Recommended reading

GeForce MX550 Edges Over Ryzen 9 APU In New Benchmark

News
By published

Vega is still going strong

Announced in December of last year, the GeForce MX550 is Nvidia's upcoming entry-level graphics card for mobile devices. Hardware detective Tum_Apisak has unearthed the first benchmark of the GeForce MX550, showing its potential performance compared to the existing market competition.

Despite the official announcement, a lot of mystery still surrounds the GeForce MX550. From what Nvidia has confirmed so far, the GeForce MX550 still leverages the prior Turing architecture. Surprisingly, the GeForce MX550 utilizes the TU117 silicon, the same die inside the GeForce MX450. For reference, the GeForce MX450's performance is somewhere up the GeForce GTX 1050 Mobile's ballpark.

We don't know much about the GeForce MX550 other than that it supports PCIe 4.0 and incorporates GDDR6 memory. Even though the graphics card has the TU117 die at heart, it doesn't support Nvidia DLSS or ray tracing, leading us to suspect that the Tensor and RT cores are disabled. Featurewise, the GeForce MX550 supports Resizable BAR and Nvidia Optimus. However, the more recent Advanced Optimus technology is off the table.

GeForce MX550 Benchmarks

Swipe to scroll horizontally
Graphics CardAverage G3D Mark# of Samples
GeForce GTX 1650 Mobile6,9691,765
GeForce MX5505,0141
Vega 8 (Ryzen 9 5900HS)4,9689
Radeon RX 5500M3,968103
GeForce MX4503,724215

It's important to note that the GeForce MX550 submission represents a single entry in the PassMark Software database. The results for the other graphics cards are based on the average scores from multiple samples, including stock and overclocked parts. Therefore, we recommend you take the GeForce MX550 result with a grain of salt.

In terms of a generation-over-generation uplift, the GeForce MX550 delivered up to 35% higher performance than the GeForce MX450. However, the GeForce GTX 1650 Mobile was 39% faster than the GeForce MX550.

Compared to AMD, the GeForce MX550 put up a 26% higher score than AMD's Radeon RX 5500M. Unluckily, PassMark Software didn't have any data on the Red Team's more recent Radeon RX 6000M (RDNA 2) graphics cards. However, the GeForce MX550 barely beat the Ryzen 9 5900HS Zen 3 mobile APU, which wields eight Vega compute units at 2,100 MHz. The margin was less than 1%.

The GeForce MX550 won't earn a spot on the best graphics cards for gaming. However, that's okay since Nvidia's GeForce MX series positions itself as a slight upgrade over integrated graphics. In AMD's case, the Radeon Vega graphics engine continues to be a competitive performance in the mobile space.

See more GPUs News
Zhiye Liu
Zhiye Liu
News Editor and Memory Reviewer

Zhiye Liu is a news editor and memory reviewer at Tom’s Hardware. Although he loves everything that’s hardware, he has a soft spot for CPUs, GPUs, and RAM.

6 Comments Comment from the forums
  • King_V
    "Edges" out is right. That's less than a 1% difference. As it says, only one sample.

    But this performance difference, if the MX550 score valid across multiple samples, represents a difference well within margin of error for performance comparisons, no?
    Reply
  • foolproof
    The Mx450 was a great card but really expensive and difficult to find in the US. Whenever I googled MX450 laptops I'd end up on Amazon India or other Indian shops. This card will probably face the same issue
    Reply
  • artk2219
    King_V said:
    "Edges" out is right. That's less than a 1% difference. As it says, only one sample.

    But this performance difference, if the MX550 score valid across multiple samples, represents a difference well within margin of error for performance comparisons, no?

    Yep. Margins of error are usually within 2 - 3%, with any notable difference really needing to be 10% or above in most cases. Honestly id rather have the APU and the increased battery life that comes with it, its not enough of a difference to be worth it as a discrete GPU.
    Reply
  • Umfriend
    And for the umpteenth time things like this make me wonder what a quad-channel APU would do for GFX performance. I realise the mem-controller and MB (due to traces) would become more expensive but what you would save is dGPU real estate to, you know, add like a 4th M.2 slot in a 14"chassis. A storage and mempory bandwith monster, relatively speaking.

    Edit: I also wonder whether the increased cost of making a quad-channel APU is higher than the cost of including an MX550.
    Reply
  • artk2219
    Umfriend said:
    And for the umpteenth time things like this make me wonder what a quad-channel APU would do for GFX performance. I realise the mem-controller and MB (due to traces) would become more expensive but what you would save is dGPU real estate to, you know, add like a 4th M.2 slot in a 14"chassis. A storage and mempory bandwith monster, relatively speaking.

    Edit: I also wonder whether the increased cost of making a quad-channel APU is higher than the cost of including an MX550.

    AMD could totally make a chip like that, design the ecosystem to go with it, and then have the OEM's give it a hard pass, because they're trying to sell as many cheap PC's as possible, which an increased cost to produce does not Jive with. Hell, AMD has a hard enough time getting OEM's to produce products with their current chips as is, personally I think that would be an excellent idea, and it will likely not see the light of day for the reason above.
    Reply
  • Umfriend
    artk2219 said:
    AMD could totally make a chip like that, design the ecosystem to go with it, and then have the OEM's give it a hard pass, because they're trying to sell as many cheap PC's as possible, which an increased cost to produce does not Jive with. Hell, AMD has a hard enough time getting OEM's to produce products with their current chips as is, personally I think that would be an excellent idea, and it will likely not see the light of day for the reason above.
    You are probably right. On the other hand, I would not be surprised if it would outperform APU+low end dGPU solution and end up cheaper because. well,. you don;t need a dGPU and accompanying memory. If only I were a billionaire, I'd make it happen (although, if I were a billionaire, I'd probably be doing other things entirely ;)|).
    Reply