127 million double dollars? Oh my…

Sooo, hyperbole aside, I did a quick search on the site of the retailer I last obtained parts through, right before the prices exploded, and their cheapest 1x64GB RAM DDR5 stick is about £2320.

(Big assumption here that the motherboard only has four RAM slots. If you're going to cram 192GB onto it, you're going to need at least a couple of those, and preferably three, not 2x64+2x32, hence not looking at smaller capacity sticks.)

As I write, eBay has one seller with a single 1x64GB DDR5 at a quarter of that, which makes me think they must be going by pre-explosion pricing, but it also has something at well over £10k, four times the other way (and more), so at least logarithmically, the stand-alone retailer is right in the middle.

That means that this server is probably going to be well in excess of £10k when they finally ship it with full spec. If they're offering a warranty or support contract, probably around £20k. (Or £30k if they do the trick of simply changing the currency sign on a US dollar amount).

For real, tho

Max is supposed to be, well, maximum

Which is not only not the case since they had to put a "+" on top of it, but it is also, apparently, wasn't good enough, so they slapped "Pro" to boot.

Even without the "AI" crap the name is ridiculous.

Why couldn't it have been Ryzen 495? Sure, make it Ryzen AI 495 even, whatever. But Max+ Pro? Seriously, AMD? Did they let a 12 year old chuunibyo name their products? What in the f happened here?

It is targeting the local AI inference market. This is a good choice for a developer who wants one x86 machine to run models that cross the threshold that fits comfortably into 128GB of unified RAM, which is quite a few.

LLM inference, most likely

I don't see a price there, so I'm not sure how "overpriced" it is. But it's a class of product that is expensive across the board, compared to a typical desktop workstation or gaming PC.

The use cases are mostly local AI. That's very demanding and requires compromises one way or another. You want high memory capacity and high memory bandwidth most of all. You can't get both without paying out the nose, so you need to find the sweet spot for your specific use cases.

Discrete GPUs give you high bandwidth, but low capacity.

Standard desktop memory gives you high capacity, but low bandwidth (and also low compute speed since CPUs are no good for the workloads).

This type of product is a middle ground. Integrated memory gives you better bandwidth than DDR5, but not as good as a discrete GPU. Integrated GPU/NPU gives you much better performance than a desktop CPU, and can access the full integrated memory directly. Upgradeable memory is not really possible for this class of product.

The real competition for this is a tricked-out MacBook Pro. Those are faster, but they ain't cheap, and I don't think Apple even sells any 192GB Macs anymore, and when they did it was close to $10k.

The reason (besides ram apocalypse) for the price and non upgradability is the ram is very fast low latency unified memory. It's close to using VRAM for the CPU and gpu.

There's a standard for upgradable lpcamm2 but Framework didn't use it probably because it's even more impossible to source with the RAM shortage.

It's a small box that supposedly does some specific workloads well. It's a weird product for framework though.

I was trying to work out the same thing. The whole idea of Framework is a repairable and modular product in a market where most products are entirely the opposite.

Yeah, same as the old one

...and DGX Spark.

Cause you can't hear anything else?