Montreal, August 12, 2026 — This has been one of those weeks where the plumbing of the computer industry — the parts most of us never see — made more news than any shiny gadget. At the Future of Memory and Storage summit in California, Samsung showed a memory architecture that abandons a fifty-year-old assumption about where memory belongs relative to the processor. Yesterday, Microsoft shipped one of the largest security updates in the history of Windows: roughly 421 vulnerabilities patched in a single day, including a zero-day flaw that attackers were already exploiting and a DNS bug that researchers describe as potentially wormable. And on the display front, the panel makers supplying the world's televisions confirmed that 2026 flagship OLEDs will chase a once-unthinkable 4,500 nits of peak brightness. In today's edition we scan the ten stories that mattered most this week, then dig deep into the science of three of them: 3D-stacked memory, memory-corruption vulnerabilities, and the physics of ultra-bright OLED. As always, we close with practical advice you can act on — including what's actually sitting in our warehouse today.

Today's Tech Radar — the 10 stories that mattered this week

Three of these stories reward a deeper look, because each one is really a science story wearing a business headline. Let's take them in turn.

Deep dive 1 — zHBM and the end of the memory wall: why Samsung is stacking memory on top of the processor

Modern memory packages sit millimetres from the processor — zHBM would close that gap to micrometres. Photo: Vishnu Mohanan / Unsplash.

The physics problem: data is slower than math

For decades, the story of computing has been told through processor speed. But the dirty secret of modern chips is that the processor is rarely the bottleneck — the memory is. A modern AI accelerator can perform hundreds of trillions of multiply-accumulate operations per second, but every one of those operations needs operands fetched from memory, and moving a bit of data across even a few millimetres of circuit board costs more energy and time than the arithmetic performed on it. Engineers call this the memory wall: compute performance has grown far faster than the bandwidth available to feed it. In physical terms, the problem is resistance and capacitance. Every signal trace between a memory die and a processor behaves like a tiny RC circuit; the longer the wire, the more energy is needed to charge it, and the longer each bit takes to arrive. Signal integrity degrades with distance, forcing slower clock rates or power-hungry error correction.

The industry's first answer to this was High Bandwidth Memory (HBM): instead of laying memory chips flat around the processor, DRAM dies are stacked vertically — currently 8 to 16 high — and connected through through-silicon vias (TSVs), microscopic vertical copper channels drilled straight through the silicon. The stack then sits beside the processor on a silicon interposer, a slice of silicon acting as an ultra-dense circuit board. That shortened the wires from centimetres to millimetres, and HBM became the memory of choice for every serious AI chip on the market.

What Samsung actually announced

At the Future of Memory and Storage 2026 summit (FMS), Samsung showed the logical endpoint of this trend: zHBM, a concept architecture in which the HBM stack is bonded directly on top of the logic die — the processor itself — rather than sitting next to it. According to reporting by TrendForce and The Korea Times, a next-generation interface built around zHBM is projected to deliver roughly eight times the performance of HBM5, more than ten times the memory density, around three times better energy efficiency, and less than half the thermal resistance of conventional designs. Samsung's own newsroom announcement frames zHBM alongside two storage-side siblings: zNAND-O, an optically connected NAND concept, and a V-NAND roadmap that has now crossed the 400-layer barrier, as Tech Times reported.

Why does stacking on top change the numbers so dramatically? Distance, again. When memory sits directly above logic, the connection path shrinks from millimetres of interposer wiring to micrometres of vertical bonding — modern hybrid wafer bonding fuses copper pads on two wafers face-to-face at pitches of just a few micrometres, creating tens of thousands of parallel connections per square millimetre. More parallel channels at shorter distances means enormous bandwidth at lower voltage. The engineering catch is heat: processors dissipate hundreds of watts, and DRAM cells hate heat (charge leaks out of the storage capacitors faster as temperature rises, forcing more frequent refresh). Placing memory directly on the hottest component in the system is only feasible with the reduced thermal resistance Samsung claims for its bonding stack — and it is precisely why zHBM remains a concept demonstration today rather than a shipping product.

Why 400-layer flash matters to you now

While zHBM targets data centres, the V-NAND announcement touches devices you can buy today. NAND flash — the storage in every SSD, phone and tablet — has been growing vertically since 2013, when planar (single-layer) flash hit its shrink limit. Each "layer" is a full sheet of memory cells; a 400+ layer chip is like a 400-storey apartment tower for electrons, built with deposition and etching steps accurate to a few nanometres. The payoff is more gigabytes per silicon wafer, which historically translates into cheaper and faster consumer storage roughly 18–24 months after the technology debuts.

Industry context and buying advice

The reason all of this is happening now is the same reason AMD just reported $11.5 billion in quarterly revenue, Intel is raising tens of billions, and TSMC's sales jumped 45%: AI workloads are memory-hungry in a way traditional software never was. Memory — not compute — is the scarce commodity, and every memory maker is racing to sell premium AI-class parts. Analysts have repeatedly warned this pulls capacity away from consumer DRAM and NAND, putting upward pressure on the price of the storage inside laptops, phones and SSDs.

What does that mean practically? If you have been putting off a storage upgrade, the physics and the market both argue against waiting. In our own inventory, the Samsung 990 PRO 1TB NVMe SSD (in stock) is a direct beneficiary of the V-NAND scaling story — a PCIe Gen4 x4 drive whose sequential throughput saturates the bus precisely because of the multi-layer, multi-plane parallelism described above. For portable storage, the ruggedized Samsung T7 Shield 2TB portable SSD (in stock) uses the same generation of stacked NAND behind a USB 3.2 Gen 2 interface. And if you want to see the mobile side of the memory story, the Samsung Galaxy S25 Ultra 256GB (in stock, limited quantity) pairs 12GB of LPDDR5X with fast UFS storage — the same short-wire, low-voltage design philosophy, scaled down to your pocket. Have a specific workload in mind — video editing, a NAS build, a fleet of laptops? Request a free quote from our team and we'll size the storage to the job.

Deep dive 2 — Patch Tuesday, dissected: 421 fixes, one exploited zero-day, and the anatomy of a use-after-free

August 2026's Patch Tuesday is one of the largest in Windows history — patching promptly is the cheapest security measure there is. Photo: FlyD / Unsplash.

The headline numbers

On August 11, Microsoft released fixes for approximately 421 vulnerabilities — SecurityWeek counts 421 CVEs, while BleepingComputer tallies roughly 400 flaws including three zero-days, one actively exploited and two publicly disclosed. Per Rapid7's analysis, the breakdown skews toward the two most dangerous categories: roughly 176 elevation-of-privilege flaws and about 110 remote-code-execution flaws, alongside dozens of information-disclosure, spoofing and denial-of-service issues. Around 62 fixes are rated critical.

The exploited zero-day: CVE-2026-68820, explained from first principles

The flaw attackers were already using in the wild, tracked as CVE-2026-68820, lives in afd.sys — the Ancillary Function Driver for WinSock, the kernel-mode driver that underpins the Windows Sockets API. In plain language: every application on Windows that talks to a network — your browser, your email client, a game — ultimately funnels its traffic through this driver, which is why a bug here is so valuable to attackers.

The bug class is a use-after-free, and it's worth understanding because it is one of the most common ways modern software fails. Programs constantly request small blocks of memory from the operating system, use them, and hand them back ("freeing" them). A use-after-free happens when code frees a block of memory but keeps a stale pointer to it — a forwarding address to a house that's been demolished. If an attacker can arrange for the operating system to build something new at that address (a technique called heap grooming), the stale pointer now points at attacker-controlled data. When the buggy code uses the pointer again, it executes on the attacker's terms. Because afd.sys runs in the kernel — the most privileged layer of the OS — a successful exploit converts an ordinary user account into full SYSTEM control. That is why this class of bug is typically chained with a browser or document exploit: step one gets the attacker onto the machine, step two (this bug) makes them its administrator.

The wormable one: why DNS flaws frighten professionals

The most alarming fix of the month may be CVE-2026-62878, a critical flaw in Windows DNS Server described in the Qualys review and CSO Online's coverage: a stack-based buffer overflow allowing remote, unauthenticated code execution with elevated privileges — no user interaction required. A buffer overflow is conceptually simple: a program reserves a fixed-size container for incoming data, and the attacker sends more data than the container holds, overwriting adjacent memory — including the "return address" that tells the processor what code to run next. Because DNS servers routinely accept queries from anywhere, and because servers can be made to query one another, a flaw like this is potentially wormable: exploit code that can spread from server to server automatically, the mechanism behind historic outbreaks like WannaCry. Security teams are treating internet-facing DNS servers as patch-first priorities this week. Rounding out the month, a maximum-severity improper-authorization flaw in SAP Commerce Cloud's Data Hub Adapter (CVE-2026-58231) reminds us the problem isn't only Microsoft's.

What a home user or small business should actually do

First, update now: Settings → Windows Update on every PC, and don't skip the reboot — kernel patches only take effect after one. Second, remember that aging hardware is a security problem in itself: machines that no longer receive OS updates are permanently exposed to everything discovered after their end-of-support date. If you're running a small office on older, out-of-support machines, replacing them is often cheaper than one incident response. Hardware can help, too: the phishing-resistant FIDO2 standard replaces guessable passwords with cryptographic keys, and the Kensington VeriMark Guard USB-C fingerprint key (in stock) adds that protection to any modern laptop. On the mobile side, a current device such as the Galaxy S25 Ultra (in stock) ships with Samsung Knox hardware-backed security and years of guaranteed OS updates ahead of it. Not sure where your setup stands? Our technicians audit and modernize small-business fleets every week — request a free quote from our team and we'll map your exposure and options, no obligation.

Deep dive 3 — The 4,500-nit OLED: the photonics of very bright pixels

Peak brightness determines how convincing HDR highlights look — especially in bright rooms. Photo: BoliviaInteligente / Unsplash.

What a nit is, and why 4,500 of them is a big deal

A nit (candela per square metre) measures luminance — how much light a patch of screen emits toward your eye. An ordinary office monitor produces 250–350 nits full-screen. High-dynamic-range (HDR) video is mastered assuming small, intense highlights: sun glinting off water, a streetlamp at night. The more nits a display can concentrate into those highlights, the closer the image gets to how the scene looked in reality. As FlatpanelsHD reported and TFTCentral confirmed, Samsung Display's 2026 QD-OLED TV panel peaks at 4,500 nits using newly optimized organic emitter materials, and LG Display's rebranded "Tandem OLED" lineup claims the same 4,500-nit ceiling with its Primary RGB Tandem 2.0 architecture — up from around 2,000 measured nits in the best TVs of just two years ago.

Two different roads to the same brightness

The two panel giants get there by different physics. QD-OLED (Samsung Display) uses blue OLED emitters as a pump light source, then converts part of that blue light into red and green using quantum dots — semiconductor nanocrystals whose emission colour is set by their physical size, a genuinely quantum-mechanical effect (smaller dot, bluer light). Quantum-dot conversion is highly efficient and yields exceptionally pure primary colours, which is why QD-OLED can combine the brightness of each RGB component for high perceived luminance. Tandem OLED (LG Display) instead stacks multiple complete emissive layers on top of one another — Primary RGB Tandem stacks the three primaries in independent layers — so each layer runs at lower current while their light output adds up. Lower current per layer matters because OLED materials age in proportion to the current driven through them; a tandem stack is therefore both brighter and longer-lived, and LG pairs it with an anti-reflection stack rated at just 0.3% reflectivity. One honest caveat from the same reporting: panel-level peak specs exceed what finished, calibrated TVs typically measure — last year's "4,000-nit" panel yielded TVs measured around 2,000 nits in practice. Treat 4,500 as the ceiling, not the showroom number.

Buying advice: brightness where it actually counts

For most rooms — and especially for business use, where a screen may run all day near a window — sustained full-screen brightness and glare handling matter more than a fleeting specular peak. That is exactly what commercial-grade panels are engineered for. In stock right now: the Samsung 55" Crystal UHD Signage QBC (in stock), a 4K panel rated for 16/7 continuous operation; its big brother, the Samsung QM85C 85" UHD (in stock), delivers 500 nits with a non-glare coating and 24/7 duty rating — the difference between a consumer TV and a display built to stay bright for years; and for the desk, the Samsung ViewFinity S8 32" 4K monitor (in stock, limited quantity) brings wide colour and 4K sharpness to photo and document work. Outfitting a boardroom, storefront or classroom? Request a free quote from our team — we'll match panel brightness, duty cycle and mounting to your space.

Glossary of the Week

Setup at a Glance

A note on laptops: our Samsung Galaxy Book 4 units sold through and are currently out of stock, so we have deliberately not recommended one today — that is our in-stock-first promise in action. If you need a laptop now, request a free quote from our team and we'll confirm live availability or arrange a special order.

That's the science behind this week's headlines: memory moving onto the processor, four hundred patched holes in the world's most-used operating system, and pixels bright enough to sting. If any of it raised questions about your own setup — an aging PC, a storage crunch, a screen that washes out at noon — request a free quote from our team and we'll turn today's theory into a working setup. See you tomorrow.

Sources & Further Reading

Samsung Newsroom — Next-Gen 3D Memory Vision at FMS 2026 · TrendForce — Samsung Showcases zHBM at FMS 2026 · The Korea Times — Samsung introduces zHBM · Tech Times — zHBM and 400-layer V-NAND · HotHardware — Samsung zHBM 3D memory concept · SecurityWeek — August 2026 Patch Tuesday: 421 CVEs · BleepingComputer — Patch Tuesday fixes 400 flaws, 3 zero-days · Rapid7 — Patch Tuesday August 2026 analysis · Qualys — August 2026 security update review · CSO Online — WinSock zero-day and SAP flaw · FlatpanelsHD — 2026 QD-OLED panel reaches 4,500 nits · TFTCentral — OLED TV panels to reach 4,500 nits in 2026 · FlatpanelsHD — LG Display rebrands as Tandem OLED · Tom's Guide — Made by Google 2026 preview · Tech Startups — Top Tech News, August 11, 2026 · Tom's Hardware — This week, August 7, 2026. Photos: Unsplash (free commercial license).