Discussion Qualcomm Snapdragon Thread

[SarahKerrigan]

Qualcomm is in deep trouble.

This is a very strange take.

If Qualcomm is in "deep trouble" based on this, I shudder to think what the implications are for Intel and AMD.

 

[Thibsie]

View attachment 92381
View attachment 92385
View attachment 92383

Also the power efficiency of the X Elite iGPU looks bad in comparison to Apple.

More power than M2 Pro while delivering less performance?

And it's not like the GPU has been pushed to oblivion. The power curve (as presented in the Qualcomm slide), looks very healthy.

Last picture was sourced from here: https://www.notebookcheck.net/Apple...re-efficient-the-CPU-not-always.699140.0.html

So it's only their P-core which is competitive with Apple. Their E-core isn't (going by the rumour), and neither is their GPU as per the information presented here.

Qualcomm is in deep trouble.

Lol, it's their first gen. Cut them some slack...

 

[hemedans]

View attachment 92381
View attachment 92385
View attachment 92383

Also the power efficiency of the X Elite iGPU looks bad in comparison to Apple.

More power than M2 Pro while delivering less performance?

And it's not like the GPU has been pushed to oblivion. The power curve (as presented in the Qualcomm slide), looks very healthy.

Last picture was sourced from here: https://www.notebookcheck.net/Apple...re-efficient-the-CPU-not-always.699140.0.html

So it's only their P-core which is competitive with Apple. Their E-core isn't (going by the rumour), and neither is their GPU as per the information presented here.

Qualcomm is in deep trouble.

Probably that's overclocked beyond efficient curve, same GPU inside sd 8 gen 2 could match Radeon 680M while using 8W. So it's going to be fine vs current igp.

 

[Doug S]

You went from Qualcomm will take over the world to that conclusion. What about waiting for real products availability and reviews before being so affirmative?

No doubt. People keep wanting to look at Qualcomm's deliberately vague graphs and compare them with Apple Silicon tests using all sorts of benchmarks different than what Qualcomm used to try to tease out a comparison. When they are released there will be plenty of benchmark results, and they can be compared head to head.

And when it sees the light of day M4 is around the corner (and A18 GB6 rumors will already have hit Twitter) so the next round of speculative comparisons will begin.

 

[FlameTail]

3DMarkWildlifeExtreme

8G2: 22 FPS
8G3: 32 FPS
X Elite: 44 FPS

What I am unable to fathom is... why is the X Elite GPU consuming so much power compared to it's mobile cousins (8G2,8G3)?

Rumours were saying X Elite iGPU is based on 8G2, supposedly doubling the GPU of 8G2. (22 FPS ×2 = 44 FPS).

But 8G2 GPU only consumes 7.5W (as tested by GoldenReviewer. Similar figure from Geeekerwan. So doubling the 8G2 GPU should mean a power consumption of around 15W. That would make for a very efficient iGPU that's even better than the base M chip.

Instead we have the X Elite consuming 30W+ in the same benchmark, as per the official Qualcomm slide.

So in that case, perhaps they didn't actually double the transistors of 8G2 GPU (as the rumours claimed), but instead doubled the clock speed? Perhaps that's how they were able to fit this SoC into just 172 mm².

 

[FlameTail]

Lenovo is charging a whopping $300 for 5G in this Snapdragon Thinkpad. I wonder how much of that cut does Qualcomm take home?

 

[Nothingness]

There are phones with 5G that cost $100. I guess Lenovo is overcharging a lot like a certain fruity company.

EDIT: I was thinking in terms of licensing costs not chips. It's still possible QCOM modem is very expensive but certainly not that much.

EDIT2: Apple is charging $150 or $200 for 5G in the iPad and as far as I know they are using QCOM modem.

 

[FlameTail]

There are phones with 5G that cost $100. I guess Lenovo is overcharging a lot like a certain fruity company.

EDIT: I was thinking in terms of licensing costs not chips. It's still possible QCOM modem is very expensive but certainly not that much.

EDIT2: Apple is charging $150 or $200 for 5G in the iPad and as far as I know they are using QCOM modem.

I recall some insider reports said Qualcomm apparently charges Apple $80 per modem.

 

[Nothingness]

You're right, it's estimated at $90: https://wccftech.com/iphone-12-pro-costs-406-to-make/

 

[eek2121]

People seem to forget that Qualcomm historically has overhyped and under delivered. That is why I continue to be skeptical.

Also, as long as Google, the ad company, is the main Android developer I will never switch back on my phone.

For PC I am a gamer and it will take at least a decade for ARM to get real traction there.

I recall some insider reports said Qualcomm apparently charges Apple $80 per modem.

That is Apple, however. Qualcomm makes it up with volume. I suspect they charge significantly more elsewhere. I mean, it is Qualcomm after all.

 

[Nothingness]

It's quite well known that Intel and AMD never overhyped and never under delivered.

 

[SarahKerrigan]

It's quite well known that Intel and AMD never overhyped and never under delivered.

As someone who used to be part of the Itanium ecosystem, I can attest to Intel never, ever overhyping and underdelivering. They had a well-known track record of shipping every generation on time and at the promised clocks.

jk

 

[Doug S]

As someone who used to be part of the Itanium ecosystem, I can attest to Intel never, ever overhyping and underdelivering. They had a well-known track record of shipping every generation on time and at the promised clocks.

jk

I remember year after year Itanic was discussed on RWT; there were a few rah rah Itanium/Intel types (one especially I remember) who were always hyping up that the "next one" was going to be the one that blows the doors of the competition, but between delays and broken promises it was always a laugh track for the rest of us.

Obviously HP couldn't have continued fabbing their own stuff like they did until the latter days of PA-RISC (I think Intel took over fabbing with the PA-8500?) and workstation/server class RISC was doomed in the long run by x86 economies of scale, but I always wonder if HP would have been better off merely extending/improving PA-RISC 2.0 rather than going after PA-RISC 3.0 / PA-WW that became Itanium. They would have had to find someone else to fab it, maybe they could have partnered with AMD who still had their own fabs and later spun them off as Global Foundries.

The Alpha people got screwed worst of all by the Itanium transition, it would have been interesting to see what the 21464 could have done. 8 way superscalar 20 years ago...

 

[SarahKerrigan]

I remember year after year Itanic was discussed on RWT; there were a few rah rah Itanium/Intel types (one especially I remember) who were always hyping up that the "next one" was going to be the one that blows the doors of the competition, but between delays and broken promises it was always a laugh track for the rest of us.

Obviously HP couldn't have continued fabbing their own stuff like they did until the latter days of PA-RISC (I think Intel took over fabbing with the PA-8500?) and workstation/server class RISC was doomed in the long run by x86 economies of scale, but I always wonder if HP would have been better off merely extending/improving PA-RISC 2.0 rather than going after PA-RISC 3.0 / PA-WW that became Itanium. They would have had to find someone else to fab it, maybe they could have partnered with AMD who still had their own fabs and later spun them off as Global Foundries.

The Alpha people got screwed worst of all by the Itanium transition, it would have been interesting to see what the 21464 could have done. 8 way superscalar 20 years ago...

Yes, Paul's interesting opinions on IPF were always a lot. Poulson was supposed to be a Power7-killing superchip, when in reality it wasn't even close; I never saw large and consistent single-thread perf improvements over Tukwila. There's a lot of "what if" points where IPF could have gone differently - the promised Tukwila in 2008 at 2GHz looks a lot different, competitively, than what we got (Tukwila in 2010 at 1.7GHz.) This was a recurring theme for Itanium's entire history - look at how badly Montecito got nerfed from the initial claims.

The other big "what if" is Tanglewood, but that's a whole different story. I don't know that it would have changed anything long-term if it had shipped, but its cancellation was a bellwether - it marked the clear triumph of "Itanium as HP's RISC/UNIX offering" over "Itanium as 64-bit commodity." Bayshore's cancellation around the same time sealed the deal.

I think from HP's perspective, Itanium did just fine. They got to offload some percentage of the design costs for their high-end UNIX franchise onto Intel (and, indirectly, Bull, SGI, Fujitsu, etc.) I don't think staying on the PA2 horse would have really changed how things ultimately went, and I think by 1998 (if not earlier) HP would have struggled to meaningfully change the direction of how things went - they had already sold ISVs on the Itanium transition, after all. For most users, PA->IPF wasn't a painful migration, and some got to take advantage of things like Windows Server consolidation onto their Integrity infrastructure.

As for Alpha - their userbase (Tru64 + VMS) was less than half of the size of HP's (UX + MPE) and so the economics of extending Alpha vs building a PA successor were not favorable. It is not a coincidence that the three companies that stayed in the RISC/UNIX market the longest - Sun, HP, and IBM - were the top three revenue leaders (and Compaq was a very distant fourth.) 21464 was, of course, going to be all things to all people, just like all paper CPUs; I have skepticism about whether it would have been quite as spectacular as the Alpha bitter-enders of the world would like to believe.

Some Alpha users ended up feeling burned by HP's handling of Tru64, even aside from the CPU question - a lot of Tru64 features (Trucluster and AdvFS) were supposed to be merged into UX, and that never happened. The VMS and Nonstop folks got a decent deal out of all of it, at least. It helps that Nonstop is structurally relatively painless to migrate between platforms, and ultimately got an x86 port too - and due to the inevitable MIPS roadmap, it would have faced a platform transition (almost certainly to Alpha) even if Itanium had never been released.

Edited to add: Late PA was fabbed by IBM (which had a long history of doing foundry work for competitors - Unisys, among others.)

 

[moinmoin]

The Alpha people got screwed worst of all by the Itanium transition, it would have been interesting to see what the 21464 could have done. 8 way superscalar 20 years ago...

Absolutely this. Such promising tech, cut short by company politics.

 

[FlameTail]

New Bombshell leak about Snapdragon 8 Gen 4


Oryon Phoenix-L is reaching 4 GHz

 

[FlameTail]

Qualcomm should ensure OEMs use good quality thermal paste and proper fan control 👍🏻

Apple's M chips alone aren't the reason why Macbooks have such excellent thermals. The cooling system is also top notch.

 

[Nothingness]

I remember year after year Itanic was discussed on RWT; there were a few rah rah Itanium/Intel types (one especially I remember) who were always hyping up that the "next one" was going to be the one that blows the doors of the competition, but between delays and broken promises it was always a laugh track for the rest of us.

Let me guess: his nickname was "someone".

EDIT: Ha @SarahKerrigan reminded me Paul de Mone was even worse than "someone".

 

[SarahKerrigan]

Let me guess: his nickname was "someone".

EDIT: Ha @SarahKerrigan reminded me Paul de Mone was even worse than "someone".

AFAIK they are, in fact, the same person.

 

[Nothingness]

AFAIK they are, in fact, the same person.

So no worse, no better

 

[coercitiv]

Apple's M chips alone aren't the reason why Macbooks have such excellent thermals. The cooling system is also top notch.

The same can be said in reverse. One Macbook Air my wife received at work, close to the last Intel iterations, had the most pathetic cooling I have seen in a premium laptop. Since then she moved on to M models, but this implementation stuck in my mind as an example of disregard for quality and basic engineering.

Here's what Notebookcheck had to say about it:

System Noise – The fan of the MBA is annoying

Probably the biggest issue of the MacBook Air is the cooling. The processor itself is cooled passively. There is a fan inside the chassis, but it is not directly connected to the heat sink of the CPU. This fan is only supposed to improve the air circulation inside the case and help dissipate warm air.

To be fair, the fan is often deactivated when you perform simple tasks. Once you start to stress the CPU a bit more, however, which can be as simple as running multiple tabs in Safari, the fan will slowly and seamlessly increase its speed. More intense load scenarios (such as the installation of applications) or higher ambient temperatures, the noise level quickly reaches up to 41 dB(A), just like the previous model. Furthermore, the frequency starts getting annoying from 37-38 dB(A). Considering the very low performance, this is not really acceptable.

 

[Doug S]

AFAIK they are, in fact, the same person.

Yeah he had pretty much tarnished his name in that forum with his over the top love (trying hard to avoid the word that's banned in these forums) of Itanium so he left for a bit then came back later under a new name. His posting style was unmistakable, however, so it quickly became clear who "someone" was. I haven't seen him there for some time. Too bad, other than his fixation on Itanium he was a good and knowledgeable poster.

There's another (who was/is over the top about Apple) who changed from posting under his real name to something else, but his posting style was unchanged so it was immediately clear who it was. He also posts on Anandtech occasionally under a different nickname, out of respect I won't name him but since you're clearly familiar with RWT I imagine you and Nothingness know exactly who I'm referring to!

 

[ashFTW]

Yeah he had pretty much tarnished his name in that forum with his over the top love (trying hard to avoid the word that's banned in these forums) of Itanium so he left for a bit then came back later under a new name. His posting style was unmistakable, however, so it quickly became clear who "someone" was. I haven't seen him there for some time. Too bad, other than his fixation on Itanium he was a good and knowledgeable poster.

There's another (who was/is over the top about Apple) who changed from posting under his real name to something else, but his posting style was unchanged so it was immediately clear who it was. He also posts on Anandtech occasionally under a different nickname, out of respect I won't name him but since you're clearly familiar with RWT I imagine you and Nothingness know exactly who I'm referring to!

—- indeed

 

[FlameTail]

I had no idea the Macbook Air had a fan inside.

TIL.

 

[gdansk]

I had no idea the Macbook Air had a fan inside.

TIL.

The Intel versions yes. And you just know Apple was mad about it, the fan design was clearly an afterthought. Intel has probably said the Y series could be fanless but...
M1/M2 are fanless.