$5 Raspberry Pi Zero Compared To Intel's NetBurst CPUs & Newer (phoronix.com) 99
An anonymous reader writes: Curious about the performance of a Raspberry Pi Zero, Phoronix has published a number of Raspberry Pi 2 + Pi Zero performance benchmarks with paired power consumption data. They found the Pi Zero performed slower than even an Intel Celeron 320 from the NetBurst era, but that the Raspberry Pi 2 was performing between that Celeron and a Pentium 4 "C" 2.8GHz CPU from 2004. While the Raspberry Pis didn't win in raw performance, the performance-per-Watt of the Raspberry Pi 2 was 220x greater than the Pentium Northwood. The Pi Zero had an average power consumption of 2.7 Watts and the Raspberry Pi 2 was at 3.5 Watts; however, compared to newer Broadwell and Skylake processors, Intel's low-end parts delivered greater power efficiency while the Raspberry Pi had the best value.
Re:FP -- an ode (Score:3, Funny)
Oh, first poster, how great thou art,
Merrily leading us to a great conversation
Granting ceaseless discourse and insight
Sending bits into the net filled with love
Telling us the true inner thoughts of your soul
Fighting your way for meaning toward the front of the pack
Unnamed anonymous hero, I salute you
Re: My $2500 macbook can blow it out of the water! (Score:5, Insightful)
More importantly the originally raspi is based on a chip that was intended for a different purpose. The bcm2835 was first and foremost a video processor capable of hd video encode and decode.
The arm(which everyone benchmarks) has a simple role to play in the intended configuration: Run linux, so that you can write simple GUIs and send compressed video data to the videocore. 3d Graphics acceleration was probably and afterthought.
You're comparing apples and oranges because the raspi community repurposed the chip. The raspi2 arm is a bit better. The biggest advantage now is the price, but this comparison is ridiculous.
Re: (Score:2, Interesting)
My friend used to work at Broadcom, and told the following story:
Broadcom had a video decoder chip, but there was a lower limit on chip size due to the space needed to connect the terminals. So there was unused silicon; they stuck an ARM core there.
See also here about how it boots:
http://raspberrypi.stackexchan... [stackexchange.com]
Re: (Score:2)
The older systems also had more ram and pci (Score:3)
The older systems also had more ram and pci / agp / some even had pci-e.
Re: (Score:2, Funny)
Re:The older systems also had more ram and pci (Score:5, Interesting)
This is my main problem with the Raspberry Pi. It's very well suited to certain tasks. But there are many places where it falls behind even very old technology.
I had one that I wanted to use as to download my torrents. It turns out that downloading to the SD Card caused the thing to lock up because it was writing data faster than the device could handle it. I was able to get around this problem by writing to a USB stick. It no longer crashed, but there was still a bottleneck writing to disk, which caused the torrents to download significantly slower than they did on my desktop.
It wasn't even due to bad memory stick or SD card. It was similar SD card and memory sticks that I used on my tablet that allow full speed torrent downloads. But something about the architechture of the Raspberry Pi that caused any kind of extensive writing to the SD or USB to cause a CPU spike every few seconds.
These tiny ARM computers probably have enough CPU and RAM at this point to run as a desktop. But until they get proper interfaces for hooking up storage and networking, they won't be of much use to anybody.
SD memory card speeds vary widely (Score:2)
Somewhere on the Internet(tm) I recently read an article comparing SD memory card speeds with the RPi. They varied by as much as 10x. For the most part, brand-name cards did better, and IIRC, medium-sized cards tended to be faster (small ones are usually cheap, large ones are trading speed for size), but it varied a lot - as long as the card's write speed was fast enough for a typical video-camera to record in real time, that's all the manufacturer cared about, and read speeds have bigger numbers so those
Re: (Score:2)
The droids you're looking for:
http://www.midwesternmac.com/b... [midwesternmac.com]
Quite eye-opening. I generally use the Pi-branded Samsung card, which is middle of the road on this test. Perhaps I should treat myself to an EVO+ for Christmas.
Thanks! (Score:2)
Yup - that's the one. (And of course I read it just after I'd bought a couple of SD cards for my RPi and other devices :-)
Re: (Score:2, Insightful)
This is my main problem with the Raspberry Pi. It's very well suited to certain tasks. But there are many places where it falls behind even very old technology.
That's my biggest beef with it. It feels like a learning-computer to me, something for a student to use to study very specific aspects of system design. A physical counterpart to Minix. Yet it seems to be touted as something more capable than that.
What I need in low-end is beyond the capabilities of this device. I'm willing to accept the power consumption penalty of old equipment because I know that the old equipment won't let me down. The Raspberry Pi could be free but if it doesn't do what I need
Re:The older systems also had more ram and pci (Score:5, Insightful)
It feels like a learning-computer to me
Funny you should say that.
Re: (Score:2)
That's my biggest beef with it. It feels like a learning-computer to me, something for a student to use to study very specific aspects of system design. A physical counterpart to Minix. Yet it seems to be touted as something more capable than that.
-- emphasis mine
By whom? Certainly not by the Raspberry Pi Foundation [raspberrypi.org]. They describe it as a tiny and affordable computer for kids.
Re: (Score:1)
> then it is useless to me
That's OK, you are probably useless at the things that RPi is good for.
From what I have gathered... (Score:1)
The Pis use the usb bus for everything, except Video (HDMI/Composite), Audio, GPIO, and maybe the SD port. Ethernet, wifi, etc. use the USB bus, which doesn't respond well to heavy IO loads, whether due to hardware failings of the broadcom soc, design issues with the board, or power limitations of design.
Re: (Score:2, Insightful)
Why is everyone trying to build a desktop PC out of these things? As a $5 embedded platform they are massively overpowered for all sorts of projects, yet the only thing these articles ever rate it on is PC type tasks.
Maybe replace the headless linux box in closet (Score:2)
Why is everyone trying to build a desktop PC out of these things? As a $5 embedded platform they are massively overpowered for all sorts of projects, yet the only thing these articles ever rate it on is PC type tasks.
Actually they may be competitive against old desktop PCs that have been retired to the closet as headless Linux servers. Read/write the data for the device being provided to a NAS box that has been mounted. Might work for a personal/home server. I'm thinking non-media applications, source code control, documentation wikis, etc. Less power and much quieter than a repurposed desktop. Again, note a NAS box has the data, the sdcard only the operating system and configuration.
Re: (Score:2)
Not if the desktop has other I/O channels other than USB. The R.Pi only has USB, and fails hard at trying to handle multiple devices.
Re: (Score:2)
Not if the desktop has other I/O channels other than USB. The R.Pi only has USB, and fails hard at trying to handle multiple devices.
The only thing to plug in are ethernet and a USB cable to the uninterruptible power supply. My pis are older and have ethernet, which I understand is implemented via the USB controller.
Re: (Score:2)
Using both at once is what causes it to fail.
Re: (Score:2)
Using both at once is what causes it to fail.
That is not my experience. I've used a pi running raspbian as an alternative to a linux box with simultaneously active ethernet and USB ports, transferring data to/from both for weeks at a time. Also with a USB based wifi adapter rather than ethernet. The USB devices plugging in had their own power. Some folks have problems by trying to draw too much power from the pi via USB.
Re: (Score:3)
I use an RPI as my primary Internet facing server and for other tasks, and matches the performance of a rackful of old Sun equipment. And runs from off-grid solar power, ie orders of magnitude less than those Sun servers:
http://www.earth.org.uk/off-gr... [earth.org.uk]
Oh, and I can shove it in a small cupboard, rather than taking a whole room.
And I run it fanless with entirely solid-state media, so it's quiet.
So, smaller, quieter and vastly more energy efficient and cheaper and people are WHINING?
Gah
Damon
Re: (Score:2)
USB based disk and network have alot of cpu overhead. Also if there only 1 USB bus then that will be come a bottleneck.
Networking needs to be on the pci-e bus or some other cpu bus and not USB. Also maybe disk as well.
Re: (Score:2)
A lot of CPU overhead? I use USB for realtime industrial communications and what you say is true only with completely braindead implementations (either host hardware or drivers for it). From what I can tell, a lot of Linux USB drivers, for both devices and hosts, are written by people who just don't dig asynchronous, realtime, low-overhead streaming of data. It's entirely a software problem, I'd say. Any reasonable USB host will support either DMA or shared memory, so byte pushing is free.
Re: (Score:3)
But there are many places where it falls behind
I know, right? If I am spending a whole $5 on a computer, it shouldn't have any limitations.
Re: (Score:2)
I know, right? If I am spending a whole $5 on a computer, it shouldn't have any limitations.
That's what I'm trying to figure out here. What exactly were they expecting. This is like buying a scooter and complaining that it fits fewer people than a bus.
Re: (Score:3)
But until they get proper interfaces for hooking up storage and networking, they won't be of much use to anybody.
I wouldn't even think of using a Raspberry Pi for storage or networking, that's not what they're for, as you've discovered.
As a media player plugged into my TV running Kodi, it works really well however. I have another plugged into a monitor displaying a slideshow.
A desktop or even a laptop is not going to be as good for something like that because they're too big and power hungry.
Re: (Score:2, Insightful)
Re: (Score:2)
To be fair, on a 120Mb (really more like 60-70Mb on a good day with prevailing wind) connection Bittorrent can make even an SSD struggle. Pulling down 10MB/sec of small, random writes and sustaining that for tens of minutes is a pretty heavy load.
Disk I/O performance is probably the biggest weakness of the Pi. For tasks that are not disk bound though, it can actually do a pretty good job as things like a network server or low power sensor platform. Even as a home file server, using USB drives, if high perfo
Re: (Score:2)
USB is the proper interface. The driver for the SD card basically sucks.
No performance/watt on the x86 side. (Score:2)
While in some benchmarks, the Pi 2 can keep up, it's clear it's overshadowed at times by the P4s.
What I want to know is, if we have a cluster of Pi 2s that consume the same amount of power as the P4, how different are the results?
for same watts, Pi wins X 220, but pointless (Score:5, Insightful)
If you had enough Pi boards to use the same amount of electricity as the Pentium 4, the stack of Pis would have 220 times as much computational power.
One P4 runs a bit faster than a Pi, and uses a LOT more power.
Of course that fact is probably not of any practical use. There are use cases for which a Pi is the right tool for the job, there are uses for which a typical desktop is the right tool for the job, and there are use cases for which the Arduino is the right tool for the job - and there isn't that much overlap. If you need a lot of computing power, you use a powerful processor, not a bunch of Raspberry Pi boards.
The power consumption does point out that there is virtually no good use case for a P4 - it's cheaper to buy a newer CPU than to power a P4.
* unless it is rarely powered on (Score:3)
There is a footnote to my comment that there is virtually no good use for a P4. It might make sense where the machine is a) free and b) rarely powered on. I actually have such a use case; my Christmas light controller is only powered on for a few hours per year. Therefore the power savings of buying something newer may not offset the cost to do so.
Re: (Score:2)
There is a footnote to my comment that there is virtually no good use for a P4.
Yep too loud for a headless server in the closet. I actually left the even older P2 in the closet when the P4 was retired. Still have the P4 in case I have to test/debug on a 32-bit system. As you say its paid for, its here, why bother setting up a 32-bit VM for testing/debugging? FWIW, I eventually replaced the P2 motherboard with some Intel all-in-one motherboard. Approximately $75, add RAM, passively cooled. I did add a low RPM quiet fan inside the cavernous case (the old P2 box, an Antec case with a qui
Re: (Score:3)
Just run whatever headless Linux-ey things you need on the NAS box.
My (inexpensive) Asus router, for example, has many times the RAM, performance, and storage of the first multi-user Linux server I ever hung off of the Internet, at a tiny fraction of the power consumption.
missing the point - the electric bill means (Score:2)
Way to miss the entire point twice. Sure, you CAN keep using that P4, on that old motherboard. Depending on where you live and how much you use the AC, and local electricity prices, it's costing you $100-$300 every year to run.
You could instead pay LESS every year and getter BETTER performance a modern low-power, low cost chip. You're paying more and getting worse performance. You -can- do that. You -can- hit yourself with a hammer too.
Re: (Score:2)
Of course that fact is probably not of any practical use. There are use cases for which a Pi is the right tool for the job, there are uses for which a typical desktop is the right tool for the job, and there are use cases for which the Arduino is the right tool for the job - and there isn't that much overlap. If you need a lot of computing power, you use a powerful processor, not a bunch of Raspberry Pi boards.
Well, a desktop from 2004 was far from useless. I'm thinking more in the direction of the Microsoft Lumia 950 + Display Dock, what's lacking is "universal apps" but you got a big screen (1920x1200 @ 60Hz) + mouse + keyboard, a dual core 1.8GHz ARM w/H.264+HEVC hardware decoding + 3GB RAM + GPU that beats Intel Bay Trail and AMD Mullins. Sure you can't compare it to a high end desktop but it might be more than good enough for many. The same kind of people who didn't need a desktop might in the future not nee
electric and AC bill. P4 is a heater, Pi is cheape (Score:2)
> Well, a desktop from 2004 was far from useless.
It WAS not useless at the time. In fact, it was sometimes worth paying $100-$1200* per year to power it and run the air conditioning to get rid of the heat it generated.
Now, you can get similar performance from a $40 machine that uses $1-$10 of electricity. Given the choice of spending $100+ to use a P4 for a year or spending $41 to use a Pi for a year, the P4 loses.
Further, rent for apartment or office space is about $1/month or so. The P4 takes up $50
Re: (Score:2)
The whole point of these things is that they're cheap and tiny, which means building them into things easy and practical. That in turn creates a community which swaps ideas and designs for them. If you want to talk pathetic computing power, there's the Arduino; but computational power isn't the point of the thing.
Re: (Score:2)
> the stack of Pis would have 220 times as much computational power.
Is it time for that old /. meme? :-)
Imagine a beowulf of pi !
That review is fake! (Score:1)
The RPiZ is not available online anywhere in the USA or Canada, which means it's not available anywhere else in the multiverse!
Re: (Score:2)
That's not online.
Re: (Score:2)
If you want to pay more than $5 there are several eBay auctions for RPiZ running right now. Yes, Adafruit is out of stock on them, but they're not the only source, again, if you're willing to pay more than $5.
They were hot back then (Score:2)
Re: (Score:2)
I look back and wonder what the heck Intel was thinking with NetBurst. Sure, hindsight is 20/20, but how did Intel's product roadmap get THAT messed up?
And what's crazy is they still sold tons of the things.
Re: (Score:2)
Wasn't Core derived from Pentium M?
Re:They were hot back then (Score:4, Informative)
Re:They were hot back then (Score:5, Informative)
Yup, rough lineage:
P5: Pentium -> Pentium MMX -> nothing for a long time -> early Atom -> MIC
P6: PPro -> P-II -> P-III -> Pentium M -> Core/Core2 -> Nehalem/Westmere -> Sandy/Ivy -> Haswell/Broadwell -> Skylake
P68: Pentium4
Re: (Score:2)
Somebody should sticky this.
Re: (Score:2)
Mod +1 informative
Re: (Score:2)
Tualatin Pentium IIIs ran circles around the early Pentium 4s
Yup, I remember that, and I was quite puzzled by it for a little while. I had a tricked-out dual-socket PIII system that was really hard to beat for a while.
Re: (Score:3)
It's running on a Pi Zero.
"Power Efficiency" measurements misleading. (Score:1)
They're just a way to make slower chips look better when they really aren't. If it gets the job done faster, what's the real issue?
I'll take a Netburst P4 over the R.Pi any day just for spite and proper USB implementation.
Re:"Power Efficiency" measurements misleading. (Score:5, Informative)
They're just a way to make slower chips look better when they really aren't. If it gets the job done faster, what's the real issue?
Not every task needs huge computing power. If the Pi gets the job done fast enough while burning less power than the cooling fans in your P4 system, taking up a fraction of the space and only costing ~$60 (by the time you've added a case, PSU and SD card), what's your issue?
I've got an original Pi running DNS, DHCP for my home network, and a Pi2 hooked to my lounge TV as a media center frontend served by a PC in the spare room (I suspect the Pi chipset was made for set-top-box use - it can decode 1080p mp4 without breaking a sweat) - the Pi 1 struggled a bit with the i/o throughput but the Pi2 handles the necessary with ease. Dedicating a P4 to either of those tasks - or making your toy robot twice as big so it could take the weight of a P4 heatsink - would be ridiculous unless you also needed to supplement your central heating system.
Still makes a great server (Score:2)
I thought about using an old PC or server but the noise, power and space requirements, long-term reliability, cooling requirements and electricity bills are kind of off-putting.
WTF are you comparing against NetBursts for? (Score:2, Interesting)
"The Pi Zero had an average power consumption of 2.7 Watts and the Raspberry Pi 2 was at 3.5 Watts";
Ok then compare that with a 3W Intel Atom E3805 [intel.com] if you want a modern performance per watt metric. Guess what the outcome will be?
Re: (Score:1, Insightful)
> Ok then compare that with a 3W Intel Atom E3805
TDP is a heat dissipation requirement and is _not_ the power draw of the CPU or of a system.
memo: chip wars are over (Score:2)
the x86-64 won. you can have a low power one, a fast one, a multicored-one, a supercomputer of them. sad since a bit nonlinear architecture but it is what it is.
Re: (Score:2)
Yeah, but when you want low-power, fast, tiny and cheap at the same time, there's not much offered with x86-64.
Raspberry Pi in general (Score:5, Informative)
Have not played with one of these but I have several A and B+ being used daily.
One is my voip system using Nerd Vittles PIAF http://nerdvittles.com/?p=1015... [nerdvittles.com]
The other does my weather station
http://weewx.com/ [weewx.com]
The other does my BBQ controller
https://github.com/CapnBry/Hea... [github.com]
Sure there are many more uses.
The new board may save a bit in my new builds will see...
All running quite fine...
So yes they have their place, low power, and reliable, no fan.
I test all these parts (Score:5, Interesting)
Re: (Score:1)
Fortunately 10 is the final Windows version, so Intel can't cut out the support based on OS version, ever!!
Re: (Score:2)
Re: (Score:3, Informative)
Re: (Score:2)
Performance per watt comparisons are only really useful within certain brackets. For example, if you want a solar powered server then an Intel NUC is hopeless, but the RPi works well. It's only useful to compare performance per watt between the Pi and other very low power minimal SoC boards.
Power draw of the Pi Zero (Score:2)
If you are interested in the power consumption of the Pi, you should probably check out this: http://www.midwesternmac.com/b... [midwesternmac.com]
car analogy (Score:2)
Yes a bicycle uses less gasoline than a automobile, and is far more efficient. But a bicycle is not practical in every situation.
Re: (Score:1)
Yes a bicycle uses less gasoline than a automobile, and is far more efficient. But a bicycle is not practical in every situation.
But does a bike's rider emit less CO2 than a car?
Actually, I finally bothered to google that, and yes, it is better. Bike is about 200MPG and walking about 60MPG
http://physics.ucsd.edu/do-the-math/2011/11/mpg-of-a-human/
That said, if you stick 4-5 people in a Prius for a long drive, it's better than them all trying to bikes or walk there. Assuming they didn't need the eerrcise int he first place... which is rarely an issue in countries that can afford a Prius.
Re: (Score:2)
If you include the total cost of maintenance and construction, the bikes will come out way ahead.
Re: (Score:2)
Re: (Score:2)
Like any good /. car analogy, it's completely inaccurate.
The R-Pi has lots of competitors (Score:2)
This is a ridiculous yet interesting comparison. (Score:2)
Raspberry Pi was never designed for heavy workload, this is why this comparison is ridiculous - it is a bit like comparing apple and oranges. However the comparison is still very interesting, as it tells us how far our technology has advanced.