Also @The Flame how did you find this place? Lol.
Nuvia: don’t hold your breath
- Thread starter Cmaier
- Start date
I think people here were underestimating how competent these guys are just because Apple is still good, while skepticism of Qualcomm is way higher than deserved — AMD/Intel are the laggards and have been for some time. Now sure, Qualcomm can act as a brick wall against Nuvia and maybe they’re too artisanal or whatever for them, but by all evidence they are not and did not want to, whereas with Intel internal fights and catabolism is just Tuesday and reportedly part of why they didn’t sell to Intel (don’t quote me tho).
But to some degree as I pointed out, their obstacles were one-time, the lawsuit, or getting in the groove of things with the X Elite which is probably why power wasn’t as good as we expected and there are multiple rumors and speculation on that. The fact that the turnaround was so significant and catapulted them ahead of the X925 for now is big.
It’s a real ball game for now.
Things do take time, so holding your breath from 2021 wasn’t wise and Cliff’s skepticism was warranted for sure, and the initial product was disappointing for the time (just less so relative to the cost and scalability imo). So in a way he was right
.
But from the point we saw they were shipping product with a reasonable advantage on Intel/AMD and then putting them in *phones* right afterwards — that was a huge tell to me. They couldn’t put something in that was objectively worse or at least not a little better and we knew where the X925 and such were going.
But to some degree as I pointed out, their obstacles were one-time, the lawsuit, or getting in the groove of things with the X Elite which is probably why power wasn’t as good as we expected and there are multiple rumors and speculation on that. The fact that the turnaround was so significant and catapulted them ahead of the X925 for now is big.
It’s a real ball game for now.
Things do take time, so holding your breath from 2021 wasn’t wise and Cliff’s skepticism was warranted for sure, and the initial product was disappointing for the time (just less so relative to the cost and scalability imo). So in a way he was right
.But from the point we saw they were shipping product with a reasonable advantage on Intel/AMD and then putting them in *phones* right afterwards — that was a huge tell to me. They couldn’t put something in that was objectively worse or at least not a little better and we knew where the X925 and such were going.
Okay maybe I'm stretching things but Qualcomm fiscal quarters start the previous year, so maybe this leak they mean H1 2026 could be September/October 2025. Maybe? Seems unlikely since they'd probably say Q1 rather than H1, but I dunno. Even September still seems late given the cores are available now and there was a report Glymur was already being tested a few weeks ago. But maybe they want to give themselves more time to get the software/drivers/accompanying SOC hardware right? Everyone thought highly of the CPU cores themselves in the last generation, a little late though they were, but it's everything else that needs the biggest boost. So maybe they're giving themselves more time to get that stuff right? Also it's possible if/when Nvidia-Mediatek releases WoA things will improve faster on the software compatibility front.Well, conflicting rumors. You may be right now though.
Qualcomm Preparing Two Snapdragon X Elite Successors For 2026, With One SoC Rumored To Be Tested With An 18-Core CPU And Full Ray Tracing Support
Two Qualcomm chipsets that will serve as the successors to the Snapdragon X Elite with up to 18 cores are being tested for a 2026 launchwccftech.com
Some report it’s still 2026 H1 and with Oryon V2, but that IP doens’t make sense. Oryon V2 is already shipping today (well very soon) at mass production. If it launched at the same time phones did or after, it would be a generation behind.
My guess is that it’s either early 2025 or summer 2025 and V2, or it’s V3 and late 2025/early 2026.
I had to gloat, but
FWIW, literally have an iPhone 16 Pro. Apple is still ahead, and the A18 and A18 Pro are still more impressive engineering, but it’s closer than ever before.
The main reason again I am excited is I don’t think Arm can get it done with their designs to the extent Apple can and expect that momentum will continue on Qualcomm’s end for Windows and for phones.
They have area efficiency (Apple does too actually), scalability, performance per watt and in the next generation I think they’ll get a lot closer on IPC, and with their clocking and efficiency much like Apple, I expect that means huge ST for Windows laptops without blowing power or area up as much as AMD and Intel, and next time they’ll have real E Cores.
I very much agree the addition of M-cores should be a big big boost - and the possibility of even S-cores if they have those too (although M-core die area seems very good already so I don't think they'll actually need S-cores, it's just the nomenclature is suggestive of those). Also glad to see two different SOCs being developed. They needed that as well. The original Elite was trying to cover too much ground with one SOC and no small cores. Doesn't seem to be any word on the GPU sizes for the two dies? But a 6+6 CPU could be a fanless M4 competitor - if not then maybe they'll have a binned version of it, say a 4+6
, that is appropriate for fanless designs. Qualcomm (and Microsoft!) have got to execute better though - no more canceled dev kits, more native apps, better Prism, better dev support, better drivers. I do agree though that the Qualcomm CPU designers are doing their part.
That lawsuit is getting worse though. I very much doubt ARM and Qualcomm will go full nuclear, but that could still be a dampener going forwards.
The Flame
Power User
- Joined
- Oct 28, 2024
- Posts
- 60
Cristiano Amon hinted during this year's Snapdragon Summit Day 1 Keynote that they have some big things to be announced at next year's Snapdragon Summit. My guess;
Snapdragon 8 Elite Gen 2
Snapdragon X Elite Gen 2
Both with next generation Oryon G3 cores (Pegasus-L and Pegasus-M), and crafted on TSMC N3P process.
______
Small nitpick: I think it's best we call it Oryon G3 (Oryon Gen 3), not Oryon V3 (Oryon Version 3).
The 'V' naming scheme came from that leaked Dell 2023 slide, and we now know that's clearly wrong.
Qualcomm is officially calling them Oryon 1st Gen, Oryon 2nd Gen, etc...
Yeah I just have a hard time seeing it be V2 in September next year. But here’s the thing. Even if it is, it will still be a very, very high performance and good laptop and they have frequency/power headroom.
But I really expect V3. If Glymur is being tested right now then based off some kind of 8-12 month range for their generation time then V3 makes total sense
I very much agree the addition of M-cores should be a big big boost - and the possibility of even S-cores if they have those too (although M-core die area seems very good already so I don't think they'll actually need S-cores, it's just the nomenclature is suggestive of those). Also glad to see two different SOCs being developed. They needed that as well. The original Elite was trying to cover too much ground with one SOC and no small cores. Doesn't seem to be any word on the GPU sizes for the two dies? But a 6+6 CPU could be a fanless M4 competitor - if not then maybe they'll have a binned version of it, say a 4+6
it will be 6+6 and 12+6 by all reports yeah. And the Graphics will be good this time, with bigger sizing and a bigger bus for the 12+6 model.
The Oryon first generation without small cores was just really impressive though RE battery life. Think about what QC is going to do to Intel with an arch upgrade on sheer performance and even more battery life, probably still with reasonably competitive area and cost. I actually think we will see a breaking point and the gaps will be worse than now where they are similar thanks to LNL being more costly and big etc.
Agree but I mean realistically Nvidia will be the thing that pushes that through along with QC getting the GPU hardware and drivers to C+ or B-, right now I’d give it a fail in X Elite.
Agreed, but I mean it seems like it’s just negotiation. There’s not much to dampen — they’ll do bit higher royalties and still save money a one time payment or the whole show is over and QC calls it quits. Kind of an all in all out thing imo
These look very similar to my own results and conclusions from similar analyses:
CB R24
Die shot analyses 1 and 2
Though as @mr_roboto pointed out, I got fooled by Intel's renaming of L1d to L0 and yet keeping a seemingly larger L1d which is more like L1.5d around.
Really? On Linux or Windows too?
My worry is that Qualcomm's partners will be reticent to release designs until the negotiations are done, potentially pushing back the release. Their partners won't want to release designs that could potentially get yanked even if that possibility is small.
I suppose if they are waiting for Gen 3 cores to release that could explain it, but there was a report that Glymur at least was already being tested. That could be wrong of course. It's just seems like 2026 is an awful long time to wait.
Yeah, but the 8 Elite has been tested for like 9 months now in leaks, seems plausible to me they’re already on G3, but who knows.
The Flame
Power User
- Joined
- Oct 28, 2024
- Posts
- 60
Development timeliness for smartphone SoCs are shorter than that for PC SoCs.
IIRC, silicon needs to be sampling to laptop OEMs atleast 12 months before it hits the market. Whereas, for smartphones it's about 6 months.
There's rumours that Intel is already testing Panther Lake, and that chip is slated for a 2025Q4 release.
I expect Snapdragon X Elite Gen 2 will be announced in 2025 Snapdragon Summit (October 2025), with laptops hitting the market in early 2026H1.
Jimmyjames
Elite Member
- Joined
- Jul 13, 2022
- Posts
- 1,094
Weird story. I wonder what is going on here? It will be interesting to see what happens with other phones as retail units come out.
www.androidauthority.com
Our first hands-on with a real-world Snapdragon 8 Elite phone reveals a hot mess
We've got our hands on one of the first Snapdragon 8 Elite phones, and benchmarks reveal some very concerning results.
www.androidauthority.com
It’s probably Xiaomi-specific. QC doesn’t do this sort of thing in software, but OEMs will and Android lets them do more than they otherwise couldWeird story. I wonder what is going on here? It will be interesting to see what happens with other phones as retail units come out.
Our first hands-on with a real-world Snapdragon 8 Elite phone reveals a hot mess
We've got our hands on one of the first Snapdragon 8 Elite phones, and benchmarks reveal some very concerning results.
View attachment 32388
View attachment 32389
Yep. This is probably the IPC upgrade GWIII mentioned. Frankly, it’s incredible the turnaround they had to 8 Elite.
Also I suspect this means the laptops will have the IPC upgrade & V3 tbh. Would be surprised if they don’t. Not for launching in 2025 Winter to really H1 2026. The difference will be that 3-9 month delay whereas more smartphones will launch immediately afterwards in October.
God the copium about this is so funny because it’s going to get worse. I kept saying that, while it isn’t certain obviously and companies can BS — those phone numbers — 4-4.4GHz casually without big IPC gains suggested huge power changes and yeah pretty much even more than I expected. When you look at GB6 curves, 8 Elite is doing 2500 at 3W or 3000 at 5.5W.
As for SpecInt (better indicator to some degree, FP is irrelevant largely), just like the X925 it hit an ~ 8 at 4W which is really the peak of the curve tradeoff. The 8.8-9 at 6-7.5W is insane and not worth it.
For reference Intel’s Lunar Lake does about an 8 in Geekerwan’s tests in SpecInt but at ~ 13-15W ish (again platform.)
Removing on-package from the phone and widening the bus would push the numbers up slightly, but not that much here, and we know Intel ain’t doing On-Package anymore with PTL so lol.
Suspect something about the fabric was a rush on the X Elite, power delivery is what I’ve heard. Arch and cache changes don’t at all save power like that they made (L1 reduction just allows higher clocks without blowing up area or other tweaks to it I think).
As for SpecInt (better indicator to some degree, FP is irrelevant largely), just like the X925 it hit an ~ 8 at 4W which is really the peak of the curve tradeoff. The 8.8-9 at 6-7.5W is insane and not worth it.
For reference Intel’s Lunar Lake does about an 8 in Geekerwan’s tests in SpecInt but at ~ 13-15W ish (again platform.)
Removing on-package from the phone and widening the bus would push the numbers up slightly, but not that much here, and we know Intel ain’t doing On-Package anymore with PTL so lol.
Suspect something about the fabric was a rush on the X Elite, power delivery is what I’ve heard. Arch and cache changes don’t at all save power like that they made (L1 reduction just allows higher clocks without blowing up area or other tweaks to it I think).
That said, while I think Qualcomm is really going to give us Store Brand (and that’s kind of too harsh at this pace but IMO no one else has) Apple SoCs next year and on —
I think the most impressive chip this year is the A18 & the A18 Pro, but specifically the A18.
Why?
Apple’s area actually isn’t that ridiculous at all if you include AMD/Intel whole cluster area instead of cherry picking private L2 for one core without L3, which is stupid due to their setups which inherently differ.
But even so, A18’s 8MB of L2 — same as the A14 — and 12MB of SLC — less than the A14’s 16MB — showed us just how far they can stretch it and be just fine for general ST and a phone. The results probably wouldn’t even be statistically significant across enough Spec & GB tests in terms of power.
Even the M2/3/4 show this basically, 16MB of L2 among 4 P cores, 4MB for E cores and an 8MB SLC. Pretty consistent. That’s…. Not that much.
Only thing I’d say is I suspect if you got some variant programs going back and forth in memory where the cache was occupied actively on the A18 Pro vs A18, you would probably be able to tease out some kind of hit to power/performance on an application if the other core was running hot with a larger working set (with 8MB vs 16MB L2 and 12 vs 24 MB of SLC.) Whether or not it’s enough to matter in a phone is a separate issue, and I suspect the answer is broadly no, and yes for some specific tasks (See media manipulation on the nicer iPhones) where Apple thinks the extra cache helps. Has to be something specific.
I think the most impressive chip this year is the A18 & the A18 Pro, but specifically the A18.
Why?
Apple’s area actually isn’t that ridiculous at all if you include AMD/Intel whole cluster area instead of cherry picking private L2 for one core without L3, which is stupid due to their setups which inherently differ.
But even so, A18’s 8MB of L2 — same as the A14 — and 12MB of SLC — less than the A14’s 16MB — showed us just how far they can stretch it and be just fine for general ST and a phone. The results probably wouldn’t even be statistically significant across enough Spec & GB tests in terms of power.
Even the M2/3/4 show this basically, 16MB of L2 among 4 P cores, 4MB for E cores and an 8MB SLC. Pretty consistent. That’s…. Not that much.
Only thing I’d say is I suspect if you got some variant programs going back and forth in memory where the cache was occupied actively on the A18 Pro vs A18, you would probably be able to tease out some kind of hit to power/performance on an application if the other core was running hot with a larger working set (with 8MB vs 16MB L2 and 12 vs 24 MB of SLC.) Whether or not it’s enough to matter in a phone is a separate issue, and I suspect the answer is broadly no, and yes for some specific tasks (See media manipulation on the nicer iPhones) where Apple thinks the extra cache helps. Has to be something specific.
Also feel pretty vindicated saying Lunar Lake wasn’t impressive at all architecturally for the cost involved. The party looks so funny lol.
Intel can only pull that BS for as long as X86 software is a big deal for developers creators and gamers on Windows. It will erode thanks to MS, QC and soon & most importantly Nvidia next year, just taking time.
New ASIO low-latency audio drivers for Arm64 Windows is huge, relatedly.
Intel can only pull that BS for as long as X86 software is a big deal for developers creators and gamers on Windows. It will erode thanks to MS, QC and soon & most importantly Nvidia next year, just taking time.
New ASIO low-latency audio drivers for Arm64 Windows is huge, relatedly.
The Flame
Power User
- Joined
- Oct 28, 2024
- Posts
- 60
Well, Qualcomm has to deliver a big upgrade with Oryon G3 to make X Elite G2 competitive.
The CPU (it's performance and efficiency) is really going to be their differentiating factor.
GPU isn't Qualcomm's strong point on PC. Intel/AMD/Nvidia are going to have better GPUs. (Oh yeah, there was a rumour that the Nvidia laptop SoC is coming in September 2025)
NPU could be anyone's ball. Pretty much impossible to guess how that will go next generation.
Qualcomm does have some strengths such as their mobile-derived ISP and their connectivity solutions (BT/WiFi/5G), but these will simply be cherries on top. The cake has to be made from the Oryon CPU.
The CPU (it's performance and efficiency) is really going to be their differentiating factor.
GPU isn't Qualcomm's strong point on PC. Intel/AMD/Nvidia are going to have better GPUs. (Oh yeah, there was a rumour that the Nvidia laptop SoC is coming in September 2025)
NPU could be anyone's ball. Pretty much impossible to guess how that will go next generation.
Qualcomm does have some strengths such as their mobile-derived ISP and their connectivity solutions (BT/WiFi/5G), but these will simply be cherries on top. The cake has to be made from the Oryon CPU.