Nuvia: don’t hold your breath

exoticspice1 said:
@RockRock8 I'll like to apologise for that rude post calling you a robot. Sorry, I should've known better. Hope we can start on a better footing from now on. I don't know if you will read this but I thought I rather do it here.
Click to expand...
Hi, I'm sorry for the delayed response.

I just want to say not only thank you for reading through what I felt and thought and feedback, but taking the feedback and be willing to apologize. It might sound ridiculous, but this means a lot to me. As I said in my original message, I'm not planning to comment for too much longer as I was actually planning to only comment about Gurman regarding that whole BS debacle with executives retiring and stuff. I never meant to continue commenting much past that, so please don't take it personally.

Nevertheless if this is the last time we chat, please keep being willing to be open to new users and feedback. Your willingness to consider what I said then apologize means a lot. This comment really brought some hope for me ❤️❤️😁
 
MerryCherry said:
Qualcomm is moving at a snail's pace. They announced the chips 6 months ago, yet the devices are still not out.

Meanwhile in that span of time, Apple announced and released M5/M5 Pro/M5 Max Macbooks and also dropped the nuke (Macbook Neo).
Click to expand...
They aren't the only ones. I've noticed a lot more Panther Lake-H chips being reviewed than I've seen actually available to buy - though I've started to see them trickle out now. Maybe not quite as slow as Qualcomm, but not exactly fast either (especially since Intel made a lot of noise about Panther Lake-H and 18A products being out at the end of 2025). Most of the reviews when they came out didn't even have prices attached to the devices.

This was NotebookCheck's note on the Asus they reviewed:

Price and Availability​

The ExpertBook Ultra with Panther Lake is expected to ship in Q2 2026. Asus has not announced any prices at the time of publishing.
Click to expand...

And that was pretty common.
 
Last edited:
casperes1996 said:
Sometimes it feels like they announce really early so they can show benchmarks against current gen competition even if they won't be out until the competition has moved ahead another notch
Click to expand...
While I don't disagree :), it's also that they have their own trade show in September and they seem to want to use that to announce all their devices regardless of when they are actually out.
 
HDUvhxjaoAA0i1y.jpeg

Image source

I was curious how Qualcomm's performance cores compared to Apple, so I did some search and the results surprised me.

According to Geekerwan's SPEC2017 numbers for the 8 Elite Gen 5, Qualcomm's performance cores have about 2/3 the performance of the prime cores. Assuming that carries over to Geekbench;

Oryon Gen 3Clock SpeedGeekbench 6
Prime4.6 GHz~3800
Performance3.6 GHz~2500

Comparing that to Apple;
M5 CPUClock SpeedGeekbench 6
Super4.6 GHz~4300
Performance4.4 GHz~3000
Efficiency3 GHz~1700

Apple's performance cores are 20% faster than Qualcomm's. That appears to be thanks to it's ~20% faster clock speed, which means the IPC of both is nearly identical.

Now, the remarkable thing is that Qualcomm's performance core is only 4 wide, compared to Apple's 7 wide.

Takeaway is that the Oryon Performance cores are much stronger than I assumed. Could Qualcomm replicate Apple's strategy with the M5 Pro/M5 Max in a future Snapdragon X CPU? Easily. They already have the hardware in the bag.
 
Last edited:
MerryCherry said:
View attachment 38400

I was curious how Qualcomm's performance cores compared to Apple, so I did some search and the results surprised me.

According to Geekerwan's SPEC2017 numbers, Qualcomm's performance cores have about 2/3 the performance of the prime cores. Assuming that carries over to Geekbench;

Oryon Gen 3Clock SpeedGeekbench 6
Prime4.6 GHz~3800
Performance3.6 GHz~2500

Comparing that to Apple;
M5 CPUClock SpeedGeekbench 6
Super4.6 GHz~4300
Performance4.4 GHz~3000
Efficiency3 GHz~1700

Apple's performance cores are 20% faster than Qualcomm's. That appears to be thanks to it's ~20% faster clock speed, which means the IPC of both is nearly identical.

Now, the remarkable thing is that Qualcomm's performance core is only 4 wide, compared to Apple's 7 wide.

Takeaway is that the Oryon Performance cores are much stronger than I assumed. Could Qualcomm replicate Apple's strategy with the M5 Pro/M5 Max in a future Snapdragon X CPU? Easily. They already have the hardware in the bag.
Click to expand...
I thought Oryon’s cores were clocked higher. Up to 5 ghz I believe. Happy to be corrected. I don’t think IPC is equal between Apple and QC although they are much closer than anyone else.

Edit. Sorry I misread. I thought you were talking about the Prime cores. It’s late here!
 
MerryCherry said:
View attachment 38400

I was curious how Qualcomm's performance cores compared to Apple, so I did some search and the results surprised me.

According to Geekerwan's SPEC2017 numbers, Qualcomm's performance cores have about 2/3 the performance of the prime cores. Assuming that carries over to Geekbench;

Oryon Gen 3Clock SpeedGeekbench 6
Prime4.6 GHz~3800
Performance3.6 GHz~2500

Comparing that to Apple;
M5 CPUClock SpeedGeekbench 6
Super4.6 GHz~4300
Performance4.4 GHz~3000
Efficiency3 GHz~1700

Apple's performance cores are 20% faster than Qualcomm's. That appears to be thanks to it's ~20% faster clock speed, which means the IPC of both is nearly identical.

Now, the remarkable thing is that Qualcomm's performance core is only 4 wide, compared to Apple's 7 wide.

Takeaway is that the Oryon Performance cores are much stronger than I assumed. Could Qualcomm replicate Apple's strategy with the M5 Pro/M5 Max in a future Snapdragon X CPU? Easily. They already have the hardware in the bag.
Click to expand...
Not sure I understand your closing argument. Apple’s Super are much faster than Oryon’s Prime. Apple’s Performance are much faster that Oryon’s Performance. It’s not like QC can simply turn up the clocks on the performance cores to catch up.
 
Jimmyjames said:
I thought Oryon’s cores were clocked higher. Up to 5 ghz I believe. Happy to be corrected. I don’t think IPC is equal between Apple and QC although they are much closer than anyone else.

Edit. Sorry I misread. I thought you were talking about the Prime cores. It’s late here!
Click to expand...
Clarification; The numbers for Oryon above are based on the mobile chip (8 Elite Gen 5), which has a peak ST frequency of 4.6 GHz.

The PC part goes upto 4.7 GHz for the X2 Elite, or 5 GHz for the X2 Elite Extreme.
Screenshot6667r.jpg


Cmaier said:
Not sure I understand your closing argument. Apple’s Super are much faster than Oryon’s Prime.
Click to expand...
+5% doesn't count as "much faster", surely?

CPUClockGeekbench6
M5 Max4.6 GHz4200+
X2 Elite Extreme5 GHz4000+
Cmaier said:
Apple’s Performance are much faste that Oryon’s Performance. It’s not like QC can simply turn up the clocks on the performance cores to catch up.
Click to expand...
That's fair. +20% is indeed a substantial difference.
Uyy7uTube.jpg

Source (Geekerwan)
Apple's Performance cores peak at 5W+ (according to the image I posted previously), whereas Qualcomm's peak at about 3 Watts (according to the above graph). So it does seem like Qualcomm has headroom to push clocks and raise performance. Though it's not going to be as simple as turning a knob, and they'll have to rearchitect their core to clock higher.
 
MerryCherry said:
Source (Geekerwan)
Apple's Performance cores peak at 5W+ (according to the image I posted previously), whereas Qualcomm's peak at about 3 Watts (according to the above graph). So it does seem like Qualcomm has headroom to push clocks and raise performance. Though it's not going to be as simple as turning a knob, and they'll have to rearchitect their core to clock higher.
Click to expand...
It’s not very easy to do that. Assuming they don’t switch to a new process, it’s likely that they would have to add pipe stages in order to clock higher. When you add pipe stages, you (1) need bigger reorder buffers; (2) complicate the scheduler; (3) you pay a higher penalty each time you have a cache or TLB miss, an exception, etc. Of course, you can try to increase cache sizes to compensate, or make the memory bus wider, or try and improve your branch prediction algorithm, etc. But each of those things means you are burning more power and increasing the size of units, which means some critical paths might be physically bigger, which means your cycle time might increase (slowing your max clock speed), so you can compensate by adding more pipe stages…

Nobody is increasing performance 20% on the same process without substantial redesign. Typically, when we did what we called “spins” (new versions of chips on the same process for bug fixes, performance improvements, etc), we could maybe get 2-5% clock improvement without having to really mess with things.
 
RockRock8 said:
I think M5 Super core has reached 4350+ consistently barring indexing
Click to expand...
Seems kind of odd that the S core, nothing more than a renamed P core, gets slightly better SC. Probably because the Pro and Max configurations have more memory bandwidth and/or perhaps expanded L2s?
 
Yoused said:
Seems kind of odd that the S core, nothing more than a renamed P core, gets slightly better SC. Probably because the Pro and Max configurations have more memory bandwidth and/or perhaps expanded L2s?
Click to expand...

Better SC as compared to what?
 
Yoused said:
Better than the base M5 in the old MBP, which has S cores but they were called P cores back when.
Click to expand...

That’s pretty common across Apple generations - including the M4. Not 100% sure what’s going but my suspicion is that even though Apple ST doesn’t draw much power Apple still gives devices with the most robust cooling capabilities that little bit of extra time at the highest clocks. This is similar to when Geekerwan uses Liquid Nitrogen to cool an iPhone/iPad to bump up Geekbench scores.

You can see the pattern here:

Mac Benchmarks - Geekbench

browser.geekbench.com browser.geekbench.com
 
Jimmyjames said:
A few more X2 Elite Extreme Geekbench scores coming in now.

Geekbench Search - Geekbench

browser.geekbench.com browser.geekbench.com
View attachment 38445
Click to expand...
Reviews must be coming soon. :) Based on this measure, the X2 Elite Extreme competes in performance with a binned M5 Pro, that makes sense based on CPU/GPU core counts/performance and memory bandwidth. Of course its primary competition will x86 PC laptops like Arrow Lake HX and Panther Lake-H and Strix Point/Halo. Let's see what pricing & reviews bring and if WoA has improved enough.
 
dada_dave said:
Reviews must be coming soon. :) Based on this measure, the X2 Elite Extreme competes in performance with a binned M5 Pro, that makes sense based on CPU/GPU core counts/performance and memory bandwidth. Of course its primary competition will x86 PC laptops like Arrow Lake HX and Panther Lake-H and Strix Point/Halo. Let's see what pricing & reviews bring and if WoA has improved enough.
Click to expand...
I wanted to do a violin plot comparing this to the M5 or M4. the json doesn’t include actual cpu speeds weirdly. Reporting “0” for most fields that matter. I believe J Poole said QC had been working with Primate Labs.

Edit. OK he says they have issues measuring cpu frequency on WoA.
 
Last edited:
Back
Top