Dell Pro Max 16/18 Plus (2025 model) pre-release discussion — MB16250, MB18250

[pickwick]

2 hours ago, Aaron44126 said:

1. If you get an iGPU-only system, you will be able to use the HDMI port and it will be attached to the iGPU.

So, with iGPU and dGPU the table in the Dell docs mean HDMI is connected to dGPU only? I would love to have everythign connected and being usable by iGPU and only offload work of interest to dGPU, if at all.

[Aaron44126]

42 minutes ago, pickwick said:

So, with iGPU and dGPU the table in the Dell docs mean HDMI is connected to dGPU only? I would love to have everythign connected and being usable by iGPU and only offload work of interest to dGPU, if at all.

 

I think the docs are right with regard to this. I haven't tested this system, but that is how prior generations behaved (Precision 7000 line starting with 7X30, when Dell switched to using DGFF GPUs).

 

If you want everything attached to the iGPU, plug your displays in via the USB-C ports, either via a dock or directly.

  

16 hours ago, pickwick said:

I'm having two external displays, shared by a business and private laptop. One display has a free HDMI and the other one DVI-D. The Dell has native HDMI in theory, which would leave me with an additional DVI-D to be supported. There seem to exist adapters form thunderbolt/USB-C to DVI-D for small prices. I only need to support 1920*1200. Those adapters should work, shouldn't they?

 

USB-C to HDMI or USB-C to DVI (...or even USB-C → HDMI → DVI...) should work fine, and attach to the iGPU.

 

If it were me, I'd just use an USB-C → HDMI adapter as that will have more potential for future use, and then connect it to your monitor with an HDMI→DVI cable if necessary.

[MyPC8MyBrain]

to the best of my knowledge when optimus puts dGPU to sleep iGpu will function through hdmi port,
optimus  is not dell's software its nvidia's software designed specifically to switch between the two,
if hdmi is not working it has to be something else preventing the port from being used.

[Aaron44126]

1 hour ago, MyPC8MyBrain said:

to the best of my knowledge when optimus puts dGPU to sleep iGpu will function through hdmi port,
optimus  is not dell's software its nvidia's software designed specifically to switch between the two,
if hdmi is not working it has to be something else preventing the port from being used.

 

This is the case with most displays, but in these systems, the HDMI port is hard-wired to the dGPU and not the iGPU. (Unless you bought an iGPU-only system and it has the DGFF spacer card installed, whose purpose is to basically route the signal from the HDMI port to the iGPU.) So, connecting a display to the HDMI port will always cause the dGPU to engage.

[pickwick]

Thanks so far, that was pretty helpful already!

 

Does anyone do AI-stuff with the dGPU and if so, which? I've read a lot about that 8 GiB of VRAM are most likely not enough and lots of pre-built models seem to have the 3K with 12 GiB RAM as well. That should be almost as fast as 5070 Ti, which can be found in many high end gaming laptops, but more like an entry-level card. 🙂

 

Does anyone has any experience with the RTX PRO 3K? Are 12 GiB of VRAM reasonable enough or just a waste of a few hundreds of Dollars most likely?

[MyPC8MyBrain]

https://ollama.com/library/llama3.1

Minimal Requirements for Llama 3.1 8B (Quantized)
VRAM (GPU Memory)    ≥ 8 GB VRAM
System RAM    ≥ 16 GB

 

If you are interested in the larger, more powerful Llama 3.1 variants, 
the requirements scale up dramatically, moving them out of the range of most standard PCs:
Llama 3.1 70B    ≈ 35 - 50 GB VRAM
Llama 3.1 405B    ≥ 150 GB VRAM

Hardware support
https://docs.ollama.com/gpu

Compute capability (CC)
https://developer.nvidia.com/cuda-gpus