Yep!It only matters for servers. If everyone using an ordinary PC switched to an equivalent Apple/Qualcomm ARM CPU (i.e. not those of you with Threadrippers and the like that don't have ARM analogs) it would be a rounding error on a rounding error.
You want to save the planet, develop more energy efficient ways to produce cement. That would prevent more CO2 emissions than shutting off every computing device on the planet.
AMD has a gaming market with full compatibility, great drivers, and will be using glue & chiplets for cost reasons. This is also a limited market and it won’t make sense for Qualcomm to go after that with their competitive advantages (or lack thereof).That's the whole point of a 256 bit bus. It's an entirely different class of SoC.
AMD is doing it with Strix Halo...
Honestly I would love to be wrong but the rumor was X Elite 2 launches in 2026 H1. It can’t be 2027 or they’re unserious but 2026 would be about 18-24 months from release which is fine.I am not fully convinced X Elite G2 will have LPDDR6.
It all depend on the timing. When will X Elite G2 launch, and how LPDDR6 availability be by then.
If it's not LPDDR6, it will quite likely be LPDDR5T/5X-9600 which is a step up from the LPDDR5X-8533 utilised by X Elite.
The idea is that QCOM isn't doing a Halo SoC like AMD. Very unlikely that they'll move from a 128-bit memory bus. Bandwidth improvements will come from LPDDR6 or LPDDR5T, whatever is available for Pegasus product cycle.That's the whole point of a 256 bit bus. It's an entirely different class of SoC.
AMD is doing it with Strix Halo...
Adreno 830 would be a big uplift over using the Adreno 740. New uArch, higher performance, and better perf/W.Snapdragon X Elite Gen 2 Rumored To Use An Overclocked Version Of The Adreno 830 GPU, Which Will Be Paired With The Snapdragon 8 Gen 4
Qualcomm is rumored to have kicked off development work on the Snapdragon X Elite Gen 2, and it may sport the same Adreno 830 GPU as the Snapdragon 8 Gen 4wccftech.com
Though idk. Maybe it launches 2025 but I doubt it
Ehhh. I am not sure if that's fine. X Elite is enough to dig the trench but not sufficient to hold it. Next gen Intel/AMD parts will certainly be more performant than X Elite, and they are coming in late 2024/early 2025. Qualcomm should iterate fast.Honestly I would love to be wrong but the rumor was X Elite 2 launches in 2026 H1. It can’t be 2027 or they’re unserious but 2026 would be about 18-24 months from release which is fine.
And LPDDR6, and I know why you keep forgetting itEhhh. I am not sure if that's fine. X Elite is enough to dig the trench but not sufficient to hold it. Next gen Intel/AMD parts will certainly be more performant than X Elite, and they are coming in late 2024/early 2025. Qualcomm should iterate fast.
I'd like Qualcomm to release X Elite G2 no later than 2025H2, same time as 8G5. It makes sense, as 8G5 will utilise N3P node and Oryon Pegasus CPU core.
Arrow Lake is terrible garbage fwiw and while Exist50 isn’t here he’d explain if he were, and it matches what everyone else has said. Not worried about that. Performance will be “better” barely and efficiency again still crappy and it’s an expensive mess.Ehhh. I am not sure if that's fine. X Elite is enough to dig the trench but not sufficient to hold it. Next gen Intel/AMD parts will certainly be more performant than X Elite, and they are coming in late 2024/early 2025. Qualcomm should iterate fast.
I'd like Qualcomm to release X Elite G2 no later than 2025H2, same time as 8G5. It makes sense, as 8G5 will utilise N3P node and Oryon Pegasus CPU core.
This person might be part of Snapdragon Insiders as he is really hyping up the product. I would avoid these people for they have extreme biases and very outlandish opinions. Qualcomm promotes this "membership".
One more reason why people should consider moving to power efficient ARM platforms.
It's better for the planet.
Shouldn’t need to be said but yeah that guy is a drone.This person might be part of Snapdragon Insiders as he is really hyping up the product. I would avoid these people for they have extreme biases and very outlandish opinions. Qualcomm promotes this "membership".
ExactlyThe idea is that QCOM isn't doing a Halo SoC like AMD. Very unlikely that they'll move from a 128-bit memory bus. Bandwidth improvements will come from LPDDR6 or LPDDR5T, whatever is available for Pegasus product cycle.
These numbers are seriously impressive, specially the latter 3 (Web Browsing, YT Streaming/Streaming and Videoconferencing), which are the most common workloads a laptop PC will face in the hands of the average user.I think some here are underrating just how much better the X Elite will be from an efficiency standpoint and with that, battery life and *responsiveness* (you can ramp to higher frequencies without killing the battery, or the same for less). People notice that.
This slide hasn’t made the rounds yet, but it’s recent:
Look at the 58% and 89% additional battery life Qualcomm gets for web browsing and Youtube over the Core Ultra 7.
They do better with local video playback too, about 40%, which is still humongous. But local video playback is also less common than streaming, and it’s ideal for some chips because it pads over the weaknesses in their CPUs — they just gate them or run minimally — but most people aren’t using laptops for local video as much as they are web browsing and mixed light loads, streaming, or editing code or video with a final compilation, and a lot of random bursty ST.
(Also: Ignore the Teams video conferencing lmao, as I suspect that’s using the NPU for something QC exclusive.)
The Intel Core Ultra 7 155H laptop used here is the Zenbook 14 UX3405, with a 3.2K OLED display (variable refresh) and a 75Whr battery, found here or here. Notice this laptop isn’t exactly doing poorly battery life wise.
Qualcomm has two Snapdragon reference laptops, and most *likely* (if they’re reasonable which lol) they are using the Thin and Light B config for the comparison, which has a 14.5-inch OLED 2880x1800 display, and a 58whr battery.
So conditional on Qualcomm using Laptop B: the X Elite with a similar OLED display is running up the score on an Asus/Intel laptop with a decent reputation for battery life — and the Qualcomm laptop even has a 22% smaller battery.
Exactly.These numbers are seriously impressive, specially the latter 3 (Web Browsing, YT Streaming/Streaming and Videoconferencing), which are the most common workloads a laptop PC will face in the hands of the average user.
Yep. Not far from Phoenix at all on a lot of stuff in similar laptops lol (but loses under load for efficiency).And MTL isn't a terrible SoC wrt battery life either, which makes these even more eye opening😯.
Yeah. I think it'll get close ish in some ways but idk.If we compare with the little amount of info we have so far from LNL (40% scenario power reduction), indeed it will be a tough battle for LNL to match.
Indeed. I myself once enthusiastically joined the Snapdragon Insiders discord server once, hoping to become an Insider myself!This person might be part of Snapdragon Insiders as he is really hyping up the product. I would avoid these people for they have extreme biases and very outlandish opinions. Qualcomm promotes this "membership".
Ah, The Fellowship of the ARM : Exist50, Vince789, RegularCircumstances, Twelvesilverswords, -protonsandneutrons-Arrow Lake is terrible garbage fwiw and while Exist50 isn’t here he’d explain if he were, and it matches what everyone else has said. Not worried about that. Performance will be “better” barely and efficiency again still crappy and it’s an expensive mess.
Aren't the leaked LNL SKUs so far topping out at like 3 GHz?LNL is redeemable ish as I explained elsewhere but probably isn’t going to be that great, it’ll be good relative to anything else Intel has or AMD has for the 5-20W range and immediate legacy compat + Intel design wins influence means that it’ll sell. It will just sell less than they would if they hadn’t, you know, lost Apple, let AMD come back, and have Qualcomm now lol.
I doubt LNL will match QC in peak performance on MT even for the 10c part and at best they might get close for ST if they have a 4++GHz SKU and QC has a Plus SKU that tops out at 3.8GHz. Efficiency much closer to QC than MTL but still skeptical. They need a literal 50-55% performance improvement iso-power in ST at multiple points.
ARE YOU INSANE?**What QC needs for gen 2 and what I expect:
N3E/P
A power and mild IPC tweak to the big cores (say single digits IPC), maybe slightly (.2-.4 GHz?) higher clocks.
Yes, we want FAT GPUs. But the rumour says that X Elite G2 will simply use an overclocked Adreno 830....Real littles for background tasks and an area efficiency boost.
A wider faster GPU, pretty straightforward
Same for the NPU obviously.
Ooh that looks good. Where are all the tech website journalists? Parked in the Zen5 thread? Write an article about this slide!I think some here are underrating just how much better the X Elite will be from an efficiency standpoint and with that, battery life and *responsiveness* (you can ramp to higher frequencies without killing the battery, or the same for less). People notice that.
This slide hasn’t made the rounds yet, but it’s recent:
Look at the 58% and 89% additional battery life Qualcomm gets for web browsing and Youtube over the Core Ultra 7.
The legend.I think they just measured the SoC/package + DRAM and VRM's directly for the entire duration of the tests, so it really is chip-to-chip - this is Andrei Frumusanu doing it and he has the tools to do so. They only mentioned the specific laptop - the Asus Zenbook 14, because of it's RAM/wireless/etc choices and any possible packaging stuff.
Well, I am impressed.But it still looks unreal for QC - this is sort of similar to how Apple has looked or AMD when Intel was on Alder Lake.
No offense, but PTL will be probably be a class below Z6 and M5, so QCOM won't be worried with it. If Lion Cove already won't be competitive with Z5 and Snap X Elite, much less will be Lion Cove+/Cougar Cove. Unless there's some serious low hanging fruits on Lion Cove design that can be leveraged.Panther Lake
TBH QCOM has been increasing the GPU size gen over gen. 8G2 was 1536 ALU, 8G3 was 1792 ALU. 830 might as well be 2048 or even bigger.Adreno 830
What matters is the performance at the end of the dayTBH QCOM has been increasing the GPU size gen over gen. 8G2 was 1536 ALU, 8G3 was 1792 ALU. 830 might as well be 2048 or even bigger.
Yup, feel the same. Especially skeptical they’ll be able to match it (the GB6 ST performance) at 2-10W. I will give them points if it’s really close, but I’d just be surprised.Ah, The Fellowship of the ARM : Exist50, Vince789, RegularCircumstances, Twelvesilverswords, -protonsandneutrons-
Aren't the leaked LNL SKUs so far topping out at like 3 GHz?
Also yes, Intel needs to make a massive leap with LNL to catch upto X Elite efficiency;
View attachment 96694
I am skeptical they will be able to do that.