Discussion Qualcomm Snapdragon Thread

[Magio]

If they have a winner, they are in pretty good position. Competitors have nothing launching in the near future that offer anything meaningful over what they currently have. Well, Intel might.

Panther Lake will probably be very low volume for a while even for the first half of 2026 and it's unlikely to be competitive with the X Elite 2 on either CPU performance or efficiency. With third gen Xe cores it's going to be interesting to see how it does in gaming (which should remain QC's main weakness) but that's pretty much it.

For the premium thin and light market X Elite 2 has a serious shot at being *the* best chip on the market until at least 2027 when Nova Lake and Medusa Point release. If they can improve Windows on ARM (and for the love of god please improve Linux support) it might really be a winner.

 

[Covfefe]

Has Prism gotten better lately? They added AVX2 support in the testing channel nearly a year ago, and as far as I can tell, it still hasn't hit mainline Windows. I'm having trouble finding any publications that tests Windows on ARM. The best I can find is random forum posts.

I really hope that Prism gets a big update with the launch of the X Elite 2. If software compatibility is still poor it will severely impact the perception of the chip. I don't think Qualcomm will see widespread success in the PC market until Windows on ARM "just works".

 

[Panino Manino]

https://twitter.com/x/status/1967997057634844760

X Elite 2

New Era hadn't started with Elite 1?

 

[CouncilorIrissa]

iPhone18,2 vs Xiaomi 25113PN0EC - Geekbench

Parity with the iPhone in clang. Mighty impressive.

 

[DZero]

Apple is done... ST wise got caugh up and MT wise got defeated. In the same month...

 

[LightningZ71]

Not to stan for either side of the issue, but I'm mighty curious to see how the power metrics compare between those two devices... I suspect that Apple might be able to edge out the Perf/mw metrics on this, but, being within 5% for ST is a reasonable wash in most people's books and not something that you'll notice on either phone. The MT numbers are impressive, but, for phone use, that's likely a less useful thing in most cases.

 

[DZero]

Not to stan for either side of the issue, but I'm mighty curious to see how the power metrics compare between those two devices... I suspect that Apple might be able to edge out the Perf/mw metrics on this, but, being within 5% for ST is a reasonable wash in most people's books and not something that you'll notice on either phone. The MT numbers are impressive, but, for phone use, that's likely a less useful thing in most cases.

Unless you use Dex mode and a cooler to maintain temps.

Then the only real use? emulators.

 

[CouncilorIrissa]

Apple is done... ST wise got caugh up and MT wise got defeated. In the same month...

nT isn't really interesting or relevant. It's mostly a function of how many littles you shove onto the die. And this years' littles look like a big improvement for Apple at a fraction of Oryon-M's power. Even last years' ones are quite a bit more efficient.



1t score is far more interesting as far as I'm concerned.

 

[mvprod123]

https://twitter.com/x/status/1967997057634844760

X Elite 2

Another new era for PCs with garbage GPU drivers?

Apple is done... ST wise got caugh up and MT wise got defeated. In the same month...

Done in what? Qualcomm is doing its utmost to increase frequency and is unable to catch up with Apple in IPC.

 

[gdansk]

Done in what? Qualcomm is doing its utmost to increase frequency and is unable to catch up with Apple in IPC.

I'm afraid you're new to DZero posting, it's twitter core overreaction.
But I think Apple no longer has a commanding lead over Qualcomm. Performance per watt remains to be seen but many phone vendors have, to some degree, rendered that moot for end users by installing more expensive battery technology than Apple.

 

[DZero]

I'm afraid you're new to DZero posting, it's twitter core overreaction.
But I think Apple no longer has a commanding lead over Qualcomm. Performance per watt remains to be seen but many phone vendors have, to some degree, rendered that moot for end users by installing more expensive battery technology than Apple.

This.
The issue is that Apple had advantage for some period of time, but this year with SD 8 Elite Gen 5 (what a long name) this could end.
Even more if we consider that Qualcomm is now in the laptop territory and can optimize their drivers even more.

 

[Tup3x]

I would say that the CPU part is not going disappoint but the GPU...? We'll see.

 

[gdansk]

I would say that the CPU part is very unlikely to disappoint but the GPU...

The GPU will disappoint for laptops or for phones?

 

[Magio]

Done in what? Qualcomm is doing its utmost to increase frequency and is unable to catch up with Apple in IPC.

QC is absolutely catching up on IPC, because Apple's IPC has been stagnant for years now. Since 2020, the only time their P core IPC has risen by more than 5% YoY was last year with the SME bump. Their gen on gen improvements are very much clock dependent too.

Meanwhile based on early numbers 8E2 is a 10+% IPC bump without SME, close to 20% counting SME. It hasn't matched Apple on IPC yet, but by definition it's catching up.

And of course, Oryon was already industry leading on Perf per Area last year.

Apple isn't "done" because their chips are still great but it's undeniable that for the first time in a very long time they have serious competition. PPW wise Apple likely still has an edge in ST (last year 8E1 needed 1W more in ST to reach its peak score) and at the very low end of the power scale in MT too (until Oryon-S?) but those leads are getting slimmer and Oryon is improving fast.

 

[mvprod123]

I'm afraid you're new to DZero posting, it's twitter core overreaction.
But I think Apple no longer has a commanding lead over Qualcomm. Performance per watt remains to be seen but many phone vendors have, to some degree, rendered that moot for end users by installing more expensive battery technology than Apple.

A larger battery capacity will not always give you an advantage. Optimisation is important.

 

[poke01]

View attachment 130409

A larger battery capacity will not always give you an advantage. Optimisation is important.

That’s the physical sim model too.

 

[poke01]

Samsung is gonna get destroyed in the battery tests in the US.

 

[mvprod123]

QC is absolutely catching up on IPC, because Apple's IPC has been stagnant for years now. Since 2020, the only time their P core IPC has risen by more than 5% YoY was last year with the SME bump. Their gen on gen improvements are very much clock dependent too.

Meanwhile based on early numbers 8E2 is a 10+% IPC bump without SME, close to 20% counting SME. It hasn't matched Apple on IPC yet, but by definition it's catching up.

And of course, Oryon was already industry leading on Perf per Area last year.

Apple isn't "done" because their chips are still great but it's undeniable that for the first time in a very long time they have serious competition. PPW wise Apple likely still has an edge in ST (last year 8E1 needed 1W more in ST to reach its peak score) and at the very low end of the power scale in MT too (until Oryon-S?) but those leads are getting slimmer and Oryon is improving fast.

Qualcomm is stuck at the A17 level in terms of IPC. Apple achieved a 7% increase in the previous generation and a 6% increase in this generation. It is too early to talk about stagnation.

 

[gdansk]

A larger battery capacity will not always give you an advantage. Optimisation is important.

I didn't say an advantage, I said consumers will see no difference in practice. Looking at that graph I see no difference. Some QC phones ahead, some behind.

 

[DZero]

Qualcomm is stuck at the A17 level in terms of IPC. Apple achieved a 7% increase in the previous generation and a 6% increase in this generation. It is too early to talk about stagnation.

less than 10% is not a good sign. Also, Apple AND Qualcomm are reaching their max on frequencies. Going over 5 Ghz in phone chips is a no go, and they saw Crashtor Lake dissaster (one factor of the issue is the Ghz race), so no.

Still, Apple has some cards like E cores (now going 2+6 could retain the MT leadership big time) and adding more cache.

 

[mvprod123]

less than 10% is not a good sign. Also, Apple AND Qualcomm are reaching their max on frequencies. Going over 5 Ghz in phone chips is a no go, and they saw Crashtor Lake dissaster (one factor of the issue is the Ghz race), so no.

Still, Apple has some cards like E cores (now going 2+6 could retain the MT leadership big time) and adding more cache.

There are no miracles, and overcoming physical limitations is difficult. It will be interesting to see what 2nm will bring. This year, Apple has allocated significant space and resources to MTE, as mentioned in their recently published article.

 

[gdansk]

Going over 5 Ghz in phone chips is a no go

For now. A few more years - maybe even a decade if cooler minds prevail - but it'll happen.

 

[DZero]

For now. A few more years - maybe even a decade if cooler minds prevail - but it'll happen.

Intel and AMD attempted, it does not work, at least with the current materials.

 

[Hesperax]

less than 10% is not a good sign. Also, Apple AND Qualcomm are reaching their max on frequencies. Going over 5 Ghz in phone chips is a no go, and they saw Crashtor Lake dissaster (one factor of the issue is the Ghz race), so no.

Still, Apple has some cards like E cores (now going 2+6 could retain the MT leadership big time) and adding more cache.

Rumor is that this could be for next year (Hawi).

 

[Magio]

Qualcomm is stuck at the A17 level in terms of IPC. Apple achieved a 7% increase in the previous generation and a 6% increase in this generation. It is too early to talk about stagnation.

When those are their best IPC increases since 2020, yes it's time to talk stagnation.

Also, Qualcomm doesn't need to match IPC exactly. Their cores are small, built for clocks and they're already very efficient (just a smidge less than Apple's last year, probably the same thing this year). Closing the IPC gap some is good (and they empirically seem to be doing that) but for example ARM's stock big cores have better IPC and they're less efficient and less performant. IPC isn't the be all and end all.