Friday, 18 July 2025
28.6 C
Singapore
28.7 C
Thailand
15 C
Indonesia
27.4 C
Philippines

AMD Ryzen AI 300: Up to 40% faster in the latest benchmarks

AMD's Ryzen AI 300 APUs show up to 40% faster performance in leaked Geekbench tests, with major CPU and GPU improvements.

AMD’s new Ryzen AI APUs, the pinnacle of technological progress, have seized the spotlight at Computex 2024. Now, we have a sneak peek at their prowess. Three Geekbench tests have surfaced, unveiling the remarkable features of AMD’s flagship APU. The Ryzen AI 300, a testament to state-of-the-art technology, stands out with its unparalleled number of TOPS in an NPU, competitive CPU performance, and a significantly enhanced GPU. Let’s dive into the specifics.

New benchmarks show impressive results

The Ryzen AI 9 HX 370 appeared in two Geekbench tests and one OpenCL test, offering a peek into CPU and GPU performance. Initially listed as the AMD Ryzen AI 9 HX 170 with a Radeon 880M GPU, AMD updated its naming scheme at the last moment. The HX 170 is now the HX 370, and the GPU should be the 890M, not the 880M. Despite these discrepancies, the results are precise.

In the Geekbench 6.3 test, the Ryzen AI 9 HX 370 scored 2,544 in the single-core and 14,158 in the multi-core tests. For reference, the Ryzen 5 7600 desktop CPU scored 2,739 in single-core and 12,287 in multi-core. This clearly shows that the APU excels in multi-threaded tasks, surpassing the Ryzen 5 7600 while slightly trailing in single-core performance. Notably, the Ryzen AI 9 HX 370 features 12 cores and 24 threads, twice as many as the Ryzen 5 7600, which explains its robust multi-core performance. When compared to its predecessor, the Ryzen 9 8945HS, the new chip leads by about 20% in multi-core tasks, highlighting its significant performance enhancements.

The OpenCL test further solidified the impressive performance of the Ryzen AI 300 APUs. The Radeon 890M GPU (erroneously labelled 880M) scored a staggering 41,995 points, matching the desktop version of Nvidia’s GTX 1650 Ti and the mobile version of the RTX 2050. This score outshines the previous generation’s Radeon 780M by an impressive 40%, leaving us with a sense of wonder about the untapped potential of these APUs.

Considering power efficiency

It is essential to consider power efficiency when comparing these benchmark results. The new APUs do not have a fixed TDP and can scale from 15 watts to 54 watts. The exact TDP for these tests is unclear, but the performance is even more impressive, assuming it is on the lower end.

Although it is early, it seems AMD has achieved significant generational improvements. The AI mobile APUs, in particular, could demonstrate a more substantial performance leap than the upcoming Zen 5 desktop processors. Both the APUs and the CPUs are scheduled for a July 2024 launch, so it will not be long before we can experience these advancements firsthand.

Hot this week

Grok chatbot now barred from referencing Elon Musk or calling itself Hitler, says xAI

Grok chatbot no longer allowed to reference Elon Musk or call itself Hitler, as xAI enforces new rules to stop offensive replies.

Google unveils new AI tools for marketing on Search and YouTube at Southeast Asia event

Google launches new AI tools across Search and YouTube at GML SEA to help marketers boost creative output, reach, and ad performance.

Vivo joins the foldable race with new X Fold5 featuring ZEISS cameras and a huge battery

The Vivo X Fold5 offers foldable power with ZEISS cameras, a 6,000mAh battery, and multitasking software, all in a slim and lightweight design.

Southeast Asia’s enterprise AI evolution enters a new phase

To mark AI Appreciation Day 2025, we gathered insights from senior technology leaders across Southeast Asia to explore how enterprises are scaling AI responsibly and strategically.

Nintendo Switch 2 hits all Best Buy stores on July 17 with limited stock

Best Buy stores will restock the Nintendo Switch 2 on July 17, coinciding with Donkey Kong Bananza’s launch, but only in limited supply.

Grok chatbot now barred from referencing Elon Musk or calling itself Hitler, says xAI

Grok chatbot no longer allowed to reference Elon Musk or call itself Hitler, as xAI enforces new rules to stop offensive replies.

Tenable uncovers remote code execution flaw in Oracle Cloud Code Editor

Tenable discovered an RCE flaw in Oracle Cloud Code Editor that allowed attackers to exploit Cloud Shell; the issue has since been fixed.

Garmin introduces Descent S1 buoy to enhance dive communication and safety

Garmin launches the Descent S1 Buoy in Singapore to improve diver tracking, messaging and safety through advanced sonar technology.

Salesforce expands Hyperforce services in Indonesia with local data residency

Salesforce brings local data residency and AI-driven services to Indonesia with the expansion of Hyperforce and new platform tools.

Related Articles

Popular Categories