If you've been on the fence with whether to buy a NAS (network-attached storage) or not—let me show you a different way. ORICO's latest 9948RC3 PRO DAS (direct-attached storage) system offers great transfer speeds, RAID, and more, all without a pesky network connection or custom operating system to deal with, and I actually really enjoyed using it.

Buying used enterprise storage is one of those things that makes perfect sense when you look at the price and becomes more complicated once you actually put the hardware into a machine. Enterprise HDDs that originally cost a small fortune can appear on the used market for surprisingly little money. For a homelab, where storage capacity is always useful and budgets are rarely unlimited, it is difficult to ignore those prices.

Proxmox is remarkably forgiving when you use it the way it was designed to be used. With it, you can have a functioning homelab in an afternoon, but that ease and simplicity can also encourage a lot of mistakes.

A NAS is one of the cheapest and most convenient ways to keep your data backed up. It’s also fairly reliable, but while HDDs and SSDs can last for years, they will fail eventually. When they do, you could lose everything stored on them, so here are some cheap ways to back up your NAS before that happens.

WinRAR is probably one of the most common third-party apps found on countless Windows PCs around the world, and for good reason. While it's technically paid software, you can use it for as long as you like without paying a dime, it handles a wide selection of archive formats, and it resides in the Windows context menu, so you don't even have to open it to use it.

Storage vendors have been calling things “unified” for years. Sometimes that means one architecture. Sometimes it means several products sharing a management screen and a marketing budget. The term has accumulated enough baggage that a little skepticism is warranted. HPE Alletra Storage MP B10000 Release 6, announced in May, is now generally available. It brings block and file onto the same disaggregated scale-out architecture and OS code base, with independent scaling of performance and capacity and native ransomware detection across both. That’s important because enterprise storage is becoming less of an array decision and more of an architectural one. HPE was recently named a Leader in the 2026 Gartner® Magic Quadrant™ for Enterprise Storage Platforms, which provides useful context, but a quadrant is still a snapshot. GA code is where the architecture has to prove itself. Unified, but not universal The B10000 is not intended to become the answer for every file workload in the data center. Massive-scale, performance-oriented unstructured workloads such as AI, analytics, EDA and media remain better suited to HPE Alletra Storage MP X10000. The aim with the B10000 is narrower and, arguably, more useful: consolidating block and adjacent file workloads onto a common architecture without forcing customers to scale controllers and capacity together. That distinction matters. Vendors have offered “unified” storage before, but the harder question is what is actually shared underneath: protocols, scaling model, software stack and operational experience, or simply the management layer. The architecture is the product The B10000 separates compute from capacity with a “shared-everything” architecture, allowing each to scale independently rather than requiring customers to buy fixed controller-and-media increments. Release 6 extends that model across block and file while retaining a common operating environment and management plane. The interesting part is therefore not simply the addition of file protocols. It is whether customers can consolidate block and the file workloads that naturally sit alongside them without recreating the infrastructure silos they were trying to remove. Disaggregation also changes the economics. If performance and capacity requirements grow at different rates, the infrastructure can follow those curves separately rather than making one drive the other. AI moves closer to operations Release 6 also extends HPE’s AI-driven operations model. Within HPE Data Services Cloud Console, autonomous health assessments can evaluate areas such as capacity and performance, while Storage Copilot provides natural-language interaction, recommendations and the ability to take actions. Agentic Support Automation pushes that model further. Unlike legacy support approaches that depend on known issue signatures, predefined rules or thresholds, it is signature-less: AI agents continuously analyze operational behavior to detect emerging anomalies, investigate likely causes and help drive remediation—even when the specific issue has not been seen before. Capacity exhaustion is a useful example. Total system capacity can appear stable and generate no alert while a single service begins consuming capacity at an unusual rate. Traditional monitoring may not react until a threshold is crossed; agentic support automation can recognize the behavioral shift earlier, connect it to the underlying operating condition and help initiate corrective action before the change becomes a disruption. That is probably the more useful test of the increasingly fashionable word “agentic.” The question is not whether AI appears in the interface, but whether the software can move from describing an infrastructure problem toward helping resolve it. Cyber resilience becomes layered Ransomware defense follows much the same architectural logic. The B10000 now provides native real-time detection across block and file, while Zerto adds protection at the application layer and StoreOnce provides integrated data protection and recovery. No single feature solves ransomware. The more interesting model is multiple detection and recovery mechanisms operating at different points in the stack. Storage competition moves beyond the array None of this means the array is dead. Someone still has to store the bits. But the competitive unit is getting larger. Controllers, media and IOPS still matter. Increasingly, so do the operating model, cyber-resilience architecture, scaling economics and the ability to support multiple data types without multiplying infrastructure silos. The term “unified storage” has been stretched pretty far over the years. What matters now is whether the unification extends below the management layer and into the architecture itself. For more information, visit HPE Alletra Storage MP B10000 web page. Contributed by HPE

Most people shouldn't worry about the majority of BIOS settings, since leaving them at their default values makes the most sense. This applies to most SSD-related BIOS settings as well. That said, there's a chance the previous owner messed with them if you bought a used PC, or that one friend who likes to tinker with PCs tweaked your motherboard's BIOS settings without your knowledge.

The relentless rise in memory prices may finally be losing steam, as PC and smartphone buyers reach the limits of what their budgets will allow them to pay. Taiwanese market watcher Trend Force estimates that rising DRAM contract prices drove a 59.5 percent quarter-on-quarter jump in industry revenue to $154.73 billion in calendar Q2. The analyst now expects conventional DRAM contract price growth to "moderate" to 13 to 18 percent quarter-on-quarter in Q3. That reflects a shift in demand away from high-capacity RDIMMs toward lower-capacity products, as well as "the limited ability of PC and smartphone customers to absorb further price increases." That 13 to 18 percent is still a punishing increase by any normal standard. Supplier inventories remain at historic lows, TrendForce warns, and total bit shipments are expected to grow modestly. In other words, prices are climbing more slowly and supplies are likely to remain tight for the forseeable future. The forecast follows a warning from market intelligence biz Context that European PC shipments are set to decline even more sharply as rising component costs push up system prices. Context forecasts a year-on-year decline in laptop shipments of 6.4 percent in Q3, followed by a 20 percent fall in Q4. The outlook is even worse for desktops, with shipments expected to decline around 20 percent in Q3 and almost 30 percent in Q4. Despite declining shipments, PC makers have continued to bring home the bacon, as higher prices have so far more than compensated for the fall in volumes. Earlier this year, The Register reported how Lenovo boasted it had sidestepped the fallout from the memory crisis by shifting its focus to premium devices. Context says corporate buyers are extending refresh cycles where possible and acquiring new PCs only when upgrades are unavoidable. "The PC refresh cycle has not disappeared, but the economics around it have changed," said Context senior analyst Marie-Christine Pygott. "For the channel, the opportunity is increasingly about helping customers prioritize where refreshes are essential, where devices can remain in service for longer and where investment in newer, more capable systems makes commercial sense." As for the AI PCs that manufacturers hoped would trigger a mass upgrade wave, Context says the end of Windows 10 support was the principal driver of recent buying activity, as The Register noted at the start of this year. While AI-capable PCs account for an increasing share of systems entering the European supply chain, this is mostly because AI functionality is becoming standard in newer devices. TrendForce notes that the three major memory makers continue to focus on high-end products for the AI market. Samsung benefited from its early move into HBM4 mass production, while SK hynix had the highest proportion of HBM in its total bit shipments among the three leading suppliers. Micron is prioritizing higher-priced server DRAM. That has left second-tier manufacturers such as Nanya, Winbond, and PSMC to meet demand for mature-process products such as DDR4 and DDR3. ®

Deleting files is easy, but making sure those files cannot be recovered later requires a different approach. A normal file deletion operation usually removes filesystem references while leaving the underlying data in place. That allows operating systems to delete files quickly, but it also means recovery software can sometimes retrieve information from previously used storage space.

I’ve been running a Plex server for a long time, which means I’ve made my fair share of amateur mistakes. Storage is probably one of the biggest things plaguing my Plex server, and I have definitely made all three of these mistakes at least once (if not multiple times).

What if the organization and storage benefits of tool shadowing could be had and improved with a modular, semi-automated process? Tracefinity attempts that by generating custom Gridfinity bins from photos of tools, and has quite a few nifty features that are worth a look.

The basic workflow is this: place one or more tools on a sheet of paper, take a photo, then upload the photo and have the system trace and save the outline and add it to a private tool library. When one is ready to create some bins, use the library of saved tool outlines to generate custom Gridfinity layouts.

If you’re unfamiliar, Gridfinity is a modular system of standardized bins and baseplates designed with 3D printing in mind, making it an ideal match for highly-customized organization tasks and a particularly natural fit for a tool-tracing system like this one.

The idea of taking a photo of a tool and generating a custom bin is a compelling one, and a couple years ago we covered a project that did just that. Tracefinity seems like a natural evolution of the idea, and includes handy features like easy design adjustments, optional magnet holes, and we really like the concept of a tool library from which individual tools are scanned once then later selected to create specific, project-based layouts.

Tracefinity takes advantage of new software capabilities like machine learning to improve and streamline the tracing process, but that doesn’t mean it relies on any external services. It can be entirely self-hosted and by default uses a local, CPU-friendly object detection model for tool tracing. There is an option to provide a API key to use Google Gemini instead, but it’s not required. It can come in handy for especially complex tool outlines or dealing with non-ideal source photos, however.

Good news if you want to keep your NAS boxes on FreeBSD – two new contenders just stepped up, with updated forks of the cancelled TrueNAS Core. FreeBSD has a strong reputation as a stable server OS, and integrates highly reliable ZFS filesystem right into its kernel. The snag is that Linux is rather more widespread, meaning more people know their way around a Linux box – but Linux knowledge doesn’t transfer well over to FreeBSD. The Reg FOSS desk just found out about two different new dedicated NAS OS based on FreeBSD – and saliently, based on the current FreeBSD 15 codebase, which was released in December 2025, with version 15.1 arriving in June. FreeCORE-15.0-U1 appeared on August 27th. It’s the first update for FreeCore 15.0, which appeared just a week earlier. The FreeCORE project website describes it as “TrueNAS® CORE — continued”, and says: “FreeCORE carries the TrueNAS CORE 13.3 system forward as an independently maintained operating system on FreeBSD. “15.0-U1 is stable. TrueNAS CORE 13.3 systems upgrade straight to 15.0 in place, then continue on the project’s update train.” The reason that anyone needed to come along and carry the TrueNAS Core system forward is that developer and vendor iXsystems cancelled its FreeBSD-based NAS OS in 2024. This followed two years after it introduced its new TrueNAS SCALE NAS OS, which is based on Debian Linux. To Register sister site Blocks&Files, the company explained no one was being marooned by the change. Some TrueNAS Core users – and this vulture is one of them – were less than delighted by the change. OpenZFS works fine on Linux, and it’s regularly updated, but it must be loaded as a module: license restrictions mean it cannot be built into the Linux kernel. Significantly, this means that the ZFS Adaptive Replacement Cache must remain separate from Linux’s Page Cache. This is not ideal if the server is memory-constrained, and this vulture’s two TrueNAS Core boxes only have 8 GB of memory each. A year after the end of TrueNAS Core, we reported on a new project fork called zVault, which took the existing FreeBSD 13.3-based TrueNAS code and removed any parts that belonged to iXsystems. Unfortunately, zVault didn’t get much further than that. By later in the year, users were warning on GitHub BEWARE RUNNING ZVAULT – complete with block caps – and the project maintainers have gone silent. The planned upgrades to a FreeBSD 14 base never emerged. Now, FreeCORE has appeared, to pick up the slack. It’s already migrated the TrueNAS Core web UI and other code over to the base of FreeBSD 15, plus all the latest security fixes. The HackerNews discussion of the new product also turned up another new alternative: BSDnas. This appeared just days before the first release of FreeCORE: the project status page is dated 19th August 2026. Its project history tells a similar story: it’s the original TrueNAS Core 13.3 code, cleaned of iXsystems dependencies and moved to the basis of FreeBSD 15. Despite the much newer codebase, the reported memory usage seems modest: it will boot in 2 GB of RAM, and function usably with 4 GB. For contrast, the TrueNAS SCALE docs recommend at least 8 GB RAM – although they do mention “in TrueNAS 13”, which suggests the docs may be rather dated. The TrueNAS Guide website suggests 16 GB as the entry-level “floor” for the current release. The Register asked the FreeCORE project if it had any connection with the older zVault or contemporary BSDnas projects. The developer replied, telling us: "I am not affilated with ZVault nor BSDnas, and not XigmaNAS either. This is an entirely independent fork." They told us that it's a solo project that they have been working on since January. They had a FreeBSD 15.0 build by April, and they made around 50 test builds before the release. For completeness, and for those who’d like a more tried-and-tested tool with more history, we would be remiss if we did not mention XigmaNAS. Previously called NAS4Free, this was forked from an older version of FreeNAS after iXsystems acquired it. Its Wikipedia page says XigmaNAS was forked from FreeNAS 8 back in 2011. The latest version on the Sourceforge download page is based on FreeBSD 14, and was last updated in October 2025. The original creators of FreeNAS went on to build the Debian Linux-based Open Media Vault. The latest version is 8.5.7, released just last weekend. We have an older version running on a testbed server here, and for the record, it seems quite content in just 6 GB of memory. ®

Tape is still not dead, but it’s not in rude health either after shipments slipped by 16 exabytes in 2025. That number comes from the LTO Program – the alliance of HPE, IBM, and Quantum Corp, that together tends the Linear Tape-Open (LTO) Format that accounts for around 85 percent of all tape sales. The Program delivers an annual data dump describing the previous year’s shipments, and this year’s effort landed on Tuesday with news that 160.3 exabytes worth of LTO tape went out the door during 2025, nine percent down from 2024’s 176.5 exabytes. The LTO Program pointed out that 2024 represented an all-time high, and that 2025’s shipments are the second-highest result for the format. The trio quoted storage analyst Tom Coughlin as saying the reason for the 16 exabyte slip was “the evolving international trade environment” which he thinks “contributed to more cautious purchasing behavior, with businesses across a wide range of industries adjusting their sourcing and inventory decisions accordingly.” “We suspect that if this uncertainty is resolved the continued strength of the tape storage market will be even more impressive, particularly as constrained power availability for data center buildouts favors including tape, which consumes less power than any other storage media,” Coughlin said. The LTO Program’s announcement also made much of the fact that Q1 2026 shipments are 57 percent higher than for the same period last year. “The strong start to 2026 confirms that many customers are fully committed to growing their tape infrastructure, leveraging the technology to complement other primary and secondary storage platforms,” the LTO trio asserted, before anticipating “2026 will further strengthen the tape market with another record-setting year.” The org attributed the early 2026 upswing to the debut of a new 40TB tape cartridge that’s compatible with the drives that handle the 30TB cartridges that were previously the best LTO could muster. Tape’s big advantages are high capacity and low power consumption. The downside is painfully long restore and retrieval times. Users must also store tapes carefully. Hyperscalers have therefore challenged the medium with archival-storage-as-a-service offerings that offer virtual tapes to deliver interoperability with tape backup tools. Providers of those services are cagey about exactly what storage medium they use, but it seems most employ large-capacity hard disk drives – which are currently in short supply because hyperscalers need them for AI applications. Even so, that situation may not create a moment for tape because just two companies – Sony and Fujifilm – make LTO tape cartridges, and shortages are not unknown. ®

Keep your files organized, secure, and accessible with 500GB of lifetime cloud storage for one low price.

The post Get 500GB of lifetime cloud storage for just $40 appeared first on TechRepublic.

Keep your files organized, secure, and accessible with 500GB of lifetime cloud storage for one low price.

The post Get 500GB of lifetime cloud storage for just $40 appeared first on TechRepublic.

Bitcoin Magazine

Coldcard Wallet Flaw Exposes Years of Bitcoin Seeds After $70M in BTC Stolen

The popular Bitcoin hardware wallet Coldcard product, made by Coinkite, is at risk following a $70 million hack.

Coinkite on Thursday admitted that its Coldcard Mk3 model was affected following the hack and advised users to move their funds. Then, on Friday, the company said that users of the later hardware devices Mk4, Mk5, and Q should also take precautions. 

Hackers on Thursday were first able to drain funds from 1,196 Bitcoin addresses because their private keys were not generated using sufficient entropy — or randomness. 

Since then, a total of 1,082.65 Bitcoins have disappeared from wallets, according to data from Galaxy Research and engineers at payments company Block. 

While Coinkite has not admitted that the hack is linked to their wallets, the company has said that a wallet seed generation bug in Coldcard products meant the hardware’s true random number generator wasn’t actually being used on certain firmware versions. 

Coinkite and other engineers in the Bitcoin space are still investigating reportedly ongoing drains still happening at the time of writing.

What actually happened 

A firmware bug in Coldcard Mk3 devices (starting with version 4.0.1 in March 2021) caused seed generation to fall back to a weak software PRNG instead of the hardware true random number generator, producing seeds with only ~40 bits of entropy rather than the intended 128.  This made private keys for many single-signature wallets (especially those created without dice rolls or a strong BIP-39 passphrase) predictable enough for attackers to brute-force.

A total of 594.5 Bitcoins worth over $35.7 million at today’s prices were moved to a new address from single-signature addresses on Thursday. 

NEW: Over 594 BTC worth $38 million was stolen from Bitcoin hardware wallet Coldcard users.

The attacker then moved around the BTC and consolidated 562 BTC into this address below.

Users are urged to review the company's official security guidance as soon as possible. pic.twitter.com/exD2Wax7CY

— Bitcoin Magazine (@BitcoinMagazine) July 31, 2026

More wallets were later drained, according to blockchain analysts, with the total now over $70 million. 

Various affected users shared their experiences on social media, with one saying that their Bitcoin had not been moved since 2021, and all of a sudden was swiped. 

Bitcoin engineers have since said that Coldcard products — specifically the Mk3 models — had  “faulty entropy in wallet generation,” meaning they did not use real randomness to create a seedphrase. 

What to do 

Developers in the Bitcoin space have since urged users to move their funds if they used a Coldcard. Coldcard has issued guidance for users to take, which can be found here.

The first post was the advisory and what users should do.

This second post has the technical details: what actually went wrong, why our reviews missed it, the impact across Mk3/Mk4/Q/Mk5, and what we changed.https://t.co/HshUxevCl3

( current evaluating Mk3 firmware release ) https://t.co/Yfdx4XcztA

— COLDCARD (@COLDCARDwallet) July 31, 2026

Coinkite first said that their Mk3 models were affected but then on Friday said that those who did not use sufficient entropy to create a seed — in this case, 50 dice rolls — should generate a brand-new seed on the updated device. Others have warned to ditch Coldcard completely to be sure their funds are safe. 

“Everything is fucked,” wrote Kevin Loaec, CEO of Bitcoin security company, Wizardsardine. 

“Every single mnemonic generated [via a Coldcard] since 2021 will be public in the next few days,” Loaec warns.

This post Coldcard Wallet Flaw Exposes Years of Bitcoin Seeds After $70M in BTC Stolen first appeared on Bitcoin Magazine and is written by Mathew Di Salvo.