Support NeoGAF

[SolidQ]

a 3050 at 9.098/2=4.55 TFlops outperforming the 9.754 TFlops 5700 XT RDNA 1 by 13.5%

Fucking Nvidia magic

No magic there 5700XT, just demolish RTX 3050

 

[GigaBowser]

DF full hypocritical here. Dave absolutely calling them out.

metaldave just put DF outa busirness ahahahahaha

 

[FalconPunch]

metaldave just put DF outa busirness ahahahahaha

Username: Gigabowser, video creator username: SuperMetalDave64..... Biased users seeks out biased news source..... Could we really expect anything less from Nintendo "fans"? I think not.

 

[Luminoth-4545]

Looking ahead and assuming there is a Switch 3 and Nintendo stays with Nvidia, I think Switch 3's Gpu architecture will be based on the eleventh generation (Feynman).

Nvidia mobile GPUs have progressed through several generations, each with its own name and features.

• Kepler (Apr 2012):
  The first generation, known for its high performance and efficiency.
  
  
• Maxwell (Feb 2014)
  The second and third generations, offering improvements in power efficiency and performance.
  
  
• Maxwell 2.0 (Sep 2014)
  The Third generation, increased cache size and the introduction of new technologies like Dynamic Super Resolution and VR Direct.
  
  
• Pascal (May 2016):
  The fourth generation, introduced with the GeForce 10 series, notable for its support for DirectX Raytracing.
  
  
• Volta (May 2017):
  The fifth generation, known for its AI capabilities and performance in professional applications.
  
  
• Turing (Sep 2018):
  The sixth generation, marked by the introduction of RT and Tensor Cores for real-time ray tracing and AI acceleration.
  
  
• Ampere (May 2020):
  The seventh generation, featuring improved ray tracing and AI capabilities, and is the architecture for the GeForce RTX 30 series.
  
  
• Ada Lovelace (Oct 2022):
  The eighth generation, further refining ray tracing and AI capabilities.
  
  
• Blackwell (Mar 2024):
  The ninth generation, expected to introduce significant advancements in compute capability and potentially the largest architectural redesign since the GeForce 256.

The next generation of NVIDIA GPU architecture after Blackwell is codenamed Rubin, named after astronomer Vera Rubin. It is expected to be released in the second half of 2026. Following Rubin, NVIDIA's next-next GPU architecture, codenamed Feynman, is planned for release in 2028.

Here's a more detailed breakdown:

• Rubin (2026):
  This architecture will be used in AI GPUs and is expected to see significant performance improvements, including a move to HBM4 memory for increased memory bandwidth.
  
  
• Feynman (2028):
  While Rubin will be a significant step forward, Feynman is expected to bring even further advancements.

 

[mnkl13]

Curious. I plan to run a lot of tests on the machine, punish a review, etc. All of these will earn me a small amount of money to offset the cost of the device.

You thought you had something there…. Sorry to rain on your parade.

ohh really, why running some tests if you already know everything about it. you already reviewed it here for us. no, no, i still have something here... because if that is really te reason to preorder it, it's even worse than i thought

 

[FalconPunch]

ohh really, why running some tests if you already know everything about it. you already reviewed it here for us. no, no, i still have something here... because if that is really te reason to preorder it, it's even worse than i thought

Lol, touch grass bud.

 

[MrA]

No magic there 5700XT, just demolish RTX 3050

well it does have twice the rops and tmus , there is more to a graphics card than just floating point performance