Samsung faces limitations in using its own Exynos chips in its new flagship series Galaxy S26The key factor was the ongoing contract with Qualcomm, which obliges the company to use Snapdragon processors in most regions. As a result, Exynos 2600 will appear in only a third of devices, while the rest will receive Snapdragon 8 Gen 5 Elite.
Samsung initially planned to return to a more widespread use of Exynos, promoting its 3nm architecture and new AI engine. However, the current agreement with Qualcomm does not give Samsung the freedom to choose the platform for all markets. Furthermore, a clear preference among buyers for Snapdragon only strengthened this strategy.
Devices with Exynos 2600 will be available in limited territories—most likely, South Korea, India, and some European countries. Major markets, including the US, Canada, and China, will again receive Snapdragon processors. This confirms that even with its own processor, Samsung must consider legal and market constraints.
Despite this, the company is not abandoning further development of Exynos. The limited implementation of the chip in the S26 is seen as a test platform for refining the new architecture and collecting feedback before moving on to the next generation.
In this way, The contract with Qualcomm remains the main barrier On the way to complete independence of Samsung in the field of mobile processors.
Despite a sharp rise in the price of RAM, leading manufacturers do not plan to increase DRAM production volumes for the consumer market. According to TrendForce, companies prefer to invest in development of HBM chips for AI, rather than expanding capacity for standard memory. Investments are aimed at optimizing process technologies, increasing density through stacked designs, and producing HBM, which requires more silicon wafers.
Samsung, SK hynix and Micron in 2025, they will spend a total of about $54 billion for capital expenditures, but almost the entire budget will be spent on HBM development. Although the DRAM sector is showing 14% investment growth, the volume of memory bits produced will not increase. Analysts expect that The memory shortage will persist until at least 2027, as demand for AI infrastructure continues to grow.
At the same time, experts warn about ten-year "price apocalypse" amid the AI bubble, pointing out that a real return on investment in infrastructure is only possible with annual AI revenues reaching $650 billion. However, manufacturers are in no hurry to expand factories — it's expensive, risky, and time-consuming. Instead, DRAM lines will be reoriented toward HBM production, which is more logical from a profitability standpoint, but worsens the situation with consumer memory.
Renesas Electronics announced "brain" for server memory new generation. This chip is called RCD (Registered Clock Driver), and its main task is to manage signals on high-speed RAM strips in data centers. Simply put, It's a "traffic light" that keeps data moving quickly. and without failures.
The new sixth generation RCD (Gen6) was the first in the industry to be able to operate stably at speeds 9600 MT/s. It 10% faster previous generation (Gen5), which "accelerated" to 8800 MT/s.
This upgrade critical for AI serversAccording to the vice president of Renesas, the "explosive growth of AI" and the increase in the number of cores in processors require "unprecedented bandwidth" memory.
Samsung has already confirmed, who is "delighted" with the new chip and will use it in their future DDR5 modules for AI workloads. Mass production of these chips will begin in first half of 2027, and that's when we'll see server memory sticks on the market that operate at such breakneck speeds.
The Vivaldi team released a major update to your browser, Vivaldi 7.7This release addresses several long-requested pain points from power users and focuses on control and convenience.
The main "feature" was "smart" synchronizationPreviously, syncing tabs between different PCs was a "flat list." Now Vivaldi 7.7 allows transfer entire Workspaces и Tab Stacks from one computer to another. The entire structure (stacks, organization) is completely preserved. This is ideal for those working from home and in the office.
The second major change was unified start pagePreviously, widgets (weather, RSS, mail) and Speed Dials (speed dial) were separate. In version 7.7, they can be mix freely on one screen, creating a personalized "dashboard".
Also introduced in Vivaldi 7.7 new "Performance" section in the settings. There you can configure it in detail. Memory Saver (Memory Saving) — how aggressively the browser will "put inactive tabs to sleep" to save RAM. You can add exceptions (for example, for pinned tabs).
Intel has released an update XeSS SDK 2.1.1, which brought important innovation for owners of laptops based on processors Core Ultra (Meteor Lake)Now they Integrated graphics officially support XeSS-FG (Frame Generation) - frame generation technology.
This came as a surprise, as it was previously thought that XeSS-FG needed hardware XMX cores required (analogous to Nvidia's tensor cores), which are present only in discrete Arc cards and the new Lunar Lake integrated graphics. In processors Meteor Lake doesn't have these XMX cores. (simpler DPAS units are used there).
With the release of SDK 2.1.1, Intel essentially removed this strict requirementcreating "lightweight" AI model frame generation. This opens the door for significant increase in FPS on millions of already released Meteor Lake laptops, and in the future, perhaps on competitors' cards that do not have XMX.
In addition, the update fixes a number of bugs in XeSS-SR (upscaling), including rare flights on Vulkan and the validation error that occurred on AMD video cards.
Amid the big announcements of the Steam Frame, Steam Machine, and a new controller, fans are left with one question: "Where's Steam Deck 2?" In a new interview, Valve engineer Pierre-Loup Griffe finally explained why the portable version didn't come out and when to expect it. According to him, Valve will not release a successoruntil I'm confident in a "worthwhile enough" performance upgrade.
Griffe said the company "not interested" in the release of Deck 2, which would be all "20, 30, or even 50% more powerful", but would have the same battery life. Valve wants a "more noticeable leap" (a little bit more demarcated) than just a small increase.
The main problem and The battery remains the sticking pointThe original Steam Deck (2022 LCD) is known for its "massive battery life issues" - for example, God of War"ate" a full charge in just 90 minutes.
The engineer confirmed that there is currently a market "there is no such SoC (chip)", which could provide the power boost Valve needed. и efficiency. Despite fans' hopes of seeing an upgrade by this holiday season, Steam Deck 2 will definitely not be released this year..
AMD held its Financial Analyst Day this week, where it shared an updated roadmap and early performance data for its future chips. The highlight of the presentation was a slide dedicated to server processors. EPYC "Venice" on Zen 6 architecture.
According to official data, the new generation will offer more than 70% improvement in productivity and efficiency compared to the Zen 5 "Turin" generation. AMD also promises more than 30% increase in flow density.
This 30% increase in density confirms previous rumors. It is achieved by moving from 192 cores (in Turin) to 256 cores and 512 threads in "Venice" (a 33.3% increase). The remaining performance gain (+70% minus +33% cores) will be achieved through IPC, clock speeds, and other architectural improvements.
The entire Zen 6 lineup, including client chips (codenamed "Olympic Ridge" for desktops and "Medusa" for laptops), will be manufactured TSMC's latest 2nm process technologyThis node uses Nanosheet transistors (GAA), which are 10-15% faster at the same power.
It's important to note that AMD achieved a 70% increase not in AI tests, but in standard server benchmark SPECrate 2017 INTThis demonstrates a significant improvement in classic computing tasks. Rumors also indicate that Zen 6 may move to 12-core CCDs, bringing 24-core processors to the desktop Ryzen segment.
Journalists got access to the prototype Valve's new Steam Machine and shared detailed photos of its interior. The pictures show just how extremely dense layout used by engineers to fit powerful hardware into a tiny 3.8-liter case.
The whole structure is "sandwich" of components, where there is almost no free space. All engineering is built around huge 120mm fan with custom blade geometry that blows air through the body. Valve engineers said they spent "a huge amount of time" on the aerodynamics.
Above the motherboard (with 6-core AMD Zen 4 and GPU RDNA 3) is located massive radiator with heat pipes. Directly under the board is located Chicony's specially designed power supply, which also acts as a shielding element.
Because of such density Upgrade is almost impossibleUsers will be able to replace only the M.2 SSDThe RAM is installed in the form of standard laptop modules (SO-DIMM), but accessing them is extremely difficult. Valve also confirmed that will publish CAD files for customizing magnetic panels of the case.
A new, yet-to-be-announced laptop has been spotted in the Geekbench database. Samsung Galaxy Book6 ProWhat's interesting is its internals: this is the first test of Intel's next-generation top processor. Core Ultra X7 358H from the line Panther LakeJudging by the results, Intel is preparing a colossal leap in integrated graphics performance.
The processor has 16 cores and 16 threads (probably 4 P-cores + 8 E-cores + 4 LPE-cores, without Hyper-Threading) and operated at a frequency of up to 4.78 GHz. But the main thing is the integrated graphics. In the test, it is defined as Intel Arc B390, which confirms the use of the new architecture Battlemage (Xe3).
In the OpenCL test, this integrated graphics showed a stunning result in 57,001 score.
To understand the scale, this is almost twice as much (+92%) more than iGPU в Lunar Lake (about 29,000 points) and 38% faster iGPU в Arrow Lake (around 41,000 points). According to TechPowerUp and iXBT, this performance puts the integrated Panther Lake processor on par with discrete graphics cards primary-middle class, such as NVIDIA GeForce RTX 3050 (both mobile and desktop).
Intel published data on its blog about "significant" increase in productivity in PostgreSQL 18.0, which was achieved thanks to implementation of AVX-512 supportThe new version of the popular database, released in September, now uses modern SIMD instructions for calculating checksums (CRC32C).
Engineers from Intel, AWS, and other companies participated in the development of the patch. Intel called it a "national example" of using modern instructions for acceleration of critical database primitives.
According to Intel's benchmarks on Xeon processors, the performance gains over the old method (SSE4.2) are enormous. The graphs show that when working with buffers of varying sizes (from 256 to 32768 bytes), the speedup ranges from 1.57x to 5.16x. Peak performance gains reaches 5.60x when working with a buffer size of 4096 bytes.
The most important thing in this news is that This is not an exclusive Intel feature.The report specifically emphasizes that this same AVX-512 support Fully works on AMD Zen 4 processors (and newer), including server EPYC and desktop Ryzen.
