Unreal Engine 5 Nanite running on Snapdragon 8 Elite (official demo):
Likely Google's new Tensor for '25 or '26. Amon alluded to out-licensing the design at one point in an interview I can't find, and some Chinese rumors indicated that clocks were around 5 Ghz, far out of line for an ARM TLA design...There were some new documents posted in the Qualcomm vs ARM lawsuit (the countersuit from Qualcomm against ARM). Lots of interesting details in this new submission from Qualcomm.
In particular, Qualcomm stated that a potential customer (a "AI and Ecosystem Company") that relies on a "competitor's chip" for "substantial processing needs" was about to agree to have a "custom chip" designed based on the "Qualcomm's custom-built CPU". But the company has now refused to "finalize a termsheet" after the Breach Letter was published (the Bloomberg leak from November).
Qualcomm specificaly asks for damages from the business opportunity with the "AI and Ecosystem Company" in its lists of remedies. So, it should be a somewhat substantial customer and opportunity for Qualcomm.
Any guesses what this "custom chip" would be and who the "AI and Ecosystem Company" is?
5 ghz?Likely Google's new Tensor for '25 or '26. Amon alluded to out-licensing the design at one point in an interview I can't find, and some Chinese rumors indicated that clocks were around 5 Ghz, far out of line for an ARM TLA design...
No that's normal.5 ghz?
Sounds like a lot even for Oryon or M4.
With the way they’re pushing frequency on the 8 Elite, I’d expect 4.6GHz in the 8 Elite Gen 2.5GHz if it's in the Laptop chips sounds possible. For instance, we don't know how high would Oryon V2 clock in a Laptop chip, possibly around 4.6GHz or so, the next chips could clock 5GHz with certain design tweaks + N3P + maybe some good binning. Of course, maybe then an IPC increase is out of the question or very small or requires a considerable amount more of transitors.
“We want to be able to have a portfolio that can go into every price point with a Copilot+ PC,” Amon said outside Qualcomm’s booth on the CES showroom floor in Las Vegas. “We're not stopping here. We're just starting with [that] portfolio of products.”
In particular, Amon said, Qualcomm is looking to expand into the high-performance PC graphics market, something that would put the company into direct competition with heavyweight Nvidia (NVDA).
Qualcomm Data Center team is developing High performance, Energy efficient server solution for data center applications.
...
Our focus is on developing server-class high performance CPU solutions that are highly optimized for the needs of the server product.
Having bigger (whatever you put here) than yours ?I don't get the MT SCORE obsession in phones.
Outside of some specific workloads, what is the point for vast majority of people?
Or die shrinks on logic allow for higher clocks and they’re already running about a 4 cycle L1 latency anyway, and can hit 4.5GHz on gaming phone chips with Oryon V2. Apple is doing 4.4GHz on M chips with a wider bigger design and I doubt that frequency is a fundamental limitation as opposed to the highest possible bin for every M chip so they don’t have to frequency bin (which they generally do not, aside from one time with the A15 in iPad Minis).5GHz if it's in the Laptop chips sounds possible. For instance, we don't know how high would Oryon V2 clock in a Laptop chip, possibly around 4.6GHz or so, the next chips could clock 5GHz with certain design tweaks + N3P + maybe some good binning. Of course, maybe then an IPC increase is out of the question or very small or requires a considerable amount more of transitors.