Walter T. Baugh

Cerebras Systems, RANOVUS Inc. win $45 million US Army deal to speed up chip connections. Andrew Feldman Jean-Philippe Fricker Gary Lauterbach Sean Lie Andy Hock Dhiraj Mallick Bob Komin Naor Penso Vinay Srinivas Hagay Lupesko Natalia Vassilieva Angela Yeung Hamid Arabzadeh Georg Roell #us #usarmy #cerebras #invest #chip #ranovus https://lnkd.in/eW9WTgEG

2

Most of us carry mobile phones and really don’t understand the technology with in. Here is an insight into a portion of it….. Welding the aluminum wires on the edges that reliably (without failing over time) connect this design to the I/Os of the package is a challenge that most don’t understand.

3

Spot on, Marin. This is the DeepSeek playbook applied to the quantum stack. It’s an ecosystem land grab disguised as open-source generosity. By giving away the integration layer, they aren't just winning users—they are capturing the foundational abstractions that will dictate how hardware speaks to software for the next decade. My advice to any lab or university tempted by the "free" price tag: approach with extreme caution. Review this only on a segmented VLAN or, ideally, a completely air-gapped environment. When an OS sits at the systems-integration level, it has total visibility into your QPUs, cryostats, and control electronics. If that OS is "calling home," you aren't just losing data; you’re surrendering the integrity of your entire physical stack. We know the cardinal rule of the digital age: If the product is free, your data (and in this case, your sovereignty) is the payment. We need to establish a Western "Sovereign Floor" for quantum integration before "free" becomes the global default. It differs little from the recent CBS interview regarding the crossing of that "red line"—once you lose control of the supply chain and the integration layer, you've already lost the territory. Enough thinking. Let’s build.

1

MP Materials - 8-K Confusion I was trying to understand a bit more about the "partnership" between MP and DoD, particularly the definition of the "NdPr" base price of $110/kg. So I had to get a copy of the 8-K document filed with the SEC - and that's where the confusion starts because there have been several versions filed with the SEC, the first one I found having a July 9th date. I don't have the time or the inclination to compare the different versions so just remember that you might find one that is different. I think all versions are on the MP website. Anyway in Exhibit 10.6 there are a number of definitions including the "Asian Metal Market Price" - which is to be the reference price - is the mid market price for a ton of NdPr Oxide. So that would suggest oxide is the reference $110/kg price. But then there is a definition of "NdPr Calculation Principles" which is how they come up with the NdPr Kilogram Equivalents included in the various "Sold Products". So maybe its $110/kg equivalent - which would make sense if those products contain some level of NdPr. The formulas for this - which would be very helpful to check are in a "Schedule A" which is not part of the document. If you have found a stray Schedule A, let me know where I can find it. In "NdPr Products" various price definitions have the letters (w), (x), (y), and (z) - some of which are redacted which doesn't help with clarity. Another puzzling section in this long paragraph says that MP can source feedstock (concentrate perhaps) from a third party, but what they sell won't count as a "Sold product" if the NdPr Kilogram equivalent (Kge) cost in that feedstock was more than $110. This suggests they could buy in oxide for say $80 Kge and then DoD would guarantee them $110 if they just sold it again at $80 - or even at a loss. Interesting business model. If you can help me with my confusion I would appreciate it greatly.

1

1 Comment

AM vs. FM Antennas: The Hidden Tech Behind Your Radio 📻⚡ Did you know AM and FM radio signals use completely different antennas? The choice depends on frequency, range, and even terrain! Here's a breakdown of how these antennas work and why engineers choose them: AM Antennas: Built for Distance 1️⃣ Monopole (Vertical) Antenna Tall structure (often 100+ feet!) Pros: Excellent long-range coverage, simple design Cons: Huge size, needs proper grounding 2️⃣ Loop Antenna Compact coil used in portable AM radios Pros: Small and affordable Cons: Weak signal for distant stations 🔹 Why it matters: AM’s low frequencies (kHz range) need massive antennas but can travel hundreds of miles, especially at night! FM Antennas: Optimized for Clarity 1️⃣ Dipole Antenna Classic "T-shaped" design Pros: Great for local FM reception, easy to install Cons: Limited range (line-of-sight) 2️⃣ Yagi-Uda Antenna Directional "arrowhead" shape (boosts signal in one direction) Pros: High gain, reduces interference Cons: Complex setup, not omnidirectional 3️⃣ Ground Plane Antenna Vertical rod with horizontal radials Pros: 360° coverage, compact Cons: Lower gain than Yagi 🔹 Why it matters: FM’s high frequencies (MHz range) deliver crisp audio but need more antennas to cover the same area as AM. Key Differences Feature AM Antennas FM Antennas Frequency Low (kHz-MHz) High (MHz) Size Very large Compact Range Long-distance Local/regional Design Monopole, loop Dipole, Yagi Why This Matters for Engineers AM: Ideal for emergency broadcasts and rural areas (but needs real estate!). FM: Perfect for high-fidelity music in cities (though requires more towers). Next time you tune in, remember: the antenna is half the magic! 🎶 #EverythingRF #DynamicEngineers #RadioTech #AntennaDesign #WirelessCommunication #STEM

1

Your obsession with MHz is a distraction. Throughput is what matters, not clock speed. We see a 200 MHz MCU and a 100 MHz MCU and assume the 200 MHz one is “better”. That assumption is lazy and often wrong. What if the 200 MHz chip takes 10 cycles to execute an instruction, while the 100 MHz chip does it in 4? The “slower” chip is actually faster. I have seen a 48 MHz Cortex-M0+ outperform a 180 MHz Cortex-M4 on a very real, I/O-heavy task because the M0+ had a more efficient GPIO toggle path and a faster bus to the peripheral. The raw clock speed was irrelevant. Stop staring at the marketing number on the front of the datasheet. Look at the benchmarks. Look at instruction cycle counts for the operations you actually care about. Look at bus architecture, peripheral latency, and memory access patterns. A slower, more efficient architecture will routinely beat a faster, bloated one in the real world. 🔥 What is the best example you have seen of a “slower” chip completely outperforming a “faster” one in a real application? #EmbeddedSystems #FirmwareEngineering #Microcontrollers #RealTimeSystems #PerformanceEngineering #EmbeddedSoftware #HardwareDesign #EngineeringReality #BenchmarksOverMarketing #SystemsThinking

341

41 Comments

US Approves $110M Sale to Expand Iraq’s Military Satellite Communications The US has approved a $110-million sale to expand Iraq’s military satellite communications network. Washington’s approval covers additional Very Small Aperture Terminals (VSATs) along with modems, hubs, L-band tactical satellite manpacks, spare parts, and long-term technical support. The package builds on an earlier foreign military sale valued at $46 million that was implemented below the congressional notification threshold. Rather than introducing new weapons, the deal is aimed at strengthening Iraq’s ability to move information securely across Iraq’s Ministry of Defence. 

18

1 Comment

Contrary to the myth, there is a smartphone assembled and sold in the USA running a Pure FOSS operating system and BIOS by a privacy-focused company and not by Big Tech.

1

Interesting, if NVIDIA, Apple, AMD, etc are already filling up TSMC’s capacity, not sure how designing their own chip helps unless they use a different foundry. Wouldn’t it be something if they made a deal with Intel to use 14a!!! I bet they plan on using Grok 4 to help with chip design. https://lnkd.in/gFVZGtUW

18

1 Comment

Another great White Paper, by our team member, Prashant Singh. 🚀 Designing high-bandwidth switch fabric SoCs? Socionext America’s latest white paper is a must-read for engineers pushing the limits of data center performance. Discover how to: *Minimize implementation risks *Achieve first-pass silicon success *Accelerate time-to-market with simulation-driven design Packed with insights on signal integrity, SerDes optimization, and thermal management, this guide is your blueprint for building next-gen SoCs that scale to 102.4 Tbps and beyond. #Wearehiring #Semiconductorjobs #SFBayArea #TPM #ASICdesign #AdvancedASICsolutions #PhysicalDesign #PackageDesignEngineers #DataCenterMarketing

5

First world problem.... my gigabit ethernet is the bottleneck in my internet speed: Gemini: 2. The "Over-Provisioning" Gain (~200+ Mbps) This is the secret most people don't know. Most ISPs (like Comcast/Xfinity, Spectrum, AT&T) secretly "over-provision" their plans by about 20% to ensure you never complain about getting less than advertised speeds. A "1 Gbps" plan is often actually configured to allow ~1200 Mbps at the modem level. With 1 Gbps Gear: You never see this bonus speed because your physical port caps you at 940 Mbps. You are leaving ~260 Mbps on the table. With 2.5 Gbps Gear: You suddenly see speed tests of 1150–1200 Mbps. The Catch: The "Chain" Rule To get this extra speed, every single link in your local chain must be 2.5 Gbps capable. If just one device is 1 Gbps, the whole system slows down to match it. You need: Modem: Must have a 2.5 GbE (or higher) port. (e.g., Arris S33, Netgear CM2000). Router: Must have a 2.5 GbE WAN port AND a 2.5 GbE LAN port. (Many routers only have it on the WAN side—check carefully!). Switch (if used): Must be a 2.5 GbE switch. Mac/PC: Must have a 2.5 GbE adapter (Your Mac mini M2 has a 10GbE option if you upgraded, or standard 1GbE. If you have the standard 1GbE model, you'd need a USB-C to 2.5GbE dongle). Summary: Switching to 2.5 GbE is the only way to get everything you pay for (plus the free buffer) on a gigabit plan.

On this day in Telcom history. 1947 – Invention of the Transistor On December 23, 1947, John Bardeen, Walter Brattain, and William Shockley at Bell Telephone Laboratories demonstrated the first solid-state transistor, a tiny semiconductor device that could amplify and switch electronic signals. This invention revolutionized electronics and laid the foundation for modern computers, mobile phones, and all digital communications technology. #telecom #history #movingmountains #letsgo

3

1 Comment