AMD is making significant strides in supporting Linux by updating its drivers for the forthcoming RDNA 3+ and RDNA 4 graphics architectures. According to reports from Phoronix, the latest Linux kernel driver patches include substantial changes and enhancements aimed at ensuring optimal support for the new GPUs upon their release.
Key updates to the AMDGPU and AMDKFD drivers include improvements in HDCP, ODM, RAS, and Devcoredump enhancements, code cleanups, and the introduction of new features such as VCN activity reporting via sysfs, support for SMU 14.0.1 and SMUIO 14.0.2, updates to DCN 3.1.4 and DCN 3.5.x, fixes for SR-IOV, suspend and resume operations, among numerous other technical advancements.
Additional modifications have been made in request and error handling, improved synchronization of page table freeing with TLB flushes, refactoring, and fixes in code handling GPU resets, and fixes for the secondary graphics pipeline on GC 10.x and 10.3.
These updates highlight AMD's commitment to providing superior Linux support compared to alternatives such as MESA's RADV Vulkan driver, which has become a more open and popular option in recent years due to its openness and extensive community support. AMD is not only aiming to enhance the technical aspects of its product support but also to strengthen its market position by giving users more reasons to choose their products for Linux-based systems.
Due to export restrictions imposed on China, NVIDIA has encountered challenges in processing return merchandise authorization (RMA) requests for its flagship GeForce RTX 4090 GPU. According to the latest reports from Chinese media outlet HKEPC, the company is offering direct refunds in cases where the shipment of defective units to Taiwan is rendered impossible due to political restrictions.
This applies to graphics cards produced by both Taiwanese and local manufacturers. The decision for refunds was made following the US bans aimed at limiting China's technological advancements. While repairs are conducted locally in Hong Kong and China for simple issues such as fan replacements, more complex cases now result in a full refund of the purchase price.
NVIDIA has officially stated that the cause of instability in gaming experienced by many users lies with the 13th and 14th generation Intel processors. According to NVIDIA representatives, there have been numerous complaints over the past few months about memory errors and crashes during shader compilation related to these processors.
The company urges users encountering such issues to direct their complaints directly to Intel. For its part, Intel acknowledges the problems and reports that it is working on analyzing them. As a temporary solution, a reduction in voltage is proposed, but Intel advises users not to independently alter processor settings to ensure stability. Such actions could lead to violations of warranty obligations and other unforeseen consequences.
In light of the current situation, NVIDIA emphasizes the importance of promptly resolving these issues, as the stability of computer systems directly affects the user experience of their products.
Recent leaks from RedGamingTech reveal that the upcoming AMD Ryzen 9000 series significantly outperforms its predecessors. The Ryzen 9 9950X, for instance, is reported to be 45% faster than its Ryzen 7000 counterpart.
Sources note that both the 16-core Ryzen 9 9950X and the 12-core Ryzen 9 9900X show substantial performance gains, attributed to increased frequencies and the new Zen 5 architecture. In Cinebench 2024 benchmarks, the Ryzen 9 9950X reportedly scores around 170 in single-core and 3000 in multi-core tests, marking a 40-45% improvement over the Ryzen 9 7950X.
This leap in performance is due to both the higher clock speeds and architectural enhancements that increase the instructions per cycle. This aligns with rumors of a significant boost in the new generation’s capabilities. The previous generation of Ryzen processors also saw noticeable performance increases, but the Ryzen 9000 series promises an even more significant jump, highlighting AMD’s trend of substantial improvements in its CPU offerings.
Insider reports suggest AMD's strategic move to continue the RDNA 3+ graphics architecture across its APU lineup until 2027. This long-term plan will commence with the upcoming Strix Point APUs, set for release later this year. Revealed at the AI PC summit in China, the RDNA 3+ architecture, an optimized iteration of RDNA 3, is expected to feature in the integrated graphics of these APUs. While details are sparse, enhancements could include increased compute units and architectural optimizations.
Strix Point APUs from AMD, boasting RDNA 3+ technology, are anticipated to offer up to 16 compute units in mono designs and up to 40 in chiplet configurations, substantially outperforming current Phoenix and Hawk Point APUs. Future iterations may see further expansions in compute units and clock speeds, particularly within the upcoming Zen 5 chip variants.
The adoption of RDNA 3+ for an extended period mirrors AMD’s previous strategy with the Vega iGPU architecture, which spanned several Ryzen APU generations, from Ryzen 2000G to Ryzen 5000 series. This consistent approach underscores AMD's commitment to long-term stability and improvement in its graphics solutions.
Moreover, alongside Strix Point, APUs like AMD's Sound Wave have been spotted, likely featuring the latest Zen cores but maintaining or updating the RDNA 3+ graphics core. This indicates AMD’s intention to ensure continuity and refinement in its graphics technology over the coming years, reinforcing the company's position in the competitive graphics market landscape.
Intel is streamlining its processor lineup, transitioning from the Xeon Scalable series to the more straightforward "Xeon 6" moniker. This change heralds the introduction of the Sierra Forest and Granite Rapids processors, equipped with E-cores and P-cores, respectively. The inaugural Xeon 6 series is poised to deliver enhanced performance and efficiency.
The P-core Granite Rapids processors will offer advanced AI capabilities, including support for the MXFP4 data format, which notably reduces next-token latency by 6.5 times compared to prior generations, and can handle models with up to 70 billion parameters like Llama-2.
Conversely, the E-core Sierra Forest processors will provide a fourfold improvement in performance per watt and a more than double increase in rack density compared to earlier generations, aiding in reducing power consumption and advancing sustainability objectives.
Intel plans to launch the Sierra Forest processors in Q2 2024, followed by the Granite Rapids CPUs later that year.
The upcoming Computex 2024 is set to be a landmark event for AMD, as the company is preparing to unveil its new generation of Ryzen processors built on the Zen 5 architecture. The lineup is expected to be extensive, featuring models such as the Ryzen 9000X/X3D (Granite Rapids), Ryzen 9055HX (Fire Range), and Ryzen 9050H (Strix Point), along with other products codenamed Kraken Point and Sonoma Valley. These processors, designed for desktops, will be compatible with AM5 motherboards.
The interest from motherboard manufacturers, particularly ASUS, in the Zen 5 platform is evident from the BIOS updates supporting AMD Zen 5 for AM5 platforms. The new Platform Management Drivers (PMF) also hint at the imminent launch of the Ryzen "Zen 5" desktop CPU family.
Furthermore, the recent ASUS chipset driver updates, which listed the Ryzen 9000 processors, confirm preparations for the new generation's launch. Although it remains unclear whether these processors are for desktop or mobile platforms, it's highly likely that they are based on the Zen 5 architecture.
The surge in AI interest is reshaping SoC development, with firms like AMD and Intel prioritizing Neural Processing Units (NPU) over other core components. Driven by Microsoft’s push for enhanced AI capabilities, both AMD and Intel have adjusted their SoC strategies, elevating NPUs' role.
Discussions on Anandtech forums reveal that AMD's Strix Point APUs, set for release later this year, have undergone significant changes to incorporate a sizeable AI Engine block for improved NPU performance. This shift resulted in reduced System-Level-Cache, impacting potential gains in CPU and iGPU performance. Meanwhile, Intel is channeling investments into its next-gen chips like Arrow Lake, focusing on the AI segment.
As long as the AI hype persists, AMD and Intel are expected to continue funneling resources into enhancing NPU speeds, possibly sidelining the advancement of other features in the process.
Intel has introduced new Core Ultra "Meteor Lake PS" and Core "Raptor Lake PS" CPUs, tailored for edge computing - a network design that processes data at the periphery of the network, closer to the source of data. This approach reduces latency and enhances processing efficiency. These processors, compatible with LGA 1851 and LGA 1700 sockets, are engineered for AI-centric tasks and support Intel's Arc integrated GPUs and AI Boost NPUs.
The PS series is bifurcated into two categories: the high-performance Core Ultra based on the Meteor Lake architecture, and the Core series founded on the Raptor Lake architecture, both aiming to optimize edge computing performance. The Core Ultra "Meteor Lake PS" series represents Intel's venture into socketed versions for edge devices, signifying a strategic expansion of Intel's CPU offerings for edge computing environments.
Moreover, Intel has expanded its Arc GPU lineup for edge applications, emphasizing its commitment to enhancing computational power and efficiency at the network's edge.
NVIDIA is rumored to be gearing up for the launch of its next-gen gaming GPUs, the GeForce RTX 5090 and RTX 5080, in the fourth quarter of 2024. These new graphics cards, part of the upcoming "Blackwell" architecture, are anticipated to follow the RTX 40 "SUPER" series which debuted earlier this year with the Ada Lovelace architecture.
The RTX 5090 and RTX 5080 are expected to enhance performance and value, targeting high-end and top-end market segments first. This strategic launch is believed to boost shipment volumes for NVIDIA's board partners and increase the average selling price of GPUs, as gamers move towards higher-end models.
Speculations suggest that the RTX 5090 will utilize the GB202 die, while the RTX 5080 will be based on the GB203 die, which is rumored to be half the size of GB202. Both GPUs are anticipated to feature the latest GDDR7 memory interface, supporting up to a 512-bit bus width.
The end of 2024 as the launch timeframe aligns with NVIDIA's historical release patterns, though the exact dates may vary depending on market dynamics.
