Nuvia: don’t hold your breath

It’s funny how much Intel is throwing at this just like I said. They’re beating QC on Perf/W ST in one benchmark — SpecInt.


Also:

IMG_8142.png

IMG_8141.png


IMG_8138.jpeg

Ignore the Apple comment, they are basically the same thing here. Just think of these as clusters per se or tiers.


This Geekerwan measurement with the Galaxy Book 16 in *Linux* Andrei said was not advised for power reasons both in idle (shows as you see — a .6W platform floor for Intel or so and a 1.8-2W floor for Qualcomm, but there’s a caveat *). Intel wins perf/W ST in SpecInt, Qualcomm wins the SpecFP results but by much more.

Given the GB16 is like the worst implementation out there I think this looks fine tbh.
*I believe his measurements were right in this specific case with the Lenovo or Asus Lunar Lake VRM monitoring vs The Galaxy Book VRM, but I am confident this includes some Samsung idiocy on a platform level whatever that may be or Linux like Andrei said.

Why?

Because A) Andrei but B) we have the XPS 9345 with Lunar Lake and The X Elite built to largely identical standards and power efficiency targets and both have similar web browsing and offline video playback in the 26-29 hour range under ideal constraints.

If you take Geekerwan’s idle floor at face value for the chips themselves and motherboard platforms in general it doesn’t make sense, .6 or so watts for Intel vs 1.8 for Qualcomm as an inherent mobo trait is going to be very influential in idle for video offloading even marginally in web browsing. QC’s probably isn’t perfect and I don’t mind them not being at Apple levels broadly but this is fishy.


In the one head-to-head test, what we actually see both from Dell’s own XPS 9345 stats (on Windows… too) is the Lunar Lake and Qualcomm X Elite laptops (display controlled etc, FHD) are similar (upper 20 hours) and — in laptopmag and tomsguide automated web browsing testing with plenty of idle, here are the results between them:

IMG_8136.jpeg
IMG_8137.jpeg




So I think the idea Linux and the Galaxy Book 16 sucks is almost certainly correct. It simply would not be possible for Qualcomm to get 26-29 hours of video playback in a 50-55whr battery just like Intel with Lunar Lake in the same laptop, much less match or beat them in two entirely separate web browsing tests in the same laptop if their overhead was that much higher (unless you claim a really inefficient active power from Intel which, yes okay but in video playback? And right when they’re using e cores for like 80% of tasks? Lol no.)

Anyway.

I feel pretty reasonable saying it’s clear who’s punching above their weight and who likely has more in store. QC isn’t even using E Cores nor N3 yet which would push power down meaningfully and battery up and as the 8 Gen 4 is likely to show, the standard frequencies achievable with reasonably dense N3B and N3E is pretty solid — if the X Elite were on N3B, I bet starting clocks would be 3.8-4GHz, peak probably 4.7 permitting timing but they had some slack it seems. It’s weird people think Lunar Lake is anything but a temporary lull in mindshare and revenue.

It's not at the top vs Apple, and it certainly isn’t cost competitive. Lunar Lake Graphics is nice but not a huge deal in context of the long arc here.
 
Huh


At the other place people were saying what a fantastic deal this was compared to the mini … apparently it was a bit too good of a deal. Overall, Qualcomm’s launch execution has been pretty lackluster.
 
Last edited:

As I’m sure many of you know I dislike when chip companies continually switch between different processors/devices to make different claims, but that aside Qualcomm seems more correct here than Intel’s marketing (who did the same thing), especially Intel’s misleading SPEC results. Intel doesn’t win on performance “period” and certainly not on performance/W. Of course, Qualcomm still has compatibility issues, a weak GPU, only a single SOC (with multiple bins), and lackluster developer support and overall execution.

Qualcomm’s second generation chips can improve some of these aspects intrinsically, but others will need work from more than just their hardware teams.
 

As I’m sure many of you know I dislike when chip companies continually switch between different processors/devices to make different claims, but that aside Qualcomm seems more correct here than Intel’s marketing (who did the same thing), especially Intel’s misleading SPEC results. Intel doesn’t win on performance “period” and certainly not on performance/W. Of course, Qualcomm still has compatibility issues, a weak GPU, only a single SOC (with multiple bins), and lackluster developer support and overall execution.

Qualcomm’s second generation chips can improve some of these aspects intrinsically, but others will need work from more than just their hardware teams.
Agree that QC is well ahead of Intel/AMD. I do find QC’a hyperbole very off-putting however. They didn’t fulfill their performance promise last year and now we”re back to unsubstantiated claims.

Do you know if they are supposed to announce their second gen laptop chips soon?
 
Agree that QC is well ahead of Intel/AMD. I do find QC’a hyperbole very off-putting however. They didn’t fulfill their performance promise last year and now we”re back to unsubstantiated claims.
Now that both sets of devices are out, as the Tomshardware author said, from reviews they can confirm most of what Qualcomm put out in this press conference wrt to Lunar Lake, just with the caveats that I also mentioned above. Basically any cherry picking and exaggeration done by Qualcomm wrt to Apple in the pre-release of Snapdragon was matched or surpassed by Intel wrt to Qualcomm in the Lunar Lake presentation.
Do you know if they are supposed to announce their second gen laptop chips soon?
They just announced new phone chips.


3nm node, but no description here if they are new Oryon cores or ports. It reads like just a port from 4nm to 3nm (because you’d think that they’d emphasize new cores), but I remember some of the early benchmark leaks were suggestive of a small iso-clock performance uplift. So it’s unclear if we’re getting v2 anytime soon.

Also, of minor interest, apparently the Dimensity 9400, which uses standard ARM cores, is likewise dropping efficiency cores. This is an odd trend on non-Apple flagship SOCs. I’m not quite sure what to make of it.

Edit: Although it should be said it is a little different where for Snapdragon all the cores are Oryons, just two at higher clocks, while the Dimensity 9400 has a mix of X925, X4, and A720 cores (and the A720 are more similar to Apple’s E-cores in performance). Which means in reality, the Dimensity chips have really only dropped the littlest 500-series cores. So I guess it’s not quite a comparable statement.
 
Last edited:
Now that both sets of devices are out, as the Tomshardware author said, from reviews they can confirm most of what Qualcomm put out in this press conference wrt to Lunar Lake, just with the caveats that I also mentioned above. Basically any cherry picking and exaggeration done by Qualcomm wrt to Apple in the pre-release of Snapdragon was matched or surpassed by Intel wrt to Qualcomm in the Lunar Lake presentation.
Oh I’m sure Intel did! I also saw a chart QC put out where they compare their highest score to the average A18 score.
They just announced new phone chips.
Yes, I watched some of the presentation. If I understand, these are newish cores designed for mobile. I’m wondering if a Oryon v2 will be previewed soon?
 
Back
Top