www. O S N E W S .com
News Features Interviews
BlogContact Editorials
.
Qualcomm announces the details of the Snapdragon 855
By Thom Holwerda on 2018-12-05 23:09:47

Today is the second day of Qualcomm's Snapdragon Technology Summit in Maui, and while yesterday was all about 5G and a teaser for its new chipset, today is all about the Snapdragon 855. The new chipset is built on a 7nm architecture, promising faster speeds, better battery life, and improved connectivity.

[...]

But as far as general performance goes, Qualcomm says that its Kryo 485 cores will offer a 45% boost, and the Adreno 640 GPU will show a 20% increase. With the firm's Snapdragon Elite Gaming Platform, gamers will be able to play in HDR with physically based rendering (PBR).

If these numbers hold up - only independent benchmarking will tell - this will go a ways to closing the wide gap with Apple's current offering at least partially.

 Email a friend - Printer friendly - Related stories
.
Post a new comment
.
wht gap?
By unclefester on 2018-12-06 11:45:22
The only 'proof'of a huge Apple speed advantage is the completely worthless Geekbench results. The same extraordinary claims were also made about the A10 using Geekbench. The results could never be replicated on any other benchmark or real world test,

Apple has a long history of making bogus performance claims - all the way back to the Motorola G3.
Permalink - Score: 2
.
RE: wht gap?
By kuiash on 2018-12-06 12:15:52
A12 is the current generation and, AFAICT, some if not most of the performance claims hold up. Anandtech gave it a good shake down.

https://www.anandtech.com/show/13...

Even included SPEC2006.

What are we comparing to what?!
Permalink - Score: 2
.
RE: wht gap?
By avgalen on 2018-12-06 12:31:26
Benchmarks in general aren't worth much to me anymore and I have a very low opinion about GeekBench in particular. However, there is no denying that the A12 and especially A12X provide much higher performance compared to the 845. Almost every app/game benchmark between Android and iOS follows this scenario:
1) Small apps start ridiculously fast everywhere
2) Big apps/games start much faster on a recent iOS device, caused by faster CPU/GPU
3) Recent Android devices* are better at taskswitching, caused by larger amounts of RAM
4) Video-editing is much faster on a recent iOS device. If this is caused by a faster CPU/GPU or special "co-processor" is unknown to me.

In the past it was unclear if the speed difference came from lower resolutions on iPhones, but since iPhone X it is clear that their CPU's are basically a generation (1 year) ahead.

In reality it seems like we have reached a point (about 2 years ago?) though where CPU/GPU performance is not a limiting factor for the software that we actually use anymore. If I were given a phone with a 845 or 660 I probably couldn't tell them apart without resorting to benchmarking. We need the OS and peripheral hardware to start making use of all this power before we need more power and more powerful software.

Windows-on-Arm is 1 such platform that is in need of more performance, both for native execution and especially for x86 emulation. For phones/tablets the CPU/GPU is pretty much overkill

* of course I am talking about the fastest flagships here since we are comparing them to recent iOS devices
Permalink - Score: 3
.
RE: wht gap?
By viton on 2018-12-06 17:28:04
The results could never be replicated on any other benchmark or real world test
This is a lie. A12 shines in every benchmark including industry standard SPEC.

In real world it can even outperform desktop i7 PC at specific tasks.

Edited 2018-12-06 17:43 UTC
Permalink - Score: 2
.
RE[2]: wht gap?
By Brendan on 2018-12-07 03:09:22
Hi,

> In the past it was unclear if the speed difference came from lower resolutions on iPhones, but since iPhone X it is clear that their CPU's are basically a generation (1 year) ahead.

As far as I know, Apple's CPUs are significantly larger (e.g. a lot more silicon consumed by caches, etc) and significantly more expensive to manufacture because of that; and they win "raw performance" comparisons because they lose "performance per dollar" comparisons.

> In reality it seems like we have reached a point (about 2 years ago?) though where CPU/GPU performance is not a limiting factor for the software that we actually use anymore.

Well, no.

The reality is that the relationship between clock speed and power consumption is non-linear (e.g. "quadradic-ish", like doubling the clock speed causes four times the power consumption); and the relationship between clock speed and performance is a lot more linear.

What this means is that you shouldn't reach a point where performance is "more than necessary", because manufacturers can reduce performance a little to increase battery life a lot.

Also note that various benchmarks encourage poor engineering. For example, Geekbench has explicit delays to allow the CPU to cool down and doesn't take power consumption or battery life into account, so bad manufacturers (that clock CPUs higher than necessary at the expense of battery life) get higher scores than good manufacturers (that have a sane balance between performance and battery life).

- Brendan
Permalink - Score: 3
.
RE[3]: wht gap?
By avgalen on 2018-12-07 09:59:30
> As far as I know, Apple's CPUs are significantly larger (e.g. a lot more silicon consumed by caches, etc) and significantly more expensive to manufacture because of that; and they win "raw performance" comparisons because they lose "performance per dollar" comparisons.
Performance per dollar comparisons cannot be made because Apple doesn't sell their CPUs. This is what I could find about cache and size:
Snapdragon 845:
L1 : 64KB + 64 KB
L2 : 256 KB
L3 2 MiB Cache
L4 3 MiB Cache system for CPU, GPU, DSP
Soc size : 12.4 x 12.4mm = 153.76 mm2
Apple A12:
L1 cache : 128 KB instruction, 128 KB data
L2 cache : 8 MB
The die size of the A12 is 83.27 mm2
I don't know how to compare "die size" to "Soc size", but based on previous knowledge about caches and performance and based on the A12 using 7nm proces I would say that "The A12 is clearly faster, but not necessarily pricier to produce"

> The reality is that the relationship between clock speed and power consumption is non-linear (e.g. "quadradic-ish", like doubling the clock speed causes four times the power consumption); and the relationship between clock speed and performance is a lot more linear.

What this means is that you shouldn't reach a point where performance is "more than necessary", because manufacturers can reduce performance a little to increase battery life a lot.

You are entirely correct and I think you will enjoy this article: https://arxiv.org/abs/1401.4655
It is clear that it isn't as simple as "run the code on max frequency so it finishes quicker and can go back to idling". IPC, leakages, 7/10/14 nm processtechnology, the quadradic-ish curve you mentioned all play a role. My comment was purely about needing more performance, but you correctly pointed out batterylife as a directly linked factor
Permalink - Score: 2
.
RE[2]: wht gap?
By avgalen on 2018-12-07 10:22:11
Adding a source is always good: https://www.anandtech.com/show/13...
Permalink - Score: 2
.
Unremarkable
By bram on 2018-12-07 19:06:34
GPU does +20% in two years.
How is that catching up to apple?
That's what, 8 percent or so per yr?
Not impressed.
Permalink - Score: 1

Post a new comment
Username

Password

Title

Your comment

If you do not have an account, please use a desktop browser to create one.
LEAVE SPACES around URLs to autoparse. No more than 8,000 characters are allowed. The only HTML/UBB tags allowed are bold & italics.
Submission of a comment on OSNews implies that you have acknowledged and fully agreed with THESE TERMS.
.
News Features Interviews
BlogContact Editorials
.
WAP site - RSS feed
© OSNews LLC 1997-2007. All Rights Reserved.
The readers' comments are owned and a responsibility of whoever posted them.
Prefer the desktop version of OSNews?