AMD has just made some exciting announcements that could potentially shake up the tech industry. This launch is a little unexpected as they have just revealed the new Ryzen 9000 series CPUs featuring new Zen 5 architecture, which is expected to trash Intel's flagship processors in both gaming and productivity. They also have a new updated chipset for the motherboards that support these CPUs with USB 4.0.
There are also some refreshed AM4 CPUs that offer better value, but that's not all. AMD also has some new chips for the laptops that they are squarely marketing at the AI functionality, which means they can handle tasks like voice recognition, image processing, and machine learning.
And if you're looking for some powerful graphics cards, AMD has you covered with its new Radeon PRO GPU and Instinct accelerators, which are designed for professional and enterprise applications. Some of these are actually rather exciting and others fill the gap, let’s get into it!
The highlight of the release is undoubtedly the CPUs engineered with the Zen 5 architecture. These CPUs are crafted using a 4-nanometer process, enhancing several architectural aspects.
They boast improved branch prediction accuracy and latency, higher throughput with wider pipelines and vectors as well as deeper windows size across design for more parallelism.
AMD claims that these advancements will result in an average of 16% IPC (Instructions Per Cycle) uplift compared to Zen 4. Additionally, these CPUs maintain compatibility with all current AM5 motherboards, offering a straightforward upgrade path for users of the 7000 and 8000 series CPUs.
The upcoming product range will maintain the standard offerings from 16 to 6 cores. However, there are notable enhancements in Thermal Design Power (TDP). The flagship 9950X with 16 cores retains its 170W TDP, while the 12-core variant sees a significant reduction, dropping 50W to 120W compared to the previous generation’s 7900X. Even more impressive is the 40W decrease for the 6 and 8 core models, going from 105W to 65W. This reduction is an unexpected and pleasant development. I’m eager to put these CPUs through their paces, and even more so - see how they handle overclocking.
As of now, pricing details have not been disclosed, but the release is anticipated for sometime in July.
In another welcome update, AMD has announced the extension of support for the AM5 socket until 2027, with the potential for further extension. This news coincides with the launch of the new x870 and x870e chipsets, which come equipped with several advanced features. USB 4.0 has become a standard feature on all motherboards. Additionally, PCIe Gen 5 support is now available for both GPU and NVMe on all boards, and there’s enhanced support for EXPO memory, promising higher performance. These new chipsets retain compatibility with the previous generation’s Zen 4 CPUs, allowing for a seamless transition for users upgrading their systems.
While the AM5 motherboards are receiving considerable attention, AMD continues to support the AM4 platform. In a surprising move, AMD has introduced two refreshed CPUs: the 5900 XT and the 5800 XT. The original 5900X, a 12-core processor, has been upgraded in the XT version to feature 16 cores. This enhancement could offer an attractive option for those seeking powerful yet cost-effective solutions for their workstations. At our end, the 5950X is still part of our editing rigs, and it’s performing without a hitch.
The upcoming laptop chips have undergone several name changes and will now be introduced as the AMD Ryzen AI 300 Series, deliberately bypassing the 100 and 200 series. This decision is presumably to avoid any perception of being behind Intel, which already has a 100 series. In my opinion, AMD could benefit from leveraging its AI technology to create a unique and independent naming convention, rather than engaging in a competitive naming strategy with Intel.
Regarding the chips’ features, they are prominently advertised for their NPU, or Neural Network Processing Unit, which is a highlight of the product. AMD boasts that these NPUs can deliver up to 50 TOPs of performance, with TOPs being a metric that quantifies the processing speed of an NPU by tallying the number of operations it can perform in a trillion per second.
In terms of standard specifications, these chips are built on the new Zen 5 architecture and can have up to 12 cores and 24 threads. They also incorporate the latest RDNA 3.5 graphics architecture, featuring up to 16 compute units. The initial models to be released will be the Ryzen AI 9 HX 370 and the Ryzen AI 9 365.
The upcoming laptops will feature AMD’s XDNA 2 technology, which boasts a fivefold increase in computing power and double the energy efficiency compared to its predecessor. These enhancements enable the NPUs to reach up to 50 TOPs. The introduction of the New Block FP16 NPU, supporting both 8-bit and 16-bit models, is another notable improvement. While these advancements are significant, the practical benefits for the average consumer are yet to be determined. AMD has showcased data that positions them ahead of competitors like Snapdragon X Elite, Apple M3, and Intel chips, but independent reviews and tests will be crucial for definitive assessments.
In the realm of GPU performance, the RDNA 3.5 is expected to show notable advancements over Intel’s offerings, which is anticipated to be quite intriguing upon release.
And while we are on the subject of GPU’s, AMD has unveiled the Radeon PRO W7900DS, a professional two-slot GPU, priced at $3499. It boasts a substantial 48 GB of GDDR 6 ECC memory and maintains a total board power of 295 watts. Accompanying this release is the AMD ROCm 6.1, which supports this GPU among others, and introduces multi-GPU capabilities for scalable computing tasks. This product will hit the shelves on June 19th.
The emphasis of this launch is on delivering cost-effective performance compared to NVIDIA’s offerings. For those with a tight budget who require data processing capabilities compatible with AMD GPUs, this could be a viable choice.
We’d love to hear your input on whether you’d like to see coverage of these GPUs. Are there any specific tests or applications you’re interested in for evaluation? Please let us know your preferences in the comments below!
And lastly update to the AMD’s Instinct range of accelerators which are engineered specifically for data centre use, excelling in high-performance computing, artificial intelligence, and data analytics. These GPUs are compatible with AMD’s ROCm open software platform.
In 2024, AMD plans to unveil the MI325X under the CDNA 3 architecture, featuring HBM3E memory to enhance memory capacity and bandwidth. This release is anticipated to significantly benefit organizations with heavy AI processing demands, such as Microsoft, who currently utilize the MI300X cards in their data centres. The MI325X is expected to provide 50% more memory than the MI300X and potentially twice the memory of Nvidia’s H200, with a total of up to 288GB.
It also boasts a bandwidth of up to 6 TB/s, surpassing the Nvidia H200 by 1.2 TB. An impressive setup of eight MI325X units can deliver a peak theoretical BF16/FP16 performance of 10.4 Peta Flops and a combined memory of 2.3 TB of HBM3E. This model is anticipated to be available in Q4 of 2024. Looking further ahead, the CDNA 4 architecture, expected in 2025, will introduce the next-gen system with a 3 nanometre process family, 288 GB of HBM3E memory, and added support for FP4 and FP6 operations, marking significant progress in the field.
To wrap this all up, it’s clear that AMD is not slowing down in their technological advancements. The early arrival of the new CPUs could be a strategic play against Intel’s expected updates later this year, or it might have been AMD’s plan all along. Either way, the Zen 5 chips are definitely something I look forward to testing.
Comments