As said earlier, none of this matters because ARM pulled their IP license 🤣
Got a tip for us?
Let us know
Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
Qualcomm Says New Snapdragon 8 Elite SoC Has 'World's Fastest Mobile CPU'
- Thread starter MacRumors
- Start date
- Sort by reaction score
Not yet. They've shown intent, but it's not a done deal. There is zero chance they will follow through before QC tries for a restraining order, or some other legal maneuver. This will not end quickly, and it will probably end in some sort of negotiated settlement. ARM doesn't really want to take those chips off the market, after all. They just want more revenue.
Probably right but no one is going to start new business with them now which is where the damage is. Look at the windows dev kit mess as well and this and its not looking good.
Building an SoC isn’t what it used to be. You can assemble a lot of licensed IP into a design and throw it at a fab using their SC if you have a customer and enough capital to bankroll a tape out and validation. You’ll get a mid to high performance device out of that in single millions quantities. There are plenty of companies out there doing exactly that already. Qualcomm were trying to push the edge and didn’t make it.
Actually on windows dev kit, they really screwed that up. I mean totally. Absolute disaster of a platform. I’m surprised they shipped. Half the stuff at Adafruit is better put together and that’s mostly garbage.
Are you having a laugh if you have certain iPhones it can’t even show you the weather animations in the weather app a strong powerful chip can’t even do that.
Dangit, I had a much longer reply to this typed up that an errant keyboard stroke destroyed. Short version:
Many people have replied pointing out there are already products that do this on the market if you want one, and they aren't even new, but no one really cited much in the way of specific examples of the reality that we're not "about at the point", we passed that point years ago.
My current home desktop, a four-year-old, top-of-the-line, final generation 27" iMac with a 10-core i9, is as an everyday driver slower than my iPhone 16 Pro, it just has a lot more RAM and a better GPU. And that is a nice computer by "those who don't need a high power computer"; it's actually quite solid even for those who do.
Until a month ago, my work desktop was a 2018 Mac Mini with a 6-core i6, which was getting a bit slow but was still serviceable and is still in active use as a general purpose machine. Not only can my current phone run rings around it, but as a light-duty computer a 3-year-old iPhone 13 is similar or better specc'd in every way--CPU, GPU, storage--other than having less RAM and worse cooling for number-crunching.
We have non-technical users still on ~4-year-old Dell Latitude 7400s with a 4-core i7 CPU that is not far off in CPU grunt from the iPhone 11 that was new at the same time, and is comparable to a high-end Android phone from 2 years ago.
And I know of plenty of low-end users (or those at low-tech companies) still using laptops far worse than any of the above--all of those were reasonably nice to very-expensive computers when they were new, not a crappy econo-box with a budget CPU and limited RAM.
Point of all that being, mobile CPUs in general are years past the point that a "regular" user could use it as a docked desktop without any complaint at all, and with iPhone CPUs being 2-3 years ahead of what an Android CPU could do until this new Qualcomm chip, even a several-year-old iPhone is faster than I suspect the significant majority of consumer Windows computers still in use, forget something like a cheap Chromebook.
That there aren't more dockable phones on the market isn't a technical limitation, at all, it's just doesn't seem to be something the majority of users are all that interested in.
Its not 10% slower in single core, most of the time is around 5-7% slower.
Yeah but as some apple fans said, efficiency is what matters and the 8 Elite's CPU matches A18 Pro's CPU efficiency at any power stage, in some situations is even slightly better, for example at 10W power usage.
So its objectively a good showing. Not to mention thta CPU is just one component of the SOC, the GPU, NPU, ISP, Modem, Wifi, Bluetooth are also top class in efficiency and performance.
You are trying to diminish Qualcomms achievements, detailed reviews are already out, the 8 Elite is a world class SOC in both performance and efficiency. An incredible achievement in the realm of SOC design and as one of Qualcomm's guys said on the stage, next year should be even more interesting.
A few comments back you said 8 Elite runs at significantly higher clocks, not it runs at lower clocks, which one is it?
Also the 8 Elite still beats the 18Pro in perf at the same power usage, not to mention it achieved its peak multicore score with at most 25% higher power. Taking in consideration the efficiency curve, the 18 Pro at the same power as the 8 Elite wouldn't be able to match the peak multicolore score anyway.
Source? The only verified score I’ve seen is Qualcomms CRD device. Those always score significantly higher than real world scores. I would guess when devices hit, they will be around 3000 in Geekbench. The overwhelming number of A18 scores are over 3300.
Source? The Elite is not as efficient generally. I believe in multi core it can be because QC lower clock speed and have more cores.
This moves into fandom now.
Not true. Maybe it happened once and I keep seeing this being repeated like its the norm.
One Plus 13 achieves similar single core scores with slower RAM and Samsung will use a version of the chip that will be clocked higher for example. So it's like I've said 5-7%.
Yeah in general(the SoC itself and all its components, GPU, Modem, Bluetooth, Wifi, NPU, ISP etc.) should easily be more efficient.
Also the 8 Elite CPU definitely doesn't run at lower clocks. The extra 2 cores excuse is really funny.
It's just the truth even if you don't like it.
Last edited:
Once? Lol. They boasted about having the fastest single core with the X Elite last year at over 3200 and to this date, not a single one has posted anywhere near that. The top out at 2900-2950
Again…source?
Source? You can’t claim proof from a device which doesn’t exist yet, and as I have shown, plenty of A18 scores are over 3400-3500. So no. Not 5-7%. The highest A18 score is over 3600. That’s over 10% more
It is was the truth, you’d have evidence. Your last post sounded like an official Qualcomm press release.
It runs at higher clocks at full power, which is one way it achieves better multi-core performance. It runs at lower clocks when we fix performance at the same level, which is how it can achieve marginally lower power in that particular case. We need to be careful when placing observations within the relevant context.
There is no doubt that 8 Elite is a world class SoC. I just don't think that it is very impressive from the technology standpoint when compared to the industry leader. Frankly, I would expect better performance from eight cores running at the advertised clocks.
It is hard for me to see the Oryon core as an innovative technology. Its design and performance characteristics are very similar to Firestorm/Avalanche. Given that these cores have been designed under the same technical leadership, it seems to me that the Nuvia team has recreated the work they did at Apple and ported it to a smaller node. It is difficult to me to see this as a new product. I would like to see some substantial innovation from them. Maybe the next core will bring it to the market, which would be great. We can talk about it when and if it happens.
Oh no no, you post the source, because the articles published after Oct 2023 with Qualcomm's benchmark numbers definitely don't show what you claim.
The OnePlus 13 isn't real? It was on the table of a Chinese YouTube channel.
The evidence is all over the internet, literally, but you already know that. The fact that you don't like it doesn't mean its not accurate.
Or you want to say that improvements of 45%, 40%, 44%, 33%, 62% aren't something remarkable for a year on year improvement? when was the last time Apple achieved something like this?
Last edited:
Not in a while, but several times.
Geekbench 3:
- A6 was 226% YoY over A5
- A7 was 100.1% YoY over A6
- A9 was 55.7% YoY over A8
- A6 was 165.5% YoY over A5
- A7 was 67.8% YoY over A6
- A9 was 63.7% YoY over A8
- A10 was 50.1% YoY over A9
- A9 was 72.5% YoY over A8
And how do you specifically know at what clocks it runs at lower power?
Also if Apple would add 2 more cores(that are constantly active and use power) you think the CPU won't use in general more power?
Overall its not impressed? or just the CPU and most likely just single core. Because you most likely talk about that, but generalize to the entire SOC, which overall without a doubt is impressive vs the A18Pro.
Why exactly? its just a tiny phone chip that scores higher than a lot of laptop chips. Just because they are 8 cores? Or you think apple with 8 cores could achieved higher scores? Even if their 6 cores aren't actually more efficient? And adding 2 more cores won't help.
Yeah, for you.
The second gen Oryon core in the 8 Elite definitely isn't identical to what first Nuvia design(of which we know very little technical design details anyway).
I haven't seen any proof that they are a copy of Apple's cpu cores arhitecture, better yet, my understanding is that in terms of the size of the cpu core themselves they have an advantage of Apple's cores.
I would like to see that from apple at least what I saw now from Qualcomm, unfortunately I gave up on this idea long ago.
On the same architecture that would be very impressive. Which is why it never happens. The 8 Elite uses technology from a startup Qualcomm purchased. Large gains are to be expected.
Again, context matters. Under any other circumstance, this would be an absolutely fantastic result. But we are talking about an ex-Apple team who founded their own company and claimed they can deliver the fastest CPU core, a claim supported by Qualcomm who acquired them. So expectations are high. As of today, they still haven't reached that goal. Of course, their tech is massively better than what Qualcomm used before. That is not what people are looking at, however.
I am extrapolating from the available evidence. If they were to at the same high clock it would mean that the per-clock performance is poor. Hence, the clock must be reduced by a decent margin, and that's a more advantageous location at the power-performance curve.
Of course it would use more power. It's about balance and tradeoffs. Right now we have Elite 8 outperforming Apple in MC by 10% while drawing 25% more power. Is that a good tradeoff for a phone? That is up to your judgement.
We were talking about the CPU, so I was specifically focusing on CPU cores. We can discuss other IP blocks if you want. I don't find the GPU particularly interesting from the technical perspective (it uses 128-wide fp16 ALUs, which gives it an edge in mobile games and synthetic benchmarks, but is not suitable for complex compute workloads), and the NPU so far fails to impress as well. The latter might be software support though.
If Apple added 2 more E-cores they would likely achieved 10-15% higher performance at the cost of 0.5 more watts. But why would they? I don't see a practical need for having that level of multicore performance in a phone. Optimizing performance for a given power level is much more important. And 8 Elite does not convince me here.
The size of the structures, number of ALUs, instruction timings etc. are very similar to Firestorm. There are some differences, of course. But you don't get such similarities by coincidence.
I don't get it. What did you see from Qualcomm? They bough a successful startup and turned it into a feasible product. That shows business acumen (although they are plagued by various issues such as their legal battle with ARM and a failure to properly position their X Elite laptops). I don't see much technical acumen in this. This is not a technology that Qualcomm has developed.
In the meantime, Apple released two new CPU microarchitectures with improved IPC and efficiency, shipped user-programmable matrix coprocessors, and built the world's first GPU with a unified cache architecture. I'd say these things are quite impressive. Then again, I am a tech enthusiast, I care about interesting technology. Maybe we have different priorities.
What on earth are you talking about. Do you think that Qualcomm doesn't get test devices for their chips from manufacturers? I am not aware of Apple overstating their CPU performance, please provide me a reference.
Are you working for Qualcomm or are you their bot?
Qcom just bought a bunch of engineers, something Apple does quite often. Do you remember Intel's modem division? Or their AI start-ups shopping spree?
The Oryon core in the 8 Elite has been developed at Qcom with Qcom's resources.
Realistically ARMs new large core is maybe even more impressive and could probably achieve similar clocks with the same 3nm HP libraries QCOM uses for the 8 Elite.
Man these sounds like a bunch of weak excuses to diminish Qualcomm's achievements and the results themselves.
According to what available evidence exactly? Its funny you don't want to admit its just an assumption.
Yes its a good tradeoff because the CPU was design to scales that way and the devices themselves are designs around these characteristics. What's important is that its not slower at similar power levels to Apple's chip.
I see how numbers have to be questioned and turned on all sides when apple doesn't have the largest ones, because when they have, such details don't matter.
That's just what you think about the GPU, but it sounds just like an excuse without anything behind, "oh the GPU is designed to run great in benchmarks and games, its terrible otherwise".
What NPU impresses you and why? QCOM probably has the current fastest one on the market, but that's not impressive, quite weird.
Also you avoided to mention anything about the Modem, or ISP, an important components in any smartphone.
Fascinating so according to you for 0.42% more power they can achieve up to 15% more performance. That doesn't make any sense.
Here's the excuse. Of course it doesn't convince you, and it most likely never will, no matter what
Source that compares this? It again sounds like an assumption.
What most saw, but hey some have glasses tinted in a certain way.
From where exactly do you know about what Qualcomm developed and did, when it comes to this product? You obviously like to assume so I'm curious.
Are they really "new microarchitectures"?
Most likely not.
Of course, but also wrong.
The world's first GPU with a unified cache architecture was NVIDIA's Ampere architecture, introduced with the NVIDIA A100 GPU in May 2020.
And the world's first GPU with a unified memory architecture was NVIDIA's Maxwell architecture, introduced in 2014.
Its not that, something else, for me it's obvious.
Last edited:
I am surprised that I have to explain such trivial details. Isn't it obvious that the CPU will be running at lower clocks to achieve lower performance? S8E is what, around 15% faster in multi-core, right? So to match the lower MC performance of A18 Pro you would need to lower the clocks, by approximately 15%, right? The rest of course is a guess. We don't really know the core activation pattern or the clocks these CPUs sustain in MC workloads. My guess judging by previous data on Oryon is somewhere in the ballpark of 3.3-3.5 Ghz for the prime cores.
At any rate, it is well known fact in computing that you can save power by using more lower clocked cores. That's how GPUs work after all.
Qualcomm GPUs have been inspected in a lot of detail, and their compute performance is well know. Please read in-depth analysis by well-informed tech sources like ChipsAndCheese, which discuss the limitations of Qualcomm GPU designs. I am looking forward to see the compute performance of the new GPU, Qualcomm has a lot to improve here, and I hope they did some work in this area.
Qualcomm claims to have the fastest NPU on the market, but in actual industry-standard benchmarks the NPU performance is not very good. It is one thing to advertise XXXX TOPS, it is quite another one to use it.
An example from related domain: Qualcomm's GPUs feature some very impressive peak FLOPS numbers since they pack a lot of ALUs behind one scheduler. But many real-world workloads cannot utilize such wide architecture efficiency. There is a reason why GPU industry has converged at 32-wide wavefronts.
That is fair. I don't know much about these things and they are not my primary interest. It is very much possible that Qualcomm is an industry leader in these components.
Apple has the best E-core design currently on the market. They have 1/3 of performance of the P-core but consume 10-20 less power. If I remember correctly, M4 is around 0.5 watts at peak, and 0.25-0.3 watts in multicore operation. Of course, that is according to Apple's internal power-reporting tools, which might be inaccurate. We lack good tooling to properly estimate power, unfortunately.
I just don't find it useful. Qualcomm certainly does, since they build the thing, and their customers probably find it useful too. I don't. I also don't understand why one would buy Alpine hiking shoes to do a light walk in the part, but there are people who do that as well.
What I said is based on the work of people lime Dougall Johnson, who reverse-engineered details of Apple's microarchitecture, the documentation published by Apple and Qualcomm, and clang patches that describe CPU-specific optimizations. You are welcome to look up these courses and do your own comparison.
That is a weird thing to say. The IPC is different, the dispatch width is different, the timing of certain instructions is different. It is obviously an evolution of Firestorm rather than a complete redesign (like Lunar Lake was), but M4 for example is quite a significant departure from M2.
That is not what I mean by "unified cache". Apple's GPU designs forego shader register files and specialized storage, merging everything into a single unified L1. What's more, that storage can spill to higher levels of the memory hierarchy . There is no other GPU on the market that can dynamically allocate shader registers or spill them to L2.
Maxwell features unified virtual memory, not unified physical memory. If I remember correctly Intel has been building UMA systems since at least 2012.
What is interesting is that Apple GPU design, despite being UMA, does not offer unified virtual memory. This is a jarring limitation for this of us who want to do complex compute on Apple GPUs and I am not quite sure what to make of it.
It’s worth noting for others reading this thread that this poster never backs up their claims. Are you defintely sure that they didn’t claim to have a Geekbench score over 3200?
What they actually provided was a model that topped out at 2979
SAMSUNG ELECTRONICS CO., LTD. Galaxy Book4 Edge - Geekbench
Benchmark results for a SAMSUNG ELECTRONICS CO., LTD. Galaxy Book4 Edge with a Snapdragon X Elite - X1E84100 - Qualcomm Oryon processor.
browser.geekbench.com
And whose score distribution looks like this:
As I said. QC misled massively.
And they didn’t show any measurement of Geekbench single core scores. Although the Spec tests might be tough for you to swallow.
The image below is the only one I have seen showing single core score. It’s their reference device (as usual), not a retail one. As usual with QC, they play shady games. Quoting their top performance with the average A18 one.
It’s important to remember this discussion began when you disagreed that there was a 10% delta between the A18 and the 8 Elite. Let’s assume the score for the 8 Elite is legitimate. What is the difference between the two. It’s 13%
A18 = 3646 https://browser.geekbench.com/v6/cpu/8335634
8 Elite = 3221 ?????
I am happy to come back and check this when we have more devices in the world. We can compare top scores or geometric mean.
I literally don’t know that and if you literally did, you’d literally show it, but you literally can’t because you literally don’t.
It’s the fact that you can’t show evidence that means it.