@malika_andrews Your interview style is fire. The little subtle compliments to make the player feel like it’s a big moment. The beaming energy. Nobody on your level.
NBA Finals!
@kellabyte Yeap! We have a desktop app to facilitate syncing your sessions + skills. Also, before we ship, we'll put our entire OS up on github. Stay tuned!
I passed 50K followers couple weeks ago. When I returned in 2025 I was at 48K
Special moment for me cuz I don’t use Twitter to promote my name or any product
I have the best followers ever. I’ve over achieved in my career cuz of you teaching me & I will write a thank you soon
Nvidia NX1 – probably the most overhyped product of this summer
Nvidia NX1 is basically a laptop version of Nvidia GB10 aka Nvidia DGX Spark, which was already released last Autumn 2025. Both have identical amount of CPU cores (20), 6144 CUDA cores and share 256-bit memory bus with ~273 GB/s memory bandwidth when using 8333 MHz LPDDR5X.
Performance wise this is a good combination and very power efficient, but when it comes desktop use it is more or less similar to AMD Ryzen AI Max 395+. Gaming performance will likely be also very similar, but with the fact the AMD supports far more games natively thanks to native x86 instruction set, while Nvidia NX1 is of course ARM and most Windows games don’t yet support ARM.
NOTE: I personally own both AMD Ryzen AI Max 395+ and Nvidia GB10 computers and I like them both, but use for different purposes.
The gaming performance on AMD’s integrated 8060s iGPU is surprisingly good and still beating the very latest Intel offerings, with the only downside being that it doesn’t do ray tracing that well, other than that you can expect it to reach playable framerates (60 Hz or more) on 4k resolution on older games easily, plus still have enough underutilized GPU and CPU that you can run agentic AI loads same time while you are gaming.
I’m expecting Nvidia NX1 gaming performance vary per game: it has far more CUDA cores than Nvidia RTX 5070 mobile, in fact the same amount as the desktop variant of RTX 5070. This implies plenty of raw power, but it will be limited with slower memory bandwidth and possibly not so well optimized drivers thanks to using ARM instead of x86-64.
For AI users Nvidia NX1 or GB10 will be better than AMD AI Ryzen, because of slightly faster memory bus (250 vs 273 GB/s), but also far more compute. In terms of decode the difference is small, only about 12%, but prefill (prompt processing) should be 5 times faster or more. NX1 and GB10 should be able to handle image and video generation also much faster than AMD AI Ryzen and beating Apple M5 Pro as well, but Apple still holds crown in decode token/s speed thanks to higher memory bandwidth (token generation is memory constrained task).
The big question mark is the pricing. Leaked information suggests than NX1’s weaker siblings, which have less compute, will come with 32 GB or 64 GB RAM configurations and pricing will be 3000 or 4000 euros, which is pretty steep, considering you can still get Asus ROG Flow Z13 13" with 128 GB LPDDR5X for 3 399,99 €, which is cheaper than any Nvidia DGX Spark clone, which are usually 5000 euros or more. NOTE: These are Finnish prices, but similar ones could be found elsewhere in Europe.
Those people are expecting to get Nvidia DGX Spark equivalent computer with 128 GB RAM + 4 TB SSD, for less than 5000 euros in a nice laptop form factor with high quality display and keyboard as well, will be disappointed. The pricing will approach Apple M5 prices.
The prices of SSD and memory are going up. This means that most Nvidia NX1 laptop will likely feature a smaller SSD e.g. just 1 TB, and only 32 GB RAM. This could be perfect for normal users, but storage space could be tight for heavy gamers and RAM won’t be enough for AI users.
The bigger question mark is software. Currently the best software support for Nvidia GB10 comes from hobbyists and they have coding and releasing their software stack of DGX Linux (variant of Ubuntu 24.04 LTS). If Nvidia NX1 ships with Windows, it will be extra pain to get all the AI stuff work flawlessly with it. Dual boot to Linux could be good, but diskspace might be tight with 1 TB SSD.
Software wise Apple currently enjoys the most mature local AI stack. Thousands of hobbyists have optimized MLX kernels so well that even Nvidia DGX Spark is still well behind in optimization.
The big story here is that GPT 5.5 (high/xhigh) outperforms claude-opus-4.8 (max/xhigh) by 20.7% succeeding on 12 additional tasks!
More impressive: GPT is roughly half the cost and twice as fast.
OpenAI is back in the game. Overall, this competition is healthy for the industry. I'd love to see a third player rise to the top of the leaderboard!
dbt Fusion is the Python-to-Rust rewrite that started at SDF Labs before it was acquired by dbt Labs. After a huge refactoring effort in the last months, we are announcing dbt Core 2.0 – the open-source slice of the Fusion codebase that is continously updated w/ Copybara.
The best database providers of the future won’t be metal vs object storage.
Betting your entire future on a single storage orientation is not how this industry will play out.
Someone will build a truly multi tiered database and this will become standard like CPU L0/L1/L2/L3
One thing I admire about Jensen is he positions his company to excel in many orientations.
He doesn’t bet that all AI compute belongs in the cloud. He doesn’t bet that all AI compute belongs local.
He ships Vera CPU to data centres.
He ships Spark CPU to homes.
@cwisecarver I think simple page caches strategies are not enough because that means you must make several requests before your set is in cache memory and it only speeds up repeated requests.
High paying customers expect their dollars to provide perf on the first request.
@cwisecarver I feel like some things are key:
- Writes need to perform like NVMe up until some threshold until it can’t (NVMe filled? Or eviction to object storage)
- This is a big one: high tier paying customers paying for high perf data needs to primarily be served from NVMe
@2big2faill I’m talking about present day. PC users were given almost the exact same message about the Surface and Snapdragon era.
But I totally agree with you! If the 10 wide IPC architecture from Vera trickles down it’s a similar trick Apple innovated on with the M series IPC.
In minutes I believe the Nvidia N1X ARM CPU is being presented as a “new era”.
I’m a little confused why in comparison to some 2-3 year old CPUs.
Maybe I’m missing something though?
@Jefflovesleafs@sidseixeiroshow@Sid_Seixeiro Yeah Rich Eisen is maybe one of the few folks still with a pleasant to watch and analytical sports show.
Also guests still want to visit because it’s thoughtful but not toxic.
@Jefflovesleafs@sidseixeiroshow@Sid_Seixeiro Barely anyone talks about the actual games anymore. Or coaching strategies or plays or decisions.
ESPN first take half of the time is top 10 players of all time debates. Did this micro event change the GOATS?!
Fox’s The Herd is Colin acting like a GM. 5 mins about the game.
@Jefflovesleafs@sidseixeiroshow@Sid_Seixeiro I can only imagine what it’s like living or playing there. Most sports media is in TO obviously so we all see it.
But who would want to grow, fail, improve in that market? Very few.
Also not really original content to just be mad and rage at poor performances.
@davepl1968 If I was Windows CEO I think it’s time for Windows X research project.
It feels like a Unix world now. Windows is always getting open source tools last.
DirectX and Windows on Linux. No more WSL. No more hacks running Linux VMs under the hood.