Prost from the Hofbräuhaus 🍻

My best investment the entire year so far

New video: youtu.be/g3xpRXD0I30

Next-gen Intel microchips are almost build like lego. I love the modular tile-based design of Panther Lake & Clearwater Forest. Video coming soonTM.

Next-gen ASML wafers are smaller, but very yummy.

The one and only OG 🍏

🍏 I did it, and it wasn’t even planned. Upgraded from a SE2 to a 17 Pro 👇

New Apple 🍏 toys, this time M3 & M4. Now only M5 is missing in the collection. If someone already dropped their new M5 iPad Pro or MBP, let me know 🫣

I counted the Intel 18A FOUPs per second (FPS) and calculated the yield. It’s definitely playable ;) #Inteltechtour

New video: How AMD is re-thinking Chiplet Design youtu.be/maH6KZ0YkXU

Apple A19 SoC chip analysis based on images by ChipWise: 2 P-cores with 8MB shared L2$, 4 E-cores with 4MB shared L2$, a 8-core NPU (Apple calls it 16-cores), 2x 6MB System Level Cache (SLC) and a 5-core GPU. All on TSMCs N3P and smaller than the previous A18.

MediaTek claims their new 3nm Dimensity 9500 SoC has the worlds 1st "compute-in-memory" NPU. If that's actually real CIM, this is a breakthrough.

Rubin CPX looks very similar to GB202, at least in the image Nvidia released. Could be very well a full GPU, with raster engines, PCI-Expess and display outputs. While unlikely, it might be how the RTX 6090 will look like.

Over 600K views in seven days is crazy and a new high score for me 🚀 Thanks for the support and to everyone who is watching!

My autocorrect switches “table” to “TSMC”. I think I might be too deep in the hole…

New video collab with SemiAnalysis youtu.be/dhqoTku-HAA

I’ve got exciting news! From now on you’ll see regular collabs between HighYield and SemiAnalysis to bring you more content & deeper insights into everything silicon. No worries, High Yield won’t change, still running my own show. There will be plenty of general semiconductor content like always.

New video: Intel Lunar Lake Close-Up youtu.be/pMhHBbKtWkE

Something is cooking 🍎

My last video just crossed 150k views, 75 of which came from YouTube notifications. Just a FYI why subscriber count doesn't mean anything anymore.

What a nice "birthday" present 🙌

New video: The most important company you've never heard of youtu.be/hDo5P578wJI

New setup - next video coming soon! Sorry for the long wait, I moved to a new city. PS: ignore the strange bald dude

ASML's High-NA EUV scanner might be the most complex machine in the world.

TSMC is building a new chip design center here in Munich! It's supposed to open in Q3, so quiet soon. If anyone has a contact at TSMC, I'd be happy to visit! #TSMC

Got a new toy working on my upcoming Strix Halo deep-dive. It's crazy how fast the Flow Z13 is! Crushes my desktop 5800X3D, even on battery.

Living the dream at the imec FAB in Leuven (this might be an experimental 0nm 🧇)

Just released by @fritzchensfritz.bsky.social Arrow Lake-S die shots projected on the full package. Full resolution: www.flickr.com/photos/13056...

The definitive Intel Arrow Lake deep-dive (pictures by @fritzchensfritz.bsky.social) youtu.be/wusyYscQi0o

Just one more teaser...🏹🌊👀

Question: what are you looking at? Bonus question: can it run Crysis?

Soon™ 🏹🌊👀

In light of today' news, here's a high-res die shot of AMD's Venice CPU. Courtesy of @fritzchensfritz.bsky.social

AMD taped-out Zen 6 "Venice" in TSMCs N2 process node, which drops FinFETs and instead used new Gate-All-Around transistors. It usually takes around 1 year from tape-out to release, so Zen 6 could in theory be a 2Q 2026 product! www.amd.com/en/newsroom/...

RX 9070 XT – RDNA4 Transistor Secrets youtu.be/u8cfrJTdo0E

Happy end! The new 25.3.2 optional driver seems to fix the problem. Now I'm getting 140+ FPS (with hair strands on high)! Used DDU again to reinstall. Didn't work with the 25.3.1 though. Maybe my system doesn't like this one.

I'm getting very low FPS in RE:4 with my RX 9070 XT. Playing in UWQHD w/o RT and I'm at 60-80 FPS, sometimes dipping below 60. I've played it before on a RX 6800 (same settings) and got more FPS. CPU is a 5800X3D. According to reviews I should be at >100FPS. Any ideas?

Just found my old sysprofile.de account, created in Summer of 2006. What a trip on 💾 lane!

Anyone knows what AMD means with "Infinity Fabric" on the Navi 48 schematic? How is Infinity Fabric used there?

Nvidia's GB300 is massive with four Blackwell Ultra GPUs (288GB HBM3e each), two Grace CPUs (512GB SOCAMM LPDDR5X each, 480GB usable) and four Connect-X 8 SuperNICs. (pic by Ryan Shrout)

If one Nvidia Blackwell is now "two GPUs" because it has two dies, does that mean one AMD EPYC Turin is 16 CPUs, because it has 16 CCDs? My take: Jensen knows how much the next-gen chips will cost, so he's preparing the "but you get multiple GPUs" narrative.

I can't upload video over 60s here on Bluesky, so here's a link to my post on Twitter: x.com/highyieldYT/... A guide on how to enable FSR4 in Cyberpunk 2077 on a RX 9070 (XT) RDNA4 GPU with OptiScaler. Spoiler: Full Path Tracing with >60 FPS on 1440p@FSR4 Perf. is possible!

I couldn't help but notice that in this slide from AMD's RDNA4 presentation, FSR4 AI Upscaling seems to be active on a RX 7800 XT 🧐

Initial Navi 48 chip analysis. AMD really outdid themself when it comes to density. Lot's of transistors in a small space.

With 53.9 billion transistors on just 356.5mm², AMD's Navi48 achieves 151 million transistors per mm². On TSMCs N4. That's crazy and I think more than what even TSMC claims is possible.

Nvidia should start making 4090s again. I'd buy a new one for $1500 over a $2500 5090 any day.

Nice.

I'm so sorry, but it had to be done 😥

This is how I'm spending my Sunday. #silicongang

#AMD Vega 64: the last gaming GPU with HBM2 memory. 2x 4Hi stacks, each with 4GB and a 1024-bit interface for a total of 8GB @ 2048-bit. Radeon VII was already “prosumer” focused.