Discussion Qualcomm Snapdragon Thread

Page 162 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

SpudLobby

Senior member
May 18, 2022
991
684
106

If it’s true 8 Gen 4 is running at around 4GHz or more (and again even just 4GHz is high in this case given Oryon right now) doing so under 6W for GB and SpecInt, and is architecturally mostly similar enough to Oryon & on N3E —

Then we know something was f*cked up with X Elite’s power delivery or physical design and fabric.

Moving to N3E will give them a bit of headroom like + 5-10% more frequency iso-power from currently but that’s not enough to get them into shape with what they’re currently doing lol. Absolutely not enough.

Anyway if they can pull off A17 Pro ST at similar power without using more (P core anyway) L2 [16MB] & SLC [24MB] than Apple, that’s amazing and means something really was sloppy about X Elite. But right now I don’t have those expectations if I had to gamble, just not impossible that they improve some lowish hanging fruit they blew up.
Well this aged well


Though they made architectural changes too but still
 

FlameTail

Diamond Member
Dec 15, 2021
4,238
2,592
106
Another possibility is they have more than one configuration and that X Elite Gen 2 will have both an M4/5 and an M4/5 Pro competitor. The leaked Dell roadmap mentioned an "Oryon V2" and "Oryon V2 UI", the latter of which targeting a 40W TDP and shipping in the next gen XPS 14 which will also feature an Intel PTL-P SKU with 12 Xe3 cores. And the year after has Oryon V3 appearing in an 80W flavor in the XPS 16 which could be a Max competitor.

Maybe Oryon V2 UI is just a higher TDP version of Oryon V2, I'm not actually sure what the rumors are on that front yet, but if instead its another chip then your option C with higher clocks would fit right in there.
Mahua and Glymur;
My source just updated me with XE G2 roadmap which are supposedly launching in H1 2026. There are two SoCs planned:

1. SoC codenamed Glymur:
  • Higher tier than Hamoa
  • 18 cores CPU (6L+6L+6M), cluster of six
  • 192-bit LPDDR5x (don't know final speed yet)
  • Full DX12U features including hardware RT

2. SoC codenamed Mahua
  • Hamoa successor
  • 12 cores CPU (6L+6M), cluster of six
  • 128-bit LPDDR5x (don't know final speed yet)
  • Full DX12U features including hardware RT
That's all I know atm, at least we have an idea what Qualcomm are planning in 2026. Again, treat this as rumor until real thing happens.
 

Magio

Member
May 13, 2024
104
111
76
Given the 12 full Oryon cores, I see the X Elite as M3 Pro competition. Whatever the V2 replacement for that chip is, beating the base M GPU is not a worthy goal if they plan to not support discrete GPUs. They are cutting off a large section of the market that would appreciate the multithreaded performance if it only has base thin and light GPU power.

Yeah on the CPU side the X Elite was more of an M3 Pro competitor but the lackluster GPU even compared to M3 base made it an unbalanced comparison. I think that's part of why Qualcomm targeted their marketing at M3 base.

GPU is definitely where X Elite gen 2 needs to impress and I think going with a base and pro SKU is one way to do that.

Mahua and Glymur;

Glymur sounds like what you'd want and expect from a higher tier X Elite, let's hope the GPU matches the CPU this time. (And that Windows on ARM improves and/or that Linux support arrives more promptly.)
 

Raqia

Member
Nov 19, 2008
63
32
91
Well this aged well


Though they made architectural changes too but still
Rumor has it they didn't implement granular DVFS for the original SDX Elite and just power gated entire clusters. They also reduced the L1i to 128kb, presumably for efficiency, and they have a bit of a deficit in L2 capacity to the Apple A18 CPU cluster; both places they could boost for Glymur...

I think the SDX Elite would have been much more striking had it been released in '23 as was their original plan (as Amon said on prior interviews...) but their dispute with ARM and the redesign of relevant blocks took a while.
 
Last edited:

FlameTail

Diamond Member
Dec 15, 2021
4,238
2,592
106
They also reduced the L1i to 128kb, presumably for efficiency,
Yeah, I found that quite intriguing. I guessed they did it in order to keep L1i latency constant. Because Oryon-L clocks higher than Oryon, the memory latency in terms of clock cycles would regress (unless they made an improvement in Oryon-L to the absolute latency in terms of nanoseconds).
and they have a bit of a deficit in L2 capacity to the Apple A18 CPU cluster; both places they could boost for Glymur.
Looking at the M4 Pro with it's 10P+4E CPU, it seems to me that Glymur (12L + 6M) will be more of an M5 Pro competitor, not M5 Max competitor, CPU wise.

On the GPU front I am not sure. Glymur has 192 bit LPDDR5X, whereas M4 Pro has 256 bit LPDDR5X. So I think M5 Pro could outstrip Glymur in terms of GPU performance.

I find Glymur to be a bit unusual. It will be Qualcomm's halo chip for the Snapdragon X2 generation, no doubt. But I feel like they did not fully commit to the idea. Why not go for a 256b memory bus, to put in an even larger GPU and to head-to-head with M5 Pro/Strix Halo ?

But there's the dilemma of who will but it for the large GPU. The Adreno X2 GPU architecture will be an improvement over X1 no doubt, but it will be still behind Nvidia/AMD for gaming and Nvidia/Apple for professional workloads.

Mahua on the other hand looks like it's shaping up to be an excellent SoC for thin-and-light laptops. 6L + 6M CPU and a good GPU hopefully....

Again, this is going by the leak- assuming it is true. It might not be.
 
Reactions: Ghostsonplanets

Raqia

Member
Nov 19, 2008
63
32
91
Yeah, I found that quite intriguing. I guessed they did it in order to keep L1i latency constant. Because Oryon-L clocks higher than Oryon, the memory latency in terms of clock cycles would regress (unless they made an improvement in Oryon-L to the absolute latency in terms of nanoseconds).

Looking at the M4 Pro with it's 10P+4E CPU, it seems to me that Glymur (12L + 6M) will be more of an M5 Pro competitor, not M5 Max competitor, CPU wise.

On the GPU front I am not sure. Glymur has 192 bit LPDDR5X, whereas M4 Pro has 256 bit LPDDR5X. So I think M5 Pro could outstrip Glymur in terms of GPU performance.

I find Glymur to be a bit unusual. It will be Qualcomm's halo chip for the Snapdragon X2 generation, no doubt. But I feel like they did not fully commit to the idea. Why not go for a 256b memory bus, to put in an even larger GPU and to head-to-head with M5 Pro/Strix Halo ?

But there's the dilemma of who will but it for the large GPU. The Adreno X2 GPU architecture will be an improvement over X1 no doubt, but it will be still behind Nvidia/AMD for gaming and Nvidia/Apple for professional workloads.

Mahua on the other hand looks like it's shaping up to be an excellent SoC for thin-and-light laptops. 6L + 6M CPU and a good GPU hopefully....

Again, this is going by the leak- assuming it is true. It might not be.
I think Glymur is well sized for what software it is likely to be running: NOT AAA games.

There is a quite a bit of enthusiasm for running games through emulation on ARM platforms as evidenced by plenty of twitter and youtube posts as well as development on the ARM branch of Valve's Proton, but just like general Windows on ARM prior to this year, there's not enough support for gaming for Valve to field a Steam Deck 2. Until more devs make ARM builds of games on Windows (or Linux), this is likely to be a non-issue; when they do they can cart out the big automotive SoC with its rumored 512-bit memory interface...

Another possibility, they release discrete Adreno GPUs into the shark infested waters of nVidia and AMD (as Intel bravely did) to battle test their GPU driver and improve adoption. (Quick thought: a good value add may be to release a GPU+NPU SoC on the same card riding a massive memory interface with some kind of SLI-like option for enthusiasts to tie multiple cards together. With their low power pedigree, this is much more feasible than with power guzzling nVidia cards.)
 
Last edited:

naukkis

Senior member
Jun 5, 2002
962
829
136
They also reduced the L1i to 128kb, presumably for efficiency

Did they? I only see Qualcomm powerpoint where they say prime cores have 192KB L1i$ and efficiency cores have 128KB. Data caches still pretty much unknown size.
 

FlameTail

Diamond Member
Dec 15, 2021
4,238
2,592
106
Did they? I only see Qualcomm powerpoint where they say prime cores have 192KB L1i$ and efficiency cores have 128KB. Data caches still pretty much unknown size.
That Qualcomm slide is the only one which suggests Oryon-L has 192 KB L1i cache. That's probably a typo, because;

(1) Geekerwan did a microarchitectural analysis and found that Oryon-L has 128 KB L1i.


(2) Die shot of Snapdragon 8 Elite from Kurnal suggests the following cache sizes;

Oryon-LOryon-M
L1i128 KB64 KB
L1d128 KB64 KB


(3) There's another Qualcomm slide showing 128 KB L1 capacity;
 
Reactions: SpudLobby

FlameTail

Diamond Member
Dec 15, 2021
4,238
2,592
106
The first phones with Snapdragon 8 Elite have landed in the market!
Another upgrade this year is a 5,400mAh silicon battery, which is a significant increase from the Xiaomi 14’s 4,610mAh capacity.
The Pro model earns its moniker thanks to several premium additions, starting with a 6,100mAh battery. That’s a notable upgrade over the 14 Pro’s 4,880mAh battery.
Love to see the much improved battery capacities in the upcoming generation of smartphones. Apparently this is enabled by new battery technology of adding silicon/carbon.
 

FlameTail

Diamond Member
Dec 15, 2021
4,238
2,592
106

Some people are speculating based on this news that they might be preparing to bring 8 Elite to laptops. I think that's highly unlikely.

I think it's more probable that they are putting the Adreno 830 out there, so they can already start testing it and building the drivers. So Snapdragon X2 with Adreno 8 series based GPU can have a decent GPU experience at launch.

Redditor comment:
The news is Qualcomm is enabling support for 8 Elite in mainline Linux (ie. the version desktop Linux runs on, not the version bundled in Android) already, which AFAIK they haven't really done for their mobile chips before. On paper, this would make it much easier to get something like an ARM version of SteamOS running on 8 Elite devices than it would have been on prior Qualcomm phone chips, which would make actual PC gaming possible.
Few caveats though, as the Linux support Qualcomm is talking about is far from complete at this stage and due to how phone bootloaders work just having support from the chip doesn't mean a phone will actually be capable of installing a Linux distro. Plus while x86 > ARM translation is already possible on Linux even for games including some intensive ones, today there isn't really a distro that makes this a straightforward process. Rumors that Valve is working on SteamOS for ARM, but it's not there yet.
 
Reactions: moinmoin

FlameTail

Diamond Member
Dec 15, 2021
4,238
2,592
106

Die Area of the various components of the 8 Elite

ComponentArea (mm²)
SoC124
Oryon-L core2.2
Oryon-M core0.88
12 MB L25.57
GPU21.9
ISP9.21
NPU9.18
DPU3.87
VPU5.86
64 bit LPDDR5X3.66
Modem12.0

DPU = Display Processing Unit
VPU = Video Processing Unit (Encode/Decode)

Measurements are based on the dieshot published by Kurnal, and his labelling of the sub-components. Also, I cannot vouch that these numbers are 100% accurate, so treat then as estimates.
Based on 8 Elite, we can construct a hypothetical X Elite 2;

8 Elite X Elite 2
CPU22 mm²

2L + 6M
12 MB + 12 MB

Oryon G2
 ARMv8
44 mm²

6L + 6M
18 MB + 12 MB

Oryon G3
ARMv9 + SME
GPU22 mm²

12 CU
3 slices
12 MB GMEM
Adreno 8
44 mm²

24 CU
6 slices
24 MB GMEM
Adreno 8 (X2)
NPU9 mm²18 mm²
ISP9 mm²6 mm²
VPU6 mm²12 mm²
DPU4 mm²

1 × 4K
1× 8K
8 mm²

1 × 4K
3 × 8K
Modem12 mm²0
Memory controller4 mm²

64 bit LPDRR5X
8 mm²

128 bit LPDDD5X
SLC 8 MB8 MB
Total SoC124 mm²

Getting the sum of the added component areas of the hypothetical X Elite 2 SoC, and adding it to the 8 Elite, we get 176 mm².

The X Elite 2 will also have some additional I/O, so that would add about 5 mm². But X Elite 2 will be on N3P, which is 4% denser than the N3E used on 8 Elite. So they cancel each other out.

For ~175 mm² N3P, they can make a great laptop SoC with an efficient and powerful GPU.
 

jdubs03

Golden Member
Oct 1, 2013
1,079
746
136
That's in INT. FP IPC should be better.
True but still, it’s quite close (about 6% higher using peak frequency for comparing both Int and Fp of Phoenix-L to the X4 in the 9300+).

On another note, the scores represented there in terms of pts/GHz for all seem to be higher than if one were to divide by peak frequency (dividing by peak frequency helps Phoenix-L). I’d be curious to know what the frequencies were when these scores were reached.

Well, good luck getting X4 to run over 4 GHz. IPC is just part of the equation.
Indeed, but when comparing it like for like, the X4 at least efficiency wise does pretty well. The proper comparison is the X925 anyway, and we know what’s efficiency compared to that is less on both Int and Fp.
 

Hesperax

Member
Nov 13, 2023
31
66
51
True but still, it’s quite close (about 6% higher using peak frequency for comparing both Int and Fp of Phoenix-L to the X4 in the 9300+).

On another note, the scores represented there in terms of pts/GHz for all seem to be higher than if one were to divide by peak frequency (dividing by peak frequency helps Phoenix-L). I’d be curious to know what the frequencies were when these scores were reached.


Indeed, but when comparing it like for like, the X4 at least efficiency wise does pretty well. The proper comparison is the X925 anyway, and we know what’s efficiency compared to that is less on both Int and Fp.
For Phoenix-L:
 
sale-70-410-exam    | Exam-200-125-pdf    | we-sale-70-410-exam    | hot-sale-70-410-exam    | Latest-exam-700-603-Dumps    | Dumps-98-363-exams-date    | Certs-200-125-date    | Dumps-300-075-exams-date    | hot-sale-book-C8010-726-book    | Hot-Sale-200-310-Exam    | Exam-Description-200-310-dumps?    | hot-sale-book-200-125-book    | Latest-Updated-300-209-Exam    | Dumps-210-260-exams-date    | Download-200-125-Exam-PDF    | Exam-Description-300-101-dumps    | Certs-300-101-date    | Hot-Sale-300-075-Exam    | Latest-exam-200-125-Dumps    | Exam-Description-200-125-dumps    | Latest-Updated-300-075-Exam    | hot-sale-book-210-260-book    | Dumps-200-901-exams-date    | Certs-200-901-date    | Latest-exam-1Z0-062-Dumps    | Hot-Sale-1Z0-062-Exam    | Certs-CSSLP-date    | 100%-Pass-70-383-Exams    | Latest-JN0-360-real-exam-questions    | 100%-Pass-4A0-100-Real-Exam-Questions    | Dumps-300-135-exams-date    | Passed-200-105-Tech-Exams    | Latest-Updated-200-310-Exam    | Download-300-070-Exam-PDF    | Hot-Sale-JN0-360-Exam    | 100%-Pass-JN0-360-Exams    | 100%-Pass-JN0-360-Real-Exam-Questions    | Dumps-JN0-360-exams-date    | Exam-Description-1Z0-876-dumps    | Latest-exam-1Z0-876-Dumps    | Dumps-HPE0-Y53-exams-date    | 2017-Latest-HPE0-Y53-Exam    | 100%-Pass-HPE0-Y53-Real-Exam-Questions    | Pass-4A0-100-Exam    | Latest-4A0-100-Questions    | Dumps-98-365-exams-date    | 2017-Latest-98-365-Exam    | 100%-Pass-VCS-254-Exams    | 2017-Latest-VCS-273-Exam    | Dumps-200-355-exams-date    | 2017-Latest-300-320-Exam    | Pass-300-101-Exam    | 100%-Pass-300-115-Exams    |
http://www.portvapes.co.uk/    | http://www.portvapes.co.uk/    |