Sidebar Sidebar
Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.

Other Fastest CPU EVER and in ANY ... A14 Bionic

Coffee_Time

Coffee_Time

Cancelled
Original poster
Nov 22, 2017
718
342
LITERALLY...
Snapdragon 865 = 15 trillions operations per second
A14 = 11 trillions
In ANY and EVER ... with world's toughest glass ... Gorilla Glass Victus
Capture+_2020-10-14-00-32-38.png
Capture+_2020-10-14-00-34-10.png
 
Chompineer said:
Ok, and?

If you’d like a snapdragon device, you get the janky android experience, which is slow.

Also that appears to be stating the Neural Engine alone is 11 trillion operations.
Click to expand...
Also Snapdragon AI engine alone... Something must be wrong with Apple's numbers... Maybe they were referring to transistors...
 
What about A13 Bionic, up until today it was WELL ABOVE snapdragon 865.... fastest in any phone my.... 🤮
 
  • Like
Reactions: snipr125
Chompineer said:
Ok, and?

If you’d like a snapdragon device, you get the janky android experience, which is slow.

Also that appears to be stating the Neural Engine alone is 11 trillion operations.
Click to expand...

My Moto G Power will disagree with you. I won’t say it is iPhone 12 fast, it is on par with iPhone 8 I have. And it only cost 299 Canadian.

Android is no longer janky or slow.
 
  • Like
Reactions: mi7chy, ct2k7, barondebxl and 1 other person
Aoligei said:
My Moto G Power will disagree with you. I won’t say it is iPhone 12 fast, it is on par with iPhone 8 I have. And it only cost 299 Canadian.

Android is no longer janky or slow.
Click to expand...
Up until yesterday A13 was smashing Snapdragon 865, yet his big brother can't match it's 15 trillions operations per second... 😂
 
You’re comparing two different parts of the cpu to one other. Neural engine computations figure on the A14 is not what you need to compare to the overall computation figure on the snapdragon.

In fact the trillions calculation of the snapdragon should be compared to the 6 cores of the a14 not the 16 core neural engine as it is designed specifically for a different type of calculation.

You would need the trillions of calculations figure for the 6 core portion of the a14 I believe.
 
Last edited:
  • Like
Reactions: bobmans
From the highlighted stuff in the original post the Qualcomm “pushes 15 trillion operations per second whereas the A14 is capable of “up to 11 trillion operations per second” but are there any accepted standards for this stuff? Do we know we are comparing anything close to like-for-like? Unless we know what an “operation” is in each case it’s like comparing battery bank A containing “50 batteries” with battery bank B containing “40 batteries” and declaring A the winner until you find out that the batteries in A are all AAA batteries and the batteries in B are all huge car batteries.

If there are established benchmarks that define what “operations” are being measured like they have on CPUs then fine but if the neural net/tensor processing world doesn’t have that, or if both companies aren’t running the same benchmark suite, then I’m not convinced we can draw any conclusions whatsoever from this.

Nice to see strong competition though, and maybe Qualcomm does actually hold the performance crown at the moment but I would want more detail to be confident on that. Ultimately this competition is good for all of us though. Just think what the next 10 years will bring.
 
  • Like
Reactions: Stridr69
JulianL said:
From the highlighted stuff in the original post the Qualcomm “pushes 15 trillion operations per second whereas the A14 is capable of “up to 11 trillion operations per second” but are there any accepted standards for this stuff? Do we know we are comparing anything close to like-for-like? Unless we know what an “operation” is in each case it’s like comparing battery bank A containing “50 batteries” with battery bank B containing “40 batteries” and declaring A the winner until you find out that the batteries in A are all AAA batteries and the batteries in B are all huge car batteries.

If there are established benchmarks that define what “operations” are being measured like they have on CPUs then fine but if the neural net/tensor processing world doesn’t have that, or if both companies aren’t running the same benchmark suite, then I’m not convinced we can draw any conclusions whatsoever from this.

Nice to see strong competition though, and maybe Qualcomm does actually hold the performance crown at the moment but I would want more detail to be confident on that. Ultimately this competition is good for all of us though. Just think what the next 10 years will bring.
Click to expand...
Apple A14's NPU pushes 11 trillion operations/second
The Snapdragon 865's Qualcomm Hexagon Tensor Accelerator pushes 15 trillion operations/second
I don't know if you can directly compare the NPU with the Tensor Accelerator, these are parts of a way bigger chipset (the chipset being A14/Qualcomm865 and the part being the NPU/Tensor accelerator) and I don't even know if you can directly compare these parts.
 
  • Like
Reactions: nill1234 and rawCpoppa
bobmans said:
Apple A14's NPU pushes 11 trillion operations/second
The Snapdragon 865's Qualcomm Hexagon Tensor Accelerator pushes 15 trillion operations/second
I don't know if you can directly compare the NPU with the Tensor Accelerator, these are parts of a way bigger chipset (the chipset being A14/Qualcomm865 and the part being the NPU/Tensor accelerator) and I don't even know if you can directly compare these parts.
Click to expand...

This was what I was trying to say earlier. I probably messed up even after editing. I agree. Can’t compare different parts of a cpu designed to do different things.
 
  • Like
Reactions: bobmans
Gorilla Glass Victus.......

Saw a YouTube demonstration of both the iPhone 11 and snapdragon with this gorilla glass victus and the victus broke. So much for the worlds toughest gorilla glass. They both broke because Glass is glass. As far as the trillions of operation per second, I can care less. My Xs has never let me down, speed wise, so the new one will be just as good. Posting this means nothing to me and many others.
 
If you pay attention the comparison is strictly of the AI engines of each CPU. Ai 1 vs Ai2. Snapdragon crushes it. But what amuses me is that up until yesterday A13 was the best processor on Any planet in this galaxy, YET his bigger brother ashamed it by unveiling it's secret. So we can reasonably assume that A13 has half of those trillions. Probably on par with sd855. Apple fan boys should open their eyes...
"In any ever" 🤭
 
The problem here is the definition of 'operation'. I think the term can be used rather flexibly.
 
Seen this mentioned a couple of times, so I'll ask... what type of operation? Half-precision? Full-Precision? Are we talking about scalar operations or matrix? What's the metric here?
 
Anyway, I would like to discuss about A13 more than A14. I repeat once again
Given the analogy that A14 is has 11 trillions, how much is left for A13? Also, didn't A13 was crushing any android snapdragon up until yesterday, without having numbers revealed of its calculations power??? 😂 Thx to it's new brother who leaked the number...
 
Coffee_Time said:
Idk, ask Apple 😂
Click to expand...
Could just as well ask Qualcomm, because they're not specifying it either...


Coffee_Time said:
Given the analogy that A14 is has 11 trillions, how much is left for A13? Also, didn't A13 was crushing any android snapdragon up until yesterday, without having numbers revealed of its calculations power??? 😂
Click to expand...
To be clear, when you say A13 or A14, you mean only the Neural Engine of the SoC, right? I'm not sure why you say numbers are not revealed. The A14 has 11TOPs inference throughput on the NE, the A13 has about 6TOPs and the A12 about 5TOPs. Nothing new here, those are the numbers. What they actually mean is another story.
 
Chompineer said:
Ok, and?

If you’d like a snapdragon device, you get the janky android experience, which is slow.

Also that appears to be stating the Neural Engine alone is 11 trillion operations.
Click to expand...

That’s a very ignorant comment. You must have had an android experience from 2009 or something.
 
  • Like
Reactions: shadyman
GrumpyCoder said:
Could just as well ask Qualcomm, because they're not specifying it either...



To be clear, when you say A13 or A14, you mean only the Neural Engine of the SoC, right? I'm not sure why you say numbers are not revealed. The A14 has 11TOPs inference throughput on the NE, the A13 has about 6TOPs and the A12 about 5TOPs. Nothing new here, those are the numbers. What they actually mean is another story.
Click to expand...
It's not Qualcomm that says "FASTEST IN ANY PHONE EVER ON THIS UNIVERSE BY 50%". IT'S APPLE
 
I think apple is talking about the cpu and gpu when they mean the fastest SoC, which it really is..
 
bobmans said:
Apple A14's NPU pushes 11 trillion operations/second
The Snapdragon 865's Qualcomm Hexagon Tensor Accelerator pushes 15 trillion operations/second
I don't know if you can directly compare the NPU with the Tensor Accelerator, these are parts of a way bigger chipset (the chipset being A14/Qualcomm865 and the part being the NPU/Tensor accelerator) and I don't even know if you can directly compare these parts.
Click to expand...
You're right. No you can't. You need to compare them only in Geekbench. Which is not an OPEN SOURCE software so that we can see what's written in their's code. Forget about other benchmarks. Only Geekbench. 😂
 
You must log in or register to reply here.
Register on MacRumors! This sidebar will go away, and you'll see fewer ads.
Top Bottom
Back