It’s a 12 core that has somewhat better performance than M3/M4. I don’t doubt the 4+4 can go fanless, but won’t that be in devices that have no hope of competing in the premium thin and light space? They can keep the snappy UI and web browsing that single thread allows, but they will get crushed in everything multithreaded, won’t they?the what.
A 4+4 with lower clocks would go for 8-10W SDP just fine.
Hamoa is 12 big chungus cores (at pretty fast o'clock), it was never inteded to be a 10W part.
That Strix run wasn't at peak clocks.This is faster than Strix benchmark that came out today XPS 13 is a thin and light laptop as well.
This is faster than Strix benchmark that came out today XPS 13 is a thin and light laptop as well.
Why isn't anyone doing a single core cluster containing two P-cores and four E-cores with shared cache because that could really reduce core to core latency and context switching latency?Just don’t expect huge gains moving forward. I echo the sentiment some others have brought up: we appear to be gradually hitting some kind of bottleneck, regardless of ISA.
What's the use of a 2P+4E cluster?Why isn't anyone doing a single core cluster containing two P-cores and four E-cores with shared cache because that could really reduce core to core latency and context switching latency?
It would reduce the overhead of thread migration between the hybrid cores in case of mixed workloads (fire up the P-cores for quick bursts of ST workloads that finish in a few seconds and continue to do your normal long duration workloads on the E-cores that don't really need 100% CPU utilization).What's the use of a 2P+4E cluster?
Doesn't make much sense. That applies only to Intel style E-cores, which are in sufficient numbers and are utilised for MT throughput.It would reduce the overhead of thread migration between the hybrid cores in case of mixed workloads (fire up the P-cores for quick bursts of ST workloads that finish in a few seconds and continue to do your normal long duration workloads on the E-cores that don't really need 100% CPU utilization).
It could help them reduce the die size by cutting out the dedicated cache for the E-cores. Or they could have bigger caches in the same die size.In Apple/Qualcomm CPUs, the E-cores are purely there for efficiency. High MT throughput workloads run on the P-cores.
Ye but no.
It seems that ultimately, the real selling point of Copilot+ PCs is the Snapdragon chips (and all the benefits that come with it), not the AI features.
Except they would only create a gaming laptop or a professional workstation with that large a battery and both of those need RTX GPU otherwise won't sell.X Elite + 99 Whr battery
Would be a holy combination.
I prefer reviews like that where they don't try to cover up the flaws with shiny marketing points.
AI is never going away. It, in fact will be the only way the overwhelming majority of people will use compute. There is not much qualitative that can be done on current devices and Qualcomm doesn’t yet have the software stack to even reach current hardware potential.Ye but no.
If they hang all this AI blarg on it and that bubble eventually pops, it will impact Windows PC sales, but likely WoA most.
This is probably true, i.e., they could have avoided the (allegedly) mess. But this is very insensible of how the Intel business works.I don't want QC to fail but the hype was just too much and they were stupid enough to ignore the advice of experienced OEMs and went with their higher cost boards, adding to the platform costs.
If the leaks are real, their price is already considerably lower.The sensible thing for them would be to scratch their lofty ambitions and achieve mass market penetration through reasonable and lower prices than the competition. Otherwise it's going to be the chicken and egg problem for native WoA application/games developers (why should they expend extra effort for low sales volume?).
He has a point with Microsoft's magic thinking about ARM being what they need to compete with Apple.Lengthy video by Wendell discussing "what when wrong this time for Windows ARM".
Qualcomm had the privilege of seeing how others did it to at least not make some mistakes, but they still don't have a Dev Kit? Why?
Is like there's still no plan from both Qualcomm and Microsoft. Why?
I do not contend that AI will go away, only that the overhype of it simply to sell more computers with product offerings of dubious value to the consumer is likely to come crashing down at some point.AI is never going away. It, in fact will be the only way the overwhelming majority of people will use compute. There is not much qualitative that can be done on current devices and Qualcomm doesn’t yet have the software stack to even reach current hardware potential.
Current Snapdragons will be able to do much more in a few months than they can now, the software stack is incoming. It will still be nothing compared to what devices in 2030 can do.