[GUEST ACCESS MODE: Data is scrambled or limited to provide examples. Make requests using your API key to unlock full data. Check https://lunarcrush.ai/auth for authentication information.]

[@lafaiel](/creator/twitter/lafaiel)
"Wonder how much better M5 will be for gaming Apple's gen X dynamic caching gave A19 a huge boost in games with complex shaders"  
[X Link](https://x.com/lafaiel/status/1978130541888016778) [@lafaiel](/creator/x/lafaiel) 2025-10-14T16:07Z 2599 followers, 5659 engagements


"@digitalix Apple's DGX spark literally It has temu tensor core same or higher bandwidth"  
[X Link](https://x.com/lafaiel/status/1978285459026378894) [@lafaiel](/creator/x/lafaiel) 2025-10-15T02:23Z 2598 followers, XXX engagements


"Recreated the graph because the previous one had errors. This one is accurate"  
[X Link](https://x.com/lafaiel/status/1978476487465738549) [@lafaiel](/creator/x/lafaiel) 2025-10-15T15:02Z 2600 followers, XXX engagements


"NVIDIA DGX Spark vs Apple M4 Pro Updated chart since previous one was wrong"  
[X Link](https://x.com/lafaiel/status/1978479957228880095) [@lafaiel](/creator/x/lafaiel) 2025-10-15T15:15Z 2600 followers, XXX engagements


"@Techtee7 @LeakerApple Samsungs market cap is about one-tenth of Apples yet it has 280000 employeesaround XX% more than Apple. Youre misunderstanding how business and market cap work: a bigger valuation doesnt mean a larger company in headcount or operational scale"  
[X Link](https://x.com/lafaiel/status/1979486935484928319) [@lafaiel](/creator/x/lafaiel) 2025-10-18T09:57Z 2623 followers, XX engagements


"@G_melo_ding what's the speed gain of N3P over N4"  
[X Link](https://x.com/lafaiel/status/1979451489358876945) [@lafaiel](/creator/x/lafaiel) 2025-10-18T07:36Z 2638 followers, XXX engagements


"Samsung's $2000 XR headset set to be revealed on the XX this month is about 4x slower in both CPU/GPU compared to the $3500 Vision Pro They'll never justify that $2000 price"  
[X Link](https://x.com/lafaiel/status/1978496665280507960) [@lafaiel](/creator/x/lafaiel) 2025-10-15T16:22Z 2643 followers, 2338 engagements


"Thinking back I was super skeptical when Apple started caring about Mac gaming at WWDC. Honestly I didn't expect them to get this far. They've made some real progress And I didn't think path tracing render would get this good. Now Apple's right up there with Nvidia"  
[X Link](https://x.com/lafaiel/status/1979709249656926280) [@lafaiel](/creator/x/lafaiel) 2025-10-19T00:40Z 2643 followers, XXX engagements


"Qualcomm just acquired Arduino and launched the new UNO Q powered by the Qualcomm QRB2210 Arduino has basically transformed from a simple controller board into a Linux-based edge-AI computer. The CPU is now hundreds of times faster with GPU and ISP added and it even includes a Qualcomm NPU for on-device AI. Through the new App Lab you can easily load built-in lightweight AI models like object or voice recognition and run them right on the board. It also jumps from a tiny XX KB SRAM to X GB LPDDR4 RAM runs Debian Linux and even supports Docker. So instead of the old XX MHz Cortex-M4 to read"  
[X Link](https://x.com/lafaiel/status/1975628071677141100) [@lafaiel](/creator/x/lafaiel) 2025-10-07T18:23Z 2630 followers, 13.2K engagements


"Hype around DGX Spark died like this With its limited memory bandwidth the results werent surprising. Even with a higher bandwidth the form factor wouldn't really allow for higher power draw or better cooling anyway It feels like a niche product and frankly Im skeptical whether NVIDIA can even supply enough GB10 chips. Yet somehow ASUS Dell HP and Acer all joined in on production makes you wonder if that was NVIDIAs pressure at play. Does NVIDIA actually plan to manufacture these in volume Im not so sure. But this is still interesting teaser for NVIDIAs first real consumer ARM SoC the N1x If"  
[X Link](https://x.com/lafaiel/status/1978305527949267001) [@lafaiel](/creator/x/lafaiel) 2025-10-15T03:42Z 2626 followers, 9738 engagements


"Ray tracing is XX% faster Probably thanks to 2nd gen dynamic caching. Might compete with NVIDIA-based laptops for Blender or Redshift 3D rendering. Cinebench GPU score will likely increase significantly"  
[X Link](https://x.com/lafaiel/status/1978485032831578142) [@lafaiel](/creator/x/lafaiel) 2025-10-15T15:36Z 2630 followers, 1362 engagements


"apple is a company that handles everything from product design to the kernel os app store ide programming languages compilers sdk and ux/ui literally the a to z of computing. unlike qualcomm the number of people they can dedicate to chip development is limited. their silicon team is much smaller than qualcomms"  
[X Link](https://x.com/lafaiel/status/1979490965900648824) [@lafaiel](/creator/x/lafaiel) 2025-10-18T10:13Z 2630 followers, XX engagements


"@Techtee7 @LeakerApple they have other things to perfect. also soc is delicate and crucial aspect for qualcomm too"  
[X Link](https://x.com/lafaiel/status/1979492364084519067) [@lafaiel](/creator/x/lafaiel) 2025-10-18T10:18Z 2632 followers, XX engagements


"what AI upscaling actually does"  
[X Link](https://x.com/lafaiel/status/1977849595058356683) [@lafaiel](/creator/x/lafaiel) 2025-10-13T21:31Z 2643 followers, 149.5K engagements


"Hold on the iPhone XX Pro has a 6-core GPU. That means latest iPhone is faster than Intel or AMD laptop processors at path traced render 🤯"  
[X Link](https://x.com/lafaiel/status/1979429002592030899) [@lafaiel](/creator/x/lafaiel) 2025-10-18T06:07Z 2643 followers, 10.6K engagements


"Its kinda wild that many AAA games on Mac/iPad/iOS are built on custom engines that had to be ported from scratch - really shows Apple and studio's effort Death Stranding (Decima) Assassins Creed (Anvil) Baldurs Gate X (Divinity) Hitman (Glacier) Cyberpunk 2077 (REDengine) Control (Northlight) Crimson Desert (BlackSpace) Resident Evil (RE Engine)"  
[X Link](https://x.com/lafaiel/status/1979444856767574267) [@lafaiel](/creator/x/lafaiel) 2025-10-18T07:10Z 2643 followers, 5164 engagements


"Btw Samsung calls it LPW"  
[X Link](https://x.com/lafaiel/status/1979701180524884441) [@lafaiel](/creator/x/lafaiel) 2025-10-19T00:08Z 2643 followers, XXX engagements


"@ISAQUES81 @ZONEofTECH Samsung's superb AI bbb"  
[X Link](https://x.com/lafaiel/status/1979723189040435316) [@lafaiel](/creator/x/lafaiel) 2025-10-19T01:36Z 2643 followers, XXX engagements


"@lackynamber1234 @vbvishappy Thats simply because your gaming laptop severely limits performance when running on battery. Its been old issue with x86 laptops"  
[X Link](https://x.com/lafaiel/status/1979787310754627992) [@lafaiel](/creator/x/lafaiel) 2025-10-19T05:50Z 2641 followers, XX engagements


"Samsung wants to make premium stuff but is stuck using low-end chips worse than $XXX budget smartphones. Why Like Meta and Chinese companies Samsung relies on Qualcomm and Qualcomm isn't releasing new chips"  
[X Link](https://x.com/lafaiel/status/1978496760541540855) [@lafaiel](/creator/x/lafaiel) 2025-10-15T16:22Z 2643 followers, XXX engagements


"Apple M5 that fits inside 5.1mm fanless iPad beats the M1 Ultra Quadro RTX 5000 and RX 9060 XT in Blender rendering performance"  
[X Link](https://x.com/lafaiel/status/1979420641125372187) [@lafaiel](/creator/x/lafaiel) 2025-10-18T05:33Z 2643 followers, 198.9K engagements


"And low power rivals like Intels Lunar Lake and AMDs Strix Point are crushed M5 (yet still not low power enough to fit in fanless tablet)"  
[X Link](https://x.com/lafaiel/status/1979425650160341364) [@lafaiel](/creator/x/lafaiel) 2025-10-18T05:53Z 2643 followers, 11.8K engagements


"is AMD's shitty performance bcs of AMD's lousy support or just hardware limits is it still low in Cinebench 2024 GPU too"  
[X Link](https://x.com/lafaiel/status/1979435988834488439) [@lafaiel](/creator/x/lafaiel) 2025-10-18T06:34Z 2643 followers, 9966 engagements


"Tesla is losing massive trust in Korea right now because of bms_a079 error. According to data Tesla Korea sent to a lawmaker XX% of the 2018 Model X had this error. For the 2020-2021 Model Y XX% of the cars had the problem. That's pretty much a vehicle defect. If your warranty is up the battery replacement costs $15K$25K. The main problem is you're not getting a new battery. Lately they've been swapping them with remanufactured batteries so some users are experiencing the exact same error all over again. What's crazy is that as of last month Korea was the 3rd largest market for Tesla in the"  
[X Link](https://x.com/lafaiel/status/1979479791184249009) [@lafaiel](/creator/x/lafaiel) 2025-10-18T09:28Z 2643 followers, 2045 engagements


"@unsatchmo To be fair the M5's memory bandwidth is 153GB/s which is lower than all the products in the screenshot except for the M3 Pro"  
[X Link](https://x.com/lafaiel/status/1979678481454883096) [@lafaiel](/creator/x/lafaiel) 2025-10-18T22:38Z 2643 followers, XXX engagements


"Just to add the mobile HBM mentioned there is called LLW DRAM and I believe it's currently supplied by SK Hynix and used in the Vision Pro. Both Samsung and SK Hynix are aiming for mass production in 2026. I wonder if it will be used in Chinese phones or Galaxy phones first or if it will be implemented in iPhones first"  
[X Link](https://x.com/lafaiel/status/1979687058898928114) [@lafaiel](/creator/x/lafaiel) 2025-10-18T23:12Z 2643 followers, XXX engagements


"LLW (mobile HBM) is a leading candidate for Apples 2027 iPhone; heres a brief overview LLW (aka LPW often called mobile HBM) still stacks memory on top of the AP like PoP-LPDDR but the connection method and bus philosophy are different PoP-LPDDR today : The memory package has tiny solder balls (BGA) on its underside. you heat-melt those balls to bond the memory to the AP package below. Smartphones typically use a 64-bit data bus. (btw package has actually hundreds of contacts because besides data (DQ) it also needs DQS/clock/command-address plus lots of power/ground pins) Energy per bit:"  
[X Link](https://x.com/lafaiel/status/1979699382909112748) [@lafaiel](/creator/x/lafaiel) 2025-10-19T00:01Z 2643 followers, 1284 engagements


"When Tensor first came out I liked Google's approach. You could tell from the interviews with Google engineers back then that they understood the benefits of having two big cores for load balancing and efficiency and they criticized designs that only used one big core. Unlike other competitors they didn't prioritize benchmarks and like Apple they focused on designing for the end-user experience. But looking at it now I'm a bit doubtful about whether the Tensor chip has any relative advantage in performance or efficiency even after switching to TSMC's N3E process"  
[X Link](https://x.com/lafaiel/status/1979821078437675278) [@lafaiel](/creator/x/lafaiel) 2025-10-19T08:05Z 2643 followers, XXX engagements