Ryzen AI Halo

The latency change ain't caused by the interconnect or packaging, it's caused by the use of lpddr with really loose timings. Point here, the product focuses on bandwidth for the igpu and not cpu gaming perf. It's a mobile workstation design that focuses on raw compute and not games That's a sea of w

You've got to be joking, the Steam machine is practically just valve dumpster diving for parts AMD has been rotting on shelves the 7600M and whatever laptop GPUs is in the steam machine is very weak. Can't compete with the Strix halo at all.

Not most accounts, it is objectively significantly faster than legion go 2. I just finished up my review of the 1X player X2 with the G3E and I'll have that published tomorrow. Long story short unless the game is extremely preferential towards AMD, there's no way that the 890 M can compete. It's a f

Your premise is actually based on a misunderstanding of the architecture. The Ryzen AI Max+ 395 and 388 (AMD 'Strix Halo' family) do not use a multi-CCD layout with cross-latency penalties like desktop chips. Both the 395 and 388 are single-die, monolithic APUs. Because there is no second CCD, there

I went pc but still have 2 Xbox consoles. I chose an AMD Strix Halo platform since it’s about as good as a 3060 class and does 4k 120 fine. My biggest grove is the tone ring as it’s running on Ubuntu Plasma. I tried SteamOS and Bazanite but it failed to work with Dolby 7.1. So now I am in a Linux ti

I have a series X. Love it. I just got a AMD Strix Halo machine with 128GB. Way way more than you need for gaming. All I can say is you can get a massive gaming experience moving to a Strix halo machine. Yeah it’s not a 5080 class GPU, but who cares when you are playing on a couch 10ft away. Strix H

AAA gaming and locally run AI would require either a full desktop setup or a higher end APU like AMD Strix Halo. Look for systems with Ryzen AI Max+ 395 with at least 32GB, preferably 64GB of RAM. A Ryzen HX370 system with LPDDR5X can run these tasks, but you will have to compromise with lower setti

the future IMO would be laptops with strong iGPUs (whom therefore can use all its RAM as vRAM), stuff like AMD Strix Halo or even RTX Spark Bc it can get all the vRAM for both gaming and local LLMs 😁 (since the best medium size models like Qwen 3.6/Gemma4/Nemotron are around 20-25GB compressed rn, s