Results are in: Qualcomm’s Snapdragon X Elite goes toe-to-toe with Apple’s new M3 Pro processor

[Windows Central]

The latest benchmarks are out on Apple's new M3, M3 Pro, and M3 Max processors. Here's how they stack up against Qualcomm's forthcoming Snapdragon X Elite for Windows laptops. The results will surprise you.

Results are in: Qualcomm’s Snapdragon X Elite goes toe-to-toe with Apple’s new M3 Pro processor : Read more

 

[naddy69]

"In short, not only does Qualcomm beat the M3 in multicore..."

Why are you STILL comparing to the M3? That is only 8 cores! Of COURSE the 12 core X Elite beats it. That is nothing to brag about.

The only comparison that matters is the 12 core M3 Pro. Of COURSE the 16 core M3 Max beats the X Elite.

And BTW, which M3 Pro are you comparing to? There is an 11 core M3 Pro available in the 14" MacBook Pro. The M3 Pro in the 16" MacBook Pro is 12 cores.

Based on the fact that the earlier claims were about "trouncing the M2 in the 13" MacBook Pro" - which is 8 cores AND 1.5 years old! - I am skeptical of these claims also.

 

[Daniel Rubino]

"In short, not only does Qualcomm beat the M3 in multicore..."

Why are you STILL comparing to the M3? That is only 8 cores! Of COURSE the 12 core X Elite beats it. That is nothing to brag about.

The only comparison that matters is the 12 core M3 Pro. Of COURSE the 16 core M3 Max beats the X Elite.

And BTW, which M3 Pro are you comparing to? There is an 11 core M3 Pro available in the 14" MacBook Pro. The M3 Pro in the 16" MacBook Pro is 12 cores.

Based on the fact that the earlier claims were about "trouncing the M2 in the 13" MacBook Pro" - which is 8 cores AND 1.5 years old! - I am skeptical of these claims also.

RTFA

 

[ShinyProton]

In the end, these numbers only indicate both processors are from the same generational breed - which is already an immense feat for Qualcomm.

I was already pretty impressed by the 8cx Gen 3 abilities. Now, the X Elite will push this to an even greater level.

 

[Daniel Rubino]

In the end, these numbers only indicate both processors are from the same generational breed - which is already an immense feat for Qualcomm.

I was already pretty impressed by the 8cx Gen 3 abilities. Now, the X Elite will push this to an even greater level.

Agree. The "vs Apple" stuff is good for context (and is fanbait, no doubt), but Qualcomm doesn't actually need to beat Apple at all. They just need to beat Intel and AMD, which it looks like they will.

 

[bio]

Let’s see the results in real world usage because afaik Windows for ARM still doesn’t shine in x86 emulation and we all knows this is really important support in the Windows world (way more than in the Apple ecosystem).

 

[Village_Idiot]

Qualcomm: We have a chip that can compete in the ARM CPU space.

Intel: Meh.. . We don't think it's going to be much of a threat.

Me: I'm thinking Intel is about to let themselves get kicked in the nuts.

 

[dkstrauss]

Qualcomm: We have a chip that can compete in the ARM CPU space.

Intel: Meh.. . We don't think it's going to be much of a threat.

Me: I'm thinking Intel is about to let themselves get kicked in the nuts.

Apple: We have an M1 chip that will replace Intel.

Intel: Meh.. . We don't think it's going to be much of a threat.

Me: Why is Intel doubled up over there in the corner crying in pain?

 

[bradavon]

Why no mention of x86 emulation?

It wasn't peformance and battery that made M1 a success, it was MacOS' x86 translation layor support. It would've tanked just like WoA if current apps in 2019 were slow.

Then when that wasn't a problem people were also thrilled with peformance and battery.

What's Microsoft doing about this?

It seems MS is hoping enough apps will be ported to ARM it won't matter but that's never going to happen. Windows has even older x96 apps in use than MacOS ever had.

Let's hope the Snapdragon X really does make this irelevant. It's a tall ask though isn't it.

 

[bradavon]

Agree. The "vs Apple" stuff is good for context (and is fanbait, no doubt), but Qualcomm doesn't actually need to beat Apple at all. They just need to beat Intel and AMD, which it looks like they will.

Agreed but but how so are Qualcomm beating Intel and AMD?

They're not in a real world sense.

Microsoft arent addressing the elephant in the room. This possibly all means nothing if 86 emulation is still a problem.

It's never been a raw power issue, not really. Windows on ARM tanked because of how Windows handles x86 apps.

 

[Daniel Rubino]

Let’s see the results in real world usage because afaik Windows for ARM still doesn’t shine in x86 emulation and we all knows this is really important support in the Windows world (way more than in the Apple ecosystem).

While relevant, I'll say this: Previously, 8cx performed like an 11th Gen Core i5 U-series, but under emulation, it dropped below that, so you 'felt it.'

With Oryon performing better than an Intel 13th Gen HX processor (literally a modified desktop CPU) and matching M3 Pro through sheer power, you're not going to notice the impact.

This is true of Rosetta2, too, for Apple. But it started with M1, which was already miles ahead of 8cx, so it was never an "issue."

To quote MacRumors: "Rosetta 2 running x86 code appears to be achieving 78%-79% of the performance of native Apple Silicon code."

Apple's Rosetta2 performance wasn't a zero-sum game, either. But sheer power overcomes that performance deficit. It has improved since M1, just like Windows x86 emulation has improved since Windows 11 came out.

 

[Daniel Rubino]

Qualcomm: We have a chip that can compete in the ARM CPU space.

Intel: Meh.. . We don't think it's going to be much of a threat.

Me: I'm thinking Intel is about to let themselves get kicked in the nuts.

Maaaybe. Don't discount Meteor Lake, yet. We'll get benchmarks for that this January. It's a major shift for Intel, and it's far from finished as Lunar Lake goes much further.

 

[Daniel Rubino]

Why no mention of x86 emulation?

It wasn't peformance and battery that made M1 a success, it was MacOS' x86 translation layor support. It would've tanked just like WoA if current apps in 2019 were slow.

Then when that wasn't a problem people were also thrilled with peformance and battery.

What's Microsoft doing about this?

It seems MS is hoping enough apps will be ported to ARM it won't matter but that's never going to happen. Windows has even older x96 apps in use than MacOS ever had.

Let's hope the Snapdragon X really does make this irelevant. It's a tall ask though isn't it.

Because it wasn't tested.

Also, x86 support has improved every year. Indeed, we're confident that the VERSION of Windows 11 that these reference design PCs were running is unreleased code from Microsoft to support Oryon (everyone will get this update, likely in the spring as part of v-next).

It's not as if x86 apps don't run on Windows on ARM, they do. But the performance hit under 8cx made it feel worse than Apple's M1 processor because they are miles apart in performance abilities.

See my other comment, which I'll repeat here:

While relevant, I'll say this: Previously, 8cx performed like an 11th Gen Core i5 U-series, but under emulation, it dropped below that, so you 'felt it.'

With Oryon performing better than an Intel 13th Gen HX processor (literally a modified desktop CPU) and matching M3 Pro through sheer power, you're not going to notice the impact.

This is true of Rosetta2, too, for Apple. But it started with M1, which was already miles ahead of 8cx, so it was never an "issue."


This is true of Rosetta2, too, for Apple. But it started with M1, which was already miles ahead of 8cx, so it was never an "issue."

To quote MacRumors: "Rosetta 2 running x86 code appears to be achieving 78%-79% of the performance of native Apple Silicon code."

 

[Daniel Rubino]

It's never been a raw power issue, not really. Windows on ARM tanked because of how Windows handles x86 apps.

Please explain. As someone who runs ARM PCs regularly, it ABSOLUTELY is about raw power. You can't compare emulation on the 8cx vs. M1/M2/M3 since those Apple chips are like 10x as powerful.

Rosetta2 did have a performance impact even on M1, as noted by MacRumors (using unoptimized GeekBench).

To quote MacRumors: "Rosetta 2 running x86 code appears to be achieving 78%-79% of the performance of native Apple Silicon code."

It just "feels" better since M1 had so much headroom. Now imagine what that would be like on x86 with an M3-Pro competitor.

What's wrong with x86 emulation? Do you have apps that are not working under it? If so, which ones I can test/validate and hopefully let Microsoft know.

 

[bradavon]

Please explain. As someone who runs ARM PCs regularly, it ABSOLUTELY is about raw power. You can't compare emulation on the 8cx vs. M1/M2/M3 since those Apple chips are like 10x as powerful.

What's wrong with x86 emulation? Do you have apps that are not working under it? If so, which ones I can test/validate and hopefully let Microsoft knows.

For context my only PC is an OG Surface Pro X and I'm holding out for a premium clamshell laptop running Windows on ARM (there aren't any) so I like the platform. I really like it in fact. No fans, consistently good battery life. But it's not ready primetime today.

What you've just explained may make my next point invalid but I'll explain my point. The day to day performance of 8cx is fine for most users, most users don't need Core i7's for instance. Technies may say they do, but they don't.

But sluggish apps "you feel" due to emulation is absolutely what has held the platform back.

But you're saying you feel that sluggishness should be removed with more raw power. So fair enough.

I tend to ignore the Benchmarking section of reviews incidentally. I find my Mum's Core i3 (with 8GB/SSD) perfectly acceptable performance wise. Windows on ARM today would be fine for her needs (very lights needs, Office) but there aren't many decent ARM laptops in the £400 range.

Facebook Messenger is sluggish but otherwise the apps I use run well.

But for the platform to succeed sluggish apps need to be nippy too.

The only apps I'd like to run that I can't, are Google Nearby Share or Samsung Quickshare. Obviously I wouldn't use both but neither run on ARM .

Annoying but not dreak breakers.

I don't get why Microsoft isn't doing a deal with Google or paying them incidentally. I don't care about Chrome personally but lack of ARM Chrome is hurting the platform.

No Google Drive or Dropbox is a problem too.

 

[diogene7]

While relevant, I'll say this: Previously, 8cx performed like an 11th Gen Core i5 U-series, but under emulation, it dropped below that, so you 'felt it.'

With Oryon performing better than an Intel 13th Gen HX processor (literally a modified desktop CPU) and matching M3 Pro through sheer power, you're not going to notice the impact.

This is true of Rosetta2, too, for Apple. But it started with M1, which was already miles ahead of 8cx, so it was never an "issue."

To quote MacRumors: "Rosetta 2 running x86 code appears to be achieving 78%-79% of the performance of native Apple Silicon code."

Apple's Rosetta2 performance wasn't a zero-sum game, either. But sheer power overcomes that performance deficit. It has improved since M1, just like Windows x86 emulation has improved since Windows 11 came out.

May you have any rough estimate of x86 code emulation performance on Windows on ARM ?

I think I read on a French website (macg) was claiming that on Windows on ARM, the efficiency of x86 code emulation was a bit below 50%

While relevant, I'll say this: Previously, 8cx performed like an 11th Gen Core i5 U-series, but under emulation, it dropped below that, so you 'felt it.'

With Oryon performing better than an Intel 13th Gen HX processor (literally a modified desktop CPU) and matching M3 Pro through sheer power, you're not going to notice the impact.

This is true of Rosetta2, too, for Apple. But it started with M1, which was already miles ahead of 8cx, so it was never an "issue."

To quote MacRumors: "Rosetta 2 running x86 code appears to be achieving 78%-79% of the performance of native Apple Silicon code."

Apple's Rosetta2 performance wasn't a zero-sum game, either. But sheer power overcomes that performance deficit. It has improved since M1, just like Windows x86 emulation has improved since Windows 11 came out.

By any chance, may you have any rough estimate of x86 code emulation performance on Windows on ARM as November 2023 ?

I think I read in ~2021 on a French website (https://www.macg.co/) that, at the time, x86 code emulation by Microsoft was a bit less than 50% efficient, compare to ~80% with Rosetta2 on a Mac.

Could we hope to see Microsoft getting closer to 70% and also nearly eliminate bugs and apps that doesn’t work ?

How much improvement has been done on all that as indeed, it is critical for any hope of success that the transition seems (completely) seemless, and that as much as possible users don’t notice anything different… (as is the case on Mac with Apple silicon)

 

[Village_Idiot]

Maaaybe. Don't discount Meteor Lake, yet. We'll get benchmarks for that this January. It's a major shift for Intel, and it's far from finished as Lunar Lake goes much further.

True. I can see Qualcomm's being popular in the business laptop market. I know of many use cases where a lightweight Windows laptop with M365 running the Snapdragon X Elite and an integrated unlocked 5G capability would work.

The only thing holding ARM based Window PCs back is the app and program availability. ARM CPU manufacturers need to start leveraging their combined muscle to get app and program developers to support Windows on ARM.

In any case, if Intel blows this off, it will get a hard kick in the nuts.

 

[Village_Idiot]

Apple: We have an M1 chip that will replace Intel.

Intel: Meh.. . We don't think it's going to be much of a threat.

Me: Why is Intel doubled up over there in the corner crying in pain?

Intel isn't that concerned about Apple's M series. Just like its A series iPhone and iPad CPUs the M series is Apple only. When Apple dumped Intel as its CPU chip provider, it didn't really put much of a dent in Intel's market share. Qualcomm and nVidia will be competing directly against Intel (and AMD) in the Windows PC market, if Snapdragon X Elite actually lives up to Qualcomm's claims, it has a good chance of cutting significantly into Intel's market share. Qualcomm nor nVidia are not a little startups that Intel can easily ignore.

 

[Chris G5]

Does Apple literally claim that the M2 MAX SoC is rated at 36W TDP? If so, they are being very creative. Thermal solutions properly implemented must cope with the peak current draw of a processor not some averaged power state at which (near) full performance can be delivered. Intel has done the same thing over the years quoting TDP (PL1) as the 'power draw' of its CPUs. A sounder rating is TDP (PL2). That is probably the number that Qualcomm is quoting. Adrei Frumusanu, the guy who provided all of the curated SDX Elite benchmarks is publicly known to consider PL1 a joke and a scam. Daniel Rubino could be right that other sources of power consumption are also included in Qualcomm's numbers but that wouldn't be a standard way to do things. Anything beyond the SoC isn't under Qualcomm's control.

The 80W laptop config is clearly meant to allow limited comparisons between the the SDX Elite and the M2 MAX/M3 MAX. I don't think we will see an SDX Elite laptop in that configuration any time soon but the comparison is interesting at least on the CPU side. At 23W the SDX Elite acquits itself very well against the M3 - the Oryon cores will do even better once evident Windows overheads are dealt with - and looks alright against the M3 Pro as well. The SDX Elite appears to give the M3 a run for its money and that makes it better than 95% of laptop CPUs out there. And, the early signs are that Qualcomm's chip has the edge on energy efficiency and is the better all round part.

In 2024, I think we will see a 23W SKU and a fanless SKU for the SDX Elite. Something to compete with the M4 MAX might come out in 2025. I would hope to see a power rating on that chip that is less than 80W.