Electron752
Posts: 142
Joined: Mon Mar 02, 2015 7:09 pm

VC core vc ARM benchmark

Wed Mar 22, 2017 7:15 pm

Hi,

I found out that the VC core has a open source compiler now. Has anybody every attempted to run any kind of benchmarks that compare the performance of a program compiled for VC vs a program compiled for ARM?

I would be curious how the performance of the two differ.

It is interested that the VC core appears to not have a full MMU.

jamesh
Raspberry Pi Engineer & Forum Moderator
Raspberry Pi Engineer & Forum Moderator
Posts: 22733
Joined: Sat Jul 30, 2011 7:41 pm

Re: VC core vc ARM benchmark

Wed Mar 22, 2017 8:52 pm

No MMU, runs at up to 400Mhz IIRC, although does have 2 cores and an interesting vector unit (assembler only). Slower than the ARM core on Pi2 and Pi3.
Principal Software Engineer at Raspberry Pi (Trading) Ltd.
Contrary to popular belief, humorous signatures are allowed. Here's an example...
"My grief counseller just died, luckily, he was so good, I didn't care."

User avatar
Gavinmc42
Posts: 3172
Joined: Wed Aug 28, 2013 3:31 am

Re: VC core vc ARM benchmark

Sat Mar 25, 2017 1:36 am

VC4 runs slower but is it deterministic in timing?
That would make it suitable as a real time co-processor.
Is there any data on the VC4 timings and cycles?
Might have to read the stickies again, had not though of this before.
Hmm, QPU's have access to GPIO or are they just number crunchers?

Now that I have cluster hat, sort of need to know this stuff now.
More brain stretching to figure out how to use all those alu's.

VPU compiler tick, QPU assembler tick, what else is needed?
Pi Zero - 1 ARM cores, 2 x VPU's, 12 x QPU's (48 ALU's?).
What benchmark to use?
I'm dancing on Rainbows.
Raspberries are not Apples or Oranges

jamesh
Raspberry Pi Engineer & Forum Moderator
Raspberry Pi Engineer & Forum Moderator
Posts: 22733
Joined: Sat Jul 30, 2011 7:41 pm

Re: VC core vc ARM benchmark

Sat Mar 25, 2017 6:34 am

Gavinmc42 wrote:VC4 runs slower but is it deterministic in timing?
That would make it suitable as a real time co-processor.
Is there any data on the VC4 timings and cycles?
Might have to read the stickies again, had not though of this before.
Hmm, QPU's have access to GPIO or are they just number crunchers?

Now that I have cluster hat, sort of need to know this stuff now.
More brain stretching to figure out how to use all those alu's.

VPU compiler tick, QPU assembler tick, what else is needed?
Pi Zero - 1 ARM cores, 2 x VPU's, 12 x QPU's (48 ALU's?).
What benchmark to use?
Vpu runs an rtos, multithreaded, fairly responsive. But not easy o write code for, especially the quads. And cannot really add code to current firmware without access​ to source code.
Principal Software Engineer at Raspberry Pi (Trading) Ltd.
Contrary to popular belief, humorous signatures are allowed. Here's an example...
"My grief counseller just died, luckily, he was so good, I didn't care."

User avatar
Gavinmc42
Posts: 3172
Joined: Wed Aug 28, 2013 3:31 am

Re: VC core vc ARM benchmark

Sat Mar 25, 2017 7:41 am

Vpu runs an rtos, multithreaded, fairly responsive. But not easy to write code for, especially the quads. And cannot really add code to current firmware without access​ to source code.
But say I did not want any graphics, hdmi, jpeg stuff I could use the start_cd.elf
And then get the VPU ThreadX RTOS to load and run some QPU code?

Just got a Cluster hat and I'm trying to figure out how to max the QPU's and use them for something.
No idea what yet :?:
Just that a Pi plus Cluster hat with 4 Zero's all running Jessie seems so ordinary.

Some clues here?
https://github.com/Terminus-IMRC?tab=repositories

This looks interesting
https://github.com/Terminus-IMRC/py-videocore

Mostly a long term project, first cluster is for a video camera system, but it got me thinking.
The hat was not too expensive and Zeros are not that expensive just rare in qty :lol:
Now if they sold the Cluster hat plus 4 Zeros they might sell 3-4 times as many :lol:

At work a phd student is working on a baby scanner with 16 cameras, that's only 4 Cluster hats.
Still some serious number crunching to get a 3D image. Do it on the GPU's?
I'm dancing on Rainbows.
Raspberries are not Apples or Oranges

jamesh
Raspberry Pi Engineer & Forum Moderator
Raspberry Pi Engineer & Forum Moderator
Posts: 22733
Joined: Sat Jul 30, 2011 7:41 pm

Re: VC core vc ARM benchmark

Sat Mar 25, 2017 9:13 am

Don't underestimate the difficulty on rewriting the entire firmware, with no access to HW docs, given there is 50 or more mans years of work in that firmware blob.

On the other hand, check out the bare metal part of the forum, where writing VPU code is a thing.
Principal Software Engineer at Raspberry Pi (Trading) Ltd.
Contrary to popular belief, humorous signatures are allowed. Here's an example...
"My grief counseller just died, luckily, he was so good, I didn't care."

jamesh
Raspberry Pi Engineer & Forum Moderator
Raspberry Pi Engineer & Forum Moderator
Posts: 22733
Joined: Sat Jul 30, 2011 7:41 pm

Re: VC core vc ARM benchmark

Sat Mar 25, 2017 9:17 am

If you want something less ordinary, try fixing this bug...

https://github.com/raspberrypi/firmware/issues/673
Principal Software Engineer at Raspberry Pi (Trading) Ltd.
Contrary to popular belief, humorous signatures are allowed. Here's an example...
"My grief counseller just died, luckily, he was so good, I didn't care."

User avatar
Gavinmc42
Posts: 3172
Joined: Wed Aug 28, 2013 3:31 am

Re: VC core vc ARM benchmark

Sat Mar 25, 2017 9:43 am

Well I hope I can interest 100 people so we can do it in less than 6 months ;)

It's not high on my list and I spend plenty of time on the baremetal forum checking others progress.
More a theoretical, "it could be done but you have to be crazy like Hove to try it" :lol:

Maybe a nudge to any Profs with loose students :D
Having hardware like the Cluster hat, just makes it a software issue.
"Just' in this case means you have the hardware, now do the other 99.9% of the work.

http://www.dictionary.com/browse/genius ... rspiration
I actually had my own company decades ago called 1%, I was good at that, it was the other 99% I had trouble with, too much like hard work. Sometimes fresh graduates can do it in 6 months but you need good ones and you never tell them how hard it is.

I have no illusions about how hard it would be which is why I added another order of magnitude to 99.9% which probably should be 99.99% :lol:

And no thanks, I'll skip trying to solve that one.
To much Linux and C code, I'm a hardware guy.
The hardware works so it is the software monkeys turn.
Already helped solve a baremetal network issue last year that was a USB/NTP timeout issue.
Took weeks.


.
I'm dancing on Rainbows.
Raspberries are not Apples or Oranges

jamesh
Raspberry Pi Engineer & Forum Moderator
Raspberry Pi Engineer & Forum Moderator
Posts: 22733
Joined: Sat Jul 30, 2011 7:41 pm

Re: VC core vc ARM benchmark

Sat Mar 25, 2017 11:20 am

Sorry that should say 500 man years - 50 people for ten years approx. And they had access to the HW designers and documentation.
Principal Software Engineer at Raspberry Pi (Trading) Ltd.
Contrary to popular belief, humorous signatures are allowed. Here's an example...
"My grief counseller just died, luckily, he was so good, I didn't care."

User avatar
bensimmo
Posts: 4072
Joined: Sun Dec 28, 2014 3:02 pm
Location: East Yorkshire

Re: VC core vc ARM benchmark

Sat Mar 25, 2017 11:56 am

How does the cluster of 4 Zero's on a Cluster hat compare to the similarly priced Pi2/Pi3 with the faster 4 cores.

User avatar
Gavinmc42
Posts: 3172
Joined: Wed Aug 28, 2013 3:31 am

Re: VC core vc ARM benchmark

Sat Mar 25, 2017 1:10 pm

Sorry that should say 500 man years - 50 people for ten years approx
Now you tell me :shock:
Ok, pushing it down the list a few spots, maybe off the bottom ;)

How many of those 50 are still around?
Retired or promoted to management?
HW guys gone too?
Some are at RPF and there is enough to fix bugs.
What happened to the rest? No VC5 so moved to something else?

Ok, Cluster hat will be used for camera stuff, not a supercomputer takeover the world with my Pi's plan.
Wonder if that's why not much recently added to Herman's github.
You have scared them all off 8-)

Will spend the time on learning AArch64 and Neon instead.
I will let RPF change the GPU in future versions but I will get upset if you change the CPU from ARM :lol:
I don't have that many spare neurons, have to make them last another three years till that youth pill makes it to the public.
Wonder if it will fix 50years of inhaling lead solder fumes? They need guinea pigs?
How does the cluster of 4 Zero's on a Cluster hat compare to the similarly priced Pi2/Pi3 with the faster 4 core
No idea, but 4 Zero's can get me a 4 camera system, 5 if I put one on the Pi3.
For vision stuff some preprocessing on the Zero's then a core on the Pi3 for each camera to do post processing. The big trick is getting enough Zero's :lol:
With 4 Zero's you get 4 more GPUs, but I have just been told it will be very hard to use them :(

I think I will leave it to someone else to figure out how to use all 240 QPU's.
That in itself might make it tempting for someone/s to spend a lot of time finding out how to use them.
Or they just will take the easy way and use a Nvidia TX PCB.

I have a rekindled interest in AI, what Jetpac did with Deep Belief blew away what I thought could be done. Imagine a Cluster Hat with 5 times that power?
Now how about a CM3 cluster with 2 cameras per CM.

https://github.com/jetpacapp/DeepBelief ... yPiLibrary
This has a compiled .so file in it.
I suspect this guy knows quite a bit about of the internals of the VC4 GPU or he is really, really smart which is why Pete Warden got headhunted by Google?
Interesting blog
https://petewarden.com/

Since I'm a hardware guy most of this is way above my maths skills.
With the Cluster and bootcode.bin and start.elf on the Zero's SD.
The kernel.bin can be made, compiled on the 3 and then it can turn the Zero on and load the kernel into ram.
This would allow fasting coding and testing.
With 4 Zero's this could mean 4 times faster development.

The guys at work are worried I will get AI working just enough to make the Pi's make themselves smarter. Bootstrap brains :evil:
Maybe Pete will see this post and accept the challenge :roll:
I think Deep Belief takes 3 seconds, running 5 times faster would be interesting.

But a Pi3 with 4 cores running 64bit code and using NEON might be just as fast as the GPU method and open.
Pi 64bit OS's are scarce which is why most of my time is spent learning Pascal/Ultibo/Aarch64/NEON.
I probably have a chance then of understanding the code if I write it.

Now vision and AI stuff works fine with 8 bit data, using the 128bit NEON/SIMD registers by the 64 bit ARMs core it should all be faster than the 32bit GPU stuff which run at a slower clock anyway.

But I am a hardware guy so hate to see silicon being wasted.
So 64bit, Neon and QPU's all working at max, that should be smoking, literally, might need water cooling :lol:
Can you imagine a 6 to 10 pack cluster of CM3's, it would fit in your hand and weight nearly nothing.
Would need copper or ally sheet laminations for cooling.
I'm dancing on Rainbows.
Raspberries are not Apples or Oranges

Electron752
Posts: 142
Joined: Mon Mar 02, 2015 7:09 pm

Re: VC core vc ARM benchmark

Sat Mar 25, 2017 4:12 pm

This is really cool. I don't what to think actually.

I might be able to contribute to setting up an access point and such. Do these things really do USB 3? I never actually tried benchmarking the USB ports.

User avatar
Gavinmc42
Posts: 3172
Joined: Wed Aug 28, 2013 3:31 am

Re: VC core vc ARM benchmark

Sun Mar 26, 2017 2:40 am

No USB3 only 2.
The Pi chips have a single USB2 port, this goes to a LAN9514 USB to Ethernet chip that also has 4 USB 2 ports.
The Zero's and the A+ don't have this ether/hub chip and so only have the one USB 2 port.
USB2 limits the Ethernet speed and there is no SATA either.
But to be fair the BCM2835 in the Zero probably only costs about $1.
A bit more for the BCM2837 in the new2 and 3's, guessing around $4-5

Still for a 'toy' computer that is a lot of bang for buck, and one can never have enough toys :lol:
I think I lost count at around 16, with the Zero W I got last week, my personal collection is maybe around 20 now :o
It would be much more if Zero's were easier to get, thankfully RPF are drip feeding them to us.
Never had a real use for the Zero's until the W and Cluster hat came out. now I need more of both model Zero's :lol:
I'm dancing on Rainbows.
Raspberries are not Apples or Oranges

jamesh
Raspberry Pi Engineer & Forum Moderator
Raspberry Pi Engineer & Forum Moderator
Posts: 22733
Joined: Sat Jul 30, 2011 7:41 pm

Re: VC core vc ARM benchmark

Sun Mar 26, 2017 10:41 am

I do find it amusing when people refer to a quad core A53 device with a fairly powerful GPU as a toy. They clearly didn't grow up in the era of 32KB BBC Micros running a 2MHz single core 6502.
Principal Software Engineer at Raspberry Pi (Trading) Ltd.
Contrary to popular belief, humorous signatures are allowed. Here's an example...
"My grief counseller just died, luckily, he was so good, I didn't care."

User avatar
DougieLawson
Posts: 35381
Joined: Sun Jun 16, 2013 11:19 pm
Location: Basingstoke, UK
Contact: Website Twitter

Re: VC core vc ARM benchmark

Sun Mar 26, 2017 10:56 am

jamesh wrote:I do find it amusing when people refer to a quad core A53 device with a fairly powerful GPU as a toy. They clearly didn't grow up in the era of 32KB BBC Micros running a 2MHz single core 6502.
They should see the toys that went to the moon with Aldrin, Armstong and Collins. The average digital wrist watch has more power than all of their computers (including the ones at NASA's Houston mission control center).
Note: Having anything remotely humorous in your signature is completely banned on this forum.

Any DMs sent on Twitter will be answered next month.

This is a doctor free zone.

User avatar
bensimmo
Posts: 4072
Joined: Sun Dec 28, 2014 3:02 pm
Location: East Yorkshire

Re: VC core vc ARM benchmark

Sun Mar 26, 2017 11:25 am

jamesh wrote:I do find it amusing when people refer to a quad core A53 device with a fairly powerful GPU as a toy. They clearly didn't grow up in the era of 32KB BBC Micros running a 2MHz single core 6502.
Probably not, but in today's world where Tablets and Nintendo Switch 'Arm based' toys with a lot more power than the Pi. There is no wonder the Pi can be classed as a toy.
It is cheap to play with, often bundled with fun toy 'child' targeted fun to make.
The fact it's a computer is by and by now. They're common place everyday throw away things.
Toys are after all generally educational fun things wrapped up so they don't know they are learning by play.

Of course please use the Pi for buisiness and serious stuff, for them it's not a toy but a tool.

User avatar
Gavinmc42
Posts: 3172
Joined: Wed Aug 28, 2013 3:31 am

Re: VC core vc ARM benchmark

Sun Mar 26, 2017 11:54 am

I do find it amusing when people refer to a quad core A53 device with a fairly powerful GPU as a toy. They clearly didn't grow up in the era of 32KB BBC Micros running a 2MHz single core 6502.
Did too, did too, did too.
But it was a Z80 4Mhz Microbee, with 32KB Sram ;)
I built my second Microbee from all CMOS chips to reduce power.

Compared to the rest of the PC's out there, Pi's are toys, just small, damn useful ones.
And that "fairly powerful GPU" is mostly undocumented and therefore unusable for stuff that does not fit the mold.
If I want to emulate the Apollo computer on a single QPU how do I do it?

And running those A53's in 32bit mode!! well would you have run the Beeb in 4 bit mode only?
I lived and breathed 8,16,32 bit cpus, RTFM many times, they were bedside reading, did stuff people said was not possible only because I read the manuals and knew all the hairy arse issues and thought outside the square.

RPF says VC4 is 500 man years and they had access to the designers and software.
So what? To me that just means we cannot do weird stuff and push the limits because RPF says this is the sandpit and outside there be dragons.

How can we find out what this 'toy' can do without looking under the hood and taking off the limiter.
It will remain a toy, but it's a damn useful one and will probably keep me busy for another decade or so;) Without complete documentation how else can it be treated?
It does what it was designed to do, if you want to do something else, tough xxxx.

I suspect that complete documentation does not exist anyway, the way Broadcom spins chips no one could find time to document them completely anyway.
It is a way for me to peek into the embedded 64bit ARM world that the World now seems to run on now.

And is there really a undocumented DSP buried in there or not.
Stepping down off my high horse now ;)
I'm dancing on Rainbows.
Raspberries are not Apples or Oranges

jamesh
Raspberry Pi Engineer & Forum Moderator
Raspberry Pi Engineer & Forum Moderator
Posts: 22733
Joined: Sat Jul 30, 2011 7:41 pm

Re: VC core vc ARM benchmark

Mon Mar 27, 2017 8:10 am

bensimmo wrote:
jamesh wrote:I do find it amusing when people refer to a quad core A53 device with a fairly powerful GPU as a toy. They clearly didn't grow up in the era of 32KB BBC Micros running a 2MHz single core 6502.
Probably not, but in today's world where Tablets and Nintendo Switch 'Arm based' toys with a lot more power than the Pi. There is no wonder the Pi can be classed as a toy.
It is cheap to play with, often bundled with fun toy 'child' targeted fun to make.
The fact it's a computer is by and by now. They're common place everyday throw away things.
Toys are after all generally educational fun things wrapped up so they don't know they are learning by play.

Of course please use the Pi for buisiness and serious stuff, for them it's not a toy but a tool.
Are modern tablets really that much more powerful? They tend to be quad core ARMs running at 1.2 or so GHz. The Pi is right up with that level of performance. Only lacks in memory in comparison.

The Switch is interesting, appears to be ARM BigLittle, quad A57's and quad A53's. I doubt they can all run at the same time, too much heat. Although they do run at a lower frequency than the Pi.
Principal Software Engineer at Raspberry Pi (Trading) Ltd.
Contrary to popular belief, humorous signatures are allowed. Here's an example...
"My grief counseller just died, luckily, he was so good, I didn't care."

User avatar
Gavinmc42
Posts: 3172
Joined: Wed Aug 28, 2013 3:31 am

Re: VC core vc ARM benchmark

Mon Mar 27, 2017 8:31 am

Most of the Octo cores are 4 x A57 + 4 x A53.
But that's top of the range $800-1000 Smartphones.

4 x A53's in the Pi3's do get warm, need heat sink and fan when I do bare metal in summer;)
Angry Birds was a breakthrough mobile game, have we seen anything that on a Pi yet?

It should be possible, but what is stopping it?
Lack of GPU documentation, no tools except ones you have to roll on your own?

Ruler flick Raspberries at the School teacher, kids would love that game :lol:
Do I really have to use Scratch to make a game like that?
What is out there right now that can do it?
As in serious CS grade stuff without needing to use assembler.

Most of those Smartphones have "better" GPUs, PowerVR, Mali, Adreno.
Where is the VC4 SDK?
I'm dancing on Rainbows.
Raspberries are not Apples or Oranges

User avatar
bensimmo
Posts: 4072
Joined: Sun Dec 28, 2014 3:02 pm
Location: East Yorkshire

Re: VC core vc ARM benchmark

Mon Mar 27, 2017 8:56 am

From memory the older Tegra K1 is faster on some slower on others (first nvidia shield, MiPads, Chromebook 13, though the nexus 9 used a 64 varient of that).
But since there is an expensive development board there are benchmarks.

Search and err phoronix website has some
The Switch with the TK1 (also newer shield, Pixel C and of course expensive development board) give benchmarks, it's fast. How much is the keplar GPU? I've not read that far. And how much the Switch has been clocked back I don't know.

Unfortunately the switch didn't get the TX2, who knows it may in time.

But take that 'switch toy' and take a fraction of the cost, bundle it with some fun and you can see how they are classed as you in a modern kids world.

Especially compared to the effective cost of that BBC today. :-)

jamesh
Raspberry Pi Engineer & Forum Moderator
Raspberry Pi Engineer & Forum Moderator
Posts: 22733
Joined: Sat Jul 30, 2011 7:41 pm

Re: VC core vc ARM benchmark

Mon Mar 27, 2017 9:02 am

Gavinmc42 wrote:Most of the Octo cores are 4 x A57 + 4 x A53.
But that's top of the range $800-1000 Smartphones.

4 x A53's in the Pi3's do get warm, need heat sink and fan when I do bare metal in summer;)
Angry Birds was a breakthrough mobile game, have we seen anything that on a Pi yet?

It should be possible, but what is stopping it?
Lack of GPU documentation, no tools except ones you have to roll on your own?

Ruler flick Raspberries at the School teacher, kids would love that game :lol:
Do I really have to use Scratch to make a game like that?
What is out there right now that can do it?
As in serious CS grade stuff without needing to use assembler.

Most of those Smartphones have "better" GPUs, PowerVR, Mali, Adreno.
Where is the VC4 SDK?
Whilst the internals of the GPU are closed, there is an industry standard API to the features called OpenGLES. So anyone including rovio could port their code to use that.

The new driver will provide OpenGL in userland which is more familiar for desktop developers.

Whether people develop for that? Who knows, they didn't for OpenGLES.

EDIT TO ADD: Of course, almost all the features on the GPU are available via industry standard API's. OpenMAX for codecs and camera (although the MMAL library sits over the components to make then more easy to use), OpenVG for 2D graphics, OpenGLES for 3D graphics. Anyone who thinks this closed GPU stuff is stopping them using the features on the GPU is wrong- these API's have been there since the launch of the Pi. There's no need for a SDK, all the API's are there and open.

Eric's new driver simply moves the 3D system to ARM space and up it to full OpenGL, plus various other bits and pieces.
Principal Software Engineer at Raspberry Pi (Trading) Ltd.
Contrary to popular belief, humorous signatures are allowed. Here's an example...
"My grief counseller just died, luckily, he was so good, I didn't care."

User avatar
bensimmo
Posts: 4072
Joined: Sun Dec 28, 2014 3:02 pm
Location: East Yorkshire

Re: VC core vc ARM benchmark

Mon Mar 27, 2017 10:29 am

Gavinmc42 wrote:Most of the Octo cores are 4 x A57 + 4 x A53.
But that's top of the range $800-1000 Smartphones.
My Kirin 650 octacore is 4xA53 + 4xA53 and the phone was £130 (2.0/1.7GHz)

User avatar
Gavinmc42
Posts: 3172
Joined: Wed Aug 28, 2013 3:31 am

Re: VC core vc ARM benchmark

Tue Mar 28, 2017 1:47 am

But that's top of the range $800-1000 Smartphones.
My Kirin 650 octacore is 4xA53 + 4xA53 and the phone was £130 (2.0/1.7GHz)
Should have said top of the range name brand Smartphones, not phones beer companies give away :lol:
It only has A53, not A57, A72's

Jamesh,
I know about OpenGLES etc but unless you have done it before it is not easy.
Anyone can grab the Android SDK/PowerVR SDK and do stuff without reading an awful lot of docs.
Just follow the examples and make stuff.
It usually takes longer to download and install than getting the first example working.

There is a lot of VC4/Pi stuff around but not in one nice easy to use form.
Anyone would have to start at Herman's web page and follow all the links.
You need a CS degree to understand most of it.

Pete Warden can do stuff no one did before on the VC4 but he lives and breaths GPU's and had lots of experience with other GPU's.
.
Noshbar on the baremetal Ultibo forum has been playing with this
http://www.kevinbeason.com/smallpt/
And has ported it to Pascal.
http://iwasdeportedandalligotwasthislou ... smallpt_tp

I am no GPU code monkey but when I see code like this

Code: Select all

   Vec operator*(double b) const { return Vec(x*b,y*b,z*b); } 
   Vec mult(const Vec &b) const { return Vec(x*b.x,y*b.y,z*b.z); } 
Even to me, that screams GPU.
Ray tracing is a good use of the parallel QPU/VPU is it not?

API's are not SDK's.
SDK's are nice friendly wrappers to the API, the VC4 is not friendly, it is a horrible little beast.
And it needs taming, but open range not in a cage.
I think I remember Eben saying even he had to make a lease to control OpenMax? :lol:

Only a CS geek would say FFT's are fun.
https://www.raspberrypi.org/blog/new-qp ... assembler/
To me, a self taught coder, they are horrible things with maths in them ;)

If Vulkan does VideoCore(Buckley's?) then I might go chew on someone else :lol:
But that just make the hardware an extra layer away.

I will probably just have do it the hard way and read all TFM's and try coding.
But I only have a limited amount of time left, this is supposed to be a hobby not an obsession ;)
I'm dancing on Rainbows.
Raspberries are not Apples or Oranges

User avatar
bensimmo
Posts: 4072
Joined: Sun Dec 28, 2014 3:02 pm
Location: East Yorkshire

Re: VC core vc ARM benchmark

Tue Mar 28, 2017 8:48 am

It's a Huawei P9 lite.
Not bargain basement stuff.

It fast enough not sure why i would need anything more powerful just yet. Played with more powerful S7/iPhones and can't see, other than for showing of, you would need it.
They're no smoother or quicker to me.
It honestly puts my tablets and older phones to shame.
Easily copes with games, the web and databases, voice recognition and mapping with no struggle.
And more in the league of Pi3 price comparison when you take into account, power, input, outputs and screens.

User avatar
Gavinmc42
Posts: 3172
Joined: Wed Aug 28, 2013 3:31 am

Re: VC core vc ARM benchmark

Tue Mar 28, 2017 9:10 am

Actually like my old flip phone with 2 week battery life, but it was retired because it was only 2G :(

Don't want more power, want more battery life.
Want Pi that run from body heat, should be able to run a few with the temp I'm at now.

Any SP's with 1 week battery life, running Linux?
What am I saying I want Redox OS, er time to go lie down.
I'm dancing on Rainbows.
Raspberries are not Apples or Oranges

Return to “Bare metal, Assembly language”