Discussion Qualcomm Snapdragon Thread

[DZero]

Extremely cheap consumer devices are basically using Ryzen 5500U or Intel Core i3 1215U currently. Both are extremely capable SoCs for the price you can buy devices with them.

It's indeed a wild leap over what we used to have 5 to 10 years back. In the following years, we'll see even more low-end love with QCOM Canim, AMD Sonoma Valley and Kraken 2, Intel PTL U. Absolutely good stuff for cheap price.

Don't forget that Mediatek and nVIDIA might enter soon if that so, low end market will be completed.

 

[Ghostsonplanets]

Don't forget that Mediatek and nVIDIA might enter soon if that so, low end market will be completed.

Can't talk about MTK but Nvidia SoC isn't meant to be cheap.

 

[FlameTail]

So far, none of the reviews of Snapdragon X laptops have evaluated how good Qualcomm's Fastconnect 7800 WiFi7 + BT5.4 solution is.

 

[DZero]

Can't talk about MTK but Nvidia SoC isn't meant to be cheap.

Indeed, nVIDIA ones will be hella expensive, but MTK ones might be another story. They might try to be the new bottom of the barrel and well... if they are competent enough, it will be taking the marketshare

 

[ikjadoon]

So far, none of the reviews of Snapdragon X laptops have evaluated how good Qualcomm's Fastconnect 7800 WiFi7 + BT5.4 solution is.

Notebookcheck did test its Wi-Fi 6E performance. I imagine few reviewers have a solid Wi-Fi 7 methodology, though.

Wi-Fi 6E was quite good, near Intel's AX211 & Apple's Broadcom Wi-Fi 6E solution. These are just peak iperf3 rates, so we'll likely need more real-world tests with a myriad of APs, BT clients, etc. to know compatibility.

But, IIRC, Qualcomm uses similar Wi-Fi IP in their Android SoCs, so compatibility should be great.

He was not joking when he said it was "easy" taking their money just to make a video with the computers. Poor Qualcomm fools!

I disagree that SKUs are "customer-facing" in any way. Qualcomm wants to pull an Apple here and have X Elite, X Plus, and X (and more) plus some "don't care about what you're getting exactly". They were just terrible at delivering it.

Qualcomm's nonsense SKUs are 100% consumer-facing in every way. They're everywhere: OEMs use the SKUs.






Apple is genuinely simpler: Mx [optional modifier] [optional core count]. So the M4, M3 Pro 11-core, etc.

//

Qualcomm copied the mentality of current Intel and AMD desktop CPUs: a lot of nonsense branding and then the CPU's actual name.

Qualcomm Snapdragon® X Elite X1E-78-100
AMD Ryzen™ 9 9950X
Intel® Core™ i9 processor 14900K

This is unfortunately par for the course for Qualcomm whose naming scheme history is adorned with such marketing victories like the:

1. Snapdragon 850 Mobile Compute Platform
2. Snapdragon 8c
3. Snapdragon 8cx Gen 3
4. Snapdragon X Elite X1E-78-100T

New generation → new naming scheme (not unlike Intel laptop names, but that's for another thread).

Edit: copyright symbol and unlike typo

 

[ikjadoon]

I posted this on Ars, but thought it'd be good to post here now that we have official NBC GB6.2 1T tests:

CPU	GB6.2 1T Pts	Peak 1T Freq. (GHz)	Pts / GHz	Pts / GHz %
Apple M4	3715​	4.400​	844​	118.7%​
Apple M3 Pro (12C)	3138​	4.056​	774​	108.8%​
Apple M2 Pro	2663​	3.504​	760​	106.9%​
Apple M1 Pro	2409​	3.220​	748​	105.2%​
Qualcomm X1E-80-100	2845​	4.000​	711​	100.0%​
Arm Cortex-X4 (8G3 Galaxy)	2287​	3.390​	675​	94.9%​
Arm Cortex-X3 (8G2 Galaxy)	2107​	3.360​	627​	88.2%​
Arm Cortex-X2 (8+G1)	1806​	3.200​	564​	79.3%​
Intel i7-14900K	3243​	6.000​	541​	76.0%​
Arm Cortex-X1 (G3X G1)	1596​	2.995​	533​	74.9%​
Intel i9-12900HK	2611​	5.000​	522​	73.4%​
AMD 7950X	2975​	5.700​	522​	73.4%​
Intel i5-1355U	2595​	5.000​	519​	73.0%​
AMD 7840U	2562​	5.100​	502​	70.6%​
Intel i3-1215U	2082​	4.400​	473​	66.5%​
AMD 6800H	2063​	4.700​	439​	61.7%​

Note: The Arm Cortex-X4, X3, X2 + Apple M4 scores are passive. I picked active cooling for the rest so that we can see the uArches at their best.
Note: These are wildly different systems, even within 1 uArch: different cooling (1T throttling exists!), different caches, different clocks.

Sources: here and here.

Can't remember the last time we had four high-performance CPU uArches for consumers (that buy laptops). Some observations:

• Even with at least the same Chief Architect, Oryon shipped w/ lower "IPC" (here Pts / GHz in GB6.2 1T) than the Apple M1.
• The Oryon core is now the highest IPC uArch shipping on Windows by a wide margin.
• AMD Zen4 & Intel Raptor Cove are reaching Arm Cortex's IPC from 3-4 years ago.
• Apple's M4 made the largest uArch improvements since M1, even at just +12.8% IPC vs M1.
• Qualcomm mini-PCs would be very interesting, especially with good Linux compat. Even embedded appliances like routers / firewalls / NAS / mini servers. With sometimes a 10X in 1T perf (as many use old Atom / ancient Arm cores), Oryon would be a slam dunk.

 

[Nothingness]

Can't remember the last time we had four high-performance CPU uArches for consumers (that buy laptops).

Interesting comparison, thanks.

But (you knew the 'but' was coming, didn't you? ), you use deskptop Intel/AMD CPUs in the comparison.

EDIT: I quickly picked 7945HX3D and 14900HX and their pts/GHz is similar to their desktop counterparts. I should repeat that with lower frequencies SKU but I feel lazy.

 

[igor_kavinski]

Apple's lies exposed?

That's an ARM CPU with FASTER RAM and Lenovo's margin must be included in that upgrade price.

But yeah, Applers will go, BUT BUT BUT!!!!!

 

[FlameTail]

Snapdragon X Plus and Snapdragon X Elite?

Great! Nice names! I have no issue with that.

What I take issue with is this:

There is nothing inherently wrong with this SKU numbering system. Perhaps you could argue that the numbers such as 84, 80, 78... were arbitrarily chosen.

The real issue is the kind of SKU segmentation they have chosen to go with.

1. The vast difference between X1E-84-100/X1E-80-100 and the X1E-78-100

All three have 12 CPU cores. But the 78 has no Core Boost whatsoever! The maximum clock speed is 3.4 GHz, while the 80/84 top out ay 4.0/4.2 GHz respectively. This is a huge difference because it affects the single-core performance.

This makes the X1E-78 more similar to the X1P-64 than the X1E-80/X1E-84. Like the X1E-78, the X1P-64 does not have any core boost whatsoever.

By releasing the X1E-78 without Core Boost, Qualcomm has devalued the X Elite, and blurred the lines between the X Elite/X Plus segmentation.

I think Qualcomm should not have made the X1E-78-100 in that form. They should have given the X1E-78 a Core Boost of 3.8 GHz. If the yields do not permit that, then disable 2 CPU cores and sell it as an X Plus instead.

2. Introducing Purwa die SKUs under the X Plus branding.

So far, all the Snapdragon X SKUs unveiled have been based on the Hamoa die. But we know Qualcomm has a smaller die coming up in the pipeline called Purwa

Purwa has less PCIe lanes, an inferior video encoder/decoder, and tops out at 8 CPU cores, according to this leak;

Exclusive: Here's what Qualcomm didn't tell you about the Snapdragon X series

The upcoming Snapdragon X1 will be available in five different models: see their specs and performance here.

www.androidauthority.com

The other parts like the GPU/Display engines may also have been cut.

Back when Qualcomm unveiled the Snapdragon X Plus 2 months ago:

The idea we got was that the X Plus shares much of the same core features as the X Elite- 45 TOPS NPU, same display support, same encoder/decoder, same ISP.... The key difference was that X Elite had 12 cores, while the X Plus had 10 cores.

So you can see how it's going to be problematic when they introduce Purwa based SKUs under the Snapdragon X Plus brand. It's going to break the knowledge consumers/enthusiasts have in their mind about the X Plus, and completely destroy any sense of coherence that Qualcomm's new naming scheme had until now.

Now, this information is based on rumours/leaks, so maybe Qualcomm will do something different when it finally unveils those SKUs.

In my opinion, these Purwa SKUs should be made to belong to another tier (a lower one than the X Plus/X Elite). They could call it the "Snapdragon X Go" or simply the "Snapdragon X" perhaps.

 

[igor_kavinski]

Snapdragon XP

 

[Hitman928]

Most in depth review I've seen of gaming experience on X-Elite so far (not all that in-depth but more than I've seen elsewhere). Besides the compatibility issues and low performance, he shows frame time graphs for all games (typically terrible, even when fps is decent) and shows multiple games having weird visual issues/artifacts with native resolution. In Cyberpunk, FSR actually cleans up most of the artifacts that were seen at native which is really interesting. The weirdest thing was in GTA5 where he accidentally touches the screen of the laptop (touchscreen) which causes a weird screen tear for an instant, but then the frame times improve dramatically after, at least for a time.

Anyway, nothing really new to conclude as we already knew it wasn't ready yet for gaming, but seeing someone actually try to play games with it, rather than just throw up bar charts, was interesting.

Edit: Also, MS should have probably allowed the fans to get louder as he says the keyboard becomes uncomfortably hot to the touch while gaming. I know the near silent angle has been pushed for these laptops, but a bit of noise is better than the surface temperature getting too hot.

 

[igor_kavinski]

Qualcomm's flagship Snapdragon X Elite in PassMark: Fails to Beat Apple's M3

But nonetheless, it demonstrates decent results.

www.tomshardware.com

Passmark no like QC. Maybe QC need to send over a blonde techie to help Passmark devs optimize for their CPU?

 

[Ghostsonplanets]

View attachment 101932

Apple's lies exposed?

That's an ARM CPU with FASTER RAM and Lenovo's margin must be included in that upgrade price.

But yeah, Applers will go, BUT BUT BUT!!!!!

There's no lie. Apple sells you an integrated hardware and software platform with great margins for them. You pay extra to be into Apple ecosystem.

 

[okoroezenwa]

View attachment 101932

Apple's lies exposed?

That's an ARM CPU with FASTER RAM and Lenovo's margin must be included in that upgrade price.

But yeah, Applers will go, BUT BUT BUT!!!!!

Has Apple ever talked about their margin before? Not sure what lies would be exposed here.

 

[POWER4]

Qualcomm's nonsense SKUs are 100% consumer-facing in every way. They're everywhere: OEMs use the SKUs.

This is not exactly Qualcomm's fault, and I imagine, if they could choose, it would not be like that. Some OEMs choose to show the SKU (thanks?).

It does not need to be like that. I give two examples. When Apple used Intel processors, they never advertised the SKU on their website. Even with the computer in your hands, they've somewhat hidden it inside System Report.

The same thing goes for the Surface Laptop. Take a look:

 

[igor_kavinski]

Not sure what lies would be exposed here.

An open lie. The price increase giving the impression that Apple has to go the extra mile to acquire this special breed of RAM.

People are fools for believing Apple gives them options. They only give reasons to avoid them and people foolishly cling to them even harder.

 

[FlameTail]

OEMs could lean into simply calling them Snapdragon X Elite or Snapdragon X Plus instead of giving the full SKU name...

IF Qualcomm didn't put such a huge variation within each tier. X1E-84 vs X1E-78 is a colossal 23% difference in ST performance.

So consumers buying laptops with X Elite want to know which version they are getting. I have seen this in many other forums/communities.

This kind of thing wouldn't happen if Qualcomm hadn't made the gimped X1E-78.

See my post in thread 'Qualcomm Snapdragon Thread':

Page 108 - Discussion - Qualcomm Snapdragon Thread

Page 108 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

forums.anandtech.com

 

[igor_kavinski]

Not sure if this was posted: https://www.laptopmag.com/laptops/s...-pc-some-apps-like-fortnite-wont-run-natively

Yes, it won't run Abode Illustrator, the software by a non-existent company!

 

[okoroezenwa]

An open lie. The price increase giving the impression that Apple has to go the extra mile to acquire this special breed of RAM.

I mean, it doesn’t give that at all. But if that’s what you want to infer from that outrageous pricing then suit yourself. This is the Qualcomm thread though, so I’ll stop with this sideline.

 

[igor_kavinski]

outrageous pricing

Why would you call it that if you thought it reflected the truth (costs+reasonable profit margin) ?

 

[trivik12]

So Qualcomm is expected to ship 2 million X Elite/plus in 2024 and are targeting 100-200% growth in 2025. So as much as 6 million chips in 2025.

 

[ikjadoon]

This is not exactly Qualcomm's fault, and I imagine, if they could choose, it would not be like that. Some OEMs choose to show the SKU (thanks?).

It does not need to be like that. I give two examples. When Apple used Intel processors, they never advertised the SKU on their website. Even with the computer in your hands, they've somewhat hidden it inside System Report.

The same thing goes for the Surface Laptop. Take a look:

View attachment 101936

Can't say I agree. How are Qualcomm's chosen processor names not "Qualcomm's fault"?

Qualcomm is the only party that picked the processor names. "if they could choose, it would not be like that": why do you think that?

Not even Qualcomm thinks "X Elite" is the processor; "X Elite" is just a platform. The "X1E-00-1DE" is the processor name:






Qualcomm does not shy away from its processor name. Even Qualcomm's own hardware shares it happily:




It's not that QC tried to brand them as "X Elite" and then mischievous OEMs tracked down -> exposed confidential Qualcomm numbering: Qualcomm has proudly accepted its branding.

And, to be real, if OEMs made up random names for Qualcomm's SoCs, Qualcomm would likely not be happy: OEMs use what Qualcomm supplied as public-facing processor names.

//

Microsoft's choice to obscure the part number appears more like Apple Envy, tbh. Nobody, but Microsoft, uses Microsoft's 2nd set of made-up names (not even Qualcomm):

Surface Laptop Snapdragon® X Elite C12
Surface Laptop Snapdragon® X Plus C10

//

Interesting comparison, thanks.

But (you knew the 'but' was coming, didn't you? ), you use deskptop Intel/AMD CPUs in the comparison.

EDIT: I quickly picked 7945HX3D and 14900HX and their pts/GHz is similar to their desktop counterparts. I should repeat that with lower frequencies SKU but I feel lazy.

Aha, no, it's great to be double-checked. Right, I thought more of a best-case uArch comparison than a SKU comparison.

But, I just realized: I should've used the highest GB6.2 1T score by NBC, instead of NBC's average (as some systems may have sub-optimal implementations, throttling, etc.). I'll correct that will update w/ an updated table.

So with the high-end and low-end, at least in cooling & caches (not yet clocks):

AMD Zen4:
7950X: 3083 / 5.7 GHz = 540.9 (107.7%)
7840U: 2562 / 5.1 GHz = 502.3 (100.0%)

Intel Raptor Lake:
Intel i9-14900K: 3294 / 6.0 GHz = 549.0 (105.8%)
Intel i5-1335U: 2595 / 5.0 GHz = 519.0 (100.0%)

Close-ish, though the desktop counterparts (with larger caches, better cooling, etc.) do fare 6% to 8% better in IPC. I'll add these lower-end CPUs to the updated above table, too.

Thank you for the helpful note and this is interesting.

 

[FlameTail]

So,
X Elite = Platform
X1E-84-100 = Part Number

We'll go with that official terminology from now on.

 

[POWER4]

Can't say I agree. How are Qualcomm's chosen processor names not "Qualcomm's fault"?

Qualcomm is the only party that picked the processor names. "if they could choose, it would not be like that": why do you think that?

Qualcomm chose SKU names. OEMs can choose to display it to the customer or not. Intel has equally confusing (to the layman) SKU names. Apple never disclosed them, and it never stopped customers from buying MacBooks.

I don't think any company, other than Apple, would ditch SKUs in favor of naming each part a different "marketing-oriented" name. After all, as you said, they are (minor, in the grand scheme of things) variations of the same platform.

Another example: every car engine has a code. Still, you don't see your retailer referring to a BMW 320i as "This is the BMW 320i sporting the B48B20 engine.", much less the specific variant.

It's not that QC tried to brand them as "X Elite" and then mischievous OEMs tracked down -> exposed confidential Qualcomm numbering: Qualcomm has proudly accepted its branding.

And, to be real, if OEMs made up random names for Qualcomm's SoCs, Qualcomm would likely not be happy: OEMs use what Qualcomm supplied as public-facing processor names.

I understand they've branded them as X Elite for the 12-core, X Plus for the 10-core, and X(?) for the 8-core, with each variant having a weird but not unseen kind of SKU scheme. Microsoft was smart enough to notice that the average Joe doesn't care for the variant. They care if it is the i3, i5, or i7 and abstracted the SKU.

Now, I like to know the SKU, and as long as they don't go the Intel route with a thousand different SKUs, I don't think it's a terrible scheme. Two minutes at their website, and I can understand what I'm getting.

 

[Ghostsonplanets]

OEMs could lean into simply calling them Snapdragon X Elite or Snapdragon X Plus instead of giving the full SKU name...

IF Qualcomm didn't put such a huge variation within each tier. X1E-84 vs X1E-78 is a colossal 23% difference in ST performance.

So consumers buying laptops with X Elite want to know which version they are getting. I have seen this in many other forums/communities.

This kind of thing wouldn't happen if Qualcomm hadn't made the gimped X1E-78.

See my post in thread 'Qualcomm Snapdragon Thread':

Page 108 - Discussion - Qualcomm Snapdragon Thread

Page 108 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

forums.anandtech.com

QCOM should strive at doing a single SKU with a single binned one rather than the multitude of SKUs they did.

So Qualcomm is expected to ship 2 million X Elite/plus in 2024 and are targeting 100-200% growth in 2025. So as much as 6 million chips in 2025.

It seems so. Very small shipments though. WARM won't change Windows PC significantly at near term.