"Device is 100% fully functional and in Excellent physical condition. The device is in near-perfect, if not perfect condition. At a 9.5/10 or better, we regard this device as like new with minimal or no signs of wear."
This collaborative space allows users to contribute additional information, tips, and insights to enhance the original deal post. Feel free to share your knowledge and help fellow shoppers make informed decisions.
"Device is 100% fully functional and in Excellent physical condition. The device is in near-perfect, if not perfect condition. At a 9.5/10 or better, we regard this device as like new with minimal or no signs of wear."
Not everything is for me, but I don't understand the allure of this. It's like a super powerful 13" tablet with a flimsy keyboard and probably not great battery life. Its igpu is equivalent to a 3060 or 4060, and it cost over $1700 after tax? You could build a really nice desktop for that.
Not everything is for me, but I don't understand the allure of this. It's like a super powerful 13" tablet with a flimsy keyboard and probably not great battery life. Its igpu is equivalent to a 3060 or 4060, and it cost over $1700 after tax? You could build a really nice desktop for that.
it's for running A.I. locally instead of paying for cloud tokens. the 8060s shares video memory with RAM, and at 128gb of ram, you can run 80-100billion point models (the whole model needs loaded into VRAM at once). it'll be slower responses than a cloud solution, but it beats a noisy $8000-$14000 rackmount server. it just so happens to be a capable entry level gaming machine too (at a not entry level price).
it's for running A.I. locally instead of paying for cloud tokens. the 8060s shares video memory with RAM, and at 128gb of ram, you can run 80-100billion point models (the whole model needs loaded into VRAM at once). it'll be slower responses than a cloud solution, but it beats a noisy $8000-$14000 rackmount server. it just so happens to be a capable entry level gaming machine too (at a not entry level price).
also i just got one 2 weeks ago, and on default power settings it benchmarks about 15% lower 3d graphics than the same chipset in a small formfactor/embedded desktop pc, so even more limited use case.
Our community has rated this post as helpful. If you agree, why not thank Will_R
Quote
from Stinktooth
:
Not everything is for me, but I don't understand the allure of this. It's like a super powerful 13" tablet with a flimsy keyboard and probably not great battery life. Its igpu is equivalent to a 3060 or 4060, and it cost over $1700 after tax? You could build a really nice desktop for that.
The 3060 and 4060 with 12GB or 16GB VRAM don't come anywhere close to the 8060S with variable VRAM size up to 96GB. That is the allure. That's it. It's not about gaming or FPS. It's about running larger LLMs anywhere with low power/high efficiency.
It's much more sensible to compare it to a $4k-$6k Mac Studio w/ 96GB of unified memory instead of either a desktop or laptop with a xx60 series card.
The 3060 and 4060 with 12GB or 16GB VRAM don't come anywhere close to the 8060S with variable VRAM size up to 96GB. That is the allure. That's it. It's not about gaming or FPS. It's about running larger LLMs anywhere with low power/high efficiency.It's much more sensible to compare it to a $4k-$6k Mac Studio w/ 96GB of unified memory instead of either a desktop or laptop with a xx60 series card.
This right here. A comparable Mac easily sets you around 5k
Smaller llm models <40B will run at decent spends comparable to older gpus on this with full context. You can run larger models >70B parameters but it will be slow <20t/s with significantly reduced context. The larger the context window, the slower the token generation speed, so with an actually usable context size, you're looking at <10t/s. You might get faster generation with older models like GPT-OSS 120B but newer, smaller models like Qwen3.6 27B will probably perform much better in areas like coding. For the price, the 48GB macbook might be a better option if you're looking to run local llm on a portable device. Of course you'll be running linux on this device if you're focused on inferencing. However for running smaller llm models, productivity, portable gaming, image editing, light video production or all of the above, this can't be beat
2
1
1
Like
Helpful
Funny
Not helpful
Sign up for a Slickdeals account to remove this ad.
also i just got one 2 weeks ago, and on default power settings it benchmarks about 15% lower 3d graphics than the same chipset in a small formfactor/embedded desktop pc, so even more limited use case.
Do you think a mini pc with the same chipset connected to a EGPU via oculink would outperform this?
it's for running A.I. locally instead of paying for cloud tokens. the 8060s shares video memory with RAM, and at 128gb of ram, you can run 80-100billion point models (the whole model needs loaded into VRAM at once). it'll be slower responses than a cloud solution, but it beats a noisy $8000-$14000 rackmount server. it just so happens to be a capable entry level gaming machine too (at a not entry level price).
this thing can only hit 40 tokens/s when loading small model, if you put a 70b model in to the ram, it will be 5 t/s or so, I won't be serious about running AI on this.
The 3060 and 4060 with 12GB or 16GB VRAM don't come anywhere close to the 8060S with variable VRAM size up to 96GB. That is the allure. That's it. It's not about gaming or FPS. It's about running larger LLMs anywhere with low power/high efficiency.It's much more sensible to compare it to a $4k-$6k Mac Studio w/ 96GB of unified memory instead of either a desktop or laptop with a xx60 series card.
No it does not compare to that all. I had it. It's garbage and why it dropped in price 2 weeks after it came out.
No it does not compare to that all. I had it. It's garbage and why it dropped in price 2 weeks after it came out.
Those 2 (this tablet and any Mac) are more comparable in what they can do but agreed for not comparable as to how quickly they can do it. You simply can't load even a 30B model onto a 3060 or 4060 at all.
this thing can only hit 40 tokens/s when loading small model, if you put a 70b model in to the ram, it will be 5 t/s or so, I won't be serious about running AI on this.
I have the 64GB version (framework desktop) and running Gemma 4 26B at 60tok/s it's pretty good for simple AI tasks that don't require deep thinking. It's certainly not up there with frontier models, but the progress local LLM models have been making is quite impressive. Once ChatGPT/Anthropic start charging API rates ($8,000/mo instead of $200/mo) local LLMs might be the only option for most people.
Like
Helpful
Funny
Not helpful
Sign up for a Slickdeals account to remove this ad.
Folks, do your homework before making a fool judgement. The AMD AI 395 Max is a 16‑core, full‑performance Zen 5 CPU, essentially a 9950X equivalent processor paired with a RTX 4070 mobile equivalent GPU and 128GB of DDR5X‑8000 RAM. When you put that into a high‑quantity, tablet‑style form factor with stunning power efficiency for both AI workloads and gaming, what do you think it's worth?
1
Like
Helpful
Funny
Not helpful
Join The Conversation
Share information with the community. Please follow our Community Guidelines and be kind!
Join The Conversation
Share information with the community. Please follow our Community Guidelines and be kind!
17 Comments
Sign up for a Slickdeals account to remove this ad.
Our community has rated this post as helpful. If you agree, why not thank Will_R
It's much more sensible to compare it to a $4k-$6k Mac Studio w/ 96GB of unified memory instead of either a desktop or laptop with a xx60 series card.
Sign up for a Slickdeals account to remove this ad.
Sign up for a Slickdeals account to remove this ad.
Join The Conversation
Share information with the community. Please follow our Community Guidelines and be kind!