Yeah, there are a bunch of weird inconsistencies in the presentation. I too would prefer to see some numbers. But it’s a marketing event, so of course there will be some cherrypicking.
Nuvia: don’t hold your breath
- Thread starter Cmaier
- Start date
Yeah, there are a bunch of weird inconsistencies in the presentation. I too would prefer to see some numbers. But it’s a marketing event, so of course there will be some cherrypicking.
Unfortunately getting numbers is going to take awhile …
If anything it’s the other way around given that these “Elite” SOCs aren’t launching any time soon. Easier to schedule an announcement event for a product that isn’t actually ready.
Very unclear and the lack of efficiency cores is almost surprising. It’s probably that they didn’t have time to develop both types of cores and licensing issues would likely prevent them from using an ARM one.
Jimmyjames
Elite Member
- Joined
- Jul 13, 2022
- Posts
- 1,094
I’m wondering what Apple think qualifies as “scary fast”?
Alright alright I’m updating my prediction in the other thread to the M3 extreme being released.
Jimmyjames
Elite Member
- Joined
- Jul 13, 2022
- Posts
- 1,094
This slide from Arstechnica is interesting
So their 2024 12 P core SoC is 50% faster than Apple’s 2022 4 P core + 4 E core Soc?
That seems less impressive.
So their 2024 12 P core SoC is 50% faster than Apple’s 2022 4 P core + 4 E core Soc?
That seems less impressive.
- Joined
- Sep 26, 2021
- Posts
- 6,330
- Main Camera
- Sony
There’s a lot about all these numbers that doesn’t make sense. Guess we’ll have to wait to see real silicon.
Aaronage
Power User
- Joined
- Feb 26, 2023
- Posts
- 144
Seems like Qualcomm went with an Intel/AMD style boost model with these chips?
Not a big fan of that to be honest, I like Apple’s approach. M chips will maintain peak single thread and multi thread performance indefinitely if thermals allow, even if the GPU is loaded (seems they only ever throttle back when the ambient temperature is too high for the cooling solution to cope).
It’s not the end of the world, but it’s something I wanted them to copy.
Wondering if the peak 4.3GHz boost is only possible in short bursts
Not a big fan of that to be honest, I like Apple’s approach. M chips will maintain peak single thread and multi thread performance indefinitely if thermals allow, even if the GPU is loaded (seems they only ever throttle back when the ambient temperature is too high for the cooling solution to cope).
It’s not the end of the world, but it’s something I wanted them to copy.
Wondering if the peak 4.3GHz boost is only possible in short bursts
Jimmyjames
Elite Member
- Joined
- Jul 13, 2022
- Posts
- 1,094
Another slide. Qualcomm are very sneaky!
50% better multi core perf than M2, but…at 50 watts?!
Chart from @AppleBites10476 on Twitter.
50% better multi core perf than M2, but…at 50 watts?!
Chart from @AppleBites10476 on Twitter.
Last edited:
Yeah it’s more similar to an Mx Pro/Max CPU in terms of well … everything. Which makes the comparison with multicore with the M2 even odder and as you and @Cmaier already pointed, with those ST scores and that number of cores, why are the multicore numbers not better? Lack of E cores? Something feels off.
- Joined
- Sep 26, 2021
- Posts
- 6,330
- Main Camera
- Sony
Maybe they can’t run all the cores at once for very long due to thermals.
Or maybe they are faking benchmarks.
I would think they’d be doing this in a test bench under ideal conditions where thermals wouldn’t be an issue but maybe not.
I hope not. That won’t last long if they are.
The only other option other than these that I can think of off the top of my head is that Apple’s memory system and fabric is fantastic. Multicore scores especially benefit from it. It’s possible that could explain some of the proportional differences between single core and multicore scores between the Elite SOC and the M2 Pro/Max. Maybe there is a problem at the cache level? Or the fabric? So getting data to all the cores is harder for the Elite SOC vs M2 Pro/Max. The RAM itself in the Elite is intrinsically higher bandwidth than the M2 but maybe the memory controller isn’t as wide? I’m just just trying to think of an area where Apple is known to shine and they may not be quite as capable which would explain some of the oddities we’re seeing. Dunno.
- Joined
- Sep 26, 2021
- Posts
- 6,330
- Main Camera
- Sony
If there are no efficiency cores, then they may get hotspotting from the performance cores being too close together. Just a horrible theory, based on too little information.
I would encourage everyone interested in the topic to watch the relevant parts of Qualcomm's keynote — it's on their website and on YouTube. Since it comes with speaker commentary it's probably more useful than chasing down screenshots at news sites.
Let me try to organise the claims we have so far (with some subjective notes from my side):
- Elite X has 12 P-cores, organised as three clusters of four cores (just like Apple Silicon)
- Elite X has x86-like Turbo-Boost up to 4.2Ghz, but only for two cores — other cores in the same cluster must be idling for this to work. The "regular" frequency (unclear whether burst or sustained) for multi-core operation is 3.8Ghz.
- Elite X is 15% faster than M2 Max at peak single-core GB6 (presumably when running at 4.2Ghz), or can match M2 Max in single-core GB6 while consuming 30% less power. Interestingly, they don't mention power consumption at the peak, that 4.2Ghz boost must come at a significant cost
- Elite X is 50% faster than base M2, and slightly faster than i9-13900HK, which would put the GB6 multi-core score at around 15500, 3-5% higher than M2 Max. The power consumption however is ~ 50 watts, 25% higher than M2 Max
Overall, it appears that Oryon's IPC is comparable to M2 Max, and that they can run at 3.6-3.7Ghz at lower power than Apple, which is a great achievement. On the other hand I am a little puzzled by their power consumption claims relative to some other chips, it is possible that they are measuring power consumption in some way that gives them advantage (for example, mainboard power instead of CPU core power). It was also reported earlier that Qualcomm had some problems with their power management unit, we will have to see whether this is true I suppose. Qualcomm also said that they will be releasing more technical info in the coming days, so maybe we will get a better idea of their performance and power consumption claims.
Let me try to organise the claims we have so far (with some subjective notes from my side):
- Elite X has 12 P-cores, organised as three clusters of four cores (just like Apple Silicon)
- Elite X has x86-like Turbo-Boost up to 4.2Ghz, but only for two cores — other cores in the same cluster must be idling for this to work. The "regular" frequency (unclear whether burst or sustained) for multi-core operation is 3.8Ghz.
- Elite X is 15% faster than M2 Max at peak single-core GB6 (presumably when running at 4.2Ghz), or can match M2 Max in single-core GB6 while consuming 30% less power. Interestingly, they don't mention power consumption at the peak, that 4.2Ghz boost must come at a significant cost
- Elite X is 50% faster than base M2, and slightly faster than i9-13900HK, which would put the GB6 multi-core score at around 15500, 3-5% higher than M2 Max. The power consumption however is ~ 50 watts, 25% higher than M2 Max
Overall, it appears that Oryon's IPC is comparable to M2 Max, and that they can run at 3.6-3.7Ghz at lower power than Apple, which is a great achievement. On the other hand I am a little puzzled by their power consumption claims relative to some other chips, it is possible that they are measuring power consumption in some way that gives them advantage (for example, mainboard power instead of CPU core power). It was also reported earlier that Qualcomm had some problems with their power management unit, we will have to see whether this is true I suppose. Qualcomm also said that they will be releasing more technical info in the coming days, so maybe we will get a better idea of their performance and power consumption claims.
Last edited:
This is my confusion. Even ignoring the potential for 15% faster in single core, it goes from matching the M2 Max at significantly lower power in single core to only beating it by 3-5% in multicore with 50% more P-cores albeit no E-cores and requiring 25% more power. A priori, that seems very odd.
If anything 12 P-cores with the stated single core stats should be able to beat the brakes off of an 8+4 design in multicore either in performance or power. And yet it doesn’t. Why not should be very interesting!
I mean Apple’s E-cores are very good, in fact, best in class, but even so they are still little cores. These aren’t Intel’s “E-cores” which are really midsize cores.
It is odd, isn’t it? That’s what makes me suspect that they might be measuring package power and not core power. But even then it’s weird. Let’s see if we will have more info in the coming days.
1st off, thank you for your compilation. However, what do you mean by x86 turbo boost?
There’s definitely something else going on with those numbers aside from power sampling. We’ll figure it out.
- Joined
- Sep 26, 2021
- Posts
- 6,330
- Main Camera
- Sony
I think he meant “x86-style.”
Sorry, I meant “x86-style TurboBoost”. It’s not as big dynamic range as recent x86 CPUs, but still almost 0.5GHz.
- Joined
- Sep 26, 2021
- Posts
- 6,330
- Main Camera
- Sony
Hah! My dumb guess had some merit. Probably explains why their multi-core benchmark doesn’t scale so well - instead of running all cores at a sustainable clock, they run half of them at boost.
I also suspect their single core benchmark won’t map great to the real world for similar reasons.