Tag: Zen 6

  • AMD Zen 6 Leak: 50% Boost in 3D V-Cache and Core Count

    AMD Zen 6 Leak: 50% Boost in 3D V-Cache and Core Count

    Key Takeaways

    1. AMD plans to increase the CPU core count per CCD to 12 for Zen 6 CPUs, a 50% increase from the 8-core CCDs in Zen 5.
    2. Zen 6 CPUs will feature larger L3 cache sizes, with around 48 MB per CCD for client models and up to 128 MB for server models.
    3. The potential total core count for the flagship Zen 6 chip could reach 24 cores and 48 threads.
    4. Future Ryzen 7 X3D CPUs might offer 12 cores and a total of 144 MB of L3 cache, significantly enhancing gaming performance.
    5. Current information is speculative, and final specifications for Zen 6 CPUs are yet to be confirmed.


    Thanks to a lot of reports that have come out recently, we now have a clearer picture of what AMD is planning with their Zen 6 CPU architecture. The main areas AMD is looking at include increasing the CPU core count, enlarging the L3 cache, and raising clock speeds, which should provide a significant performance boost compared to Zen 5.

    Core Count and Cache Size Updates

    Regarding the increase in CPU core counts and L3 cache sizes, there’s now more evidence about AMD’s intentions. HXL shared information on X, indicating that AMD plans to raise the core count per CCD to 12 for the Zen 6 CPUs. This marks a 50% increase compared to the current Zen 5 Ryzen 9000 CPUs, which utilize 8-core CCDs. Therefore, the flagship Ryzen 9 9950X uses two 8-core CCDs. If this trend continues, the Zen 6 flagship chip could potentially feature as many as 24 total cores and an impressive 48 threads.

    Plans for Epyc Server CPUs

    Moreover, AMD might also be working on 32-core CCDs for their Epyc server CPUs. To fit 32 cores on a single CCD, the cores will likely need to be notably smaller than those found in the client Zen 6 CPUs. Thus, it’s expected that there will be 32 Zen 6c cores per CCD rather than the standard Zen 6 cores.

    As previously mentioned, Zen 6 CPUs will also feature more L3 cache compared to their Zen 5 predecessors. HXL reports that the client “Olympic Ridge” Zen 6 SKUs will have around 48 MB of L3 cache per CCD, while server chips could see a massive 128 MB L3 cache per CCD.

    Comparison with Current Models

    For context, the Ryzen 9 9950X provides 32 MB of L3 cache per CCD, resulting in a total of 64 MB. Thus, the increase in L3 cache could mean that the dual-CCD Ryzen 9 Zen 6 flagship could reach 96 MB total. This is quite significant for average users, and a per-CCD L3 cache of 48 MB is also substantial for gamers.

    The Ryzen 7 9800X3D, which is currently the fastest gaming CPU available, has one CCD with 8 cores, 16 threads, and 32 MB of L3 cache. With a single 64 MB 3D V-Cache stack positioned below the CCD, the total L3 cache amounts to 96 MB. If Zen 6 indeed offers 48 MB of L3 cache per CCD, then the successor to the Ryzen 7 9800X3D could potentially have 112 MB of total L3 cache, assuming AMD maintains the same 3D V-Cache amount.

    Potential for a New Gaming Champion

    This is where things become really intriguing. Kepler_L2 responded to HXL’s post, suggesting that AMD might also raise the total 3D V-Cache from 64 to 96 MB. When combined with the 48 MB of L3 cache for a Zen 6 CCD, a future Ryzen 7 X3D CPU could feature 12 cores and a whopping 144 MB of L3 cache.

    In summary, if AMD does launch a Zen 6 X3D chip with 144 MB of cache, 12 cores, and higher clock speeds than the Ryzen 7 9800X3D, Team Red might just hold onto the gaming performance title for another generation. However, this information is still speculative and based on rumors. The exact specifications of the final Zen 6 CPUs remain to be seen.

    Source:
    Link

  • Zen 6 CPU Cores Outperform Intel E-Cores in Power Efficiency

    Zen 6 CPU Cores Outperform Intel E-Cores in Power Efficiency

    Key Takeaways

    1. AMD’s Zen 6 “Olympic Ridge” processors may feature a maximum core count of 24, utilizing new 12-core CCDs.
    2. Each CCD will have an increased L3 cache of 48 MB and could reach clock speeds up to 6 GHz with TSMC’s N2X process.
    3. The addition of two Zen 5 LP cores in the Olympic Ridge CPUs promises energy-efficient performance, achieving 65-75% IPC and 50-60% clock speeds compared to full Zen 5 cores.
    4. Zen 5 LP cores are expected to have clock speeds between 2.5-3.5 GHz and consume 1 W or less, making them more efficient than Intel’s E-cores.
    5. The launch of Zen 6 processors is anticipated in the second half of 2026, so current leaks should be viewed with skepticism.


    It’s apparent to those keeping an eye on the latest Zen 6 leaks that AMD is gearing up for some major enhancements in their Zen 6 lineup. In particular, the “Olympic Ridge” desktop processors are rumored to see a maximum core count boost to 24, thanks to the introduction of 12-core CCDs. Additionally, there’s an increase in L3 cache, now totaling 48 MB for each CCD. The new Zen 6 CCDs will utilize TSMC’s advanced N2X process, which might enable the boost clock to hit a remarkable 6 GHz. The IOD chiplet is expected to use either the N3P or N6 process.

    Noteworthy Changes

    One intriguing detail that might be missed is the anticipated addition of two Zen 5 LP cores in the Olympic Ridge desktop CPUs. According to Moore’s Law Is Dead, these low-power cores are more than just a gimmick. The leak suggests that these Zen LP cores perform well while also being energy-efficient.

    In conversations with Tom, a source from AMD indicated that the Zen 5 LP cores can achieve approximately 65-75% of the IPC and 50-60% of the clock speeds compared to the full Zen 5 “Classic” cores. These performance assessments were reportedly done using the 4 nm process. Therefore, if AMD opts for N3P for the IOD die, the performance of the Zen 5 LP cores in Olympic Ridge CPUs could vary.

    Impressive Performance

    Based on these findings, MLID predicts that Zen 5 LP cores could reach clock speeds between 2.5-3.5 GHz with an IPC that falls “between Zen 3 and Zen 4.” What’s even more remarkable is that achieving this level of performance demands minimal energy.

    According to one of AMD’s internal presentations shared by MLID, the Zen LP cores are categorized as “Client” products aiming for “Peak Core Efficiency.” The presentation shows that each Zen 5 LP core consumes 1 W or less, making them more efficient than Intel’s E-cores.

    In simpler terms, the Zen 5 LP cores not only provide solid performance to enhance multi-core tasks but also do so while being economical with power usage.

    Caution Ahead

    Even with these exciting updates, it’s crucial to remain cautious, as the Zen 6 processors won’t launch until the second half of 2026. Therefore, it’s wise to view MLID’s insights with some skepticism.

    Source:
    Link


  • AMD Zen 6 CPUs Expected with Up to 96MB L3 Cache on Non-X3D Model

    AMD Zen 6 CPUs Expected with Up to 96MB L3 Cache on Non-X3D Model

    Key Takeaways

    1. The upcoming AMD Zen 6 processors may feature a high-end desktop variant with up to 24 CPU cores, surpassing the previous limit of 16 cores.
    2. AMD plans to increase the L3 cache per CCD to 48 MB, potentially resulting in up to 224 MB of L3 cache for a 24-core CPU with dual V-cache.
    3. A single CCD may house 16 Zen 6 cores, with configurations available for 32-core variants totaling 128 MB of L3 cache.
    4. Future laptop models might combine Zen 5 and Zen 6 CCDs, as suggested by the Strix Point development.
    5. The next-generation Epyc server processors are expected to benefit from the increased cache and new core configurations.


    Now that the AMD Zen 5 series is largely released, the whispers about the upcoming Zen 6 processors are becoming more common. Initially, it appears that the highest-end desktop variant (possibly named Ryzen 9 11,950X) will surpass the previous limit of 16 CPU cores. A previous leak hinted that this number might reach up to 24 cores. A recent report from Weibo provides further details on this topic.

    Increased Cache Capabilities

    AMD is set to boost the L3 cache per CCD to 48 MB. Therefore, a 24-core CPU utilizing 2 Zen 6 CCDs would feature a total of 96 MB of L3 cache. If a 64 MB 3D V-cache tile is added to one of the CCDs, this would result in 160 MB of L3 cache, and if both CCDs have one, that would be an impressive 224 MB. However, it’s still too early to make guesses about what the Zen 6 X3D will offer, though it would be a perfect opportunity to introduce a dual V-cache CPU for consumers.

    New Core Configurations

    Additionally, AMD intends to fit 16 Zen 6 cores within a single CCD, accompanied by 64 MB of L3 cache. There is a 32-core variant with 2 of these CCDs totaling 128 MB of L3 cache, though it remains unclear which lineup it belongs to. If Strix Point is indicative of future developments, some laptop models may feature a combination of Zen 5 and Zen 6 CCDs. Also, this configuration is expected to appear in the next-generation Epyc server processors, which will also gain from the increased cache.

    Source:
    Link


  • AMD Ryzen AI 400 Medusa Point Laptop Chips to Feature RDNA 3 GPU

    AMD Ryzen AI 400 Medusa Point Laptop Chips to Feature RDNA 3 GPU

    Key Takeaways

    1. AMD’s Zen 6 series will be branded as Medusa, with Medusa Ridge for desktops and Medusa Point/Halo for portable devices.
    2. Medusa Point laptops will feature an unknown version of RDNA 3 iGPUs, likely RDNA 3.5, but will not upgrade to RDNA 4.
    3. The Video Core Next (VCN) for Medusa Point may also be limited, affecting performance compared to Nvidia and Intel GPUs.
    4. AMD has confirmed that RDNA 4 will be exclusive to desktop discrete GPUs, explaining the choice of RDNA 3.5 for Medusa Point.
    5. Medusa Halo might benefit from newer GPU technology, but there are concerns about reusing the RDNA 3.5 GPU from the previous Strix Halo generation.


    A recent leak suggests that AMD’s Zen 6 series will operate under the Medusa name, featuring Medusa Ridge for desktop processors and Medusa Point/Medusa Halo for portable devices. It would be logical to think that the new Zen 6 CPU cores would come with an updated RDNA integrated GPU (iGPU). However, according to a new leak from Weibo, this may not be true for laptops.

    Medusa Point’s iGPU Limitations

    Laptops under the Medusa Point label will apparently be using an unknown version of the RDNA 3 iGPUs, likely falling into the RDNA 3.5 category. Unfortunately, these devices won’t receive the highly sought-after RDNA 4 upgrade, which introduces hardware-based FSR 4.0 and other features that come with the new architecture. As a result, the Video Core Next (VCN) may also be limited to the older version, which could create issues as it lacks many of the advanced capabilities of its Nvidia and Intel counterparts.

    AMD’s Confirmation on RDNA Exclusivity

    In a prior statement, AMD made it clear that RDNA 4 would be reserved only for desktop discrete GPUs, so the choice to utilize RDNA 3.5 for Medusa Point isn’t particularly shocking. Meanwhile, RDNA 5 (or UDNA, depending on who you ask) is still planned for the desktop Ryzen 11,000 series components. This brings us to the eagerly anticipated Strix Halo successor—Medusa Halo. A newer generation GPU could significantly benefit from its additional memory bandwidth and the rumored 3D V-cache. It would be quite disappointing if AMD decides to reuse the Strix Halo’s RDNA 3.5 GPU for yet another generation.

    Source:
    Link