This all sounds so inept on Qualcomm’s part that I have trouble believing the explanation.
Qualcomm is trying to make money on the whole shebang, not on the PMICs. They’re trying to sell, for lack of a better word, a system on multiple chips. Intel and AMD do this too (see their chipsets, for example). When someone makes a laptop, the the CPU vendor accounts for some of the BOM cost, and that’s their revenue. The manufacturer is willing to pay some total BOM cost at a given performance point, and they’ll buy more units at a lower total. So the CPU vendor wants the total to be as low as possible and their portion of it to be as high as possible. Except to the extent that supplying PMICs (if done competently) diverts more of the BOM to them without increasing the total, they don’t really care whose PMICs are used in the grand scheme of things. And they certainly wouldn’t torpedo the entire product launch over PMICs.
I bet this is actually just overconfidence in their product. They designed around their proprietary PMIC, and that design is important: a CPU draws variable amounts of power, and the system will not work if the power and voltages are wrong. For bonus points, the CPU would much rather operate at lower power (skip clocks or whatever) than simply crash if an intensive workload starts and the battery plus whatever power supply is connected can’t keep up. [0]. So I bet they tried to get all this right, they designed around their PMICs, they didn’t realize the chip format was inappropriate, and they can’t fix it in time for market. Oops.
Qualcomm has never done the laptop thing for real before. Intel, AMD, and Apple have.
[0] Anyone remember Apple’s issues here?
Remember , and for those who don't know. This is semiaccurate, known for hyperbole. I can't remember a single thing they got right.
I distinctly recall my disappointment when I heard it was Qualcomm that acquired Nuvia because I knew they'd find a way to f*ck it up. This is unfortunately incredibly on brand for Qualcomm. They do the same bundling with the 5G Radios and Snapdragon SoC's. They're more of a patent law firm than a tech company.
Honestly was ever Qualcomm not a master of proprietary cash grabs? After all, even in 90s they had their own standards as opposed to GSM. It is in their DNA.
We could have been better off had any of the more "universal" chip companies won the Android SoC race (like TI, Freescale and so on)
Even after GSM they lobbied for their CDMA standard which is why 3G and its successors are based on it. Due to their parents.
It's too bad because GSM was good tech with guaranteed cell capacity and better on weak connections.
DEC's Alpha was doomed at release, because the operating systems licensed with it were a minimum of $30k just to get the install CD (for VMS or OSF/1 "Unix").
There must have been desperation among the design team, as 3rd party Alpha motherboards slowly began to appear that ran basic Linux.
However, Alpha was a glutton for power. DEC then designed the StrongARM, which amazingly worked out far better for low power applications.
I think that I can hear DEC designers screaming inside Qualcomm.
This was like listening to one of Quals engineering leaders try to explain, repeatedly, to business leadership why what they were doing was not going to work.
Popcorn worthy.
This is one of the dumbest things I've ever seen.
To sum up:
- Qualcomm mandated that manufacturers use their power supply chips with their processor
- these power supply chips are wildly inappropriate for laptops, requiring 4-6 in parallel when a single properly spec'd chip would work
- said chips are designed for space-constrained phones, and are so small they require a different type of PCB
- those PCBs can't handle the power and need to double the layer count
- cost of PCBs becomes absolutely ludicrous at this scale
- Qualcomm refuses to allow manufacturers to use a more appropriate power supply
- manufacturers literally cannot afford the design and stopped using Qualcomm
- Qualcomm is now paying manufacturers to use their chipset, and are losing money in the deal
I don't even know how to express what an absolutely terrible idea this is. The Qualcomm chips are not fit for purpose, at all. Totally inappropriate for the application. No engineer in their right mind would go for this.
Qualcomm is mandating that manufacturers eat a fuckton of cost in raw materials, drastically reduce the efficiency and battery life of their product, and charge way more than the market would pay for an underpowered, overheating laptop with a pitiful battery life. For... Reasons?
It makes no sense at all. No one is going to use this chipset. If any devices do make it to market, their terrible performance will absolutely trash Qualcomm's reputation in this space and consumers won't buy them in the future.
Absolutely insane
A bit off topic, it it keeps surprising me how many dedicated chips cellphones have.
iPhone 15 Pro has 10 different power management chips. It has one dedicated transmitter and two receiver chips. It has two different rf front end chips. Plus four chips for cellular modem, wifi/BT, uwb, and nfc. And then two chips for envelope tracking and clock generating.
I have this feeling that designs in general really are multi-chio, and no one bothers engineering alternatives. Almost everyone goes out and buys the recommended power supply chip for the chip they want to use. Qualcomm not having an actual decent power supply solution & just asking people to keep adding more copies of the known power supply unit seems both all too expectable & also ridiculous. And it just seems all too typical that no one, even the very upper end, isn't going to take the time & risk to engineer their own better support solution anyways.
https://www.ifixit.com/Guide/iPhone+15+Pro+Max+Chip+ID/16532...
It’s interesting. A big reason why Qualcomm got to making CPUs in the first place was that they could sell it as a value add to their modem chips everyone was using: now you can have both in one chip!
We have seen Apple try to make a modem of course, but it seems to have fizzled out again. Wonder why they never looked at WiFi/BT, it’s a lot easier and Intel recently sold theirs iirc.