
    Reply
  • bigdragon
    PlaneInTheSky said:
    I haven't seen any games that need more than 8GB of VRAM at 1080p, and 1080p would be kind of the target resolution for a 4050 no.
    Have you considered the scenario of VR use? The 4050 is likely to be advertised as VR-ready just like the 3050 was. VR games can readily eat up tons of VRAM. While the user's primary display may be 1080p, there is a strong possibility of them plugging in a Quest headset or similar to experience VR with the power PC computing can provide.

    Furthermore, I don't trust big game studios to keep making games that fit within 8 GB VRAM. People said the same thing about 3 GB VRAM in the Radeon HD 7950 and 7970 being enough -- it wasn't!
    Reply
  • PlaneInTheSky
    Furthermore, I don't trust big game studios to keep making games that fit within 8 GB VRAM.

    Game developers will just target what the average user has, like they always do.

    Of the top 12 GPU in use by Steam users, only 1 out of those 12 has more than 8GB VRAM.
    The average VRAM available is 6.2 GB.

    Steam's most popular user GPU:
    GTX 1650: 4 GB VRAM
    GTX 1060: 3 GB - 6 GB VRAM
    RTX 2060: 6 GB VRAM
    GTX 1050 Ti: 4 GB VRAM
    RTX 3060 (laptop): 6 GB VRAM
    RTX 3060: 12GB
    RTX 3070: 8 GB VRAM
    GTX 1660 SUPER: 6 GB VRAM
    RTX 3060 Ti: 8 GB VRAM
    GTX 1660 Ti: 6 GB VRAM
    GTX 1050: 4 GB VRAM
    RTX 3050: 8 GB VRAM
    Reply
  • JarredWaltonGPU
    thestryker said:
    The only game that leaps to mind which can use over 8GB (without mods etc) at 1080p is Doom Eternal. 8GB really needs to be the starting point on desktop though and the bus width that nvidia is playing with doesn't really leave them any options. Given that their charts still show the 30 series covering lower end I'm not sure nvidia is even planning lower desktop parts any time soon.
    From my test suite, I can assure you that Far Cry 6, Forza Horizon 5, Total War: Warhammer 3, and Watch Dogs Legion all have performance fall off at 4K with 8GB, sometimes badly. Those are just four games out of 15. At 1440p, I think most of them all do okay, but I know a few other games in the past year have pushed beyond 10GB. I think Resident Evil Village and Godfall did, for example. And yes, you can turn down some settings to get around this, but the point is that we've got 12GB on a 3060 right now, and that feels like it should be the minimum for a mainstream GPU going forward. Anything using AD106 will presumably be limited to 8GB... the same 8GB that we had standard on R9 290/290X and GTX 1070 over six years ago.

    As far as Steam goes, it tracks a very large group of users, many of whom only play on older laptops, stuff like DOTA and CSGO that will run on old hardware. The fact that the top 12 is dominated by lower cost GPUs isn't at all surprising. Fortnite hardware stats would probably look the same. But there are also roughly 5% of gamers that do have GPUs with more than 8GB, and more critically, most GPUs don't have more than 8GB. Like, the list of hardware with more than 8GB currently consists of:

    RTX 4090
    RTX 4080
    RTX 4070 Ti
    RTX 3090 Ti
    RTX 3090
    RTX 3080 Ti
    RTX 3080 (both 12GB and 10GB)
    RTX 3060
    RTX 2080 Ti
    RTX 1080 Ti
    RX 7900 XTX
    RX 7900 XT
    RX 6950 XT
    RX 6900 XT
    RX 6800 XT
    RX 6800
    RX 6750 XT
    RX 6700 XT
    RX 6700
    Radeon VII
    Arc A770 16GB

    Of those, only the 3060 and 6700 series are remotely mainstream, and Arc if we're gong to count that. But the point is we should be moving forward. Going from 8GB on a 256-bit interface to 8GB on a 128-bit interface will at times be a step back, or at best a step sideways.
    Reply
  • PlaneInTheSky
    I strongly agree with the comment about DLS 3.0.

    They're not real frames, and they introduce a delay. It adds half a frame of lag to delay the last frame and add in that interpolated frame in between.

    A lot of players with lower end GPU play at relatively low, or very low, framerates.

    For someone with a 4080 playing at 120 fps, half a frame delay is not a big deal. It's 1/240th of a second. But for someone with a low-end GPU playing at 30 fps or below, it is a big deal. It is extra lag you simply don't really want if your FPS is already low.

    The lower your framerates, the more milliseconds of lag DLSS 3.0 introduces. When a game temporarily drops to 10 fps, and you delay that latest frame even more, that just tanks response time. It's like putting the chains on an already struggling GPU.

    At 30 fps DLSS 3.0 adds 1/60th of an input delay. At a struggling 10 fps DLSS 3.0 adds 1/20th of a second of delay, that's an extra 50 milliseconds of lag, and you're definitely going to notice that, stutters in games will seem much worse.

    I truly hope Nvidia is not going to pull out the "DLSS 3.0" fanfare if they make a desktop 4050 or 4060, because that extra bit of lag DLSS 3.0 introduces is negligeable at 120fps, but is a problem at 30fps and below.
    Reply
  • thestryker
    JarredWaltonGPU said:
    From my test suite, I can assure you that Far Cry 6, Forza Horizon 5, Total War: Warhammer 3, and Watch Dogs Legion all have performance fall off at 4K with 8GB, sometimes badly. Those are just four games out of 15. At 1440p, I think most of them all do okay, but I know a few other games in the past year have pushed beyond 10GB. I think Resident Evil Village and Godfall did, for example. And yes, you can turn down some settings to get around this, but the point is that we've got 12GB on a 3060 right now, and that feels like it should be the minimum for a mainstream GPU going forward. Anything using AD106 will presumably be limited to 8GB... the same 8GB that we had standard on R9 290/290X and GTX 1070 over six years ago.
    Yeah, but you're unlikely to have the graphics power to run any of those games well at 4k with lower end 40 series which is why I specified 1080p (though I'd bet most will be 1440p capable). We're probably still going to see 6GB desktop cards too which just shouldn't be happening anymore. I certainly agree it's a problem and it is only going to get worse as we move along.

    JarredWaltonGPU said:
    As far as Steam goes, it tracks a very large group of users, many of whom only play on older laptops, stuff like DOTA and CSGO that will run on old hardware. The fact that the top 12 is dominated by lower cost GPUs isn't at all surprising. Fortnite hardware stats would probably look the same. But there are also roughly 5% of gamers that do have GPUs with more than 8GB, and more critically, most GPUs don't have more than 8GB. Like, the list of hardware with more than 8GB currently consists of:

    RTX 4090
    RTX 4080
    RTX 4070 Ti
    RTX 3090 Ti
    RTX 3090
    RTX 3080 Ti
    RTX 3080 (both 12GB and 10GB)
    RTX 3060
    RTX 2080 Ti
    RTX 1080 Ti
    RX 7900 XTX
    RX 7900 XT
    RX 6950 XT
    RX 6900 XT
    RX 6800 XT
    RX 6800
    RX 6750 XT
    RX 6700 XT
    RX 6700
    Radeon VII
    Arc A770 16GB

    Of those, only the 3060 and 6700 series are remotely mainstream, and Arc if we're gong to count that. But the point is we should be moving forward. Going from 8GB on a 256-bit interface to 8GB on a 128-bit interface will at times be a step back, or at best a step sideways.
    Do you think this capacity issue with a low bit bus can be solved with Samsung's GDDR stacking technology they'd announced some weeks back?
    Reply