Different speeds and possibly lower/higher power versions. Bin quality plays a role there. I don't think it makes sense to make two chips. The second one would have to be much smaller and at that point it would be budget mobile chip (or Chromebook chip at best).
Qualcomm Snapdragon X Elite & Plus SKU lineup leaks out - VideoCardz.com
Qualcomm preparing at least 8 variants of the Snapdragon X series Up to 12 CPU cores in X Elite. Display driver entries have revealed what Qualcomm is preparing for its Snapdragon X series launch. There are at least 8 SKUs named and at least 4 more yet undisclosed, according to this leak. The...videocardz.com
Do you guys think the Snapdragon X Plus is it's own die, or derived from the X Elite's die?
Ugh, can one company name their products in a sensible manner?
Er, didn't Qualcomm show off benchmarks where their GPU stomped AMD/Intel? But now its rated at about 4.5TF? I was expecting a GPU comparable to AMD's big APU coming later this year. Instead, its half of what Hawk Point is rated for. That's disappointing as I was hoping this chip would maybe offer some competition for gaming handhelds.
I realize that yes. But as for why, area. M3 is almost double (or 66% more) GPU cores than the iPhone 15 and M2 is double the 4/5c A15 stuff at 8/10. Apple does still raise clocks some ofc.
I don’t think Qualcomm added as many units, they added some I believe to the 740 but most of it seems like the result of basically doubling the frequency from the 660-750mhz range to 1.4GHz from the rumors, which brings us from that 2-2.2tflops range to the 4.6TFLOPs, albeit obviously perf/w will suffer a bit.
This explains what I had been grappling with a few months ago: Why did X Elite GPU consume 4x the power of the 8G2 GPU, while delivering only 2x the performance.
It explains it kind of, yeah, Adreno 740 looks like it’s what 1/4 to 1/7th of the die on the 8 Gen 2? Right? It’s already huge. Like if Qualcomm made it 66%% larger what’re they adding, 10-18mm^2?It also explains how Qualcomm managed to fit 12P cores, a 45 TOPS NPU and a performant GPU in a 172 mm² die.
Yeah it was obvious. They just thought “well we already have an efficient GPU arch and it’s good enough at 5-15W, so we’ll just add one GPU core(?) and gun it”.
Which honestly, that’s fine. Still better than Intel and AMD current gen and at worst will tie them in 10-15W stuff next gen, and I would be willing to bet still better below 10W.
And yep see they’re gonna do it again with the 830. Might have an extra few hundred ALUs but basically a supercharged one.
It explains it kind of, yeah, Adreno 740 looks like it’s what 1/4 to 1/7th of the die on the 8 Gen 2? Right? It’s already huge. Like if Qualcomm made it 66%% larger what’re they adding, 10-18mm^2?
Still would make the part a bit larger than Phoenix (like 188mm^2) assuming A740 is 20-30mm^2.
Samsung Galaxy Book Edge für 1800 Euro: Neue ARM-PCs werden teuer
Samsung bereitet derzeit die Einführung seines ersten Laptops auf Basis des neuen Qualcomm Snapdragon X Elite vor. Wir haben ein paar erste Details, wobei sich abzeichnet, dass die neuen "Apple-Killer"-Notebooks mit Windows 11 recht teuer werden.winfuture.de
1800 Euros for the X Elite in a Book 4 Edge (Edge being the Arm branding version) with a 16/512GB config.
An M3 in an MBP with that config is about 1759 Euros.
And the Intel version of this is 1900 Euros starting.
But most interesting is — what will the marginal pricing be like for more RAM/SSD?
Yeah and in practice I think this stuff does go on sale, the MSRP is kind of fake. All that matters to me is that the Intel version is like 100 euros more, tells us the Snapdragon itself probably isn't uniquely expensive or something.That should be... $1500 in the US?
“I think in this first phase, you’re going to see traditional chassis along the brand lines that these PC OEMs have already established,” McGuire commented. Following this he highlighted two products in particular: “So a Surface Pro is still a Surface Pro, and a Lenovo Yoga is still a Lenovo Yoga.”
Looking further afield, McGuire highlighted how this first phase will run through “the back to school holiday,” and then you’ll see OEMs “get into their next cycles.” What does this next cycle look like?
“It’s not just about the fact that our platforms are small and energy efficient, but it’s going to be about this idea of distributed compute,” McGuire ponders. This is something he spoke to us at length about when defining what an AI PC is — simply put, it’s not just one device, it’s multiple AI-enable devices coming together to form a anticipatory, proactive assistant in your life.
And sure, there are other more physical elements that McGuire does talk about, such as the miniaturization of the Snapdragon X Elite component meaning that laptop makers could add a bigger battery into its chassis, but he sees a future in his metaphorical crystal ball surrounding how “use cases are being distributed across different device form factors within a person’s daily life.”
Impressive wins for the Elite in Visual Studio compilation and NPU benchmark: https://www.tomshardware.com/pc-com...x-elite-x1e80100-takes-down-core-7-ultra-155h
Probably not optimized very well for WoA yet. Visual Studio is obviously going to be decently optimized coz Microsoft wants WoA to succeed and they actually develop the compiler so they know how to make it fly on ARM.Why is the gap in 7-zip so much narrower?
It looks like the Arm specific part was not updated for 2 years: https://github.com/ip7z/7zip/tree/main/Asm/arm64Probably not optimized very well for WoA yet. Visual Studio is obviously going to be decently optimized coz Microsoft wants WoA to succeed and they actually develop the compiler so they know how to make it fly on ARM.
Do we?and we know from devices that this is not the case.
Yes 8 gen 3 is more efficient than 7+ gen 3 even in devices, check video he test some games and 7+ gen 3 was meh, devices with 8 gen 3 have Excellent battery life, so it can replicate that efficiency in real life too.I don't get these graphs. By these graphs, 8G3 is more efficient then 7+G3 and any other SOC in every scenario, and we know from devices that this is not the case.
So what gives? The graphs are embellished?
The idea is that in low-load workloads, the 8G3 will consume less power than 7+ G3. But when you push them with a high load, 8G3 will consume more power than 7+ G3.I don't get these graphs. By these graphs, 8G3 is more efficient then 7+G3 and any other SOC in every scenario, and we know from devices that this is not the case.
So what gives? The graphs are embellished?