Tag: gaming GPUs

  • Nvidia to Increase Prices Due to Higher HBM4 Memory Costs

    Nvidia to Increase Prices Due to Higher HBM4 Memory Costs

    Key Takeaways

    1. Tim Sweeney warns that high prices for graphics memory threaten the premium gaming computer industry, as PC manufacturers can’t compete with AI giants like Nvidia and Google.
    2. RAM prices have surged significantly, with examples showing a jump from $240 to nearly $500 for 64GB Crucial RAM in just a month.
    3. Nvidia is projected to pay over $500 for HBM4 graphics memory by 2026, with memory manufacturers raising prices due to strong demand from Nvidia.
    4. The new HBM4 memory is expected to provide over double the bandwidth of the current HBM3E modules, but at a much higher price point.
    5. Upcoming memory technologies like LPDDR6 and GDDR7 from SK Hynix are set to offer significant speed improvements, with launches scheduled for February at the ISSCC.

    The CEO of Epic, Tim Sweeney, recently raised alarms about the premium gaming computer industry facing serious threats due to sky-high prices for graphics memory. He pointed out that PC and laptop manufacturers struggle to match the financial muscle of AI powerhouses like Nvidia, Google, and Meta, who are willing to spend a lot on their high-end GPU and AI data center initiatives. Sweeney’s concerns come in light of a surge in RAM prices, with one user noting that their 64GB Crucial RAM, which they purchased for $240 a month ago, has now skyrocketed to almost $500. At present, Amazon is offering two 32GB modules for sale, but the price is still significantly higher than what it was last October.

    Market Dynamics

    Interestingly, the $500 price tag is reportedly what Nvidia is getting ready to pay Samsung and SK Hynix for their upcoming HBM4 graphics memory in 2026. Insiders from the industry suggest that memory manufacturers are raising prices for Nvidia by as much as 100%, knowing they have the leverage. The production costs for SK Hynix’s HBM4 memory are expected to increase by 50% since it needs to have its base die produced at TSMC, and this increase will be fully transferred to Nvidia. Currently, SK Hynix provides Nvidia with its 12-layer HBM3E memory modules for around $350 each, while Samsung sells them for $100 less due to delays in certification.

    Future Pricing Trends

    By 2026, the high-end HBM4 memory for Nvidia’s AI chips is predicted to be priced in the mid-$500 range, which is more than double the cost of its HBM3E predecessor from Samsung. Insiders have indicated that Nvidia’s demand for HBM4 is so robust that Samsung Electronics has to ensure a supply, even at elevated prices. This could lead to increased costs for Nvidia’s products since the demand for its GPUs remains strong.

    In addition to the pricing of Samsung’s HBM4 memory, sources in the industry have provided updates on its specifications. Samsung has reportedly revamped the interface and stacking design, achieving a bandwidth of 3.3 TB/s for the 36GB module. The enhancements include “better signal accuracy in high-speed areas by using automatic compensation for the alignment signal (TDQS) of the channel-specific through-silicon via (TSV) path,” which is relevant for processing AI accelerator and LLM-specific data. For comparison, the existing HBM3E modules that Samsung supplies to Nvidia offer a bandwidth of 1.2 TB/s, meaning the HBM4 module will provide over double the bandwidth, albeit at a doubled price.

    Upcoming Innovations

    On top of the new HBM4 specifications, an SK Hynix spokesperson has also reiterated the specs for its LPDDR6 and GDDR7 memory. The LPDDR6 mobile DRAM modules deliver 14.4 Gb/s throughput per pin, featuring innovative low-voltage regulator technology that stabilizes the signal at these enhanced speeds. Conversely, the 24GB GDDR7 graphics memory modules are aimed at high-end gaming and AI inference, boasting speeds of 48 Gb/s per pin—three times the bandwidth of the current SK Hynix GDDR6 modules.

    The next-gen HBM4, LPDDR6, and GDDR7 memory technologies from Samsung and SK Hynix will be unveiled at the International Solid-State Circuits Conference (ISSCC) in San Francisco this February. Samsung is anticipated to start delivering HBM4 modules to Nvidia in the second quarter, on an expedited timeline at double the current price, likely leading to more expensive Nvidia GPUs by 2026.

    Source:
    Link

    Nvidia to Increase Prices Due to Higher HBM4 Memory Costs

     

    Tags: , ,
  • Nvidia 572.XX Drivers Cause RTX 3000 & 4000 GPU Instability

    Nvidia 572.XX Drivers Cause RTX 3000 & 4000 GPU Instability

    Key Takeaways

    1. Nvidia’s new RTX 5000 series drivers (572.XX) are causing significant issues for older GPU generations (RTX 4000 and RTX 3000).
    2. Game developers are recommending users with older GPUs revert to driver version 566.36 to avoid crashes and instability.
    3. Users of RTX 5000 GPUs cannot revert to older, more stable drivers, leading to frustrations with their new hardware.
    4. The RTX 5000 series is not free from instability issues, with reports of crashes, freezes, and performance drops.
    5. Nvidia appears to be prioritizing its compute and server divisions over the gaming market, impacting driver support and stability.

    Nvidia has been relatively smooth sailing with its gaming GPU drivers for a while now. Fans of Team Green often boast about the better driver stability compared to AMD’s offerings, but things have gotten a bit more complicated for Nvidia since the launch of the new RTX 5000 series. Generally, GPU manufacturers support the latest 3-4 generations, but the drivers (572.XX) that brought support for the new RTX 5000 cards are causing significant issues for older generation GPUs. As a result, game developers have started suggesting users revert to older drivers to avoid these instability problems.

    Issues Across Generations

    The RTX 5000 series isn’t immune to instability either; however, the previous generation RTX 4000 and RTX 3000 cards are experiencing more frequent crashes, freezes, artifacting, and frame drops with the 572.XX drivers. Users with older Nvidia GPUs have been voicing their concerns over various games in recent months, leading developers of inZoi and The First Berserker: Khazan to recommend using driver version 566.36 for RTX 4000 and RTX 3000 cards. Reports from Mpr_reviews on X indicate that the 566.xx drivers resolve most issues for these older cards, but this workaround comes at the cost of some DLSS features that are necessary for the latest Nvidia applications.

    Limited Options for New Users

    Sadly, owners of RTX 5000 GPUs are stuck with the latest driver versions as they cannot revert to previous ones to enhance stability, since only the new 572.XX drivers are compatible with their newer cards. Even though Nvidia has been rolling out new driver updates more frequently, the problems surrounding the RTX 5000 series, including supply and hardware challenges for both desktops and laptops, clearly signal that Team Green is not as focused on the gaming market as it once was. Instead, their compute and server divisions have been central to the surge in AI advancements over the last few years.

    Source:
    Link

    Nvidia 572.XX Drivers Cause RTX 3000 & 4000 GPU Instability
    Tags: , ,
  • Nvidia RTX 5050 and 5060 Registered as Budget GPUs in EEC

    Nvidia RTX 5050 and 5060 Registered as Budget GPUs in EEC

    Key Takeaways

    1. The Nvidia RTX 5070 Ti is launching on February 20th with a starting price of $749, while the anticipated RTX 5050 may offer a more budget-friendly option.
    2. The RTX 5050, along with the RTX 5060 and RTX 5060 Ti, has been mentioned in filings, suggesting potential for more affordable graphics cards in the RTX 50 series.
    3. The RTX 5050’s emergence is surprising as Nvidia did not previously release an RTX 4050 for desktops, with budget options being the RTX 3050 and RTX 3060.
    4. If the RTX 5050 includes DLSS 4 features, it could enhance performance significantly, but specific technical details remain unclear.
    5. The launch of the RTX 5060 and RTX 5060 Ti may face delays, and the RTX 5050 might not be available until summer if it comes to market.

    Graphics cards often pose a significant challenge for those looking to create an affordable gaming setup. Lately, the spotlight has been on the Nvidia RTX 5080 and RTX 5090, which are definitely not budget-friendly options. The RTX 5070 Ti is set to launch on February 20th, with a starting price of $749 or more. However, there are some hints from Zotac’s filings with the Eurasian Economic Commission (EEC) that the RTX 5050 might be on its way, potentially offering a more budget-friendly choice for entry-level systems.

    New Additions to the Lineup

    Besides the RTX 5050, the RTX 5060 and RTX 5060 Ti have also been mentioned in the EEC filings. It’s crucial to note that not every product that gets filed with the EEC actually makes it to store shelves. But many believe that the RTX 50 series will eventually need to include more affordable alternatives. Initially, it was thought that the RTX 5050 would be exclusive to laptops. However, Zotac appears to have different ideas, as they plan to produce a desktop version.

    Surprising Developments

    The emergence of the RTX 5050 is a bit unexpected, especially since Nvidia did not release an RTX 4050 for desktops. The RTX 3050 and RTX 3060 have been the go-to budget options from the company, priced between $249 and $330. Gamers on a tight budget have had to look toward alternatives like the Intel Battlemage series or AMD’s Radeon RX 7600. Many analysts believe a desktop RTX 5050 could be priced around $250 or less.

    Features and Specs Uncertain

    If the RTX 5050 incorporates all the DLSS 4 features from the RTX 50 series, it could greatly enhance performance through multi-frame generation. Instead of depending solely on the raw CUDA core power, lower-end graphics cards could leverage Nvidia’s AI tech to improve frame rates. Unfortunately, the Zotac EEC filing did not provide specific technical details. There’s ongoing debate regarding whether the RTX 5050 will utilize DDR6 or DDR7 memory, with rumors circulating based on laptop specifications.

    Looking ahead, the RTX 5060 and RTX 5060 Ti are more likely to be released soon. However, delays in Nvidia’s production may push their launch to later months, possibly beyond March or April. If Zotac’s RTX 5050 does come to fruition, it might not be available until the summer.

    Source:
    Link

    Nvidia RTX 5050 and 5060 Registered as Budget GPUs in EEC

     

    Tags: , ,
  • Zotac AMP Extreme Infinity Specs for 600W Nvidia RTX 5090

    Zotac AMP Extreme Infinity Specs for 600W Nvidia RTX 5090

    Although the RTX 50-series “Blackwell” gaming GPUs, including the impressive RTX 5090, have been officially revealed, there remains much about these cards that is still unclear. Leaked product listings have come to our aid, providing valuable insights into the high-end models from AIB partners.

    Specifications of the Zotac AMP Extreme Infinity

    A recent product listing has brought attention to the upcoming Zotac AMP Extreme Infinity card, which features the RTX 5090 GPU. This card boasts a whopping 600-watt TDP, which is 25 watts more than its competitors. As reported by VideoCardz, this marks the highest TDP recorded for any GeForce card so far. The listing also highlights impressive specs, including 21,760 CUDA cores and 32 GB of GDDR7 memory with a memory speed of 28 Gbps. The only notable difference is a slightly elevated maximum clock speed of 2,467 MHz.

    Launch Date and Pricing

    The RTX 5090 is expected to be available starting January 30 at a price of $1,999. However, it’s not surprising that the AMP Extreme Infinity variant will likely come with a much higher price point. Recent information indicates a mostly optimistic outlook for the RTX 5090’s performance, but this is particularly true for those who are open to utilizing AI technology to achieve significant FPS improvements.

    Source:
    Link

    Image 1
    Tags: ,