Oh no! Jeff got loose is trying to build his Pi5 based gaming PC again. Quick someone give him a raid adapter so he can build a NAS or something. Or make him bench Stable Diffusion on the pi5.
I think I'd like that little box but with a PCI-e slot instead of a built in card. On my desktop I'm using OCuLunk to add an extra PCI-e slot for VM USB shenanigans.
Hi Wendel, can you please do a deep dive explanation on what an NPU really does? 1- Is it a cheap replacement for CUDA? 2 - Who is suppose to develop for it, Intel or AMD or anyone? 3 - For example, Can i write an object recognition or face recognition application and run it on NPU if i don't have an NVIDIA cuda graphics card? 4- If i have an NPU laptop and an RTX card can i write code that can simultaneously take advantage of both CUDA AND NPU? 5 -What is the real difference between an NVIDA CUDA card capable and an NPU machine? 6 -If I don't want to pay INGRIDIA prices when deciding on a laptop but Need something similar to CUDA can i just get a NPU laptop? MY question exclude gaming scenarios. I am just talking about having simple laptop to run A.I code when your sitting in a coffee shop with no access to desktop pc. My questions are from the perspective of a developer who wants to know how NPU can benefit them when selecting a laptop that gives access to something like cuda.
@@mattrogers6646 It doesn't show in the task manager, but on AMD website, under AI capabilities it says: "AI Engine Capabilities Brand Name AMD Ryzen™ AI AMD Ryzen™ AI Available Performance Up to 10 TOPS" Maybe it's "just" the GPU, but AFAIK it was first AMD CPU that had TOPS rating. I mean it's Phoenix, which was supposed to be a bit of a trial before the CoPilot was launched. Or rather before it was delayed and this is why it barely meets minimum specifications of CoPilot. Or at least early specifications. But now I do have to check and I wonder if it isn't just using GPU to achieve this.
I have one stupid thing in my mind that i would like you to try out. Is it possible to connect an oculink gpu to an usb4 port using an usb4 to 4xpcie nvme adapter? Would be interesting to see if and how the bandwidth is different compared to thunderbold over usb4.
I got on the oculink bandwagon this past Summer, and can confirm the gaming experience is indistinguishable from a dedicated gaming PC (I was using an rx 7800xt external gpu attached via. oculink + m.2 adapter to a framework laptop with a 12th gen intel processor). At gen 4 x4 bandwidth, I was getting very solid 4K framerate on ghosts of tsushima. I just wish more devices had the built-in port!
Wendel pls ask Minisforum whenthey will release a Minisforum V4 with strix point Ryzen AI 9 HX 370 and Oculink. This will beat any handheld out there . I have been wanting a V3 but lack of oculink port on it is such a missed opportunity
Because I’m weird but I wish apple would only have 1-2 TB ports on their computers so you have additional video options. They really need 2 4lane or 1 8 lane oculink port.
I really wish someone would make an Nvida powered dock like this. It’s always AMD. I have a OneDock v2 and I’ve tested a 3060ti with my GPD Win Mini over Oculink. It’s amazing! I just love the form factor of these laptop GPU based docs. They look fantastic but DLSS is a must for me at this point. Especially on mid range laptop hardware. FSR looks like crap at 1080p. A 40 series would also have DLSS 3. Give me a laptop 4070 in a doc like this and we have a deal. I don’t know why they keep choosing AMD. They must be getting a great deal on these chips from AMD.
If only more laptops supports oculink. I'd ditch my pc in a heartbeat. The only ones i can see are the laptops for chinese market, like the thinkbook chinese edition.
Thinking about adding an M.2 Oculink adapter to my mini pc but I have a hard time to find something reliable. Many adapters I found (AE and alike) have mixed reviews. Any firsthand experience or specific adapters you can recommend? What make is the adapter you were showing in your video?
How many ssds would i be able to connect through this link? Can i make a 100tb ssd server with it? Minisforum is doing wonders right now My bd790i has 4x4x4x4 bifurcation, so 4 m.2/pciex4 slots for 4 oculink connections. Just to figure out a way to get 25 4tb drives connected via 4-6 oculink ports.
7600M? So is it worse or better than 7700S? BTW I did buy Framework Laptop 16 and unboxed and assembled it and put the videos on my channel. Sort of thanks to you - and because Framework started to ship to my country. Though I didn't buy a GPU. If it was 100W version of actual RX 7700 with 12 GB of VRAM, I would have considered it. Or even 7600XT. Funny thing is I had such a potato of a machine that I couldn't even edit my videos. So you get full uncut version with me mumbling in English 🤣. But it really was your video on them that made me decide to buy it. Expensive, but probably worth it. Also a question regarding Framework Laptop 16 - I bought it with 32GB of RAM. I wanted 96GB, but it was too expensive. I could buy Crucial memory for 40% of the price that Framework asks for their RAM. But on the Compatibility list, they only show that it is compatible with Crucial 32GB memory (basically the same model number except for the amount of memory). Did you do any tests of compatible memory on yours? Heck, I don't think the BIOS shows all the timings even. I hope I can get them, so that I can compare if that Crucial memory isn't much slower or something. And Mini-PC's are fantastic - especially in the last 2-3 years.
Oh. Also bought last 2 Optane H10 on firesale. Now I will have to think how to make them run on AMD - even if it's just for science project ;) Also thanks to you.
The problem is the price ... Minisforum DEG1 is 90euro + a 30-40euro PSU and you have a 450euro budget for the GPU to match the price of the MGA1. And you are not limited to AMD. Yes it is a more messy setup, not as compact but if you need to travel, a laptop is still the better solution overall. And 8GB of VRAM is quickly becoming a bottleneck even on 1080p gaming if you want to crank up the details.
619€ is too much for a 7600M XT. Same price with a 7800 XT or a 7900GRE it would be a instant buy. 350€ would be a correct price for what they sell IMHO.
I have one i paid $160 to use as proxmox server I love it, However when the prices of these mini PC's start approaching desktop Price's.. I can't justify why the exist. Is not even portable or does any special except that... it's mini. Someone please convince me getting these mini pc is better than a full desktop with an rtx 4060 or even 6600xt for the same price.
What's the goal of this channel? You're clearly an intelligent, capable guy but your content is so niche I can't decide who the audience is? Do you want to put all this time and effort in for 20k views per video? You gotta find a way out of the hole you're in big dawg.
Kudos to Minisforum for releasing this only after an Oculink to PCIe adapter, loving my DEG1. Other manufacturers just want to milk consumers.
Yeah, the DEG1 is pretty good, although I am seriously thinking of pulling it apart and mounting the PCB in a chassis with a HDPlex GaN PSU.
I wonder if they built this for the Pi 5? :D
Though it's not faster than Thunderbolt on the Pi...
Oh no! Jeff got loose is trying to build his Pi5 based gaming PC again.
Quick someone give him a raid adapter so he can build a NAS or something. Or make him bench Stable Diffusion on the pi5.
@@LeminskiTankscor that's very bad....
Lol
Jeff (sees a tech product) : I can run a RBP off of you
@@CR0SSJ literally Jeff ^ 😅
I think I'd like that little box but with a PCI-e slot instead of a built in card. On my desktop I'm using OCuLunk to add an extra PCI-e slot for VM USB shenanigans.
i hoped that was what this product was
Thank you, Wendellman! 👍🏼
Hi Wendel,
can you please do a deep dive explanation on what an NPU really does?
1- Is it a cheap replacement for CUDA?
2 - Who is suppose to develop for it, Intel or AMD or anyone?
3 - For example, Can i write an object recognition or face recognition application and run it on NPU if i don't have an NVIDIA cuda graphics card?
4- If i have an NPU laptop and an RTX card can i write code that can simultaneously take advantage of both CUDA AND NPU?
5 -What is the real difference between an NVIDA CUDA card capable and an NPU machine?
6 -If I don't want to pay INGRIDIA prices when deciding on a laptop but Need something similar to CUDA can i just get a NPU laptop?
MY question exclude gaming scenarios. I am just talking about having simple laptop to run A.I code when your sitting in a coffee shop with no access to desktop pc.
My questions are from the perspective of a developer who wants to know how NPU can benefit them when selecting a laptop that gives access to something like cuda.
I would be very interested with this as well. I do have 1 barely "AI capable" NPU in my house. I could try to run something on it (Ryzen 7840 HS).
Nothing yet. Never buy the promise. Still waiting for my Strix Halo desktop APU...
@@jannegreyAre you certain 7840HS has NPU? IIRC, the 8000 series of CPUs/APUs introduced it.
@@mattrogers6646 It doesn't show in the task manager, but on AMD website, under AI capabilities it says:
"AI Engine Capabilities
Brand Name AMD Ryzen™ AI
AMD Ryzen™ AI Available
Performance Up to 10 TOPS"
Maybe it's "just" the GPU, but AFAIK it was first AMD CPU that had TOPS rating.
I mean it's Phoenix, which was supposed to be a bit of a trial before the CoPilot was launched. Or rather before it was delayed and this is why it barely meets minimum specifications of CoPilot. Or at least early specifications.
But now I do have to check and I wonder if it isn't just using GPU to achieve this.
CUDA is a proprietary architechture wich means code made for it will only work on it
Thanks for covering these more niche graphics solutions. A lot of people still don’t know about OCuLink…
I have one stupid thing in my mind that i would like you to try out.
Is it possible to connect an oculink gpu to an usb4 port using an usb4 to 4xpcie nvme adapter? Would be interesting to see if and how the bandwidth is different compared to thunderbold over usb4.
Is this still the old set, Wendell? Looking forward to seeing all the hard work you, Steve and the whole L1T team did pay off.
I got on the oculink bandwagon this past Summer, and can confirm the gaming experience is indistinguishable from a dedicated gaming PC (I was using an rx 7800xt external gpu attached via. oculink + m.2 adapter to a framework laptop with a 12th gen intel processor). At gen 4 x4 bandwidth, I was getting very solid 4K framerate on ghosts of tsushima. I just wish more devices had the built-in port!
have you posted photos anywhere?
@@mbaltrusitis prowlers.com/~cdavis/PXL_20240905_223944359.jpg
But usb 4 is just open sourced thunderbolt 3 no? The reason for choosin thunderbolt 4 over 3 is numerous compatability fixes.
Wendell please ask manufacturers to push oculink x8!
He truly knows us, we all strive for fabulous.
2:51 I think you meant to say, ‘via M.2 adapter’ not ‘via USB adapter’ ?
High pass filter on that mic!
Wendel pls ask Minisforum whenthey will release a Minisforum V4 with strix point Ryzen AI 9 HX 370 and Oculink. This will beat any handheld out there . I have been wanting a V3 but lack of oculink port on it is such a missed opportunity
Wish they'd come out with PCI 5
Because I’m weird but I wish apple would only have 1-2 TB ports on their computers so you have additional video options. They really need 2 4lane or 1 8 lane oculink port.
I really wish someone would make an Nvida powered dock like this. It’s always AMD. I have a OneDock v2 and I’ve tested a 3060ti with my GPD Win Mini over Oculink. It’s amazing! I just love the form factor of these laptop GPU based docs. They look fantastic but DLSS is a must for me at this point. Especially on mid range laptop hardware. FSR looks like crap at 1080p. A 40 series would also have DLSS 3. Give me a laptop 4070 in a doc like this and we have a deal. I don’t know why they keep choosing AMD. They must be getting a great deal on these chips from AMD.
Minisforum DEG1 eGPU Dock, just bought it. and you didn't mention Thunderbolt 5. which Razer release Q1 2024.
3:00 "let's shove that in the bottom"
let's clip that
This review was all over the place
Uploaded as soon as i clocked out for break ? Thanks, L1T
I would have expected the industry to make a VERY quick transition away from OCULink to MCIO by now already.
2:33 that is not a GPD Mini. That is a GPD WinMax 2
If only more laptops supports oculink. I'd ditch my pc in a heartbeat. The only ones i can see are the laptops for chinese market, like the thinkbook chinese edition.
Using the DEG1 with my older 5700XT GPU to Minisforum UM780 XTX and loving it...
Thinking about adding an M.2 Oculink adapter to my mini pc but I have a hard time to find something reliable. Many adapters I found (AE and alike) have mixed reviews. Any firsthand experience or specific adapters you can recommend? What make is the adapter you were showing in your video?
it is cool tech but at that point i think it would be better to glue a monitor to an itx case. the size wouldn't be much different.
Are there any USB 4 eGPU products that are able to use the extra bandwidth over TB3/4?
No
2:56 He calls the M.2 to Oculink adapter a USB adapter... lol
How many ssds would i be able to connect through this link? Can i make a 100tb ssd server with it? Minisforum is doing wonders right now
My bd790i has 4x4x4x4 bifurcation, so 4 m.2/pciex4 slots for 4 oculink connections. Just to figure out a way to get 25 4tb drives connected via 4-6 oculink ports.
I've been testing my Gpu's. I don't think I can tell a difference between running my gpu at gen4- 16x or 8x. 32GB of smart access memory on tap here!
Is there a way to have 2 (or more) OCuLink connections to the same device at the same time???
Would MCIO be better in this application? Is the connector more robust?
is miniforum v3 (windows tablet) can take advantage of using this ?
What Laptop is that?
Thunderbolt 5 needs to come fast
Is out, but for Apple at the moment...
Wil still have more overhead than straight PCIe. Especially when OCulink x8 gets traction
hey wendell, you lug servers around all the time, don't you have a server jack or something? those things are heavy
Monday ramblings! Let's GOOOO!!!
Minisforum do an actual dock. Add your own oculink gpu. Come on, 8gb in 2024.
7600M? So is it worse or better than 7700S?
BTW I did buy Framework Laptop 16 and unboxed and assembled it and put the videos on my channel. Sort of thanks to you - and because Framework started to ship to my country. Though I didn't buy a GPU. If it was 100W version of actual RX 7700 with 12 GB of VRAM, I would have considered it. Or even 7600XT.
Funny thing is I had such a potato of a machine that I couldn't even edit my videos. So you get full uncut version with me mumbling in English 🤣. But it really was your video on them that made me decide to buy it. Expensive, but probably worth it.
Also a question regarding Framework Laptop 16 - I bought it with 32GB of RAM. I wanted 96GB, but it was too expensive. I could buy Crucial memory for 40% of the price that Framework asks for their RAM. But on the Compatibility list, they only show that it is compatible with Crucial 32GB memory (basically the same model number except for the amount of memory). Did you do any tests of compatible memory on yours? Heck, I don't think the BIOS shows all the timings even. I hope I can get them, so that I can compare if that Crucial memory isn't much slower or something.
And Mini-PC's are fantastic - especially in the last 2-3 years.
Oh. Also bought last 2 Optane H10 on firesale. Now I will have to think how to make them run on AMD - even if it's just for science project ;)
Also thanks to you.
The problem is the price ... Minisforum DEG1 is 90euro + a 30-40euro PSU and you have a 450euro budget for the GPU to match the price of the MGA1. And you are not limited to AMD. Yes it is a more messy setup, not as compact but if you need to travel, a laptop is still the better solution overall. And 8GB of VRAM is quickly becoming a bottleneck even on 1080p gaming if you want to crank up the details.
Wendell sorry but wtf is that red on those TVs behind you?
Neat device
so... how do I connect it to my Oculus headset???
Yea, id need to see it on more laptops before i can justify the cost of the dock
4x gen 4 is pupu yea nope fails end
Why the hell arent there 3 DisplayPort connectors?
619€ is too much for a 7600M XT. Same price with a 7800 XT or a 7900GRE it would be a instant buy. 350€ would be a correct price for what they sell IMHO.
I don't get why these exist. It costs more than a cheap gaming PC with the same GPU.
I use WiFi 6E to stream 4K90 to my VR headset and portables.
I have one i paid $160 to use as proxmox server I love it, However when the prices of these mini PC's start approaching desktop Price's.. I can't justify why the exist. Is not even portable or does any special except that... it's mini. Someone please convince me getting these mini pc is better than a full desktop with an rtx 4060 or even 6600xt for the same price.
@@chaliseTV 10GBE M.2 mini servers are great, enough speed to actually use as storage for your whole home, combine that with a mech backup drive.
What's the goal of this channel? You're clearly an intelligent, capable guy but your content is so niche I can't decide who the audience is? Do you want to put all this time and effort in for 20k views per video? You gotta find a way out of the hole you're in big dawg.
This thing is horribly over priced.
Be sure to take care of your back, do physical therapy, yoga classes, whatever you prefer. But, by all means, take care of yourself.