Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.

crazy dave

macrumors 65816
Sep 9, 2010
1,451
1,221
Traditionally thin and light notebooks are not good performers in the processing department.

Base Mx SoC are purpose built for thin and light portables.

The X Elite/Pros are targeted at a diverse range of form factors, and so it it logical to think that it will have to make compromises to meet those needs.

I think this is the reason why we are seeing GB6 benchmark all over the place recently for the X Elite/Pro SoCs.

IMHO, Apple's approach is better, as SoCs are designed based on end product needs, instead of the other way around.
I think the 2nd generation of Oryon-based SOCs will give a better point of comparison to Apple designs. Remember, it took nearly a year before the Pro and Max SOCs were released. Prior to Oct 2021 we only had the base M1. Right now Qualcomm has to rely on frequency/voltage binning but for ARM to build essentially a singular SOC around all the different product categories. They have no custom efficiency cores and can't use ARM's. The rumors are that the Snapdragon Phone SOC coming later this year/early next year will have 2nd generation P-cores and custom designed E-cores. If that's so, then 2025 Qualcomm Snapdragon SOCs may be more comparable to Apple's equivalent, especially depending on how many different SOCs they decide to design for what markets.

That the Mx SoC has 8 cores is Apple's choice. Maybe the MBA is an outlier and laptops at that price point have more than 8 cores.
There's good news there, Apple's most current M4 SOC has 10 so we can presume the future MBA will as well. ;) But the larger point is see above. Slightly shorter: Qualcomm went with the design they had available with the core they had available. I strongly suspect when they develop their own E-cores, we'll see Snapdragon SOCs with a mix of P- and E-cores, though potentially a different mix from Apple depending on what markets Qualcomm is targeting.
 
Last edited:
  • Like
Reactions: killawat

Xiao_Xi

macrumors 68000
Oct 27, 2021
1,627
1,101
The rumors are that the Snapdragon Phone SOC coming later this year/early next year will have 2nd generation P-cores and custom designed E-cores.
Where did you get this rumour?

According to Gizmochina (today):
The Snapdragon 8 Gen 4 is expected to switch to a 2+6 design, eliminating the efficiency cores altogether. It will reportedly feature two high-performance Oryon cores clocked at 4.26 GHz, along with six Cortex-A725 cores at 2.8 GHz.

I think the 2nd generation of Oryon-based SOCs will give a better point of comparison to Apple designs.
The second generation of Oryon-based SoCs would give the Nuvia team the opportunity to design a notebook-specific core. I wonder if they will have to start the design from scratch.
 

quarkysg

macrumors 65816
Oct 12, 2019
1,247
841
I think the 2nd generation of Oryon-based SOCs will give a better point of comparison to Apple designs. Remember, it took nearly a year before the Pro and Max SOCs were released. Prior to Oct 2021 we only had the base M1. Right now Qualcomm has to rely on frequency/voltage binning but for ARM to build essentially a singular SOC around all the different product categories. They have no custom efficiency cores and can't use ARM's. The rumors are that the Snapdragon Phone SOC coming later this year/early next year will have 2nd generation P-cores and custom designed E-cores. If that's so, then 2025 Qualcomm Snapdragon SOCs may be more comparable to Apple's equivalent, especially depending on how many different SOCs they decide to design for what markets.
I would think tho. that the fundamental issue remains, as in QCA have to design their SoC for a diverse set of ecosystem and therefore looses the option to optimize for a specific set of use cases, due to economy of scale.
 

crazy dave

macrumors 65816
Sep 9, 2010
1,451
1,221
Now that I'm double checking it seems like there were a host of rumors all agreeing that it would be a 2+6 configuration but differing on what those "6" were, from being just down clocked P-cores (Tom's), to ARM E-cores (yours), to fully custom designs (still looking but I did see this about month ago which makes tracking it down hard, I should get Recall 🤣).

With regards to mixing designs, I could be wrong, but my memory is that ARM had tightened its licensing around mixing ARM Cortex cores with custom designs so that was not really practical. Basically you could do it, but they had to be in different core clusters and there were issues with the fabric - essentially P to E-core communication was terrible and this was one reason why a lot of early custom designs from Qualcomm and Samsung suffered in the 2010s. However, it's been awhile since I've look it into and I'd have to recheck that though. It's possible they've changed their licensing model and my information is out of date.

The second generation of Oryon-based SoCs would give the Nuvia team the opportunity to design a notebook-specific core. I wonder if they will have to start the design from scratch.
Do you mean the efficiency-core? If they are designing one in-house, then yes. As for the Oryon-core itself then I'd say the whole point was that the Apple-like could be used for servers but it would still fundamentally be the same core. Xeons and Epycs fundamentally use the same cores as Intel Core and Radeons.
I would think tho. that the fundamental issue remains, as in QCA have to design their SoC for a diverse set of ecosystem and therefore looses the option to optimize for a specific set of use cases, due to economy of scale.
That's where multiple SOCs comes in. While their multiple SOCs still have to work for a wider range of applications than Apple's, I suspect that in future Qualcomm will have multiple SOCs targeting different large classes of devices. That is of course if the current iteration is successful.
 

Xiao_Xi

macrumors 68000
Oct 27, 2021
1,627
1,101
Do you mean the efficiency-core? If they are designing one in-house, then yes. As for the Oryon-core itself then I'd say the whole point was that the Apple-like could be used for servers but it would still fundamentally be the same core. Xeons and Epycs fundamentally use the same cores as Intel Core and Radeons.
Nuvia appears to have designed the Oryon core at 3.4 GHz, which is lower than the M3 P-core (4.05 GHz)/M4 P-core (4.41 GHz) and higher than the M3 E-core (2.75 GHz)/M4 E-core (2.89 GHz).

The Snapdragon 8 Gen 4 is expected to switch to a 2+6 design, eliminating the efficiency cores altogether. It will reportedly feature two high-performance Oryon cores clocked at 4.26 GHz, along with six Cortex-A725 cores at 2.8 GHz.
Indeed, Apple's frequencies for the P and E cores are more similar to what rumours suggest the upcoming Snapdragon 8 Gen 4 frequencies for the P and E cores would have.

It will be very interesting to see if Qualcomm manages to increase the frequency of the Oryon core to more than 4 GHz without sacrificing too much efficiency as it has done so far.
 

crazy dave

macrumors 65816
Sep 9, 2010
1,451
1,221
Nuvia appears to have designed the Oryon core at 3.4 GHz, which is lower than the M3 P-core (4.05 GHz)/M4 P-core (4.41 GHz) and higher than the M3 E-core (2.75 GHz)/M4 E-core (2.89 GHz).


Indeed, Apple's frequencies for the P and E cores are more similar to what rumours suggest the upcoming Snapdragon 8 Gen 4 frequencies for the P and E cores would have.

It will be very interesting to see if Qualcomm manages to increase the frequency of the Oryon core to more than 4 GHz without sacrificing too much efficiency as it has done so far.

Even in the current Elite, the top end model boost up past 4GHz and indeed they do use substantially more power doing so. My own rough estimates are that they use about 2x power going from 3.4 to 3.8GHz (unsurprisingly similar to the M2). Some of the Elite processors we're seeing reviews for in "Air-like" devices are binned for 3.4GHz. The base M2 P-core was roughly around here, 3.49 GHz. While this article has a lot of problems (e.g. P and E cores, but that's just one of the problems), I think the core clocks per SKU are correct:



In the article you linked and in Tom's, they're talking about clocking the phone variant of the Oryon 2 past 4GHz.

I thought Oryon runs at up to 4.3Ghz? The average frequency of M4 on multicore workloads appreads to be 3.4-3.7Ghz btw.

Jinx :)
 
Last edited:

leman

macrumors Core
Oct 14, 2008
19,520
19,670
Nuvia appears to have designed the Oryon core at 3.4 GHz, which is lower than the M3 P-core (4.05 GHz)/M4 P-core (4.41 GHz) and higher than the M3 E-core (2.75 GHz)/M4 E-core (2.89 GHz).

I thought Oryon runs at up to 4.3Ghz? The average frequency of M4 on multicore workloads appreads to be 3.4-3.7Ghz btw.
 

Xiao_Xi

macrumors 68000
Oct 27, 2021
1,627
1,101
I thought Oryon runs at up to 4.3Ghz?
Only some cores of the X1E-84-100. Some cores of the X1E-80-100 can reach 4.0 GHz, but none of the X1E-78-100 cores exceed 3.4 GHz.

CPU NAMECORES2-CORE BOOST / ALL-CORE BOOST
X1E-84-100124.2 - 3.8 GHz
X1E-80-100124.0 - 3.4 GHz
X1E-78-10012N/A - 3.4 GHz

Although notebook manufacturers do not seem to disclose the exact X Elite model, I expect most notebooks to use the X1E-78-100. So, in practical terms, X Elite could be considered a 12-core 3.4 GHz SoC.
 
  • Like
Reactions: killawat

l0stl0rd

macrumors 6502
Jul 25, 2009
482
413
Only some cores of the X1E-84-100. Some cores of the X1E-80-100 can reach 4.0 GHz, but none of the X1E-78-100 cores exceed 3.4 GHz.

CPU NAMECORES2-CORE BOOST / ALL-CORE BOOST
X1E-84-100124.2 - 3.8 GHz
X1E-80-100124.0 - 3.4 GHz
X1E-78-10012N/A - 3.4 GHz

Although notebook manufacturers do not seem to disclose the exact X Elite model, I expect most notebooks to use the X1E-78-100. So, in practical terms, X Elite could be considered a 12-core 3.4 GHz SoC.
Most actually do, the Asus one is the 78, and yes most use the 78 or 80.
Samsung at the moment seems to the only one to use the 84.
 
  • Like
Reactions: Xiao_Xi

Xiao_Xi

macrumors 68000
Oct 27, 2021
1,627
1,101
Most actually do, the Asus one is the 78, and yes most use the 78 or 80 only Samsung at the moment seems to use the 84.
You are right. Although it is sometimes a bit difficult to find, laptops do promote the exact model. I wish the marketing brading was better, so that it would be easier to know the exact model.
 
  • Like
Reactions: l0stl0rd

crazy dave

macrumors 65816
Sep 9, 2010
1,451
1,221
Only some cores of the X1E-84-100. Some cores of the X1E-80-100 can reach 4.0 GHz, but none of the X1E-78-100 cores exceed 3.4 GHz.
Similarly the M3's boost to 4.05 is only one of the cores. That's the nature of the boost.
CPU NAMECORES2-CORE BOOST / ALL-CORE BOOST
X1E-84-100124.2 - 3.8 GHz
X1E-80-100124.0 - 3.4 GHz
X1E-78-10012N/A - 3.4 GHz

Although notebook manufacturers do not seem to disclose the exact X Elite model, I expect most notebooks to use the X1E-78-100. So, in practical terms, X Elite could be considered a 12-core 3.4 GHz SoC.
This gets to @quarkysg's point: it's the same core design but binned by frequency for multiple markets. What "clock speed" would you put the 13th Gen desktop Intel Core iX chips at? Well it depends on which bin, right? i9's TVB goes to 6GHz, while i5 doesn't have TVB and its TB2.0 maxes out at 5.1GHz. For each their base clocks in the 3.2-3.5 range. The Oryon core is designed to operate from 3.4 to 4.2GHz with higher performing silicon needing less power to achieve stable clocks binned for higher end SKUs like the 80 and 84. Comparing overall microarchitecture, node, and SOC design, the Oryon Core is basically an overclocked M2 P-core - some of the M2 Max and Ultras were overclocked M2s too with higher base and boost clocks (which in the parlance of a recent thread could be Apple's "desktop" version ;)).
 

MRMSFC

macrumors 6502
Jul 6, 2023
371
381
From skimming the discussion I see an unexpected advantage for Apple vs. PC ARM designs: licensing.

If some chipmakers are prohibited from using each other’s core designs for P and/or E cores then that adds a hurdle to making the best chips.
 
  • Like
Reactions: killawat

crazy dave

macrumors 65816
Sep 9, 2010
1,451
1,221
From skimming the discussion I see an unexpected advantage for Apple vs. PC ARM designs: licensing.

If some chipmakers are prohibited from using each other’s core designs for P and/or E cores then that adds a hurdle to making the best chips.
To clarify, that was awhile ago and I'm not sure if that licensing issue still applies. It may not. My current (limited) research suggests it may no longer be true.
 

Bodhitree

macrumors 68020
Apr 5, 2021
2,085
2,216
Netherlands
Hmm, ARM may have an advantage here. X86 is limited by the licenses to just two players, Intel and AMD, while there are quite a few ARM licensees who can create their own cores without losing software compatibility.

Like Nvidia creating their own custom ARM cores.
 

Macintosh IIcx

macrumors 6502a
Jul 3, 2014
625
612
Denmark
I’m still not convinced that their power efficiency is so much better than the latest AMD and Intel laptop CPU designs that it is worth the trouble to go Win ARM over Win x86, but if the hype can help introduce a new Win CPU player, then sure, why not...
 
  • Like
Reactions: komuh

Macintosh IIcx

macrumors 6502a
Jul 3, 2014
625
612
Denmark
Of course, Qualcomm have a few months in the sun before AMD Zen 5 Strix Point and Intel Lunar Lake comes out for laptops, so they got that going for them….
 
  • Like
Reactions: komuh

Xiao_Xi

macrumors 68000
Oct 27, 2021
1,627
1,101
Qualcomm have a few months in the sun before AMD Zen 5 Strix Point and Intel Lunar Lake comes out for laptops,
According to rumours, notebook manufacturers will launch notebooks with AMD's latest SoCs next month.
 

komuh

macrumors regular
May 13, 2023
126
113
Tbh battery test are currently not so hot for x elite it seems that windows still sucks in power management department, but cpu results are very impressive probably not so much compared to new Intel and AMD launching soon but still very good for a first gen product
 
  • Like
Reactions: killawat

komuh

macrumors regular
May 13, 2023
126
113
To clarify, that was awhile ago and I'm not sure if that licensing issue still applies. It may not. My current (limited) research suggests it may no longer be true.
Nvidia is in different positions they were trying to go into server maybe even cloud business before. but now it seems they happy milking AI and server GPU spaces whole arm cpu side of things is kinda stale compared to heavy enterprise GPU push and who can blame them they like 10x thanks to AI stack and best GPUs on the market
 

deconstruct60

macrumors G5
Mar 10, 2009
12,493
4,053
Indeed, it would put the system builders on the horns of a devilish dilemma if they wished to create a system that would sell. Sub-M1 levels of ST performance is not ideal for a premium product, even if the MT scores are sustained by the 12 performance cores.

They don't need to take Sub-M1 levels of ST performance it just relax the 'revolutionary better' battery aspect. What seeing in the initial reviews is that system vendors on the more premium systems are going to "incrementally better battery" or "about the same battery" (but perhaps slightly lighter. i.e., smaller battery than Intel model) option. They didn't "have to be" slower than M1.

The 'revolutionary better battery' is a dubious path to go down anyway. AMD and Intel upcoming solutions for 2024-2025 have thrown lots of effort at improving the battery efficiency. They are not matching M-series, but substantially better than they were 1-3 years ago. Qualcomm also already tried that with previous Snapdragons. It really didn't get deep traction. It is a 'possible' mode to set the laptop to, but just not likely going to be a primary mode.
 

deconstruct60

macrumors G5
Mar 10, 2009
12,493
4,053
It may be old, but it's a legal conflict which, as far as I know, has not yet been resolved. So why do you claim it must be AMD/Intel FUD to discuss it now?

1. It was not discussed hardly at all several weeks ago. Or early 2024. It is rather too highy coincidental that this pops up right before the 'buy' button goes live.

2. Also rather convenient timing as the stories about "You aren't going to get Windows 2H24 " if have x86-64


stories also popped up about the same time.

( Do users really want to get 2H release (that usually rolled out in Sept-Oct in June ... maybe not. ), but the point is likely to just to get folks to incrementally slow down impulse buys for a short time. The hoopla about the benchmarks being really bad isn't panning out. )


AMD's response is really coming in July. Intel's toward end of Q3.

3. This case has very slowly gone to trial ( not due until Q4 this year). If Arm said "throw away the Nuvia stuff" years ago at this point and Qualcomm ignored them. That may get Arm bigger punitive damages if they win , but it seems very unlikely that an injunction is coming at this point. There would be more harm to the systems vendors than to Qualcomm at this point since they have wrapped lots more money around the SoCs (in other PC parts ) than what they paid to Qualcomm. An injunction at this point would likely more seriously harm Arm's relationship with system vendors more so over the long term , than any short term harm they could do to Qualcomm.

If Qualcomm losses and get fined $150M but made $200M, then they just pay and move on. ("Better to ask forgiveness than permission" strategy). Arm wants money more than a dead, in the trash Arm chip. If Qualcomm pays the fine and moves on the impact on consumers is about zero.

[ Wouldn't be surprising if both Qualcomm and Arm are waiting to see just how much money is going to be in the pot after shipping for a quarter or so. Qualcomm doesn't want to 'overpay' to get Arm's blessing. (they likely substantively overpaid for Nuvia in the first place. ). Arm doesn't want to miss taking as big a slice as they can get ( since arch license generates less long term revenue for this implementation. ) ]

Arm's punitive damages check just gets bigger the more Qualcomm sells. At this point, they are zero interest in impeding Oryon sales. The folks who are nevous about a successful launch are AMD and Intel.
(This has little to no impact on Apple. They already left x86-64 land and not pressed about Windows. )


4. The longer this drags out the closer get to Oryon version 2 which would be a substantively different design chip anyway ( Arm asked Qualcomm to start over and to a large extent it would be a start over with largely trace and design input data that Qualcomm had collected. ) . Leaked Dell diagram indicated Oryon 2 is coming at end of 2025. Even if Arm wins in mid 2025... appeals and delays could come after Oryon v1 has stopped shipping. Impact to consumers ... not much.


Behind money the second irritant for Arm is time to market for Qualcomm. Even Qualcomm disappeared there is still Arm alternatives coming. The sustained threat there is to the x86 folks. If the PC market doesn't grow much and there are more players for base SoC, then shares of the pie are going to get smaller. That is going to get more FUD slinging. Arm is still a year out from getting to market (e.g., via Mediatek) to the Windows market. Short term delaying articles isn't going to help them no as they are not ramping anything to ship this year.
 
Register on MacRumors! This sidebar will go away, and you'll see fewer ads.