Discussion Qualcomm Snapdragon Thread

[SpudLobby]

Not really. Lunar Lake is it's own swimlane and has no proper successor for now.

Panther Lake U is a semi-successor to Lunar Lake which inherits some of LNL ideas (Like 4P + 4LPE), but with a smaller GPU and targeting a cheaper price segment. But it doesn't use On-Package Memory afaik.

Edit: Just saw the shipping data that raises the possibility of PTL-U having PoP Memory SKUs. Interesting stuff.

I don’t believe LNL will get a proper successor. The way Pat talked about it seems like PTL-U is the successor and/or LNL is a learning experience more than a future category

 

[poke01]

Yes, it is a tradeoff. That's something the people who are always cheering when battery life is increased and think the world is going to end if it is reduced don't get.

They will say "why does Apple want the iPhone to be thin, I would rather have a device 50% thicker with double the battery life" That's what THEY may want, but that's not what I want. If I could I'd get one that's even thinner. Not because I care about thinness per se, but because I prefer it be lighter. My nearly two year old 14 Pro Max still has 60% or so battery life at the end of my day's use of it. That's down from where it was when it was new, but it could have had a smaller battery without impacting MY use of it.

People who want a thicker phone with longer battery life can buy today's iPhone (or my hypothetical thinner one) and put it in a case that builds in a battery. They get what they want without impeding my choices. That's a little harder to do with say a Macbook Air, since no one puts laptops in a "case", but the Macbook Pro exists for people who want to trade off increased weight for a larger battery.

I'm not saying the Pro Max has too big of a battery - it is too big for me but Apple is choosing it based on their overall customer base. But they don't need to satisfy everyone, because those willing to trade size/weight for battery have a perfectly good DIY option.

This is also why they introduced the 15” Air bigger battery and bigger screen for those that don’t mind the increased weight plus a much cheaper option than the 16” Pro. Also cuts into 15” sales of LG and Microsoft Surface.

 

[Ghostsonplanets]

4 cores in 2024? No thanks. Fewer cores would get overwhelmed more easily by heavily threaded applications or heavy multitasking so don't see the point unless you just want to do light tasks. On a tablet or maybe netbook, such a cluster might be fine but in a work laptop? Get the hell out of my work laptop!

Cluster. Not cores. Think of AMD Zen 2 4 Core CCX to Zen 3 8 Core CCX.

I don’t believe LNL will get a proper successor. The way Pat talked about it seems like PTL-U is the successor and/or LNL is a learning experience more than a future category

Yes, I also think the same. Specially with the shipping leak that some PTL-U SKUs will also have PoP memory.

 

[SpudLobby]

Cluster. Not cores. Think of AMD Zen 2 4 Core CCX to Zen 3 8 Core CCX.

Yes, I also think the same. Specially with the shipping leak that some PTL-U SKUs will also have PoP memory.

Yeah, fwiw even without that I think it would be — Qualcomm is clearly showing it’s not a necessity.

That said, Pat and other intel execs made comments about memory capacity and configuration that would match the shipping leak. IOW — saying “there will be options” as in “we’ll have PoP, CAMM, non-CAMM”

 

[SpudLobby]

4 cores in 2024? No thanks. Fewer cores would get overwhelmed more easily by heavily threaded applications or heavy multitasking so don't see the point unless you just want to do light tasks. On a tablet or maybe netbook, such a cluster might be fine but in a work laptop? Get the hell out of my work laptop!

Dude please try actually reading things.

 

[igor_kavinski]

Dude please try actually reading things.

Dude thank you

 

[FlameTail]

Good news for Snapdragon X laptops;

Microsoft postpones Windows Recall after major backlash — will launch Copilot+ PCs without headlining AI feature

Microsoft postpones Windows Recall after major backlash — will launch Copilot+ PCs without headlining AI feature

Windows Recall will arrive at a later date, available in preview "in the coming weeks."

www.windowscentral.com

This gives the opportunity for those laptops to really shine in reviews with their great battery life, thermals and other qualities, without being clouded by the controversial recall feature.

 

[poke01]

Good news for Snapdragon X laptops;

Microsoft postpones Windows Recall after major backlash — will launch Copilot+ PCs without headlining AI feature

Microsoft postpones Windows Recall after major backlash — will launch Copilot+ PCs without headlining AI feature

Windows Recall will arrive at a later date, available in preview "in the coming weeks."

www.windowscentral.com

This is a good thing. But one month later AMD Strix is launching…

The only supposed benefit of these QC chips will be battery life. They lost the AI feature advantage not that it was ever good.

 

[FlameTail]

The only supposed benefit of these QC chips will be battery life. They lost the AI feature advantage not that it was ever good

It still has the other AI features. Only the controversial Recall got nixed.

 

[igor_kavinski]

All this AI preaching and thousands of AI servers at their disposal and Microsoft still can't avoid making dumb mistakes. That should tell anyone how useful AI in its current form is.

Or the Recall feature development was done in a super hurry without using AI assistance. Worse if it was done USING AI assistance!

Dear Lord.

 

[FlameTail]

That 31 hour battery life claims should be investigated...

 

[FlameTail]

Adreno GPU naming scheme:

From:

AnandTech | Gallery - Qualcomm Snapdragon X CPU & GPU Architecture Slide Deck - 30 Photos

www.anandtech.com

 

[soresu]

Adreno GPU naming scheme:
View attachment 101145
From:

AnandTech | Gallery - Qualcomm Snapdragon X CPU & GPU Architecture Slide Deck - 30 Photos

www.anandtech.com

Nothing to see here really beyond the drivers and the potential to hopefully replace OGLonDX12 with Zink.

Be interesting to see if it can handle external GPU connections through USB4.

 

[FlameTail]

Oryon uarch diagram by Cardyak;

https://twitter.com/x/status/1801544455578595740

I am going to make a comparison table of Oryon, Cortex X4, and M4-P.

 

[FlameTail]

Comparison of microarchitecture of M1-P, M4-P, Cortex X925, Oryon.

Oryon looks pretty similar to Firestorm, as expected. Cortex X925's IPC is only a bit better than Oryon's, but they have spent significantly more resources (substantially wider structures) to get there. I think this really speaks to the design proficiency of the Oryon team, and what sets it apart from stock ARM cores. I am curious to see how much die area an Oryon core take up (will have to wait for X Elite die shots).
____
Geekbench 6 scores for M1-P and M4-P are based on MacOS. Geekbench 6 score for Cortex X925 is based on leakers and estimates. Geekbench 6 score for Oryon is based on Linux/Android for fairness.

Numbers were obtained from Cardyak's uarch cheat sheet.

 

[FlameTail]

@igor_kavinski

Littlestgirlonearth is saying she got 31 hours of battery by using it at 60 Hz

https://www.reddit.com/r/GalaxyBook/comments/1dd7t0v/comment/l8ji0mq

 

[igor_kavinski]

Comparison of microarchitecture of M1-P, M4-P, Cortex X925, Oryon.

I find X925 most impressive. So much performance with lower cache sizes. That IS efficient!

 

[FlameTail]

I find X925 most impressive. So much performance with lower cache sizes. That IS efficient!

I think the exact opposite. In terms of core structures, X925 is so much wider than Oryon (even wider than M4-P in some aspects). Not so efficient, I think.

 

[igor_kavinski]

Not so efficient, I think.

Caches can be pretty costly in terms of power. X925 is running at a lower speed yet performing decently. It could be the most power sipping high performance CPU at the moment. In second place is M1-P and I have to say it's scoring impressively at the lowest speed among the four. But I suspect that X925 will scale the best in multicore workloads as the core count increases.

 

[soresu]

Cortex X925's IPC is only a bit better than Oryon's, but they have spent significantly more resources (substantially wider structures) to get there

#1. This also means they have more potential for improvement by simply tweaking and optimising their current design.

#2. ARM's MO has been incremental year on year improvements on a base design since v7-A.

Just look at how much they squeezed out of the original A76/Ares design.

#3. If Oryon 2 is delayed we could have X930 and X935 before QC brings their next core to market.

 

[soresu]

I think the exact opposite. In terms of core structures, X925 is so much wider than Oryon (even wider than M4-P in some aspects). Not so efficient, I think.

Cache width vs smart use of smaller caches is a design tradeoff.

At the end of the day it may cost you less to have a super wide core with smol cache than a less wide core with beeg cache.

This may change as 3D interconnect pitch decreases, and total connections (and therefore bandwidth potential) increase making stacking <L3 cache viable and they may be able to have their cake and eat it with cheap(er) cache layers on a different node.

 

[FlameTail]

#3. If Oryon 2 is delayed we could have X930 and X935 before QC brings their next core to market.

X930 yes, X935 no.

X930 will be announced in May 2025, and come to devices in late 2025. X953 is a 2026 core.

The rumours is that Oryon V2 is coming in 2025Q3 at the earliest, and 2026Q1 at the latest. At any rate, some kind of announcement at the Snapdragon Summit 2025 is 99.99% certain.

 

[FlameTail]

Ok here's a question for the experts here:

*Let's say Qualcomm's engineers have finalized the CPU, GPU and NPU of X Elite G2.

Compared to X Elite performance,
CPU MT +30%
GPU 2x
NPU 2x

Now we need to find the bandwidth to feed all that performance. The engineers haven't decided on the memory subsystem. They are looking at 2 possibilites:

(1) LPDDR5X-9600 128 bit [153 GB/s] + 32 MB SLC.

(2) LPDDR6-10700 256-bit [228 GB/s] + 8 MB SLC

So which solution would be the best?

Adding more cache to the die increases the cost of the die, which in turn increases the price of chip, which is turn raises the msrp of the OEM laptop. But by using LPDDR5X, we can keep the cost of memory low.

Utilising LPDDR6 increases the the cost of the memory (because it will be new and expensive at the time), which will in turn raise the price of the OEM laptop. But by using less cache, we can keep the cost of die low.

(Disregard supply/availability constraints of LPDDR6).

*This is a completely hypothetical scenario.

 

[FlameTail]

@eek2121

No. ARM won't be faster. It will also (on top of not being faster) have an additional 30-50% penalty for emulating X86, which will also hurt perf/watt.

I guess you haven't heard that Qualcomm has implemented hardware acceleration for x86 emulation in Oryon?

The Qualcomm Snapdragon X Architecture Deep Dive: Getting To Know Oryon and Adreno X1

www.anandtech.com

That's why Microsoft is claiming that their new Prism emulation layer is as good as Rosetta2. I have heard that the emulation penalty is only like -15% performance.

Find me one vendor that routinely drops 15+% IPC or performance increases for CPUs without driving power consumption through the roof. I'll wait.

ARM?

They achieved 17% IPC increase in one year (Cortex X4 -> Cortex X925). I expect Oryon V2 will also bring substantial IPC improvements.



If x86 continues in this trajectory, ARM CPU vendors are going to surpass them in absolute performance. ARM already has huge IPC advantage over x86, which makes up for the clock speed disadvantage.

 

[reggie_fils_aime]

Looks like there's a boost frequency bug with early production Samsung Snapdragon laptops. I'm sure we're all shocked to learn that it's a total dog when locked at 2.5 ghz!

Snapdragon X Elite in the wild is allegedly slower than iPhone 12 — first benchmarks of Samsung Book4 Edge disappoint

Wait for firmware updates before condemnation

www.tomshardware.com