Tag: GDDR7

  • Nvidia RTX 6000D Seeks to Regain Market Share in China

    Nvidia RTX 6000D Seeks to Regain Market Share in China

    Key Takeaways

    1. Nvidia plans to release the RTX 6000D in Q3 2025, targeting two million units by year-end.
    2. The RTX 6000D is built on Blackwell architecture using TSMC’s 4 nm technology and features GDDR7 memory with 1,100 GB/s bandwidth.
    3. CEO Jensen Huang introduced the RTX 6000D during his trip to China, aiming to revive a key market for Nvidia.
    4. Export controls since 2022 have led to significant inventory losses and unrealized revenue for Nvidia, totaling $7 billion.
    5. Competitors like Huawei face challenges in AI silicon production, while Nvidia maintains a strong position in the AI GPU market with the upcoming RTX 6000D.

    Nvidia is set to release the RTX 6000D in the third quarter of 2025, with a goal of shipping as many as two million units by the end of the year. This new graphics card is based on Blackwell architecture and is manufactured using TSMC’s 4 nm technology. It features GDDR7 memory, which offers approximately 1,100 GB/s bidirectional bandwidth—these specifications are crafted to navigate the existing U.S. export limits while still nearing HBM-level performance.

    Jensen Huang’s Commitment

    CEO Jensen Huang emphasized this initiative during his third trip to China in 2023 on July 14, where he introduced the RTX 6000D and other compliant GPUs to important partners amidst tight U.S. scrutiny. This visit highlighted Nvidia’s ambition to revive a market that previously accounted for over 13 percent of its yearly revenue.

    Impact of Export Controls

    Since 2022, the company has felt the weight of export restrictions. Initial bans prohibited the A100 and H100 models, followed by their lower clock-speed replacements, the A800 and H800. The restrictions extended to the H20 and even high-end gaming GPUs like the RTX 4090 and 5090D. Due to these limitations, Nvidia has already recorded $4.5 billion in inventory losses and written off an additional $2.5 billion in unrealized revenue.

    Challenges for Competitors

    Analysts observe that Chinese competitors such as Huawei and Cambricon are still pursuing domestic AI silicon but are hindered by production constraints at SMIC, high costs, bandwidth issues, and a software gap when compared to CUDA. These challenges, along with Nvidia’s near-monopoly in the AI GPU market, position the RTX 6000D favorably if the company can secure enough GDDR7 in the latter half of the year.

    A successful launch would also enhance TSMC’s 4 nm utilization and could help Nvidia regain a considerable portion of its revenue from China, which dropped below five percent after the latest export bans. Thus, the RTX 6000D is not only a commercial opportunity but also a crucial test of how much downgraded AI hardware the U.S. regulators will permit.

    Source:
    link

    Tags: , ,
  • RTX 5050 Desktop Graphics Card Launching in July with GDDR7

    RTX 5050 Desktop Graphics Card Launching in July with GDDR7

    Key Takeaways

    1. Nvidia plans to launch the GeForce RTX 5050 in July, after delays in production.
    2. The RTX 5050 will likely be introduced quietly through a press announcement, unlike the RTX 5060.
    3. It is expected to feature a GB207-300 GPU with 2,560 CUDA cores and a TGP of 130 Watts.
    4. The RTX 5050 may utilize GDDR7 VRAM instead of the previously thought GDDR6 due to supply issues.
    5. Pricing is expected to be around $200, targeting entry-level consumers and competing with the RTX 4060 and Intel Arc B580.

    An earlier rumor suggested that Nvidia intended to unveil the GeForce RTX 5050 desktop around late April. This event, however, did not take place, likely due to delays in production and the company’s primary focus on artificial intelligence. Nevertheless, the budget-friendly graphics card remains on schedule for a launch in July.

    Upcoming Release Details

    Videocardz contacted various AIBs, who confirmed that the RTX 5050 is expected to be released in the next few weeks. In contrast to the RTX 5060, which received some attention at Computex 2025, the RTX 5050 will probably be introduced quietly through a press announcement. Additionally, its laptop version has been seen online several times, but there’s no update on its release date.

    Specifications of the RTX 5050

    The GeForce RTX 5050 is rumored to come with a GB207-300 GPU, boasting 2,560 CUDA cores and a total graphics power (TGP) of 130 Watts. At first, it was thought that the graphics card would incorporate 8 GB of GDDR6 VRAM on a 128-bit bus. However, Benchlife now indicates that Nvidia plans to utilize GDDR7 modules as Intel and AMD have been purchasing the GDDR6 VRAM supply.

    Pricing Expectations

    The exact price of the RTX 5050 has not yet been disclosed, but a price around $200 seems reasonable considering it is aimed at entry-level consumers. With only limited specifications available, it’s challenging to assess its performance accurately. However, we can anticipate that it will compete closely with the RTX 4060 or Intel Arc B580.

    Source:
    Link

    Tags: , ,
  • Intel’s Next-Gen Arc Celestial GPUs to Include GDDR7 Memory

    Intel’s Next-Gen Arc Celestial GPUs to Include GDDR7 Memory

    Key Takeaways

    1. Intel’s Arc GPUs are budget-friendly options aimed at gamers who prefer 1080p and 1440p resolutions, competing with Nvidia’s RTX 4060 series.
    2. A job posting suggests future Intel Arc GPUs may feature GDDR7 memory, particularly for the upcoming “Celestial” models, while “Druid” models may still use GDDR6.
    3. Intel’s Senior Vice President expressed confidence in the performance and software of the upcoming Arc Celestial GPUs, which will use the new Xe3P architecture.
    4. The next-gen Arc Celestial and Druid GPUs are expected to be released later, following the cost-effective Arc B580 and B570 GPUs launched at $249 and $219, respectively.
    5. There is uncertainty about Intel’s plans for high-end graphics cards in the future as the third generation of Arc GPUs rolls out.

    Intel’s Arc GPUs have shown to be good budget-friendly options aimed at gamers who prefer 1080p and 1440p resolutions. These cards come with GDDR6 memory and are in competition with Nvidia’s RTX 4060 series, although they do not compete with high-end graphics cards. As the third generation rolls out, it remains uncertain whether Intel plans to introduce high-end options, but a recent leak suggests that the new cards could use GDDR7 memory, similar to Nvidia’s latest RTX 50-series.

    Job Posting Leak

    The information originates from a supposed job listing for “Memory Tuning and Qualification – Discrete Graphics” at Intel, which was noticed and shared by Haze2K1 on X. The screenshot indicates that the job was posted on April 16 for the Toronto area in Canada. The description mentioned responsibilities to “drive and deliver pre- and post-silicon validation and characterization of GDDR6/7 on Intel’s ARC products,” hinting that future Intel Arc GPUs may feature GDDR7 memory, at least for some models. A comment on the post suggests that the Arc “Celestial” GPUs will include GDDR7, while the “Druid” GPUs might still use GDDR6 memory.

    Intel’s Confidence

    Earlier this year, during a chat with PCGamer, Jim Johnson, who is the Senior Vice President of Intel’s Client Computing Group and General Manager of the Client Business Group, reportedly expressed Intel’s confidence in the upcoming Arc Celestial GPUs, stating that they “have the discrete card and the software that is required to make it perform.” Reports also indicate that the new GPUs will utilize the Xe3P architecture rather than the earlier Xe3. Xe3P is thought to be developed in-house by Intel.

    Anticipation and Pricing

    It will be some time before these next-gen Celestial and Druid GPUs hit the shelves. The Arc B580 debuted in December of last year at a cost-effective price of $249, followed by the Arc B570 priced at $219. More GPUs from the B-series are anticipated to complete the product lineup.

    Source:
    Link

    Tags: , ,

  • Nvidia GeForce RTX 5090: No VRAM ECC for Pro Workloads

    Key Takeaways

    1. The Nvidia GeForce RTX 5090 features the new GB202 GPU with significant hardware upgrades compared to previous models like the RTX 4090 and RTX 3090 Ti.

    2. The RTX 5090 does not include the ability to switch between ECC and non-ECC memory states, a feature available in the RTX 3090 Ti and RTX 4090.

    3. ECC (Error Correction Code) memory is crucial for tasks requiring high data accuracy, such as machine learning, while regular consumers may not need it.

    4. GDDR7 memory specifications now include on-die ECC to handle increased memory densities and improve error correction capabilities.

    5. The RTX 5090 has high-performance GDDR7 memory but its support for ECC features is uncertain, relying on future updates for potential activation.

    Since the Ampere generation, Nvidia has replaced its top Titan card with the 90 series models aimed at both professionals and gamers.

    Significant Hardware Upgrades

    The Nvidia GeForce RTX 5090 features the new GB202 GPU, which shows major hardware enhancements when compared to the RTX 4090’s AD102 and RTX 3090 Ti’s GA102 GPUs. Interestingly, while the RTX 3090 Ti and RTX 4090 allowed users to change the VRAM ECC state in the driver, this function seems to be omitted in the RTX 5090.

    Understanding ECC

    Error Correction Code, or ECC, is a method that allows memory to fix itself. Memory errors can happen due to bit flips during data transfer or when errors arise in the data as memory cells discharge and recharge their energy. This self-correction is achieved through either a dedicated memory chip that checks for parity among the other eight chips (known as on-die ECC) or through the memory controller (DRAM ECC).

    Most DDR5 consumer system memory supports ECC, but not fully. DDR5 RAM is designed to spot multi-bit errors but only fix single-bit errors through its built-in checking system. Due to how DDR5 divides 64-bit memory into two 32-bit sections, DDR5-ECC RAM comes in 72-bit (32+4) EC4 or 80-bit (32+8) EC8 configurations.

    When is ECC Necessary?

    ECC memory is not often needed for regular consumer tasks. If this term is new to you, it’s likely you won’t require ECC memory. Nevertheless, ECC is crucial for mission-critical and machine learning tasks where data accuracy must be preserved throughout the entire process. Google faced significant issues back in 1999 when they neglected to use ECC memory, which severely hampered their search engine’s performance due to memory corruption.

    All GPUs that use GDDR5 and GDDR6/6X VRAM have a system for detecting memory errors called Error Detection Code (EDC). Nvidia refers to this as Error Detection and Replay (EDR), a process that requests the retransmission of bits from the memory controller after a cyclic redundancy check (CRC) is completed. EDR helps reduce pixel artifacts when VRAM is overclocked, although this may have a slight negative effect on performance.

    Features of RTX 3090 Ti and RTX 4090

    A lesser-known feature in the Nvidia GeForce RTX 3090 Ti and RTX 4090 desktop GPUs is the ability to switch between ECC and non-ECC memory states via the driver. Unfortunately, this option is missing in the new RTX 5090. Both the RTX 3090 Ti and RTX 4090 incorporate a method known as “soft ECC,” which doesn’t require a separate chip for parity. Instead, activating this feature dedicates some VRAM to act like an on-die ECC module.

    Consequently, this reduces the total available VRAM and memory speed. For the RTX 4090, the usable VRAM drops from 24 GB to 22.5 GB, with 1.5 GB allocated for ECC functions. Activating the ECC state affects performance; for instance, with ECC on in the RTX 4090, 3DMark Speed Way scores saw a 6.4% decrease, and Cyberpunk 2077 2.21 Phantom Liberty experienced about a 5% dip in average fps. The extent of performance loss varies based on the specific task.

    Advances with GDDR7

    With GDDR7, JEDEC has now included on-die ECC as part of the VRAM specifications, recognizing the higher chances of errors due to increased memory densities. GDDR7 employs on-die ECC with a protocol that informs the memory controller about the types of errors that occur. According to JEDEC, GDDR7 can fully correct 1-bit errors and completely detect 2-bit errors, although the detection for rare 3-bit errors drops slightly to 99.3%.

    Moreover, the official specifications also include command address parity with command blocking (CAPARBLK) to enhance the reliability of the command address bus. However, it remains uncertain whether Blackwell’s memory controller utilizes this on-die ECC functionality by default.

    Specifications of RTX 5090

    The RTX 5090 is equipped with 512-bit GDDR7 memory rated for an impressive 1.792 TB/s bandwidth at a rapid 28 Gbps clock, which could lead to transmission errors. Furthermore, Nvidia is promoting the RTX 5090 for AI workflows, which could gain from ECC when processing large datasets. However, Nvidia’s architecture whitepaper only mentions support for “Enhanced Cyclic Redundancy Check (CRC) for Reliability, Availability, and Serviceability (RAS),” which does not equate to ECC.

    While it would be reasonable to assume that Nvidia would activate GDDR7’s on-die ECC capability for the anticipated Blackwell workstation GPUs, it is still unknown if the ECC state toggle will be available for the consumer RTX 5090 through a future driver or VBIOS update.

    Source:
    Link

    Tags: , ,
  • New Nvidia GeForce RTX 5060 Details: 12 GB GDDR7 VRAM Option

    New Nvidia GeForce RTX 5060 Details: 12 GB GDDR7 VRAM Option

    Nvidia has finally launched its Blackwell architecture for consumers at CES 2025 last week. In summary, the company has exceeded expectations by revealing not only the GeForce RTX 5090 and GeForce RTX 5080 but also the more budget-friendly GeForce RTX 5070 and RTX 5070 Ti. There are even rumors of cheaper options coming soon in the shape of GeForce RTX 5060 cards.

    Potential New Variants

    Recent speculation suggests Nvidia might be planning to release more variants of the GeForce RTX 5060 than was previously anticipated. To put this into perspective, Intel has just rolled out the Arc B580 (currently priced at $399 on Amazon), which features 12 GB of GDDR6 VRAM. In comparison, Moore’s Law is Dead (MLID) reports that the GeForce RTX 5060 will come equipped with 8 GB of VRAM as standard, but it will utilize the faster GDDR7 standard due to its GB206 GPU.

    VRAM Capacity Insights

    The GB206-based cards are expected to have a more substantial VRAM capacity, similar to what was seen with the GeForce RTX 4060 Ti. Specifically, MLID mentions that Nvidia plans to offer a 12 GB variant in addition to the base 8 GB model, both utilizing GDDR7 VRAM and operating on a 128-bit bus. This reliable source claims both variants will hit the market in March, featuring a GB206 die measuring 183 mm². However, it is still unclear how many of the GB206’s 4,608 CUDA cores will be available in these models.

    Source:
    Link

    Tags: , ,
  • RTX 5090 Laptop VRAM Leak: 50% Increase Over RTX 4090

    RTX 5090 Laptop VRAM Leak: 50% Increase Over RTX 4090

    We shared news in October regarding Nvidia’s GeForce RTX 5090 laptop GPU, which is the top mobile graphics card in the RTX 50 Blackwell series, boasting 24 GB of VRAM. At that time, we mentioned that achieving 24 GB of VRAM on a 256-bit bus would necessitate using 3 GB GDDR7 modules. Thanks to the leaker Golden Pig Upgrade from China, we now have confirmation that the RTX 5090 laptop will indeed come with 24 GB of VRAM.

    RTX 5090’s Unique Features

    According to Golden Pig Upgrade, the RTX 5090 laptop will be the first in the RTX 50 series to utilize these 3 GB GDDR7 modules. This setup allows the GPU to have 24 GB of VRAM on a 256-bit bus. Interestingly, it seems that the RTX 5090 laptop will be the sole RTX 50 mobile card to adopt these 3 GB modules, while less powerful variants such as the RTX 5080 laptop will continue with standard memory configurations. For instance, the RTX 5080 laptop is expected to have a memory buffer of 16 GB.

    Release Timeline

    In other news, Golden Pig Upgrade suggests that gaming laptops featuring the RTX 5090 and RTX 5080 will launch in February. Meanwhile, models equipped with the mid-range RTX 5070 and RTX 5060 are expected to arrive a month later, around the end of March 2025. This aligns with CES, where manufacturers typically showcase their upcoming gaming laptops.

    Performance Expectations

    When the RTX 5090 laptop GPU is released, it may utilize the GB203 Blackwell GPU, the same chip found in the desktop RTX 5080. This GPU is thought to have 10,752 CUDA cores, a 256-bit bus, and a total TGP of 175 W. For comparison, the RTX 4090 laptop has 9,728 CUDA cores, a 256-bit bus, and a max TGP of 150 W. While the difference in CUDA cores is only about 10.5%, the architectural upgrades and increased power consumption should enable the RTX 5090 laptop to deliver a significant boost in performance.

    Anticipated Performance

    So, what kind of performance can we expect from the RTX 5090 laptop GPU? If we look at the RTX 4090 laptop’s performance, we might anticipate that the RTX 5090 laptop will perform similarly to the mid-range RTX 5070 desktop GPU, as the RTX 4090 mobile is like the RTX 4070 Super. Of course, this is a rough estimate of performance, so it’s likely that the RTX 5090 laptop could be even faster.

    Source: Link

    Image 1
    Tags: , ,
  • Massive GeForce RTX 5090 GPU Leaked with GDDR7 Chips

    Massive GeForce RTX 5090 GPU Leaked with GDDR7 Chips

    Just a day after the PCB for the eagerly awaited GeForce RTX 5090 was leaked, the same source has now shared an image of the GPU itself. This picture showcases the GB202, a graphics processor based on the Blackwell architecture, which is set to be the heart of the RTX 5090. The official reveal is anticipated to take place in early January at CES 2025.

    Specifications of the RTX 5090

    At the core of the RTX 5090 is the GB202-300 silicon, marking it as the largest gaming GPU in NVIDIA’s Blackwell series with a remarkable 21,760 CUDA cores. The physical size of the GB202 is considerable, measuring around 63 × 56 mm. The die itself occupies a substantial area of 24 × 21 mm, leading to an overall die size of 744 mm².

    Memory and Performance

    Surrounding the GPU are 16 Samsung GDDR7 memory chips, contributing to a total memory capacity of 32 GB. The combination of numerous chips, the large GPU package, and a high pin count strongly suggests a 512-bit wide memory bus. If the memory speed reaches 28 Gbps, this setup could achieve an astonishing memory bandwidth of 1,792 GB/s, representing a major advancement compared to earlier models.

    Power and Launch Details

    To power this powerful GPU, a sturdy 24-phase VRM solution is implemented. It is anticipated that the card will utilize a single 12+4-pin 12V-2×6 power connector. Given that this connector can handle 600W of continuous power, experts in the industry predict that the Total Graphics Power (TGP) for the RTX 5090 could be between 500 and 550W, or maybe even more.

    Both the GeForce RTX 5090 and RTX 5080 are set to debut in January at CES 2025, with the RTX 5080 possibly arriving on the market ahead of its more advanced counterpart. The existence of samples with board partners, along with the matching NVIDIA branding on the GPU label, further confirms the credibility of these leaks. The only detail still unknown is when production of the GPU will actually begin.

    Source: Link

    Image 1
    Tags: , ,

  • Samsung Launches World’s First 24GB GDDR7 DRAM Module

    Samsung has announced the development of the world’s first 24GB GDDR7 DRAM memory chip, specifically tailored for high-performance GPUs (graphics processing units). This innovative memory chip stands out as not only the largest in capacity but also the fastest in the market. It is anticipated that this chip will be integrated into high-end GPUs in the upcoming year, offering remarkable enhancements in both performance and energy efficiency. Additionally, Samsung was the first company to complete the development of GDDR7 chips, which is a significant milestone in the industry.

    Features of the 24GB GDDR7 Chip

    The newly introduced 24GB GDDR7 DRAM memory chip from Samsung showcases the fastest speed and the highest storage capacity available. Utilizing a technology known as Pulse-Amplitude Modulation (PAM3), this chip can achieve data transfer speeds of up to 40Gbps. This marks a 25% increase compared to the prior generation of GDDR chips, and in optimal conditions, it can reach even higher speeds of 42.5Gbps.

    Manufactured using Samsung’s cutting-edge 5th-generation 10nm process, this memory chip houses 50% more density while maintaining the same physical size as its predecessors. In addition, Samsung has enhanced energy efficiency, with a 30% reduction in power consumption achieved through clever design features, including clock control management and a dual VDD system. To minimize power waste even more, Samsung utilized power gating methods that manage power leakage effectively.

    AI and Future Applications of GDDR7 DRAM

    The GDDR7 DRAM memory chip by Samsung is designed to support demanding applications such as artificial intelligence (AI), large-scale data centers, and intensive graphics processing tasks. It is also expected to find its way into future gaming consoles, personal computers, and advanced vehicles equipped with smart technology.

    Samsung is planning to have this GDDR7 memory chip validated later in the year. Following the validation, shipments to major customers, including AMD and Nvidia, are projected to occur in early 2024. These companies are likely to incorporate this cutting-edge memory chip into their next-generation GPUs.

    In a statement, YongCheol Bae, Executive Vice President of Memory Product Planning at Samsung Electronics, expressed that after launching the world’s first 16Gb GDDR7 chip last year, Samsung has solidified its position as a leader in the graphics memory field. He mentioned that the company is committed to continuing the delivery of advanced memory products to meet the rising demand from sectors like AI.

    Source: Samsung

    Tags: , ,
  • Nvidia GeForce RTX 5090 Laptop Specs Leaked: What to Expect

    Nvidia GeForce RTX 5090 Laptop Specs Leaked: What to Expect

    A recent leak has revealed details about the GeForce RTX 5090 desktop specs. This graphics card is expected to feature 32 GB of VRAM and consume as much as 600 Watts of power. The laptop variant is also set to see an update, according to a new leak from Moore’s Law is Dead. Here’s a glimpse of how the GeForce RTX 5090 laptop might perform:

    Potential Launch Delays

    Nonetheless, the launch could be delayed until later next year since the 24 GB VRAM on a 256-bit bus necessitates 3 Gb GDDR7 modules. Before we see the RTX 5090, it’s possible that the GeForce RTX 5080 will arrive first, which had been rumored to come with 16 GB of video memory. The mid-range GeForce RTX 5070 could also feature 12 GB of VRAM, though it remains uncertain if this will be GDDR6 or GDDR7.

    Speculation on Nvidia’s Strategy

    This aligns with Nvidia’s supposed strategy regarding its desktop models, as the GeForce RTX 5080 is anticipated to debut before the RTX 5090. Sadly, we can only rely on speculation at this point since Nvidia has not provided any official comments about its upcoming graphics cards.

    Moore’s Law is Dead on YouTube has been the source of much of this information.

    Image 1
    Tags: , ,