- AMD has published official reference points for the Ryzen AI Max+ 395 APU
- Demonstrates a clear cable about the intel core ultra 7 258v
- The tests were performed in LM Studio with several LLM
The official reference points have supported the performance of “Strix Halo” AMD Ryzen Ai Max+ 395 as the most powerful “Apu X86” in the market for the computing of AI.
The AMD Ryzen AI Max+ 395 is a 16 -nuclei (32 -wire) processor with a peak of more than 50 years AI tops XDNA 2 NPU, and integrated graphics of Radeon 8060 (40 Adna 3.5 computer units for a seriously processing power for the form factor. AMD markets it mainly for its management of AI workloads, as in applications such as LM Studio.
This is evident in the published reference points of AMD for the Ryzen AI Max+ 395, which are measured in ‘Tokens per second’ and ‘Time to First Token’ in LM Studio against its competition. Specifically, we see how the new “Strix Halo” processor within the ASUS ROG Z13 flow with 64 GB of RAM is compared to a similar specification Asus Zenbook S14 with 32 GB of RAM.
The last machine has half of the unified memory and is using the APU Intel Core Ultra 7 258V with its integrated ARC graphics baked with registration at 140 V, so it is not necessarily a comparison 1: 1. However, AMD has exhibited the skill of its latest chipset in LM Studio 0.3.11 with “several LLM” with a model size of 16 GB, which shows at least double the double Effective tokens per second with Deepseek R1, Phi 4 Mini Instruction and calls 3.2 compared to its rival.
The leadership becomes more dramatic when comparing time with the first token in text models, with up to 12.2x faster as evidenced by the reference points in Deepseek R1 distill Qwen 14B, with a similar leadership of 11.3x in Phi 4 14b. However, it is not an advantage consisting of all text models, since the Ryzen AI Max+ 395 is 4x to 9x faster in flame 3.2 and other distilled models Deepseek R1.
Ryzen Ai Max+ 395 of AMD It is also claimed that it is up to seven times faster in the Sota vision models in time to the first token, this can be seen in IBM Granite Vision 3.2 2b while the chip is six times faster on Google Gemma 3 12b; However, it compares approximately half -compared to Gema 3 4b.
Powerful performance that should not be too shocking
AMD’s principal “Strix Point” APU is the head and shoulders over Intel Core Ultra 7 processor in a way that should not be surprising for those interested in AI computing. This is because Team Blue hardware was made with the lowest threshold computing in mind, and this can be seen in architectural differences when analyzing the two.
The Intel Core Ultra 7 258v has eight nuclei and eight threads with a maximum impulse clock up to 4.8 GHz and a maximum 37W TDP. In contrast, the AMD Ryzen AI Max+ 395 has 16 cores, 32 threads, an impulse watch up to 5.1 GHz and a predetermined TDP of 55W. However, this TDP is configurable up to 120 W, so it is a night and day hardware difference in the chipsets. Of course, AMD hardware was going to go on top; It is much more powerful in all areas.
Then we have to consider the two tested machines used for the reference points, the differences between the ASUS ROG Z13 flow (a laptop of the leading game) and the Asus Zenbook S14 (an ultrabook of average range). We review the last device at the end of last year, giving it a four -star article, citing the “solid performance” of the Luke Like processor. The chip debuted within this machine (and similar) in September 2024, while the AMD Ryzen ™ Ai Max+ 395 reached the scene this month.
They are not just the laptops of ia that are using the ia ryzen flagship chipset for their performance capabilities, since a mini pc myriad is using them for productivity and even the use of games. It has become a career to launch the most powerful mini PC as in mid -March in mid -May, they go to companies, such as GMKTEC and AOOSTAR, which lead the load.