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

[Marco_il_bello]

and IF the test is a fake ?

 

[dalek1234]

Just a friendly reminder that Charlie of semiaccurate infamy has been complaining about Qualcomm questionable business and marketing tactics for the last year. This initial benchmark disparity is validating his incessant criticism

Qualcomm Is Cheating On Their Snapdragon X Elite/Pro Benchmarks

Qualcomm is cheating on the Snapdragon X Plus/Elite benchmarks given to OEMs and the press.

www.semiaccurate.com

Quite an interesting article you linked. For those who don't feel like reading, here a juicy quote:

"A deep source at Qualcomm told us that the benchmarks were cheats, told us how they were cooked, and told us that Qualcomm was well aware of it."

 

[-Fran-]

The article says the chip didn't boost from 2.52GHz to 4.0GHz during the test like it should. This should be easily fixed with an update.

Edit: Just to add, compare the single core scores. (1841 vs what it should've got, 2977). That's approximately a 60% gap, which just so happens to be the same gap between 2.52GHz to 4.0GHz. This is absolutely a boost issue. AFAIK, Qualcomm's mobile chips do not use turbo boosting, so this is new for them. If the chips are ruined and can't turbo boost, then we can pull the pitchforks out, but I am still expecting a patch.

The point is simple: the numbers they're quoting in all press-releases are using some form of firmware/driver/BIOS revision which, I have to honestly ask: why are production samples not using a refined version of it?

Remember RDNA3's launch? Why are the weird nerds making excuses to defend Qualcomm here?

Come on; enough with the 'beaten wife' mentality.

Regards.

 

[bit_user]

"A deep source at Qualcomm told us that the benchmarks were cheats, told us how they were cooked, and told us that Qualcomm was well aware of it."

Qualcomm must know they will face the scrutiny of independent reviewers. So, perhaps they only rigged the benchmarks to predict what they estimated performance of the final, released product should be like? Otherwise, they're ultimately just shooting themselves in the foot - and I'd expect they wouldn't be quite so reckless.

Neither Charlie nor Qualcomm is being at all transparent, here. Some might like to believe the whistleblower, but I consider Charlie to be deficient in journalistic integrity and don't afford him that privilege.

Fortunately, the embargo will be over quite soon, so I say folks should just sit tight and wait.

 

[CmdrShepard]

It's probably closer than you think. Here's what I discovered about the Golden Cove and Gracemont cores, in Alder Lake:

You can't just look at that without also considering power consumption and thermal limitations.

 

[carlthomas]

I had a lot of respect for Tom's Hardware prior to this article. The fact that you guys have a literal link in the article that goes to your source where there's a flurry of benchmarks at a variety of ranges, with MANY at the promised single and multi core scores among the many pages of ones dropped today, shows you guys really are just after clicks and not accurate reporting.

 

[bit_user]

You can't just look at that without also considering power consumption and thermal limitations.

For the purposes that this point was raised, we can.

 

[Notton]

THG try not to succumb to yellow journalism challenge, difficulty: impossible

You are aware that this garbage coverage only lowers THG's reputation as a reliable source of information, right?

 

[kealii123]

"Review embargo hasn't lifted yet" Linus said last week that the embargo was lifting this week in his Qualcomm sponsored video

 

[JamesJones44]

Sorry, I'm not interested in your excuses for them. If you go back and look at what Qualcomm has said about their timeline for shipping products, these Nuvia/Oryon cores have slipped a couple years, at least!

In terms of performance and efficiency, the one place I'll cut them a little slack vs. Apple is that they're on an older node. These new Snapdragons are on TSMC N4 (not sure which), while Apple's latest iPad (not sure about their phones) is on N3B. Still, that means they have no excuse not to beat at least Apple's M2, which was made on TSMC N5, I'm pretty sure.

Who's making excuses? I'm simply stating facts based on a persons comment that "It's not like Qualcomm is a start up and has no past experience with building things". I even go on to say Qualcomm may be rushing them to get it out the door.

 

[bit_user]

I even go on to say Qualcomm may be rushing them to get it out the door.

And that's not an excuse, in your book? Like I said, Qualcomm has given them years of extra time. If this thing is being rushed out the door, I think it's not primarily Qualcomm's fault!

As for Nuvia, they've been working away on it for over 5 years, already!

“NUVIA was founded in early 2019 ..."

https://www.anandtech.com/show/15115/nuvia-breaks-cover-new-startup-to-take-on-datacenter-cpu-market
​

 

[JamesJones44]

And that's not an excuse, in your book? Like I said, Qualcomm has given them years of extra time. If this thing is being rushed out the door, I think it's not primarily Qualcomm's fault!

As for Nuvia, they've been working away on it for over 5 years, already!

​

“NUVIA was founded in early 2019 ..."​

​

https://www.anandtech.com/show/15115/nuvia-breaks-cover-new-startup-to-take-on-datacenter-cpu-market​

​

PA Semi, the company Apple bought to make the A series processors took years to develop their first CPU and even more years to develop the first M series CPU. PA Semi was founded in 2003, Apple bought then in 2008. PA didn't release their first CPU used by Apple until 2010! That's 7 years from founding to release. If you believe some of the writes up that claim it took Apple 10 years to create the first M1 (I'm not sure I believe all of that was specific to creating the M1 itself, but so be it). That would be 17 years of development. If you use that as a yard stick, it seems like there is potential for things to be rushed in a 5 year span.

I don't know what you think I'm saying, but it seems personal thoughts are being injected into my comment. There is no "excuse" here, I personally don't give a crap about Qualcomm, just stating why things MAY be in this state, assuming any of what was stated in this article is even ends up being true.

It seems you may have some personal axe to grind here, but respectfully, this is axe not mine.

 

[CmdrShepard]

For the purposes that this point was raised, we can.

No, no you can't, especially not for the purposes of this point.

This CPU is meant for notebooks and tablets where both power consumption and thermals are restricted and probably won't be able to sustain the performance which was demoed originally for any significant period of time.

 

[LordVile]

No, no you can't, especially not for the purposes of this point.

This CPU is meant for notebooks and tablets where both power consumption and thermals are restricted and probably won't be able to sustain the performance which was demoed originally for any significant period of time.

The full fat Elite is an 80W chip, they’re going to have to significantly scale it down for thin and lights which will put it easy below the M3.

 

[bit_user]

PA Semi, the company Apple bought to make the A series processors took years to develop their first CPU and even more years to develop the first M series CPU. PA Semi was founded in 2003, Apple bought then in 2008. PA didn't release their first CPU used by Apple until 2010! That's 7 years from founding to release.

Ha! That's a bad example, because PA Semi first made a CPU based on the POWER architecture, which launched in Q4 of 2007!

So, when PA Semi got acquired by Apple and had to make an ARM-based phone SoC, they were essentially starting over from scratch. In the case of Nuvia and Qualcomm, there was no such reset.

Anyway, unless your a Qualcomm shareholder, I don't see why you'd be peddling excuses for Qualcomm. Let the data speak for itself. And we all now know the caveat that they're using N4.

If you believe some of the writes up that claim it took Apple 10 years to create the first M1

Yeah, but Nuvia isn't starting from scratch, because they know exactly what it takes to build something like that, having just done it at Apple!

It seems you may have some personal axe to grind here, but respectfully, this is axe not mine.

Not me. I just saw you making excuses for Qualcomm that I thought were unfounded (Qualcomm pushing it out the door). No. 5 years cannot be considered pushing it out the door, when they were initially talking about 3-4. If that initial timeframe makes anyone look bad, it's Nuvia. Qualcomm would've wanted to know the production timeframe, when they did the acquisition, and shame on Nuvia if they set an unrealistic expectation they couldn't meet.

 

[bit_user]

Oh Snapdragon! The Deep Dives already went live!

https://www.anandtech.com/show/21445/qualcomm-snapdragon-x-architecture-deep-dive
​

 

[CmdrShepard]

Oh Snapdragon! The Deep Dives already went live!

​

https://www.anandtech.com/show/21445/qualcomm-snapdragon-x-architecture-deep-dive​

​

How are they allowed to trademark common word like "hexagon"? That's insane.

Also, more slides and promises? Yawn.

 

[JamesJones44]

Not me.

Yes, you.

You are trying to claim I'm saying something I'm not while injecting assumptions to support the claim.

1. The complaint and benchmark is real and not some one-off or troll
2. This isn't a Samsung specific issue. When you look at Geekbench submissions many of the benchmarks look to be at or near what Qualcomm presented https://browser.geekbench.com/search?q=Snapdragon+x+elite. Some of the Samsung submissions do look back, but most look like what Qualcomm presented. Most of the Asus models look less variable.
3. Nuvia set the expectations (you can't possible know that)
4. A product is not rushed simply because of time past. Any timeline can be considered rushing something out the out the door if a product isn't ready. I doesn't matter if it's 5 months, 5 years or 5 decades. Boeing and the 737 Max shows what happens when you rush a product even if it's taken a very long time to develop.
5. Assuming PA wasn't already working on an ARM product when Apple bought them (you can't know that)

Finally there is this...

Anyway, unless your a Qualcomm shareholder, I don't see why you'd be peddling excuses for Qualcomm

I'm not really sure how this comment makes logical sense. Speculating that Qualcomm rushed something out the door is the same as calling Qualcomm dumb for rushing a product and destroying confidence in it, thus hurting Qualcomm. An excuse by definition is trying to excuse them from the issue, this is not that, I'm calling them dumb if the did rush it out the door before it was ready.

None of those assumptions and word twists will change what I said. Pointing out that the product came from a startup and wasn't developed directly by Qualcomm and speculating Qualcomm might, if this is an Elite X problem, be pushing something that is not ready (aka calling them dumb for doing that).

 

[bit_user]

Yes, you.

Nope. Don't project or deflect... and have some self respect.

You are trying to claim I'm saying something I'm not

Okay, well I said how it came across, but if you'd like to clarify what you were saying or what you meant by it, you certainly have that opportunity.

while injecting assumptions to support the claim.

1. The complaint and benchmark is real and not some one-off or troll

You lost me. Is this something you're claiming that I've said? I never said diddly squat about the benchmark result.

This isn't a Samsung specific issue. When you look at Geekbench submissions many of the benchmarks look to be at or near what Qualcomm presented https://browser.geekbench.com/search?q=Snapdragon+x+elite. Some of the Samsung submissions do look back, but most look like what Qualcomm presented. Most of the Asus models look less variable.

Seems like you're either confused about who your arguing with or maybe you're just using me as a convenient foil for making these points.

Nuvia set the expectations (you can't possible know that)

I'm certain that when Qualcomm approached Nuvia for acquisition talks, a keen subject of interest would be the schedule for getting their cores into Qualcomm SoCs. You don't spend Billions of dollars without having a clear idea of when you're going to see a return on that investment, especially when you have activist shareholders like Qualcomm does. That means Qualcomm execs need to be answerable for their decisions and that hinges on the RoI story for the acquisition.

Qualcomm really had no incentive to overpromise, regarding the delivery date. Their schedule slippage just makes them look bad and there was no real upside to it.

A product is not rushed simply because of time past.

That's not a point I made. I said they allowed substantial schedule slippage. If they were desperate to ship something, then I'm sure it would've happened sooner. However, I trust Qualcomm is smart enough to know that it does worse damage to their business if they ship a product that underperforms because it wasn't ready, than if they wait until it actually is ready.

BTW, one clear sign of its lateness would be the node they fabbed it on. If it were targeted for release when it's actually launching, they'd have used the N3B node. Qualcomm is almost as aggressive as Apple in adopting new nodes. Apple is on N3B and so is Intel's Lunar Lake. AMD adopted N4 last year, so that gives us a pretty clear sign of when they expected it to ship, at the time they committed to a node.

Boeing and the 737 Max shows what happens when you rush a product

That's a terrible analogy, since its problems weren't simply a matter of being rushed. They paired the wrong engines with the airframe, then tried to paper over it by hacking the avionics to mask how they messed up the balance. It was fundamentally the wrong product.

Assuming PA wasn't already working on an ARM product when Apple bought them (you can't know that)

Read it and weep:

"You build in basically a roadmap, so I was thinking about what we were going to do for five years, chip after chip. We did this at Apple too when we built the first big core at Apple - we built big bones [into the design]."

"I was then at P.A Semi, and we delivered a great product, but they didn't really want to sell the product for some reason, or they thought they were going to sell it to Apple. I actually went to Apple, and then Apple bought P.A Semi, and then I worked for that team,"

(remember: this was still during Apple's POWER era, so it makes sense that P.A. Semi thought Apple would be interested in their PWRficient product)

Source: https://www.anandtech.com/show/16762/an-anandtech-interview-with-jim-keller-laziest-person-at-tesla

"On 11 June 2008, during the annual Worldwide Developer's Conference, Apple CEO Steve Jobs said that the acquisition was meant to add the talent of P. A. Semi's engineers to Apple's workforce and help them build custom chips for the iPod, iPhone, and other future mobile devices such as the iPad."

Source: https://en.wikipedia.org/wiki/P.A._Semi
​

So, Jim went to Apple, then started building the A4. After that, Apple bought P.A. Semi for its engineering team. At Apple, Jim didn't just create a core, but they made the investment in creating an architecture with "big bones" that would have longevity and support lots of enhancements to follow. That's not the sort of ambitious, long-range project a startup embarks on, when the cash is starting to dry up.

So, it's right there: they did the design at Apple.

Speculating that Qualcomm rushed something out the door is the same as calling Qualcomm dumb for rushing a product and destroying confidence in it, thus hurting Qualcomm. An excuse by definition is trying to excuse them from the issue, this is not that, I'm calling them dumb if the did rush it out the door before it was ready.

It's all a matter of perspective. If someone falls short of an objective, they could protect their ego by saying they weren't really trying. In a similar way, claiming Snapdragon X might underperform against Apple could be trying to preserve hope that they're still capable of achieving it in a future generation, since the reason they (hypothetically) didn't in this generation is due to the fact that they weren't able to put forth their full effort.

 

[caponica23]

HI there guys, i'm the reddit user.
I was the first to be suprised about this results but tought was a drivers issue. Yesterday I received a driver update trough Windows Update (because Samsung Update doesn't work) and the resullts are just the same.
For all of you who think the test is carried out in power saving mode, here you have the power saving results:

 

[bit_user]

HI there guys, i'm the reddit user.

Hello, thanks for stopping by & sharing the info!
: )

 

[cyrusfox]

How are they allowed to trademark common word like "hexagon"? That's insane.

Happens all the time especially in computing (Apple ). Its about branding, recommend you give this a view:

View: https://www.youtube.com/watch?v=p5EOwcM0v9c&t=83s

Anyone can trademark basic common words because trademarking is not about unique words, unique phrases, unique designs, trademarks are about unique identifiers for products and services and there's a huge difference there...

 

[SyCoREAPER]

I thought these were out, saw a Surface with one at my Best Buy yesterday. Guess they made an oopsy.

I find it hard to believe that this result is legit. They wouldn't hype it to the extent they have to be this bad. Windows wouldn't even boot on a ARM chip that weak.

I'll take this with an oceans worth of salt and wait for real world reviews.

 

[JamesJones44]

Okay, well I said how it came across, but if you'd like to clarify what you were saying or what you meant by it, you certainly have that opportunity.

I've clarified 3 times. I also can't understand how in your mind calling a company dump for rushing a product constitutes an excuse. IDK if there is some comprehension or some other issue we are dealing with here, but, when I explained it you, you simply doubled and tripled down on your redirect that calling Qualcomm dumb is an excuse. Though IMO your the one does the excusing your own comments with this continued nonsense.

I'm certain that when Qualcomm approached Nuvia for acquisition talks, a keen subject of interest would be the schedule for getting their cores into Qualcomm SoCs. You don't spend Billions of dollars without having a clear idea of when you're going to see a return on that investment, especially when you have activist shareholders like Qualcomm does. That means Qualcomm execs need to be answerable for their decisions and that hinges on the RoI story for the acquisition.

Yet another assumption about what you think. I don't know why you can't see the pattern, but it's pretty clear. What you think, doesn't make something fact. The truth is we have no idea what happened and so there is nothing that can be proved by this statement. It's an assumption with zero weight.

Since I can predict your behavior in back peddling, I'll say this. I'm sure there was some conversation around timelines, but since we can't be sure what was said, there is no value in trying to guess and use it as an argument. Nuvia could have said 1 year or 10 years to ship, there is no way to know, so assuming what was said proves nothing.

So, Jim went to Apple, then started building the A4. After that, Apple bought P.A. Semi for its engineering team. At Apple, Jim didn't just create a core, but they made the investment in creating an architecture with "big bones" that would have longevity and support lots of enhancements to follow. That's not the sort of ambitious, long-range project a startup embarks on, when the cash is starting to dry up.

So, it's right there: they did the design at Apple.

There is nothing in that article that gives and kind of timeline specific to the A4. However, you seem to be saying you think Apple developed and produced the A4 in almost exactly 24 months (The acquisition occurred April 2008, the A4 launched April 2010)? Or you are just ignoring that Keller was the VP of Low Power Mobile for 4 year prior to the acquisition? Seems like more of the same speculation about what you THINK someone said.

 

[bit_user]

Yet another assumption about what you think.

I know the startup game well enough and how corporate acquisitions work. Companies don't make acquisitions this big without knowing when they expect to see a return on that investment.

you seem to be saying you think Apple developed and produced the A4 in almost exactly 24 months (The acquisition occurred April 2008, the A4 launched April 2010)?

It would help if you actually knew anything about it. It wasn't a full custom design. It essentially used a tuned version of ARM Cortex-A8 that they licensed from ARM. The A5 also did something similar, utilizing a Cortex-A9. The first Apple SoC to contain an in-house core was the A6, which launched in Sept. 2012.

It's an assumption with zero weight.

It's funny how you're leveling all these attacks at me, since this whole tangent started with your completely unfounded suggestion that it was rushed. Zero evidence provided and plenty of reasons to think it wasn't.

At least I provided sources and supporting facts to my claims. All you can do is the equivalent of schoolyard taunts. That's pretty sad for someone I had previously respected.