[bloomingkales]

Jensen did say in recent interview, paraphrasing, “they are trying to kill my company”.

Those Macs with unified memory is a threat he is immediately addressing. Jensen is a wartime ceo from the looks of it, he’s not joking.

No wonder AMD is staying out of the high end space, since NVIDIA is going head on with Apple (and AMD is not in the business of competing with Apple).

[T-A]

From https://www.tomshardware.com/pc-components/cpus/amds-beastly...

The fire-breathing 120W Zen 5-powered flagship Ryzen AI Max+ 395 comes packing 16 CPU cores and 32 threads paired with 40 RDNA 3.5 (Radeon 8060S) integrated graphics cores (CUs), but perhaps more importantly, it supports up to 128GB of memory that is shared among the CPU, GPU, and XDNA 2 NPU AI engines. The memory can also be carved up to a distinct pool dedicated to the GPU only, thus delivering an astounding 256 GB/s of memory throughput that unlocks incredible performance in memory capacity-constrained AI workloads (details below). AMD says this delivers groundbreaking capabilities for thin-and-light laptops and mini workstations, particularly in AI workloads. The company also shared plenty of gaming and content creation benchmarks.

[...]

AMD also shared some rather impressive results showing a Llama 70B Nemotron LLM AI model running on both the Ryzen AI Max+ 395 with 128GB of total system RAM (32GB for the CPU, 96GB allocated to the GPU) and a desktop Nvidia GeForce RTX 4090 with 24GB of VRAM (details of the setups in the slide below). AMD says the AI Max+ 395 delivers up to 2.2X the tokens/second performance of the desktop RTX 4090 card, but the company didn’t share time-to-first-token benchmarks.

Perhaps more importantly, AMD claims to do this at an 87% lower TDP than the 450W RTX 4090, with the AI Max+ running at a mere 55W. That implies that systems built on this platform will have exceptional power efficiency metrics in AI workloads.

[adrian_b]

"Fire breathing" is completely inappropriate.

Strix Halo is a replacement for the high-power laptop CPUs from the HX series of Intel and AMD, together with a discrete GPU.

The thermal design power of a laptop CPU-dGPU combo is normally much higher than 120 W, which is the maximum TDP recommended for Strix Halo. The faster laptop dGPUs want more than 120 W only for themselves, not counting the CPU.

So any claims of being surprised that the TDP range for Strix Halo is 45 W to 120 W are weird, like the commenter has never seen a gaming laptop or a mobile workstation laptop.

[bmicraft]

> The thermal design power of a laptop CPU-dGPU combo is normally much higher than 120 W

Normally? Much higher than 120W? Those are some pretty abnormal (and dare I say niche?) laptops you're talking about there. Remember, that's not peak power - thermal design power is what the laptop should be able to power and cool pretty much continuously.

At those power levels, they're usually called DTR: desktop replacement. You certainly can't call it "just a laptop" anymore once we're in needs-two-power-supplies territory.

[adrian_b]

Any laptop that in marketed as "gaming laptop" or "mobile workstation" belongs to this category.

I do not know which is the proportion of gaming laptops and mobile workstations vs. thin and light laptops. While obviously there must be much more light laptops, the gaming laptops cannot be a niche product, because there are too many models offered by a lot of vendors.

My own laptop is a Dell Precision, so it belongs to this class. I would not call Dell Precision laptops as a niche product, even if they are typically used only by professionals.

My previous laptop was some Lenovo Yoga that also belonged to this class, having a discrete NVIDIA GPU. In general, any laptop having a discrete GPU belongs to this class, because the laptop CPUs intended to be paired with discrete GPUs have a default TDP of 45 W or 55 W, while the smallest laptop discrete GPUs may have TDPs of 55 W to 75 W, but the faster laptop GPUs have TDPs between 100 W and 150 W, so the combo with CPU reaches a TDP around 200 W for the biggest laptops.

[bmicraft]

You can't usually just add up the TDPs of CPU and GPU, because neither cooling nor the power circuitry supports that kind of load. That's why AMDs SmartShift is a thing.

[bloomingkales]

People are very unaware just how much better a gaming laptop from 3 years ago is (compared to a copilot laptop). These laptops are sub $500 on eBay, and Best Buy won’t give you more than $150 for it as a trade in (almost like they won’t admit that those laptops outclass the new category type of AI pc).

[nomel]

> since NVIDIA is going head on with Apple

I think this is a race that Apple doesn't know it's part of. Apple has something that happens to work well for AI, as a side effect of having a nice GPU with lots of fast shared memory. It's not marketed for inference.

[Tepix]

Apple is both well aware and marketing it, as seen at https://www.apple.com/my/newsroom/2024/10/apples-new-macbook...

Quote:

"It also supports up to 128GB of unified memory, so developers can easily interact with LLMs that have nearly 200 billion parameters."

[JoshTko]

Which interview was this?

[bloomingkales]

https://fortune.com/2023/11/11/nvidia-ceo-jensen-huang-says-...

I can't find the exact Youtube video, but it's out there.

[hkgjjgjfjfjfjf]

You missed the Ryzen hx ai pro 395 product announcement