Discussion Qualcomm Snapdragon Thread

[FlameTail]

I still refuse to believe that the S8G4 will score more than 3000 in single core.

I am a believer. 3000+/10000+. Let's go!

That would be out performing their flagship laptop SKUs.

Yes. It's going to be funny when Qualcomm releases a mobile SoC with stronger Single-core performance and stronger GPU, only a few months after releasing their laptop chip- the X Elite.

 

[FlameTail]

Well one thing for sure is that 4.37GHz is achieved with FinFlex.

How can you be sure?

 

[FlameTail]

https://twitter.com/x/status/1820494291510370374

It seems like Oryon is very strong in Clang.

 

[Tup3x]

I am a believer. 3000+/10000+. Let's go!

Yes. It's going to be funny when Qualcomm releases a mobile SoC with stronger Single-core performance and stronger GPU, only a few months after releasing their laptop chip- the X Elite.

Suddenly Lunar Lake feels really underwhelming too.

 

[hemedans]

Suddenly Lunar Lake feels really underwhelming too.

Probably they will use same power, current soc can scale up to 17W and probably 8G4 Wil hit 20W power usage.

 

[Tup3x]

Probably they will use same power, current soc can scale up to 17W and probably 8G4 Wil hit 20W power usage.

Random short peak aside, there's no way in hell.

 

[FlameTail]

Random short peak aside, there's no way in hell.

17 Watts baby!

 

[FlameTail]

MLID did an interview with an Alderon Games developer. He had some interesting comments to say about the GPU driver situation on Snapdragon X.

Timestamp 1:39:40

 

[hemedans]

Random short peak aside, there's no way in hell.

But those benchmarks were made at same power, sd 8 gen 3 will reach 7000 score in Geekbench when it consume 17W, probably 8G4 will reach 10K score at 17 or 20W, if LNL reach that score while Using 17W aren't they same?

At normal usage 3-6W most phones will score less than 5K

 

[ikjadoon]

The Dell XPS 13 X Elite 80 just posted Notebookcheck's longest-ever-running laptop on their Wi-Fi battery test.

https://www.notebookcheck.net/Dell-XPS-13-9345-Snapdragon-X-Elite-review-Undercutting-the-Intel-Core-Ultra-7.869106.0.html

I'm very impressed. 1500+ minutes with a ho-hum 55 WHr battery is bonkers-level battery that I only expected from industrial / 100 WHr designs.

The diversity in implementations is a bit sad; some X Elite designs did quite poorly, but it's clear that the best SoC & laptop implementations are virtually unmatched. I mean, comparing "best to best" implementations, not one Apple laptop comes even close to the XPS 13 Elite 80.

Gosh, if they just had a normal bloody keyboard (and maybe fanless), I'd really have that on a wishlist by now.

Somewhat unrelated, I'm a little disappointed that AMD did not focus on Zen5 battery life as much as I expected in their disclosures. Maybe we'll see more power-optimized designs in the next six months.

 

[jdubs03]

But those benchmarks were made at same power, sd 8 gen 3 will reach 7000 score in Geekbench when it consume 17W, probably 8G4 will reach 10K score at 17 or 20W, if LNL reach that score while Using 17W aren't they same?

At normal usage 3-6W most phones will score less than 5K

The leak for the gen 4 shows 8840. Where are they going to get an extra 13%?

 

[Hitman928]

The Dell XPS 13 X Elite 80 just posted Notebookcheck's longest-ever-running laptop on their Wi-Fi battery test.

https://www.notebookcheck.net/Dell-XPS-13-9345-Snapdragon-X-Elite-review-Undercutting-the-Intel-Core-Ultra-7.869106.0.html

I'm very impressed. 1500+ minutes with a ho-hum 55 WHr battery is bonkers-level battery that I only expected from industrial / 100 WHr designs.

The diversity in implementations is a bit sad; some X Elite designs did quite poorly, but it's clear that the best SoC & laptop implementations are virtually unmatched. I mean, comparing "best to best" implementations, not one Apple laptop comes even close to the XPS 13 Elite 80.

Gosh, if they just had a normal bloody keyboard (and maybe fanless), I'd really have that on a wishlist by now.

Somewhat unrelated, I'm a little disappointed that AMD did not focus on Zen5 battery life as much as I expected in their disclosures. Maybe we'll see more power-optimized designs in the next six months.

It's interesting that the power consumption under load is so high (and the battery life is so low) given the high battery life of the other tests. It makes me think that Dell might be frequency limiting the CPU on battery (the HP model was doing this and had terrible on battery performance as a result). Some plugged/unplugged tests would be interesting.

 

[FlameTail]

The leak for the gen 4 shows 8840. Where are they going to get an extra 13%?

 

[FlameTail]

I'm very impressed. 1500+ minutes with a ho-hum 55 WHr battery is bonkers-level battery that I only expected from industrial / 100 WHr designs.

The diversity in implementations is a bit sad; some X Elite designs did quite poorly, but it's clear that the best SoC & laptop implementations are virtually unmatched. I mean, comparing "best to best" implementations, not one Apple laptop comes even close to the XPS 13 Elite

The XPS 13 and Surface Laptop 7s seems to be the only 'good implementations'.

 

[jdubs03]

View attachment 104614

Yeah saw that. I’ll believe that when I see it in the end product.
Also not sure what that QRD is referring to exactly particularly the timing. Considering the benchmark had just run.

We already know how far the X Elite is behind the M4. Both SoCs for Qualcomm and Apple are lowered powered derivatives.

It beggars belief that the mobile version for Qualcomm is better than the laptop SKUs. But the Apple is worse. And then QCOM also improves over 30pct on multicore.

 

[SpudLobby]

RE: power

This whole discussion on power curves is idiotic. Some of you are talking about MT, when ST is what’s actually impressive and makes LNL look like a joke — and also suggests Qualcomm turned a corner.

A 2800 & 4+GHz in a phone chip at 5-6W suggests they ironed out kinks from the X Elite. Probably changed some fabric and phdes stuff + additional SRAM (SLC) reduced power — N3E will help a lot but as I mentioned a while ago, not enough to get the 2800-3000 score (or the 4-4.3Ghz) at similar IPC in a phone ST profile.


So clearly design and other changes were made without serious micro architecture changes given the IPC is the same. Looks good if true and bodes well for the X Elite G2. E cores will also be interesting.

 

[Hesperax]

Interesting speculation from this blog based on the Geekbench leak: https://gamma0burst.tistory.com/413193

part1 is the architecture and r3p4, but part1 r1p1 in the SnapX Elite system information using the Oryon architecture.

It can be said that the Snap 8G4 uses the same Oryon architecture, only the revision has changed.

 

[Hesperax]

That’s a good catch.

I checked a representative score for the S24 Ultra: https://browser.geekbench.com/v6/cpu/7201148
and the single core HDR score is 2271.
What’s even stranger is the multi core HDR score for the S24U is 7173, but this Snapdragon score has 7380.

It seems like the Chinese Leakers are blaming the lower score on llvm18 for some reason. It's hard to follow the conversation with Google Translate so I can't tell if the issue is with the compilation of Android or Geekbench.

 

[FlameTail]

Interesting speculation from this blog based on the Geekbench leak: https://gamma0burst.tistory.com/413193



View attachment 104622

Oryon is worse than Cortex X4 in some tests. Interesting.

View attachment 104624

Why is the Exynos pulling 15W+...

So according to this chart, At 10W, N3E Oryon has 20% better performance than X Elite?

That is very very substantial.

Edit: It's Linux vs Windows, so about 10% should be coming from there.

 

[FlameTail]

So 8G4 will use ~50% more power [ 6W -> 10W] for a ~50% performance increase (2200 -> 3200). I guess this is what the leakers were talking about when they say the efficiency uplift of the CPU is in the single digit range....

I am concerned about this trend in increasing ST power in smartphone SoCs. Only a few years ago, it was well below 5W (which is what a typical phone can sustain it's thermal envelope). Now it's touching 10W!

Snapdragon 8G4 will not be alone in this. I suspect the Apple A18 Pro will also be in that range, considering that M4 blew up the power consumption by +50%.

 

[SpudLobby]

or put another way it’s using 35% less power than the SDXE at the same performance, and the N4P to N3E node change doesn’t buy that, more like -15% ish.

It isn’t going to run at 10W in most devices lmfao. That will be for the 4.3GHz gaming phones and purely benchmark BS if it even gets there. The curves are just plotted because why not and they can hit that.

2500 is at around 5W, 2750-2850 low 6W. Not fantastic but not exactly terrible when the X4 was gunning 2200-2300 at 5.5+W afaict. 8g4 is roughly similar to the X925 on N3 and I suspect Qualcomm will iterate over theirs more than Arm will.

Anyway you are completely out of your mind if you think any phones with the A18 and most mass market phones with the 8 Gen 4 or next Exynos will hit 10W over the average of a GB6 ST test. Those frequencies just won’t be shipped or allowed for most — it’s engineering — and Apple certainly won’t do that either.

Spec is different and hits the memory subsystem and ram harder, always has been high, especially FP.

 

[FlameTail]

First Snapdragon X laptop with 5G?

Surface Pro 11 with 5G and Surface Laptop 7 is now available for pre-order via Microsoft's "for business" store

Microsoft is now taking orders for the "for business" models of Surface Pro 11 and Surface Laptop 7.

www.windowscentral.com

 

[DZero]

I still refuse to believe that the S8G4 will score more than 3000 in single core.
That would be out performing their flagship laptop SKUs.
Certainly it won’t reach 3250 (as discussed a couple days ago too).

Just going by the single core score of 2884, apply the frequency adjustment up from 4.09GHz to 4.37GHz, and it’s still 3085.

Assuming a 5% increase in the HDR value (the nT score increase was 2.9%), the score would be 2910. And then the frequency adjustment gets it to 3109.

Based on that score, if it can boost 4.37 then >3000 seems plausible.
But I still think it’s unlikely. If I’m wrong, I’m wrong.

Maybe the phone CPUs will consume way more TDP than expecting or a refined version of their laptop SKUs

 

[Doug S]

Snapdragon 8G4 will not be alone in this. I suspect the Apple A18 Pro will also be in that range, considering that M4 blew up the power consumption by +50%.

Only if they use HP cells in the P cores for A18P. I don't think we should make that assumption.

If they stick with HP cells it'll just have to downclock those P cores more quickly than before to keep from heating things up unacceptably, but even having a one second long burst of higher performance before dropping back would be fine for phones, because most phone loads (outside of demanding games people should be playing on a PC/console) are bursty. Might not benchmark as well, but who cares? In the end I care how fast it feels in my actual use, not bragging rights on GB6 or SPEC.

 

[Hitman928]

Some negative news re: Surface after Microsoft's earnings. Unfortunately, Microsoft had lower Surface division sales this quarter (March to June 2024) and is expecting even fewer sales next quarter (July to September 2024).

Microsoft’s cloud revenues rule again in Q4, as Surface continues to dip

Office and cloud are strong for Microsoft’s revenues.

www.theverge.com

Of course, this is the whole division's sales. Maybe without X Elite / X Plus, Microsoft's sales would've been even worse. Perhaps when the next U.S. IDC report comes in, we'll see how much Microsoft's laptops / desktops have sold, if they can break into the top 5 (which they rarely do). Tablets perhaps, then.

Saw in my feed today that, according to Mercury Research, ARM lost market share in 2Q24. Specifically for notebooks,

Arm lost 144 basis points in the notebook market, to account for 11.4% of the market in the period, down from 12.8% in the first quarter.

These are probably before X-Elite had time to get a lot of volume into the market, but if MS is predicting further Surface sales in the next quarter, that's probably not a good sign for QC Windows laptops in general.

AMD, Intel in focus as data shows gains over Arm

AMD (AMD) and Intel (INTC) were in focus on Wednesday after Citi maintained its respective ratings on the semiconductor companies. Read for more.

seekingalpha.com