Post reply

Name
Email
Subject
Message icon

[quote author=there u go link=msg=779792 date=1789980204]
[quote]The standard retail model comes with 64 GB of memory, while Thunderobot says the M7000 can support up to 128 GB. [/quote]
Good, because I was gonna write that this 4060 Laptop performing laptop would be neither here, nor there, if it only supported up to 64 GB RAM/unified memory, instead of the full 128 GB RAM, that would allow it to run SOTA AI LLM model Qwen3.8-Flash-Next (e.g. huggingface.co/unsloth/Qwen3.8-Flash-Next-GGUF/tree/main/UD-Q4_K_XL): While the quant is 111 GB, the n-gram embeddings are 30 GB and are streamed from the SSD with almost no performance impact. So it becomes 111 GB - 30 GB = 81 GB. 
On Windows: 128 GB RAM - 32 GB RAM = 96 GB RAM freely available for any AI LLM model and 81 GB fit comfortably into 96 GB RAM + decent amount of context.
On Linux (search for 'strix halo linux gtt'): 128 GB RAM - 8 GB RAM = 120 GB RAM freely available for any AI LLM model and 81 GB fit very comfortably into 96 GB RAM + big amount of context.

For 3000 bucks you can build a desktop PC that will be faster for AI:
[list]
[li]this (64 GB RAM/unified memory at 256 GB/s) / any 64 GB Strix Halo: AI score: [b]16384[/b] = 64 GB * 256 GB/s.[/li][/list]
vs
[list][li]desktop PC build using 64 GB RAM + two 5060 Ti 16GB: AI score: [b]34406[/b] = (64 GB * 89.6 GB/s (DDR5 5600 MT/s)) + (2*16 GB*2*448 GB/s (in tensor split mode the memory bandwidth doubles))[/li]
[/list]
[/quote]

Other options

Return to this topic
Don't use smileys

Verification:

Please leave this box empty:

Shortcuts: ALT+S post or ALT+P preview

Topic summary

Posted by there u go

- Today at 10:43:24

QuoteThe standard retail model comes with 64 GB of memory, while Thunderobot says the M7000 can support up to 128 GB.

Good, because I was gonna write that this 4060 Laptop performing laptop would be neither here, nor there, if it only supported up to 64 GB RAM/unified memory, instead of the full 128 GB RAM, that would allow it to run SOTA AI LLM model Qwen3.8-Flash-Next (e.g. huggingface.co/unsloth/Qwen3.8-Flash-Next-GGUF/tree/main/UD-Q4_K_XL): While the quant is 111 GB, the n-gram embeddings are 30 GB and are streamed from the SSD with almost no performance impact. So it becomes 111 GB - 30 GB = 81 GB.
On Windows: 128 GB RAM - 32 GB RAM = 96 GB RAM freely available for any AI LLM model and 81 GB fit comfortably into 96 GB RAM + decent amount of context.
On Linux (search for 'strix halo linux gtt'): 128 GB RAM - 8 GB RAM = 120 GB RAM freely available for any AI LLM model and 81 GB fit very comfortably into 96 GB RAM + big amount of context.

For 3000 bucks you can build a desktop PC that will be faster for AI:

this (64 GB RAM/unified memory at 256 GB/s) / any 64 GB Strix Halo: AI score: 16384 = 64 GB * 256 GB/s.

desktop PC build using 64 GB RAM + two 5060 Ti 16GB: AI score: 34406 = (64 GB * 89.6 GB/s (DDR5 5600 MT/s)) + (2*16 GB*2*448 GB/s (in tensor split mode the memory bandwidth doubles))

Posted by Review feedback expert

- Today at 08:03:14

Not mentioned in the article -

1. USB power input is available or not? Since its a notebook form factor, carrying a separate charger is extra hassle, can we use standard usb power delivery if supported ? And upto how many watts ?

2. Battery capacity and backup time ?

How many minutes can it run unplugged ? For example how many minutes can it run at idle or average 10% / 40 % / 100 % usage of cpu and GPU ?

3. Is battery hot pluggable (swap battery while system is on) ? How many seperate batteries are there ? Some good notebooks have 2 so you can replace one while its running.

These info are crucial for a laptop since its all about portability.

Otherwise just buy a desktop. Much more cost effective (but less portable)

From an expert individual notebook feedback provider. Contact for similar review feedbacks at [email protected]

Posted by Redaktion

- Yesterday at 17:12:45

Thunderobot's AI Master M7000 mobile workstation laptop has launched with an AMD Ryzen AI Max+ 395 processor, 64 GB of memory and an AI SSD in China. The 16-inch laptop supports Phison's aiDAPTIV+ technology and is designed to run large AI models locally.

https://www.notebookcheck.net/Powerful-laptop-launches-with-Ryzen-AI-Max-395-64-GB-RAM-8-TB-storage-and-2-5K-165-Hz-display.1403862.0.html