Discussion Qualcomm Snapdragon Thread

[The Hardcard]

Ryan Shrout battery chart:

https://twitter.com/x/status/1803088585651855435

I’d like to see Speedometer performance before and after that bend in the line.

 

[adroc_thurston]

Snapdragon X Elite (X1E-78-100) does not seem to have the boost clock

Oh yeah, but -80 aka the 4G one is available.

 

[Nothingness]

As most of the people living in the Northern Hemisphere are aware, it is cherry-picking season here.

So, expect some cherry-picked top bin Snapdragon X parts in the officially sanctioned reviews, while the vast majority of laptops on sale are going to come from low bins.

That's not what Notebookcheck did. They picked an ASUS with X1E78.

https://www.notebookcheck.com/Asus-Vivobook-S-15-OLED-im-Test-Mit-dem-Snapdragon-X-Elite-in-ein-neues-Notebook-Zeitalter.847338.0.html

EDIT: link fixed

 

[Joe NYC]

That's not what Notebookcheck did. They picked an ASUS with X1E78.

https://www.notebookcheck.com/Asus-Vivobook-S-15-OLED-im-Test-Mit-dem-Snapdragon-X-Elite-in-ein-neues-Notebook-Zeitalter.847338.0.html

EDIT: link fixed

Yeah, I noticed.

If that's what most of the Snapdragon buyers end up with is nowhere near the performance Qualcomm has been peddling for months and months.

 

[Hitman928]

Sooo, is this a game of "spot the unfair test condition"? Or are we just going to pretend that Ryan Shrout is reliable?

Intel (Acer) machine has an OLED vs. the HP (Snapdragon) with an IPS. The Intel machine's display also has a slightly higher resolution and I believe a higher refresh rate, so the display on the Intel machine is almost certainly consuming significantly more power. Acer machines are also not typically known for being configured for good battery life. Not really a fair comparison if trying to attribute battery life to the chip inside.

 

[FlameTail]

I must write a wall of text about their terrible naming scheme...

 

[Hitman928]

Yeah, I noticed.

If that's what most of the Snapdragon buyers end up with is nowhere near the performance Qualcomm has been peddling for months and months.

How it started:

How it's going:




Edit: added multi-core results.
Edit 2: Wrong CB version on MC results initially.

 

[del42sa]

First Qualcomm Snapdragon X Series laptops tested in gaming: “3D performance is disappointing”

 

[adroc_thurston]

How it started:

How it's going:

View attachment 101394

Well that's 4.3 versus 3.4.
Lmao.

The IPC is there, they just kinda lied about their binning capabilities.

 

[Hitman928]

Well that's 4.3 versus 3.4.
Lmao.

The IPC is there, they just kinda lied about their binning capabilities.

Yeah, I was just continuing with the thought of the post I quoted.

Besides the lower binned SKUs seemingly being in the majority of the designs, it also seems that QC's initial advertised TDPs have increased fairly significantly. Originally, they advertised 23W "device" TDP for getting the ~950 score in CB24 as shown in my previous post.



But in the NBC review, it seems to be updated now to a "device" TDP of 35 W to get to that performance. That's a 52% increase in TDP. Unless I'm missing something. . .

 

[adroc_thurston]

But in the NBC review, it seems to be updated now to a "device" TDP of 35 W to get to that performance. That's a 52% increase in TDP. Unless I'm missing something. . .

Yeah looks like their binning data was overly optimistic and then the reality hit.
Frankly, now I want someone to get that 4.3G devkit and measure 1t power.
Go go housefire championship.

 

[Joe NYC]

How it started:

How it's going:

View attachment 101394
View attachment 101397

Edit: added multi-core results.
Edit 2: Wrong CB version on MC results initially.

It seems like the "78" SKUs are garbage. No redeeming quality.

Edit: other than perhaps the bottom feeding Chromebook end of laptops.

 

[adroc_thurston]

No redeeming quality.

It's a competent and efficient part but by 2024 standards it's just 'ok'.

 

[gdansk]

It seems like the "78" SKUs are garbage. No redeeming quality.

Cheap would be redemption (but we're not there yet)

 

[CouncilorIrissa]

It seems like the "78" SKUs are garbage. No redeeming quality.

It is a decent part in terms of hardware all things considered, but being simply decent isn't going to cut it. It should've been straight up significantly better in native performance AND perf/w to offset emulation penalty, buggy iGPU drivers and immature platform in general.

Hopefully other designs perform better tbh.

 

[Joe NYC]

It's a competent and efficient part but by 2024 standards it's just 'ok'.

Yeah, it is not even keeping up with 2023 models - the 78 SKU.

 

[Hitman928]

Yeah looks like their binning data was overly optimistic and then the reality hit.
Frankly, now I want someone to get that 4.3G devkit and measure 1t power.
Go go housefire championship.

Have there been any announced models with the 4.3 GHz boost SKU besides the devkit?

 

[adroc_thurston]

Have there been any announced models with the 4.3 GHz boost SKU besides the devkit?

Nope, the thing is vaporware.

Yeah, it is not even keeping up with 2023 models - the 78 SKU.

Still more efficient in cinememe@w and BL than Phoenix or MTL.
Again, has technical merits but in 2024H2 the world is different.

 

[Abwx]

Still more efficient in cinememe@w and BL than Phoenix or MTL.

More efficient than MTL yes but not that Phoenix, they just match the latter at 35W while the former need 45-50W to match them at 35W.

 

[Nothingness]

How it started:

How it's going:

It was expected that the X1E78 results would not be great. But I think jumping to conclusions at this point is a bit premature. Let's wait and see how models with X1E80 and X1E84 behave.

Then the question will be if availability will be there for the higher-end SKUs. Which will confirm or infirm the binning issue.

 

[adroc_thurston]

More efficient than MTL yes but not that Phoenix, they just match the latter at 35W while the former need 45-50W to match them at 35W.

1t and overall SOC efficiency are still higher.

X1E84

Doesn't exist?

 

[Abwx]

1t and overall SOC efficiency are still higher.

As efficient than their gaming claims where they viral marketed AMD, at the end of the day they didnt win a single game bench, with the 780 being first, MTL second and their Snapdragon, or rather Snaplizard, as a distant third.

Snapdragon X Elite vs. Intel Core und AMD Ryzen im Test

Wie gut ist Windows 11 on Arm mit dem Qualcomm Snapdragon X Elite? ComputerBase hat mit dem Asus Vivobook S 15 den Test gemacht.

www.computerbase.de

 

[Joe NYC]

Is this the "End of the Affair"?

Qualcomm AI/Copilot PCs don’t live up to the hype

Qualcomm’s new AI/Copilot PCs has overwhelming hype but their actions point to a far murkier picture.

www.semiaccurate.com

 

[SpudLobby]

”garbage” is ridiculous lol — QC have the best CPU architecture in Windows here for the core and the layout/uncore. It’s simple, scales well and they haven’t even thrown in a low-voltage E core cluster (not that they need it as is but it would further improve things likely)

The GPU: I will punt on this. Thanks to both drivers and some other BS (also probably undersized, bad for compute fun too, lol) it’s just not that great. In general I think if I could have a Nvidia designed iGPU, Xe2 or RDNA3.5, or honestly even Immortalis that all seems better especially for Windows (yes I think Arm might literally more likely to build good windows drivers and their GPU is better for compute).

So yes — the AMD guys here are right the Adreno driver thing is embarrassing with the crashing etc etc.

But the CPU — even for that 78 part — great and could still compete with Strix in MT handily, and wouldn’t lose in battery life either (most likely comes out on top still). And in ST perf/W I doubt Strix is ahead either, at worst they’re likely similar. Absolute ST tho Kracken and avg Strix parts will beat the common 3.4GHz skus yes.

Obviously, tradeoffs with emulation is the reason we’re quibbling, which is valid from a market POV — consumers just want what works or a tradeoff that works out. Not saying that’s wrong. But this is still a very competent engineering effort that probably has some tweaks they could make retrospective if Charlie’s PMIC story is accurate.

 

[Hitman928]

It was expected that the X1E78 results would not be. But I think jumping to conclusions at this point is a bit premature. Let's wait and see how models with X1E80 and X1E84 behave.

Then the question will if availability will be there for the higher-end SKUs. Which will confirm or infirm the binning issue.

Granted I didn't do an exhaustive search, but I couldn't find any offerings with the X1E-84 SKU which is why I asked if there were any. It seems like there aren't any of them actually available, at this time at least.

The X1E-80 should definitely perform better, I just wasn't expecting such a cut down SKU, let alone for it seemingly to be in the majority of designs and I don't think most others were either, that's all. With the higher base and actual boost, it will also be interesting to see how battery life is effected when the X1E-80 is used. It's not what it was hyped up to be, but it is very nice to see that the ARM version of Windows is finally a good experience for at least most users (outside of gaming) and QC provided a solid chip to drive that experience. I do worry that without a killer app or some kind of killer performance characteristic (whether performance or battery life), it still won't be enough to get much market penetration. We'll see though, MS seems to be pretty keen (at least for now) in pushing these to be successful.