Nintendo conseille de ne pas jouer à la Switch 2 pendant la canicule
Nintendo a indiqué que la plage d'utilisation de la Switch 2 se situait entre 5 et 35 degrés Celsius. En cas de canicule, il faut donc être très prudent.
Nintendo a indiqué que la plage d'utilisation de la Switch 2 se situait entre 5 et 35 degrés Celsius. En cas de canicule, il faut donc être très prudent.
[Deal du jour] Depuis le début de l'été, Bouygues propose un forfait baptisé Summer Edition, avec une enveloppe de 40 Go de data utilisable dans de nombreux pays.
Cet été, Bouygues Telecom a lancé un forfait international avec de la data à l'étranger. Que vaut-il par rapport aux meilleures offres du secteur ? Nous l'avons comparé avec les forfaits internationaux de Free et de RED by SFR.
The CWWK M8 NAS motherboard, equipped with either the Intel Twin Lake N150 or N355 processor, is a compact Mini-ITX platform aimed at advanced home NAS builders and small office users looking for a cost-effective alternative to branded NAS systems. Measuring just 17 x 17 cm, it combines several high-end features such as an onboard 10GbE RJ45 LAN (via the AQC113C controller), dual 2.5GbE Intel i226-V ports, and support for up to eight SATA drives through dual SFF-8643 ports. The board also integrates two M.2 NVMe slots, a DDR5 SO-DIMM memory slot supporting up to 48GB, and a PCIe Gen3 x1 slot for modest expansion. Unlike many low-power ITX boards, the M8 includes support for Wake-on-LAN, PXE boot, and hardware monitoring, which makes it a viable candidate for 24/7 operations and remote deployment scenarios. With its efficient lane distribution—critical for balancing 10GbE, NVMe, SATA, and PCIe simultaneously—it delivers a level of I/O flexibility not commonly found at this price point, particularly in the sub-$300 range.
The CWWK M8 NAS motherboard strikes a practical balance between performance, expandability, and power efficiency, making it a compelling choice for DIY NAS builders looking for 10GbE capability without the complexity or cost of larger platforms. With support for up to eight SATA drives via dual SFF-8643 connectors, dual NVMe slots, and a DDR5 SO-DIMM socket (up to 48GB), it delivers a surprising level of storage flexibility in a compact Mini-ITX form factor. Performance across the 10GbE port is strong—achieving near-saturation read speeds and respectable write throughput—while NVMe and SATA access remain consistent thanks to a careful PCIe lane allocation strategy. Power draw remains modest, even when fully populated with drives and expansion cards, reinforcing its suitability for 24/7 deployments. However, limitations like Gen3 x1 NVMe speeds, a single RAM slot, and shared PCIe/E-Key lane usage should be considered by those seeking maximum expansion or high-end performance. Still, for its price, pre-installed CPU, and strong open-source OS compatibility, the M8 offers an unusually capable base for home servers, backup targets, or even Plex and Proxmox environments.
9.0
10GbE RJ45 port (AQC113C) with full Gen3 x2 bandwidth
Dual 2.5GbE Intel i226-V ports with wide OS compatibility
Supports up to 8 SATA drives via dual independent SFF-8643 ports
Includes 2× M.2 NVMe 2280 slots, suitable for cache or boot use
Very low power draw (~20W under load with 10g+2xM.2, ~31W idle fully populated with HDDs)
Compact Mini-ITX form factor with well-organized layout
Exceptional Price vs H/W Level
Broad OS support (TrueNAS, Unraid, PVE, Linux, Windows, etc.)
PCIe slot and M.2 E-Key share a lane—only one usable at a time
M.2 NVMe slots limited to PCIe Gen3 x1 speeds
Single DDR5 SO-DIMM slot (no dual-channel support)
Where to Buy? |
The physical design of the CWWK M8 motherboard is centered around the Mini-ITX standard, maintaining a compact 17 x 17 cm footprint that caters to space-conscious NAS builds. Despite its small form factor, the layout is methodically structured to maximize accessibility and airflow. Key components such as the dual SFF-8643 ports, NVMe slots, and RAM socket are positioned for easy cable routing and minimal overlap.
The CPU arrives pre-installed with a low-profile ball-bearing cooler, which is sufficient for the low 6W TDP of the N150 processor. There’s also a system fan header onboard with PWM support, allowing for basic thermal management in enclosed NAS chassis. The board is finished in a neutral white PCB, aligning with recent CWWK trends that blend aesthetic minimalism with function-first engineering.
Storage expansion is one of the most defining elements of the M8. It features dual SFF-8643 ports that, with breakout cables, provide connectivity for up to eight SATA III (6Gbps) drives.
These connectors are routed through independent ASM1164 controllers, each on a dedicated PCIe Gen3 x1 lane, ensuring that drive traffic is not bottlenecked through a single controller.
This separation also means users can confidently deploy SSDs or mixed SSD/HDD arrays without major performance drops under load. The board supports RAID configurations at the OS level via TrueNAS or Unraid, and is capable of delivering reliable throughput for multi-drive setups including RAID-Z, RAID5, or JBOD.
In addition to traditional SATA storage, the board includes two M.2 NVMe 2280 slots, each operating at PCIe Gen3 x1. While this limits peak performance to around 900MB/s per slot, it is sufficient for cache drives or SSD-based boot volumes, especially in NAS environments where latency and parallel IOPS matter more than raw sequential throughput. The placement of the NVMe slots, one top-side and one underside, helps distribute heat and gives builders flexibility in cooling strategy. Both slots are directly accessible, and installation doesn’t require removing other components, which is particularly useful during upgrades or replacements.
Storage scaling is enhanced through the modularity of the board’s SFF-8643 interfaces. As discussed in your review, these ports can be adapted not just to standard SATA breakouts but also to additional M.2 or U.2 devices with the correct adapter cards. This creates potential for hybrid NAS setups—using SATA for bulk data storage and NVMe for hot data or VM usage. Such versatility in drive mapping is rarely offered at this price point, and makes the board viable not only for home media servers but also for lab environments or light virtualized storage nodes.
One lesser-known but practical addition is the inclusion of a MicroSD (TF) slot on the PCB. While it’s not ideal for installing major OS platforms like TrueNAS Core, it can be useful for loading bootloaders such as Unraid or for system config backups. Importantly, the TF slot is recognized natively by most operating systems and appears as a usable storage device without requiring extra drivers. It also enables simple out-of-band recovery or local snapshot scripts in more advanced workflows. Combined with the available internal USB port, the board allows multiple low-impact boot or recovery paths to coexist alongside primary storage deployments.
The CWWK M8 motherboard is equipped with a well-rounded selection of external and internal I/O ports that support a broad range of NAS and server use cases. Most notably, it includes one 10GbE RJ45 port powered by the AQC113C controller and two additional 2.5GbE ports via Intel i226-V chips.
These networking options allow the board to operate in multiple roles simultaneously, such as high-speed file sharing over 10GbE while maintaining service management or redundancy via the dual 2.5GbE ports. The inclusion of Intel network controllers ensures wide compatibility with open-source operating systems like TrueNAS and Unraid, as well as ESXi and PVE, making it a suitable base for software-defined networks, VLAN tagging, or bonded interface configurations.
On the USB front, the M8 provides a combination of high-speed and legacy options. It includes 1× USB Type-C (10Gbps) and 1× USB 3.2 Gen2 Type-A (10Gbps) ports for external storage or fast USB peripherals. There are also 2× USB 2.0 Type-A ports located at the rear I/O and an internal USB 2.0 header, which is useful for OS boot drives such as Unraid.
Internally, the board also features a USB 3.0 header and a Type-E header, allowing front-panel USB 3.x support if the chassis includes such connectors. These ports give builders the flexibility to attach boot media, backup targets, or even USB-based UPS management tools without additional hardware.
For video output and direct display use, the M8 includes 1× HDMI 2.0 and 1× DisplayPort 1.4, both capable of 4K@60Hz output. These are connected via the integrated UHD graphics included with the N150/N355 CPU. While these outputs are generally not essential in a headless NAS environment, they provide value in cases where the system is used as a hybrid HTPC/NAS, or when diagnostics and BIOS access are needed without SSH or remote management tools. The GPU is also supported for hardware video decoding, making the board a viable base for light Plex or Jellyfin deployments that rely on integrated graphics acceleration.
Internally, the board features several headers that further expand its flexibility. Alongside the previously mentioned USB and fan headers, there’s an M.2 E-Key slot for wireless modules, which shares PCIe lanes with the x1 PCIe slot and cannot be used simultaneously. The board also includes an SD card (TF) slot which appears natively in supported OSes—suitable for bootloaders or small backup tasks.
While not suited to high-throughput use, it does provide an alternative storage option in embedded or recovery scenarios. The arrangement and accessibility of these ports are well considered for such a small form factor, ensuring that builders can access almost all essential functionality without relying on riser boards or USB hubs.
Interface Type | Details |
---|---|
Ethernet Ports | 1× 10GbE RJ45 (AQC113C), 2× 2.5GbE RJ45 (Intel i226-V) |
USB Ports (Rear) | 1× USB 3.2 Gen2 Type-A (10Gbps), 1× USB-C (10Gbps), 2× USB 2.0 Type-A |
USB Ports (Internal) | 1× USB 2.0 (boot drive), 1× USB 3.0 header, 1× USB 3.0 Type-E header |
Display Outputs | 1× HDMI 2.0, 1× DisplayPort 1.4 (both support 4K@60Hz) |
PCIe Slot | 1× PCIe Gen3 x1 (x4/x8 slot compatible, shared with M.2 E-Key) |
M.2 Slots | 2× M.2 2280 NVMe (PCIe Gen3 x1), 1× M.2 E-Key for WiFi/BT |
SD Card Slot | 1× TF (MicroSD) slot (appears as storage device) |
Fan and Headers | 1× PWM fan header, various USB/F_USB headers for front I/O |
At the heart of the M8 motherboard lies a choice between two Intel Twin Lake processors: the N150 and the N355. The N150 is a quad-core, four-thread CPU with a base architecture derived from the Alder Lake-N family, running at up to 3.6GHz and featuring a modest 6MB cache. It operates at a remarkably low TDP of 6W, making it suitable for passive or semi-passive cooling environments.
The N355, on the other hand, doubles the thread count and bumps performance further, albeit at a slightly higher price. Both CPUs are pre-soldered to the board and arrive with a compact, ball-bearing fan assembly that supports quiet, efficient cooling. These processors are not meant for heavy computation but offer enough power for file server duties, light containerization, and even modest Plex media serving—with the N150 proving capable of 4K playback in testing.
Memory support is handled via a single DDR5 SO-DIMM slot, officially supporting up to 48GB at 4800MHz. While dual-channel operation is not available, DDR5’s higher base bandwidth helps compensate for this limitation in real-world usage. The board accepts standard non-ECC modules and will clock down any faster memory to the platform’s 4800MHz limit.
For NAS and virtualization users, this constraint is acceptable, though power users may note that memory upgrades are capped to a single slot. That said, 32GB or 48GB configurations are more than adequate for common use cases like running TrueNAS Scale with Docker containers, or spinning up a few VMs in Proxmox.
The board’s PCIe lane distribution is particularly deliberate given the constraints of the Twin Lake architecture, which provides just 9 usable PCIe lanes. Despite this, the M8 balances connectivity by allocating PCIe Gen3 x2 bandwidth to the 10GbE port, ensuring full 10Gbps throughput with bandwidth overhead. The SATA controllers each receive dedicated PCIe Gen3 x1 lanes, and each M.2 NVMe slot is similarly mapped at x1 speed.
The remaining lane is shared between the M.2 E-key (for Wi-Fi/BT modules) and the physical PCIe x1 expansion slot. This means that users must choose between Wi-Fi upgrades or additional PCIe peripherals—a typical tradeoff on ITX boards, but worth noting during build planning.
From a system management perspective, the board supports UEFI-only boot modes and includes features such as Auto Power-On, Scheduled Power-On, PXE boot, Wake-on-LAN, and Secure Boot, making it suitable for remote deployment or integration into managed environments. The board includes thermal monitoring via BIOS and OS-level tools, with fan control limited to one system fan header supporting PWM. These features, while basic, are sufficient for home server use or edge deployment in micro data centers. The compact ITX layout also makes the board a candidate for embedded use in custom NAS chassis or OEM enclosures with constrained airflow or proprietary mounting.
Component | Details |
---|---|
CPU Options | Intel N150 (4C/4T, 3.6GHz, 6W TDP), Intel N355 (8C/8T, higher performance) |
Memory | 1x DDR5 SO-DIMM, up to 48GB (4800MHz), non-ECC |
Chipset/Lanes | Intel Twin Lake SoC, 9 PCIe Gen3 lanes total |
NVMe Storage | 2x M.2 2280 NVMe (PCIe Gen3 x1 each) |
SATA Support | 2x SFF-8643 (8x SATA III via breakout cables, each on ASM1164 controller) |
PCIe Expansion | 1x PCIe Gen3 x1 slot (shared with M.2 E-Key) |
WiFi Module Slot | 1x M.2 E-Key (2230) for Wi-Fi/BT (shares lane with PCIe slot) |
Boot Features | UEFI-only, Auto Power-On, Wake-on-LAN, PXE boot, Secure Boot |
Fan Support | 1x PWM system fan header, bundled CPU fan |
During benchmarking and real-world tests, the N150-based M8 motherboard demonstrated performance levels consistent with expectations for an ultra-low-power NAS platform. Sequential read speeds over the 10GbE interface approached saturation during synthetic ATTO Disk Benchmark tests, particularly with a 256MB block size, where throughput consistently exceeded 950MB/s.
Write performance, however, plateaued slightly lower, averaging between 650–700MB/s for 1GB and 4GB file tests. These figures are typical for systems utilizing Gen3 x1 NVMe SSDs and efficiency-focused CPUs like the N150, where write-intensive operations are more limited by CPU capability than disk throughput. Larger transfers or workloads involving compression will see slightly more variation, but in most scenarios, read performance remained stable and consistent.
Using a RAID 1 array of Seagate IronWolf drives connected via the dual SFF-8643 SATA ports, the board achieved average write speeds of 550–580MB/s, with occasional peaks in read performance reaching up to 800MB/s, though these were not sustained.
These results reflect the benefit of having each SATA group routed through a separate ASM1164 controller, ensuring that bandwidth isn’t choked under RAID configurations or multi-drive reads. In practical terms, this makes the board well-suited for file-serving tasks, Time Machine backups, or media library hosting, with no obvious contention across interfaces during simultaneous read/write operations.
NVMe performance was constrained by the PCIe Gen3 x1 link per M.2 slot, which limited theoretical throughput to under 1GB/s. Tests confirmed read speeds of around 720MB/s and write speeds of approximately 520MB/s in sustained transfers. While not ideal for high-performance VM storage or video editing scratch disks, these speeds are more than adequate for cache duties or container storage. Importantly, the board maintains predictable performance across both NVMe slots, and thermals were manageable under active load without throttling, thanks in part to the pre-attached CPU cooler and accessible airflow pathways on the board’s surface.
In terms of power efficiency, the system consumed approximately 19–20W under load when configured with the N150 CPU, 8GB of DDR5, two NVMe SSDs, and a 10GbE connection in active use. When idle but fully populated with four SATA drives and an expansion card installed (but unused), power draw settled at around 31.4W. This confirms the board’s suitability for 24/7 operation without requiring high-capacity PSUs or custom thermal management.
![]() |
![]() |
Test Category | Result (N150 Model) |
---|---|
10GbE Read (ATTO, 256MB) | ~950MB/s (near saturation) |
10GbE Write (1–4GB) | ~650–700MB/s |
RAID 1 HDD (SATA) | Write: 550–580MB/s, Read Peak: up to 800MB/s (occasional spikes) |
NVMe (Gen3 x1) | Read: ~720MB/s, Write: ~520MB/s |
Power Draw (Load) | ~19–20W (N150, 2× NVMe, 10GbE active) |
Power Draw (Idle, full config) | ~31.4W (4× HDD, PCIe card, NVMe, no I/O) |
Thermals | Stable under load; no active throttling observed |
The CWWK M8 motherboard delivers a rare combination of high-speed networking, broad storage expandability, and low power consumption, all within a Mini-ITX footprint. It manages to balance PCIe lane allocation across 10GbE, dual NVMe, and eight SATA drives without compromising basic performance, thanks to deliberate hardware pairing and thoughtful board layout. The use of separate SATA controllers, a well-provisioned 10GbE controller on Gen3 x2 lanes, and native UEFI support reflects a clear intent to make this a serious option for NAS enthusiasts and advanced home users. Its ability to sustain near-saturation speeds on the 10GbE connection and provide usable NVMe throughput makes it a capable base for TrueNAS, Unraid, or Proxmox environments—whether for home backup, Plex media hosting, or light VM workloads.
However, there are trade-offs. The limited PCIe expandability, single RAM slot, and Gen3 x1 constraints on NVMe performance may not meet the needs of high-end workstation builders or enterprise deployments. Additionally, the shared PCIe lane between the M.2 E-key and the PCIe slot limits simultaneous use of both interfaces, which could affect those hoping to add both Wi-Fi and a PCIe peripheral. Still, for its price point and target use case, the M8 delivers well above average. It avoids many of the bottlenecks seen in competing low-power boards and manages to do so at under $300 with a pre-installed CPU. For users building a power-efficient, high-bandwidth DIY NAS with flexible drive options and capable base specs, the CWWK M8 stands out as a strong contender.
Where to Buy? |
Pros | Cons |
---|---|
10GbE RJ45 port (AQC113C) with full Gen3 x2 bandwidth | PCIe slot and M.2 E-Key share a lane—only one usable at a time |
Dual 2.5GbE Intel i226-V ports with wide OS compatibility | M.2 NVMe slots limited to PCIe Gen3 x1 speeds |
Supports up to 8 SATA drives via dual independent SFF-8643 ports | Single DDR5 SO-DIMM slot (no dual-channel support) |
Includes 2× M.2 NVMe 2280 slots, suitable for cache or boot use | |
Very low power draw (~20W under load, ~31W idle fully populated) | |
Compact Mini-ITX form factor with well-organized layout | |
Pre-installed CPU and active cooling fan included | |
Broad OS support (TrueNAS, Unraid, PVE, Linux, Windows, etc.) |
This description contains links to Amazon. These links will take you to some of the products mentioned in today's content. As an Amazon Associate, I earn from qualifying purchases. Visit the NASCompares Deal Finder to find the best place to buy this device in your region, based on Service, Support and Reputation - Just Search for your NAS Drive in the Box Below
![]() |
![]() |
Les tensions commerciales entre Washington et Pékin sur la question cruciale des semi-conducteurs n'en finissent plus. Nvidia se retrouve cet été au cœur des soupçons : la Chine exige des « preuves de sécurité convaincantes » concernant ses puces H20, soupçonnées d’abriter des portes dérobées.
Aux États-Unis, Nintendo annonce augmenter les prix des Nintendo Switch, Nintendo Switch Lite et Nintendo Switch OLED en réaction aux conditions du marché. Les autres territoires sont épargnés pour l'instant, mais l'impact des tarifs douaniers pourraient avoir des répercussions à l'avenir.
Les tensions commerciales entre Washington et Pékin sur la question cruciale des semi-conducteurs n'en finissent plus. Nvidia se retrouve cet été au cœur des soupçons : la Chine exige des « preuves de sécurité convaincantes » concernant ses puces H20, soupçonnées d’abriter des portes dérobées.
Battlefield 6 arrive à la rentrée, mais il sera possible d’y jouer bien avant grâce aux phases de bêta fermée et ouverte prévues cet été. Encore faut-il que votre ordinateur suive : voici les configurations PC minimale et recommandée pour faire tourner le nouveau FPS de DICE.
À l'occasion de la publication de ses résultats financiers, Nintendo fait un point sur la situation de la Switch 2 deux mois après sa sortie. Tous les indicateurs sont au vert : la Switch 2, si elle suit la trajectoire actuelle, pourrait devenir la console la plus vendue de tous les temps.
À l'occasion d'un Nintendo Direct dédié aux éditeurs partenaires, Nintendo et EA Sports ont dévoilé des premières images du gameplay de FC 26 sur la Switch 2. On y découvre un jeu bien plus proche des versions PlayStation et Xbox. Si EA fait bien les choses, le jeu pourrait rencontrer un grand succès sur la console.
Selon de récentes rumeurs AMD envisagerait sérieusement de lancer une carte NPU dite "discrète" pour nos PC de bureau. De quoi s'agit-il ?
Cet article AMD envisagerait des cartes PCIe NPU pour nos PCs de bureau a été publié en premier par GinjFo.
Bon, on est le 31 juillet 2025 et dans deux jours, c’est le grand chamboulement. L’AI Act entre en application et j’ai vu passer tellement de conneries sur le sujet que j’ai décidé de vous faire un guide clair et net. Parce que non, vous n’allez pas devoir mettre “CONTENU GÉNÉRÉ PAR IA” en gros sur chaque article de votre blog.
Alors respirez un coup, prenez un café, et on va démêler tout ça ensemble. Je vous promets qu’à la fin de cet article, vous saurez exactement ce que vous devez faire sur votre site. Et surtout, ce que vous n’êtes PAS obligé de faire.
L’AI Act, c’est donc le nouveau règlement européen sur l’intelligence artificielle. Un peu comme le RGPD mais pour l’IA. Et comme le RGPD, ça s’applique à tous ceux qui proposent des services dans l’Union européenne, même si vous êtes basé aux États-Unis ou sur Mars.
Le truc important à comprendre, c’est que l’AI Act fonctionne par niveaux de risque. Plus votre système d’IA présente de risques, plus vous avez d’obligations. Pour nous, éditeurs web et créateurs de contenu, on est généralement dans la catégorie “risque limité”, ce qui veut dire qu’on a principalement des obligations de transparence.
Et c’est là que ça devient intéressant car c’est l’article 50 du règlement qui définit ces obligations de transparence, mais il y a plein d’exemptions que personne ne vous raconte.
Concrètement, si vous utilisez l’IA pour générer du contenu sur votre site, vous devez en informer vos utilisateurs. Mais attention, ce n’est pas aussi simple que “mettez une mention partout”.
Voici ce que dit précisément le texte :
Mais voilà le twist, ces obligations ne s’appliquent pas dans tous les cas car l’AI Act prévoit plusieurs cas où vous n’avez pas besoin de signaler l’utilisation de l’IA :
C’est probablement l’exemption la plus importante pour vous ! D’après l’article 50 paragraphe 4 de l’AI Act, vous n’avez PAS besoin de mentionner l’utilisation de l’IA si :
Concrètement, ça veut dire que si vous utilisez ChatGPT / Claude pour générer un brouillon d’article, qu’ensuite, vous le relisez, le modifiez, l’éditez, le corrigez et quie vous en assumez la responsabilité en tant qu’éditeur/blogueur, vous n’avez PAS besoin de mentionner que l’IA a été utilisée !
C’est énorme car cette exemption reconnaît que la relecture humaine et la responsabilité éditoriale réduisent considérablement les risques. Le texte officiel précise ainsi que cette exemption est conçue pour les cas où les textes générés par IA sont “examinés, classés et dont la responsabilité est assumée par du personnel éditorial” (source Lexology). Je trouve ça très bien car dans ce cas précis, l’IA est utilisé comme un outil sous contrôle humain, et pas un moyen automatisé capable de faire n’importe quoi.
Par contre, pour les images générées par IA, la mention reste nécessaire.
L’article 50 précise aussi que vous n’avez pas à informer les utilisateurs si c’est “évident du point de vue d’une personne raisonnablement bien informée, observatrice et circonspecte”.
En clair, si c’est évident que c’est de l’IA, pas besoin de le dire. Par exemple, si vous avez un chatbot qui s’appelle “Assistant IA” avec une icône de robot, pas besoin d’ajouter “Ceci est une IA”. C’est du bon sens.
Si votre contenu fait partie d’une œuvre “manifestement artistique, créative, satirique, fictionnelle ou analogue”, vous n’avez qu’une obligation minimale, celle de signaler l’existence du contenu généré “d’une manière appropriée qui n’entrave pas l’affichage ou la jouissance de l’œuvre”.
Traduction, si vous faites de l’art, de la fiction ou de la satire avec l’IA, vous pouvez mettre une petite mention discrète quelque part, pas besoin de gâcher votre création avec un gros bandeau rouge.
Si l’IA n’a qu’une “fonction d’assistance pour l’édition standard” et ne modifie pas substantiellement le contenu, pas non plus d’obligation de transparence.
Donc si vous utilisez l’IA pour :
Vous n’avez PAS besoin de le signaler. C’est considéré comme de l’édition standard, au même titre qu’utiliser un correcteur orthographique.
Si vous utilisez l’IA dans un contexte personnel et non professionnel, l’AI Act ne s’applique pas. Donc votre blog perso où vous racontez vos vacances n’est pas concerné sauf si vous le monétisez ou si vous avez une audience professionnelle.
Bon, parlons maintenant du nerf de la guerre. Si vous ne respectez pas ces obligations, vous risquez :
Mais avant de paniquer, sachez que ces amendes maximales sont pour les cas graves et répétés. Les autorités vont d’abord vous demander de vous mettre en conformité. C’est un peu comme le RGPD… on commence par la prévention avant la répression.
Maintenant qu’on a vu la théorie, passons à la pratique. Voici exactement ce que vous devez faire selon votre situation :
Cas 1 : Vous générez un brouillon avec l’IA puis vous le relisez/éditez → PAS d’obligation de mention grâce à l’exemption de relecture humaine (article 50.4) → Condition : vous devez vraiment relire et assumer la responsabilité éditoriale
Cas 2 : Vous publiez directement le texte généré par l’IA sans relecture → Si c’est pour informer le public : mention obligatoire → Si c’est créatif/satirique/fiction : mention discrète suffisante
Cas 3 : Vous utilisez l’IA juste pour améliorer votre texte (grammaire, style) → Pas d’obligation car c’est de l’édition standard
Pour toute image générée par IA, vous devez le signaler, SAUF si :
Comment le signaler ? Une mention dans la balise alt, dans la légende ou en bas de page suffit. Pas besoin d’un watermark géant.
Vous devez informer les utilisateurs qu’ils parlent à une IA, SAUF si c’est évident (le chatbot s’appelle “Bot IA”, a une tête de robot, etc.).
Voici donc mes recommandations pour dormir sur vos deux oreilles :
Voilà, si vous voulez creuser le sujet (et je vous le conseille), voici les liens officiels :
Voilà, vous savez tout ! L’AI Act, c’est pas la fin du monde, c’est juste un nouveau cadre pour utiliser l’IA de manière responsable et, la plupart d’entre vous n’auront que peu de changements à faire.
L’important, c’est de rester transparent avec votre audience quand c’est nécessaire mais pas besoin d’en faire des tonnes. Je trouve que l’AI Act est plus intelligent qu’on ne le pense car il fait la différence entre publier directement du contenu généré par IA et utiliser l’IA comme assistant de rédaction. Par contre, les gens mal informés sur le sujet ou bien cons risquent de vous prendre le chou donc restez zen et envoyez leur le lien de mon article.
Ah, et un dernier conseil : si vous avez un doute sur votre cas particulier, demandez à un juriste spécialisé car cet article vous donne les grandes lignes basées sur les textes officiels, mais chaque situation est unique et je ne suis pas juriste, alors mieux vaut prévenir que guérir, surtout avec des amendes à 15 millions d’euros !
The UGREEN DH3400 Plus is a 4-bay NAS solution introduced quietly into the company’s growing lineup of storage devices, appearing online without the usual fanfare. Closely related to the DH4300 Plus in both hardware and feature set, this model forms part of UGREEN’s value-tier NAS systems, designed to provide essential private cloud functionality at a lower cost. Targeted at home users, prosumers, and small offices, the DH3400 Plus offers a mix of modern storage capacity and lightweight computing through its ARM-based architecture. With support for RAID 0, 1, 5, 6, and 10, 2.5GbE networking, and up to 120TB of raw storage across four SATA bays, the device aims to provide a reasonably capable platform for tasks like file serving, multimedia streaming, and basic photo management.
However, the DH3400 Plus doesn’t just follow the pattern of other budget NAS solutions. It includes features uncommon at this price point, such as HDMI output and 10Gbps USB connectivity, which extend its use case beyond simple backups into media and even light-duty AI-powered photo organization. Its system-on-chip, the Rockchip RK3588, is more powerful than the processors seen in comparable entry-level NAS systems from other brands like Synology or QNAP. While it lacks the high-end features of premium NAS units—such as multi-port networking or M.2 expansion—it still manages to deliver a compact, energy-efficient package suited to most daily NAS workloads. This review breaks down each component of the DH3400 Plus, from its physical design and internal hardware to its software ecosystem, to help users decide whether this is the right solution for their needs.
The UGREEN DH3400 Plus stands out as a well-equipped entry-level NAS that offers considerably more than its modest price tag suggests, positioning itself as a serious contender in the growing value NAS market. Powered by the RK3588 ARM processor and backed by 8GB of LPDDR4X memory, it delivers capable performance for a wide range of NAS tasks including file sharing, media streaming, light containerization via Docker, and AI-assisted photo management. Its support for 2.5GbE networking, 10Gbps USB ports, and HDMI 2.1 output adds versatility, enabling faster-than-Gigabit transfer speeds and options for direct media playback or on-site display access. The system’s low power consumption, compact footprint, and quiet operation make it especially well-suited for 24/7 deployment in home or small office environments. AI functionality for face recognition and album generation, powered by the on-chip NPU, further enhances its value for users managing large personal photo or media libraries. However, the device is limited by its lack of PCIe or M.2 expansion, absence of link aggregation or redundant LAN, and the relatively immature UGOS Pro software, which trails behind DSM and QTS in terms of advanced features, third-party app support, and overall refinement. Still, for users seeking a reliable, efficient, and feature-packed NAS at a competitive price, the DH3400 Plus represents a compelling option that punches well above its weight—offering core functionality, multimedia capabilities, and surprising hardware strength without venturing into high-end NAS pricing territory.
7.6
Powerful ARM CPU: Equipped with the RK3588 SoC, offering 8 cores, integrated GPU, and NPU for AI workloads.
Generous (but fixed!) Memory: 8GB LPDDR4X RAM, rare in budget NAS systems, supports multitasking and Docker use.
2.5GbE Network Port: Provides faster-than-Gigabit throughput for backups, media streaming, and multi-user access.
HDMI 2.1 Output: Rare on ARM powered turnkey NAS, and enables direct media playback or NAS control at up to 4K 60Hz, uncommon in value-tier NAS units.
USB 10Gbps Ports: Dual USB-A 10Gbps and one USB-C 5Gbps allow for high-speed backups or external storage expansion.
AI Photo Management: Built-in NPU supports facial recognition and scene detection for local, private media organization.
Low Power Consumption: Efficient under load (~30W) and idle (~5W without drives), suitable for 24/7 operation.
No PCIe or M.2 Expansion: Lacks future scalability for NVMe caching, 10GbE, or other upgrades.
Single LAN Port: Only one 2.5GbE port, with no failover or link aggregation support.
Limited Software Ecosystem: UGOS Pro lacks iSCSI, VM support, and native Jellyfin, trailing behind DSM/QTS in maturity.
Buy the UGREEN DH4300 on Amazon @409 | Buy the UGREEN DH4300 on UGREEN.COM | Buy the UGREEN DH4300 on B&H |
![]() |
![]() |
![]() |
The UGREEN DH3400 Plus adopts a minimalist, vertical chassis design that blends functional cooling with a relatively compact footprint. Measuring 155mm x 155mm x 215.7mm, the enclosure is tall and narrow, allowing it to accommodate four internal 3.5″ or 2.5″ SATA drives without consuming significant desk space. Instead of traditional hot-swappable trays, UGREEN uses a toolless drive insertion mechanism where each drive slides vertically into the chassis.
This approach reduces complexity and keeps production costs lower but introduces compromises in terms of drive handling and vibration isolation. The top panel is removable and held magnetically, granting access to the internal drive bays and initial setup QR code sticker located inside the lid. Ventilation is handled by side vents, perforations at the top of the cover, and a single downward-facing internal fan near the system board.
Drive installation is simple but not entirely refined. Each of the four bays inside feeds directly into a fixed SATA backplane. Drives are inserted vertically without caddies and seat against rubber bumpers at the rear, which help reduce some of the mechanical vibration but do not provide the same structural integrity or thermal separation found in more traditional hot swap and more easily accessible tray-based designs.
The plastic inner rails feel light and somewhat fragile, especially when compared to metal-based bays found in more premium NAS models. Still, the internal structure was found to be sufficient during testing, even when populating the system with higher-capacity drives. The unit officially supports up to 4 x 30TB drives, allowing a total storage capacity of 120TB in JBOD or RAID configurations, though users will need to supply their own disks as none are included in the package.
Airflow and thermal behavior of the DH3400 Plus are managed via passive intake vents and a single active cooling fan positioned on the base of the chassis. This fan draws air from the top and sides of the NAS, passing it over the CPU and power delivery components before exhausting out the bottom. While this layout is thermally efficient for a low-power device like this, it lacks redundancy and doesn’t allow for custom fan curve profiles within UGOS Pro.
![]() |
![]() |
![]() |
![]() |
During 24-hour tests with four IronWolf drives installed, drive temperatures remained between 44°C and 45°C, while the system chassis and CPU areas stayed in the range of 38°C to 42°C under typical workloads. Though these temperatures are within operational thresholds, users deploying the NAS in warmer environments or fully populating it with high-RPM drives may need to consider ambient airflow or active external cooling assistance.
The DH3400 Plus is constructed from a mix of metal and plastic materials, with a soft matte finish and minimal front I/O clutter. This helps the device appear more neutral in office or home setups. All primary status LEDs are located subtly on the front edge, along with a power button and high-speed USB-C port. Despite its budget positioning, UGREEN has maintained an external aesthetic that doesn’t look out of place beside other premium tech. However, internal cost-saving decisions are more evident. The lack of a proper drive locking mechanism or hotswap support reinforces that this NAS is not meant for heavy-duty enterprise usage or high-frequency drive replacement tasks. It is more appropriate as a near-permanent local storage solution once disks have been installed and configured.
In terms of storage functionality, the DH3400 Plus supports multiple RAID modes, including Basic, JBOD, RAID 0, 1, 5, 6, and 10. This flexibility is notable, as several competing ARM-based NAS units limit RAID options—particularly RAID 6, which requires higher CPU and memory resources to compute parity. While software RAID performance will depend heavily on the chosen configuration and disk types, the Rockchip RK3588 SoC proved capable of maintaining consistent performance in RAID 5 across moderate workloads, including file copying, media playback, and snapshot operations. UGREEN’s system also offers automatic drive recognition and formatting through the UGOS Pro interface, allowing less experienced users to get started quickly, though without the advanced data scrubbing and repair features available on some higher-end NAS platforms.
At the heart of the DH3400 Plus lies the Rockchip RK3588 processor, an 8-core ARM-based SoC featuring a hybrid core architecture. It combines four high-performance Cortex-A76 cores with four efficiency-oriented Cortex-A55 cores, providing a significant performance uplift over most ARM-based NAS processors in the same price bracket. The RK3588 includes a Mali-G610 GPU and an integrated NPU capable of delivering up to 6 TOPS of AI performance. This allows for features like facial recognition, object tagging, and scene categorization within UGOS Pro. Notably, this chip has seen growing popularity in DIY NAS and edge computing applications due to its multimedia capabilities and low power draw. Its inclusion in a turnkey NAS platform at this price point positions the DH3400 Plus as an outlier among typical value-series offerings.
Complementing the processor is 8GB of LPDDR4X memory, which is soldered to the board and not user-upgradable. While this limits long-term scalability, 8GB is a generous amount for an ARM-based NAS and sufficient for most general-purpose NAS workloads including file sharing, media serving, AI-assisted photo sorting, and light Docker container deployment. The system also features a 32GB eMMC module, serving as a dedicated system drive for UGOS Pro. This separation of the OS from user storage pools ensures stability during boot and updates while also freeing up the full capacity of installed SATA drives for data. The lack of NVMe or SATA DOM options means the eMMC is fixed, but in practice, it performed reliably during testing without any bottlenecks.
From a thermal and power standpoint, the RK3588 platform proves to be highly efficient. Even under moderate to heavy loads—such as simultaneous file transfers, streaming tasks, and indexing—the SoC maintains stable temperatures in the low 40s Celsius range. Combined with a modest power draw of around 30–35 watts under active use (and ~4.7 watts at idle without drives), the DH3400 Plus is suitable for continuous 24/7 operation in home or office environments. While the lack of ECC memory or redundant power limits its appeal for enterprise deployment, the core hardware is well-balanced for its target audience, especially when considering how much performance and functionality is packed into a device with a sub-$400 price point.
Component | Specification |
---|---|
CPU | Rockchip RK3588 (8-core, up to 2.4GHz) |
Architecture | 4x Cortex-A76 + 4x Cortex-A55 (64-bit ARM) |
NPU | 6 TOPS AI acceleration (int4/8/16, FP16, etc.) |
GPU | ARM Mali-G610 MC4 |
System Memory | 8GB LPDDR4X (non-upgradable) |
System Disk | 32GB eMMC (internal OS drive) |
Drive Bays | 4 x SATA 3.5”/2.5” HDD/SSD |
Max Storage | Up to 120TB (4 x 30TB) |
Power Supply | 12V / 6A external adapter |
Cooling | 1 x internal base-mounted fan |
Chassis Dimensions | 155 x 155 x 215.7 mm |
The UGREEN DH3400 Plus offers a streamlined but functionally adequate set of connectivity options for a value-focused NAS. On the front of the device, users will find a USB-C port rated at 5Gbps, accompanied by two additional USB-A ports on the rear capable of 10Gbps transfer speeds. These high-speed USB ports are somewhat unusual on an ARM-based NAS and allow for faster direct-attached storage backups or peripheral integration, including UPS management or external media access. While not hot-swappable in the OS UI, the ports performed reliably in tests when mounting USB SSDs and thumb drives for quick file transfer and offline sync. Their positioning also maintains a tidy cable layout, with most high-traffic connections located on the back panel.
Network connectivity is delivered through a single 2.5GbE RJ45 port, which represents a solid step above the standard 1GbE ports found on most entry-level NAS units. This allows the DH3400 Plus to exceed typical Gigabit transfer speeds, reaching up to ~280MB/s in peak file transfer tests using large sequential data. However, the inclusion of only one LAN port means there’s no support for link aggregation or failover. This limits the device’s flexibility in multi-user environments or scenarios requiring redundancy. Given the hardware capabilities of the RK3588 platform, a second LAN port—or even a fallback 1GbE—would have been ideal. Still, for a single-user or small team deployment, the 2.5GbE connection is more than sufficient for everyday access to large media files, backups, and collaborative workspaces.
One of the most unexpected and welcome features is the presence of an HDMI 2.1 output, capable of 4K at 60Hz. HDMI on value-tier NAS systems is uncommon, and its inclusion here enables media playback, real-time NAS control via a directly connected monitor, and potential kiosk or signage applications. The UGOS Pro operating system includes a built-in theater mode to pair with this feature, although its capabilities are more basic than those found in platforms like QNAP’s HD Station. Nevertheless, for users wanting to leverage the NAS as a lightweight local media center or for on-site admin access without relying solely on browser-based control, the HDMI output adds flexibility. The only caveat is that the device lacks any form of PCIe expansion, so users cannot add additional ports or NVMe caching options later.
Port Type | Specification |
---|---|
LAN | 1 x 2.5GbE RJ45 |
USB-C (Front) | 1 x USB 3.2 Gen 1 (5Gbps) |
USB-A (Rear) | 2 x USB 3.2 Gen 2 (10Gbps) |
HDMI Output | 1 x HDMI 2.1 (4K 60Hz) |
PCIe Expansion | None |
Drive Interface | 4 x SATA III (direct backplane) |
Power Input | 12V / 6A DC barrel connector |
And for those concerned about noise, the system was surprisingly low noise, for it’s scale. It will, of course,e depend on the kind of drives you use (anything above around 12TB in a 4 disk configuration will likely be louder than the system fans, regardless), but the UGOS NAS software also has some level of fan control (low/high) too which can be adjusted if needed. Overall, the noise level was pretty good for a 4 disk NAS.
The DH3400 Plus runs UGOS Pro, UGREEN’s proprietary NAS operating system tailored for ARM-based platforms. While it doesn’t match the feature depth of more mature systems like Synology DSM or QNAP QTS, UGOS Pro has seen steady improvements and now covers most of the foundational functions expected from a modern NAS. The UI is browser-accessible and includes modules for storage management, user permissions, cloud sync, backups, and multimedia playback. During testing, the OS handled basic setup, RAID initialization, and file system formatting efficiently. Users can configure shared folders, enable SMB/NFS/AFP protocols, and even set up multi-tiered backup routines to local, USB, or cloud destinations such as Google Drive and Dropbox.
One of the more advanced features is the integrated AI photo engine, which leverages the RK3588’s NPU to organize media libraries using facial recognition, object tagging, and scene classification. This is similar in principle to what Synology Photos or QNAP QuMagie offer, though UGOS Pro lacks the same level of customization or filtering depth. The system can automatically generate albums, identify duplicates, and even create baby-focused timelines.
These features worked reliably on smaller photo sets but began to slow down when indexing larger libraries—likely a result of both memory and algorithmic efficiency. Despite this, the inclusion of AI functionality in a sub-$400 NAS is notable, especially since it runs locally and does not require cloud processing.
Application support is growing within UGOS Pro. Native apps for file access, media playback, surveillance (limited), and mobile sync are included, along with support for Docker containers, allowing users to sideload additional tools not officially available. Notably, BTRFS is supported as a file system option, enabling snapshot functionality and some degree of data integrity checks. However, there are limitations: virtualization is not supported due to the ARM architecture, and there’s currently no iSCSI target support.
Additionally, while the broader UGREEN ecosystem includes a Jellyfin app for other devices, this model did not have native Jellyfin support at launch—requiring users to deploy it manually via Docker. Features like multi-factor authentication, scheduled shutdown/start, and remote access are included, but power users may find the interface lacking compared to more polished systems.
The UGREEN DH4300 Plus delivers respectable media performance when running Jellyfin, particularly with 1080p and 4K content. During testing, the system handled native playback of 4K files smoothly and managed light transcoding tasks without excessive CPU strain, thanks to the efficiency of the RK3588’s integrated GPU and NPU. Native playback of 8K media was technically possible, but pushed the CPU usage to 70–75%, and any attempt at transcoding 8K content resulted in the processor maxing out at 100%, making it unsuitable for high-resolution real-time conversion.
While the NAS can serve 8K files over the network for compatible client playback, its ARM-based architecture lacks the raw transcoding power and hardware acceleration frameworks seen in x86-based systems with Intel Quick Sync. For most users, 1080p and 4K Jellyfin playback—both native and lightly transcoded—is handled reliably, but 8K should be considered the upper limit of what the DH4300 can manage, and only under specific playback conditions without conversion.
The UGREEN DH3400 Plus carves out a unique niche in the budget NAS landscape by delivering hardware typically reserved for higher-tier systems at a much lower price point. Its RK3588 processor, 8GB of RAM, and support for 2.5GbE networking place it well ahead of most similarly priced competitors in terms of raw specifications. Additionally, features such as HDMI output, 10Gbps USB ports, and local AI-powered photo indexing are rare to find in entry-level NAS systems. Despite its plastic-heavy internal design and lack of expansion options like PCIe or M.2, the device delivers stable performance for file sharing, media access, and low-intensity AI workloads. It is not suited for power users demanding virtual machines or advanced snapshot automation, but within its class, the DH3400 Plus presents an appealing balance between cost and capability.
That said, the software experience is still a work in progress. UGOS Pro covers the essentials and offers a visually accessible UI, but lacks the advanced features and ecosystem integration found in more mature platforms like Synology DSM or QNAP QTS. Docker and snapshot support add welcome flexibility, but the absence of native Jellyfin, iSCSI, and VM functionality limits its use in more complex environments. Still, for home users, media collectors, or small office setups looking for reliable backup, modest AI-enhanced photo sorting, and smooth 4K playback, the DH3400 Plus delivers value well beyond its price tag. While it won’t replace high-end NAS appliances, it serves as a capable, efficient, and quietly innovative option in a saturated entry-level NAS market.
Buy the UGREEN DH4300 on Amazon @409 | Buy the UGREEN DH4300 on UGREEN.COM | Buy the UGREEN DH4300 on B&H |
![]() |
![]() |
![]() |
PROs of the UGREEN DH4300 NAS | CONs of the UGREEN DH4300 NAS |
|
|
This description contains links to Amazon. These links will take you to some of the products mentioned in today's content. As an Amazon Associate, I earn from qualifying purchases. Visit the NASCompares Deal Finder to find the best place to buy this device in your region, based on Service, Support and Reputation - Just Search for your NAS Drive in the Box Below
![]() |
![]() |
Vous connaissez cette sensation quand votre imprimante vous lâche au pire moment ? Eh bien en 1980, Richard Stallman a eu exactement le même problème… sauf que lui, au lieu de râler dans son coin comme nous tous, il a décidé de déclarer la guerre à toute l’industrie du logiciel propriétaire. Résultat, ça a donné GNU, Linux, et la moitié d’Internet tourne aujourd’hui grâce à ce geek barbu qui danse le folk bulgare entre deux lignes de code.
J’vous jure, le personnage est savoureux. Il explique que dans l’Église d’Emacs, utiliser “vi” n’est pas un péché mais une pénitence et que l’exorcisme consiste à taper “M-x butterfly” !
Le XKCD à ce sujet
Après Emacs, il s’attaque à GCC (GNU Compiler Collection) en 1987 et là, coup de génie : GCC devient rapidement meilleur que tous les compilateurs propriétaires. Plus rapide, plus optimisé, supportant plus de langages et d’architectures. Les entreprises commencent à l’adopter massivement, prouvant ainsi que le logiciel libre peut être techniquement supérieur.
Bon maintenant, on va pas se mentir, Stallman c’est pas le mec le plus facile à vivre. Le personnage a ses… disons… excentricités.
Déjà, côté hygiène et convenances sociales, c’est… comment dire… spartiate. Le mec refuse catégoriquement d’avoir des cartes de fidélité (surveillance), un téléphone portable (traçage), et ne porte jamais de vêtements avec des logos ou des messages (pub). Et pour naviguer sur le web, il envoie l’URL par email à un daemon qui télécharge la page et la lui renvoie, comme ça il évite d’être tracé par du JavaScript !
Et puis il y a cette anecdote qui a fait le tour d’Internet… Pendant une conférence, Stallman s’est déchaussé en plein milieu de sa présentation, a commencé à gratter son pied, et… bon, disons qu’il a mangé ce qu’il a gratté. Devant tout le monde. Filmé. Pendant qu’il répondait aux questions sur le logiciel libre. Beuuuurk !
Quand on lui a demandé des explications plus tard, il a refusé de regarder la vidéo, disant qu’il y avait “plusieurs obstacles techniques et éthiques” qui l’empêche de visionner des vidéos sur Internet. Mais là où Stallman m’a vraiment surpris, c’est avec sa passion pour la danse folk internationale. Le mec est fan absolu de danses traditionnelles, en particulier bulgares et hongroises. Il y a même une photo mythique où on le voit jeune, en costume traditionnel bulgare, en train de danser… avec une Machine Lisp !
Sa célèbre “Free Software Song” (Chanson du Logiciel Libre) est d’ailleurs chantée sur l’air d’un folk bulgare, “Sadi moma bela loza” et les paroles sont “Join us now and share the software / You’ll be free, hackers, you’ll be free…” Même dans ses hymnes révolutionnaires, il glisse ses références à la danse folk !
Sauf que Stallman, il est pas d’accord avec cette appellation. Pour lui, c’est “GNU/Linux”, parce que le système complet, c’est GNU avec le noyau Linux. Et techniquement, il n’a pas tort car sans bash, sans GCC, sans glibc, sans tous les outils GNU, Linux tout seul c’est juste un noyau qui sert à rien.
Cette querelle de nommage, ça fait 30 ans que ça dure. Stallman insiste pour qu’on dise “GNU/Linux” à chaque fois et il corrige même les journalistes en pleine interview. Certains trouvent ça pénible, d’autres respectent sa position. Perso, j’trouve qu’il a raison sur le fond, même si l’obsession peut sembler excessive. Et à cause de lui, il y a aujourd’hui des armées de super relous qui comme lui reprennent n’importe qui sur le net qui oserait écrire “Linux” dans placer “GNU” devant.
Une autre des innovations les plus géniales de Stallman, c’est le concept de copyleft. C’est un jeu de mots sur “copyright” sauf qu’au lieu d’interdire la copie (all rights reserved), le copyleft garantit que le logiciel reste libre (all rights reversed).
Concrètement, avec la GPL (General Public License) créée par Stallman et l’avocat Eben Moglen, vous pouvez maintenant copier, modifier, redistribuer un logiciel… mais toute version dérivée doit rester sous la même licence libre. C’est ce qu’on appelle l’effet “viral” de la GPL, et l’idée est brillante : utiliser le système juridique du copyright pour garantir la liberté plutôt que pour la restreindre.
Bon, évidemment, je ne vais pas faire l’impasse sur l’un des aspects les plus controversés de Richard Stallman… En septembre 2019, Stallman a été obligé de démissionner de la présidence de la FSF et de quitter le MIT après des commentaires maladroits sur l’affaire Jeffrey Epstein et Marvin Minsky. Le mec a toujours eu des positions tranchées et une façon très littérale d’analyser les choses, ce qui l’a mis dans la sauce plus d’une fois. Ses écrits passés sur des questions de société ont également refait surface et ont provoqué un tollé.
Mais en mars 2021, surprise, il est revenu au conseil d’administration de la FSF. Ça a provoqué une nouvelle controverse, avec plus de 2400 signatures demandant son départ et des organisations comme GNOME, Mozilla et Red Hat qui ont pris leurs distances. Mais la FSF a maintenu sa décision.
Et en septembre 2023, nouveau coup dur… Stallman révèle qu’il a un lymphome folliculaire, c’est à dire un cancer du système lymphatique. Mais fidèle à lui-même, il reste optimiste : “C’est un cancer à croissance lente. Le traitement l’a mis en rémission, et je peux espérer vivre encore de nombreuses années.”
Et aujourd’hui, en 2025, qu’est-ce qui reste de la révolution de Stallman ? Et bien… quasiment tout ce qu’on utilise au quotidien !
Linux (pardon, GNU/Linux ^^) fait tourner 96 % des 500 plus gros supercalculateurs, tous les smartphones Android (3 milliards d’appareils), la majorité des serveurs web et les outils GNU sont partout. Git, créé par Torvalds, s’inspire directement de la philosophie du libre. Des langages comme Python, Ruby, Rust, des frameworks comme React, Vue.js, des systèmes comme Kubernetes, Docker… tout ça existe grâce à l’écosystème que Stallman a initié.
Même Microsoft, l’ennemi juré historique, a fini par embrasser l’open source. Ils ont racheté GitHub, contribuent au noyau Linux, et ont sorti WSL (Windows Subsystem for Linux). Si on m’avait dit ça en 1990, j’aurais cru à une blague !
À 72 ans, malgré son cancer, Stallman continue son combat. Il reste le “Chief GNUisance” du projet GNU, donne encore des conférences, écrit des articles et sa nouvelle bataille, c’est contre l’Intelligence Artificielle, car pour lui, les IA propriétaires comme ChatGPT ou Midjourney représentent un danger énorme pour nos libertés. Il prône donc le développement d’IA libres, transparentes, dont on peut examiner et modifier le code et les données d’entraînement.
Maintenant, quand je regarde l’évolution du numérique aujourd’hui, surveillance de masse, algorithmes opaques, dépendance aux GAFAM, IA boîtes noires…etc, j’me dis que Stallman avait vu juste il y a 40 ans et ses 4 libertés fondamentales sont plus que jamais d’actualité.
Alors la prochaine fois que vous utilisez Firefox, VLC, LibreOffice, WordPress ou même Android, ayez une petite pensée pour ce geek barbu excentrique et agaçant qui a préféré la liberté de tous à son confort personnel.
Sources : The GNU Project, FSF History, GNU Initial Announcement (1983), Stallman.org, Wikipedia - Richard Stallman, Washington Post - Stallman’s MIT Resignation, FSF Statement on Stallman’s Return (2021), The Register - Stallman’s Cancer Diagnosis
[Précommande] Annoncée depuis plus d’un mois, les consoles portables ROG Xbox Ally, conçue en partenariat avec Asus, arrivent bientôt. Les précommandes ne devraient plus tarder à s’ouvrir. Voici tout ce qu’il faut savoir sur leur date de sortie, leur prix, et les différences entre les deux modèles.
La Commission européenne dit s'engager à ce que l'Europe achète pour 40 milliards d'euros de puces spécialisées dans l'IA. Mais comme pour d'autres promesses liées à l'accord entre l'Union européenne et les USA, cette perspective ne semble pas contraignante.
Dans une vente aux enchères autour de l'histoire d'Apple, un Apple I fonctionnel est en vente. En plus de ça, il est signé par Steve Wozniak et Daniel Kottke ; vous pouvez l'acheter.
Avec ce tutoriel, apprenez à configurer une interface réseau physique sur un pare-feu FortiGate, avec l'interface web de FortiOS ou la ligne de commande.
The post Pare-feu FortiGate : comment configurer une interface réseau physique ? first appeared on IT-Connect.
Nintendo joue la carte de la continuité avec la Switch 2, qui reprend le concept hybride ayant fait le succès de la première génération : une console à la fois portable et de salon. Si l’esprit reste le même, cette nouvelle version apporte plusieurs améliorations. Voici les principales différences à connaître pour savoir si le saut vers la Switch 2 en vaut la peine.