Beelink SER10 Max: l’AMD HX 470 atrapa l’M4 Pro?

I've spent a bit of time testing this brand new mini PC, this Sir 10 from BeLink against last year's model, the Sir 9, and the Sir 8 before that. They'll look kind of alike, but inside they're different, because the Sir 10 has AMD's new Gorgon point ship versus the previous strict point. And after doing my developer and LLM, AI tests, I came out with something I didn't expect. Alright, so this is the BeLink Sir 10 Max, they added Max to it. It's got the Ryzen AI 9, Achex 470, codename Gorgon point, but the real shopping question here is 3 machines, not one. If you take a look at the BeLink site, last year's Sir 9, still available, still hot, and still on sale. Usually a couple hundred bucks cheaper. In fact, I've been using mine now for a while for more than a year as a media server. And as a side note, this is what it looks like if you just have it on a rack for over a year. It's got a little bit of dust, but it's not too bad. Really solid machine. The other option is the M4 Pro Mac Mini, which I also have set up here. I'm including that because the M4, the base M4, caps out of 24 gigs, which is really not that great for a lot of the workflows I'm going to show you. And also since we're comparing it to the Sir 10, which is a 32 and 64 gig configurations, and we should go to the M4 Pro. But the M4 Pro though, Apple charges some serious money for that. This is the basic M4 Pro, not even the 20 core GPU, and it's 48 gigs, not the 64 gig version. Did you even have the 64? They got ready to do that? It's good of one terabyte. 10 gigs, Ethernet, suddenly you're at $2,000, and you're only at 48 gigabytes of RAM. So the BeLink site still makes sense. You're only going up by a couple hundred bucks,

and you're away under that. It does come with one terabyte of storage. So the question isn't, is Sir 10 a good upgrade? It's should I get a Sir 10? Sir 9? Or save up for the M4 Pro, or the upcoming M5 Pro, whenever that comes. But this is a performance video, so we're going to get into that. So these days, I'm always flipping between models. GPT for research, cloud for coding, nano banana for image generation, vehicle, cling, and runway for video, six tabs, six bills, and counting. Enter chat, LM teams. One dashboard houses every top, LM, and route LM picks the right one. GPT mini for ultra fast answers, cloud-sandered for coding, Gemini Pro for massive context. They recently added Gemini 3 and GPT 5.1. The moment they dropped, create professional presentations with graphs, charts, and deep research detail content. Need human sounding copy? Humanized rewrites text to defeat AI detectors. Need visuals? Pick frontier or open source models. Nano banana mid-journey flux for images, Magnific for upscaling plus VO, WAN, and SORA for video. All built in. You also get advocacy AI depagent to pretty much do anything. Build full stack apps, websites, reports with just text prompts, and deploy them on the spot. They have advocacy AI desktop, which is the brand new coding editor, and assistant that lets you vibe code and build production ready apps. And the kicker? It's just $10 a month, less than one premium model. Head over to chat. LM.abbecause.ai or click the link below to level up with chat. LM teams. By the way, if you don't know what a gorgon is, this is a gorgon. I don't know why M.D. has to go with scary things for their chipmaps. This one will turn you into rock if you make eye contact. So don't do it. Some quick specs on this chip. We've got 12 cores and 24 threads. Zen 5 architecture plus Zen 5c 8 of those. And it boosts up to 5.2 gigahertz.

Check it out. Shared memory. Half of the memory can be shared. 29 gigabytes of my case. So AMD's marketing material obviously will include the MPO. The use cases are limited, but I'll show you how we can use that later on in the video. And the IGPU is here too. This happens to be the radion 890m. If that sounds familiar, you're right. It's the same one that was in the Serenite. More than that later. The memory here, I have 64 gigabytes. On my machine, but it does come with 32. And this is DDR5, 5600. User configurable and user upgradable, which is different than the Serenite was. You can also upgrade the storage up to 8 terabytes if you can find it and afford it. A few other life improvements like USB 4, HDMI 2.1, display port 1.4. And 4k at 240 Hertz triple monitor support. And 10 gigabit network. Ethernet. So that's pretty cool considering the Serenite only had 2.5. Now if you're so inclined, there's also nice looking orange one. Is it orange or red? I don't know, but it's got open claw pre installed on it. And they're charging a 100 bucks more for it. But look at this option right here. 96 gigabytes with 2 terabytes SSD for $2300. This 96 gigabytes for that much is actually not bad right now. I spec the 2 terabyte version of a Mac Mini. We're at $2500 bucks and we're half the ramp. But it's kind of funny that they're charging a 100 bucks more for the open claw version of this. And it's definitely not as bad as that guy on Craigslist who is charging $1500 for base Mac Mini with open claw pre installed on it. All right let's start light. The Domino 3 is a JavaScript benchmark. It's a similar real web application interactions. The kind of thing you feel every time you open a sluggish site single threaded and everyone's got a frame of reference for this one. The Serenite hit 31.5 in the seven way comparison video that I did not too long ago.

That one had a bunch of other mini PCs included. Last year the M4 Pro Mac Mini got 45.2 in that. The base M4 got BDM4 Pro, got 47.8 but they're close. Apple is kind of hard to beat in single core scores right now. They've got that market corner. And we'll see that pattern again. Now, speedometer is a JavaScript runtime web tooling benchmark is another one that I use by V8. Yeah those are the same V8 people that actually write the engine. This tested JavaScript 2 chain like TypeScript, Babel, Tercer and stuff like that. Here the Serenite came in at 20 on the TypeScript score. An 18.8 on the geometric mean. Both backminis are way ahead 35.99 each on TypeScript by the way the M4 base and the M4 Pro tie on that one. The GOMIN is where they separate a little bit. And after running the Serenite test it landed on 34.14. That's where TypeScript and 28.62 for the GOMIN. Okay hold on. Now that's 65% jump on TypeScript over the Serenite. 65 and the GOMIN is up over 50% here too. This is by far the biggest generation of our generation win I've seen so far for this machine. And look at that TypeScript score. It's within 5% of both the max. By the way I'm looking over here because that's where my computer with the chart is. So the Serenite trail de max by 75% on this test last year. This year the gap is essentially gone. The geometric mean is still behind by 11 to 13% not a clean tie there. But this mini PC has finally caught up to Apple's Mac many on a real V8 tool chain benchmark. All right the obligatory geek bench benchmark. I don't usually like to dwell on this but in this case it's important. Everybody uses geek bench to measure this stuff so might as well throw it in and last generation Serenite came in at 2889 for a single score and 1489 for multi the M2 Pro Mac

many. That's the one from it's the bigger one okay from a couple years ago. That one was 2700 and 13000. Now the M4 Pro that's the number to beat. It's got pretty decent score. 3945 for single and 15,321 for multi. You can see the ser 8 was quite a bit behind on that one. And the Serenite 10 that lands at 2993 single and 15216 multi. Pretty big leap over last generation and for the single core it's not a huge difference however the multi core compared to the M4 Pro Mac many that's the difference of less than 1%. It basically ties Apple's flagship Mac many on multi thread geek bench. Maybe this will translate to when I do compilation tests in a bit. All right it's time for the meat can't believe I just said that that's just some vegetarian might object but I don't care I eat meat I like it it's good lamb is my favorite beef but a good steak is really good all right enough python mad abroad. This is a test that I run to show how pushing all the course to the max with an interpreted language like this interpreted test like this this is a python algorithm and it's the channel signature move it pegs every single core to 100% oh yeah there it goes you can see that working quite nicely there. Now the Serenite already beat the M4 base on this 28.64 seconds versus 31.41 short of numbers better in this case. Bealing ahead of Apple silicon here folks and that's the previous generation. As an aside I also ran this on the X Elite mini PC that came out some of you might remember that well the Serenite tied with that machine which is the first generation X Elite that was the most powerful one I want to see the X2 Elite in many PCs that's going to be a killer one anyway M4 Pro was still the one to beat 23.24 seconds on that one for this test

considerably faster and the Serenite was quite a bit slower 42.73 seconds on that one so we saw a 33% jump from 08 to 0.09 that's huge how does the Serenite do? Hmm okay 0.09 was 28.64, Serenite 28.9 and nice number here this is the fastest that I got that's literally a quarter of his second difference so basically kind of the same still beats the M4 base so that part of the story holds but generation through generation nothing here so this was an interpreted multi-core test but before we get into compiled multi-core test I want to back up here and show you this project which is something I ran before on those machines as well this is a large monoribo by Victor Safkin this is an X monoribo and includes 26,000 NX JavaScript components there's a lot of components but they are very small so it corresponds to a medium size enterprise repository basically JavaScript at scale there's five next JS apps here 20 libraries 250 components each and NX provides a cache but I ran this build with data cache just to see what it's like I'll before I get your hopes up this test didn't work out so well because there's repository drifts and I just don't believe there results because it can't be like this last year Serenite we got 14.1 seconds the M4 Pro was 7.76 the M4 base was 7.36 yeah I know M4 base won again third test in a row so what I like about this test is in a subtypescript tool chain it runs through node lots of process spawning lots of file system hits apple silicon's been good of this for years so this is where I thought the Serenite has to climb right maybe not dominate but climb then Serenite landed at 2.9 one seconds and I thought come on five times faster than the Serenite I don't think so so because of the way this test is set up the NX tool itself probably

got faster and optimized and unfortunately I'm gonna have to say goodbye to this tool because yeah when you're doing benchmarks you can compare things side to side side by side over time but when you're doing real world projects then things tend to drift a little bit so it's not the same workload anymore we can't really compare to the previous builds however it builds this thing in under three seconds which is pretty incredible and that's good for both the chip the machine and the monorepo and the tool so good for them we're gonna skip this test next time though I want to move on to a benchmark that I designed this is a dot net build and this build basically generates 100,000 namespaces and classes each one with a recursive calculations and nested loops so the compiler can't just throw them away and skip the work it has to do it same workload every time year after year this is the test with a Serenite B damn for base model 91 seconds that B damn for which got 106.7 seconds the Serenite was 109 seconds and the only machine that was meaningfully faster was the M4 Pro at 66.4 seconds so this is the test for Gorgon point if Zen 5's IPC story is real and the math behind the new generation actually shows up and compiled code is gonna land the hardest year and this is also the loudest test I'm hearing the fans here here it is being built pretty much all the CPU cores are involved in this one although they don't go as hard as that matter broad test but still it's more realistic and we got 90.9 this is a difference of one tenth of a second this was supposed to show the new chip strengths basically we got the same result as last time with the Serenite still beats the base M4 so being continues to be a better choice than the base Mac many for a compiler heavy dot network but on the M4 Pro Apple Silicon leads still now again the

previous test was synthetic so I wanted to go and build a real thing on Braco that's an open source CMS it's been raw for ages it's a mature dot net project help us or else but again this is a real project on GitHub so again this gonna be variation over time it's constantly evolving they're adding features they're refactoring pulling in new dependencies so the number I might get might not be fully repeatable year after year but if you have a machine and you want to build it right now compared to my result that I'm gonna get then at least it's gonna be valuable in that way if you want to compile it yourself you can find it on GitHub right over here they're constantly updating it 11 hours ago was last update can't believe this thing has been arousal on sir 949 seconds for the compilation ser 8 161 so general virgin on a brachel was only 7% why's that well because real builds are IO heavy there's disk there's new get the file system and again Mac dominates this one M4 Pro 84 seconds M4 base model even destroys the be link 93.9 seconds there all right got done ser 10 lands at 161 seconds now hold on that's slightly slower than the ser 9 what's up and it just happened to match the ser 8 from a couple years ago on both 161 seconds so again on this one real dot net production app full release pack every project the ser 9 happened to be the best of the three here at 149 by the way I want to flag something real quick here the first time I ran this I got 217 seconds on the ser 10 and that was with Windows Defender running so I had to turn that off so if you're doing this kind of compilation you got Windows Defender running you might want to exclude the your code directory from that and the way I have my my systems set up first of all I automated build the deployment

of this package and I do have videos on how to set up a development environment on a Windows machine and a Macs so check that out it's on my channel and I recently created a script to automate that whole process too I recommend going through the whole thing just to know what's getting installed and then you can use a script on your subsequent builds if you want. It's up on my GitHub I'll link to a down below but I have made a special video from members of the channel thank you to the members by the way for supporting the channel and made a video on how to run it it's all the instructions on the GitHub though if you want to just go check it out I'll link to it. All right let's switch topics to another popular thing that's happening these days in 2026 or I don't know what could that be AI AI. AI. Nice. This thing has a GPU, a CPU and an NPU so this little box is three chips on it and they all can do AI there's a 12 core CPU radion 890M I GPU and there's a 55 top's X DNA 2 NPU AMD says all three together do 86 top's of AI work tops yeah we all know the marketing stuff now out of the box and windows the most popular tool people use to run local elements use one of those but it's the wrong one so let's do it I'm going to do Olamarun Quentin 257B do your boss and I'll do my architecture prompt which produces a pretty long output there it is results of streaming looks pretty good but let's check task manager the CPU looks peg but the GPU is not being used at all NPU 0 doesn't even show up here three accelerators in this box and one of them the regular CPU is doing all the AI work and we all want that that's going to be

slow well there is a Vulcan flag that you can set on the environment for Olamar that allow you to use the GPU so on day one the AMD AI chip you just bought is going to be doing AI the part that is in the AI part doing AI the part that is in the AI part we got a result at 14.2 tokens per second so to enable that on windows you just go to the environmental variables and set the Olamar Vulcan flag to true if there isn't one just added and then just make sure you restart Olamar all right here we go that looks like it's going a little faster doesn't it? let's take a look at the GPU and yes there's the GPU compute it's happening on the GPU now which is beautiful 100% utilization there so I run a couple of them as a test. Lama 3.23B we went from 27 tokens per second to 37.5 quant 2.5 1.5 b went from 47.5 to 68.3 tokens per second so there's some differences there so we've got the iGPU thing going but there's still one chip in this box we haven't even touched and that's this MPU and guess what the old one had it too AMD's been putting this thing on the box for two years and nobody actually benchmarks it so first try I used lemonade server they got these hybrid recipes where the MPU does the pre-fill and the iGPU does the decode those are the two stages of inference you do when comes with this little handy app let's do a high here it's kind of hard to see down here but tokens per second is 14.3 on this one this is a 1 billion parameter model we can see that a bumps the MPU just a tad bit and then it uses the GPU for the decode it's probably used my longer prompt here boom okay let me see a little bit more MPU activity that's the pre-fill and the

pre-fill happens pretty quickly most of the work happens on the GPU though because that's the decode phase so the MPU is kind of built for that stage for the pre-fill it's a different job different job different chip so I ran a longer prompt and it was a 4400 token prompt with a short reply and that's the actual pre-fill work load on quence 7b the CPU got 255 tokens per second there the Vulcan iGPU 240 and the NPU with a hybrid mode 631 tokens per second that's 2.5 times the iGPU on the same chip so if you're streaming chat use the iGPU if you're doing rag, agents or coding assistance anything that dumps a long context then you want to use the MPU and since this is a lemonade server you can actually connect it to a coding agent it's for another video perhaps you want to see that let me know in the comments down below if you do I'm curious myself because I haven't actually tested lemonade before on the channel but now I want to show you something that this hardware does that if you just read the specs you'd say no way that's not possible quen 2.5 14b that's the Q4 quant over 8 gigabytes on disk now the iGPU on this chip it shows you 4 gigs of memory just 4 and I'm gonna load 8.37 gigs of weights onto 4 gig iGPU this shouldn't work and it's loaded all 49 layers on the GPU how all right so I'm watching this in real time get counter on the right side GPU adapter memory before I loaded the model it was 0.69 gigabytes dedicated 5.26 shared and after 3.62 dedicated 17.36 shared 20.98 gigabytes of iGPU memory in use right now on a chip that device manager says there's only 4 so that's uma unified memory architecture which dm4 and the apple silicon also has but this has a 2 so when the iGPU asks for more memory

than its allocation windows just quietly grabs from system ram and says here you go hands it over and the model has no idea so prompt processing was about 50 tokens per second generation 8.88 tokens per second a 14 billion parameter model and by the way on this machine there's still like 25 more gigabytes of unified memory hetero left here so a 22 billion parameter model would fit probably even a 30 billion parameter model would fit too with q4 on the iGPU I remember that the 0.10 and the 0.9 had the exact same GPU in there so if you bought a 0.9 last year for the AI stuff your hardware is still fine update your software so quick recommendations here you should buy the 0.10 if you care about the mpu for rag and long context AI work if you want upgradable ram because 0.9 doesn't have that or if you want that 10 gig ethernet port you should probably skip it if you already have a certain 9 and you may in workload is the iGPU illa lembath save your money until the next generation which is medusa comes out in 2027 another scary creature that I wouldn't want to mess with now besides these there's actually a bunch of other mini PCs that you can consider and i made a video up here and up here you can check those out you might be interested in those thanks for watching and I'll see you next time.