Minisforum MS-S1 Max Review – The Best Ryzen AI Max Mini-PC Yet

Minisforum MS-S1 Max Performance

Interestingly, Minisforum has configured the MS-S1 Max with multiple performance modes, essentially giving owners the ability to choose how much of a tradeoff they want to make between performance and acoustics/power consumption.

Altogether, Minisforum offers three desktop modes and a single mode intended for rack mount usage – aptly named “Rack Mode.” Of these, only Rack Mode has any kind of artificial fan speed limitations, though interestingly it does not have the highest TDP.

For the desktop modes, the options are for Performance, Balanced, and Quiet, each with progressively lower TDPs and fan speed limits. Quiet in particular is very laptop-like, being tuned for quick bursts of power while enforcing a rather low average TDP limit overall – and this being why it is still quieter than Balanced mode even though the fan limit is the same. Otherwise we have Performance mode, which lets the SoC fly at the cost of noise, and Balanced mode that does what it says on the tin.

For the sake of our testing, we are running things in Performance mode to see what the MS-S1 Max can do under optimal conditions. Though it will also be assessed as such with noise and power consumption.

Given that this is the fourth Ryzen AI Max system we have reviewed so far, we wanted to take a particular look at how it compares to the other Max systems. And we found some interesting results.

Geekbench 6 CPU

In an apples-to-apples comparison with Framework’s Ryzen AI Max-based Desktop – an even larger system – the Minisforum MS-S1 Max already shows a surprising lead. While single-threaded performance is virtually identical, the system takes a 15% lead in multi-threaded performance. Making it the fastest Ryzen AI Max system we have tested under Geekbench thus far.

More interesting still, we found that we can squeeze out even more CPU performance from the system when changing the DRAM allocation for the underlying Ryzen AI Max processor. Shifting that from 32GB to 126GB – taking all but a few gigabytes of DRAM from the iGPU – we were able to add another 12% to multi-threading performance. With that said, the integrated GPU is the true star of the show of the Ryzen AI Max platform, but it goes to show what the CPU can do when given more resources for itself.

MLPerf Client 1.5

On that note, the MS-S1 Max delivers the same kind of high-performance results we have become accustomed to seeing with other Max systems under MLPerf Client 1.5, with Minisforum’s system edging out the others to claim the top spot in GPU performance as well. The Ryzen AI Max+ SoC will remain performance-bound relative to discrete offerings due to its TDP limits, memory bandwidth, and the sheer amount of hardware applied to the problem. But as far as integrated offerings go, this is very good. Coupled with 128GB of memory, this underscores the Ryzen AI Max+’s value proposition as a system with enough DRAM to run local inference on larger models.

Here, you can see the system running the 80-billion-parameter version of Qwen3 Next. The overall inference rate is not blistering fast, but it is more than fast enough to spit out text at a decent clip..

The Achilles’ heel of these systems is the memory bandwidth to the GPU. Generally, you buy these for the memory capacity and running larger models, rather than using them to run smaller models fast

Next, let us get to the power consumption and noise.