How fast are smartphones compared to desktops?

[happy medium]

I'd say Tegra 3 well outdoes the Wii's GPU, though in the end, Nvidia can't do something crazy like go for the best outright mobile GPU at the cost of highly crucial battery life. Balance is the key.

absolutely!
This guy said it best................

quote:
"Think we'll drag everything for the big switch off this, T3 has yet what it seems the fastest gpu on the androidmarknaden, and then it presents figures on 60fps here it is more than enough to push something onto a portable device. No developer will surely be so stupid that they release a game that is more demanding than the most powerful gpu on the market (android) can handle, there would be financial suicide!

Think nVidias ACE in our sleeve is its tegra zone and its good cooperation with developers, say what you like about consoles and the Apple, but if it is something they have shown abundantly clear is that if you are developing on a specific hardware, and adds much resources on optimizing one can get much better performance on a technical slower platform. Also hope that the additional cpu resources should lead to an increased use of physics and ai! It merged with nvidias support common hand controls can lead to Tegra 3 becomes a competent gaming platform yet ...

Understand, however nvidia in development terms, Tegra 3 was high however as it was to be a SOC, and Lighthouse churning cpu probably greater marketing value, one cannot have everything in a "small" and power-efficient package for an affordable price ..."

 

[Silverforce11]

It's so good to see Power VR still going strong.. one of my early 3D gpus was the Kyro II before the radeon/geforce dominance.

SGX543MP2 is beastly, never realized how good it is in comparison.

 

[aigomorla]

Phone - One legged turtle stuck in tar

Desktop - Millennium falcon

AHAHAHAHAHAHHAHAHA...

what happens if chewy forgets to fix the light speed drive is that Bulldozer then?

(sarcasm... you amd owners please dont take offense..)

 

[LiuKangBakinPie]

Tegra 2 is a dual-core design with no SIMD extensions, while the high-end Kal-El is a 4+1 core design (2+1 variants are likely to appear as well) that uses the NEON vector extensions. Kal-El’s five cores are logically identical, but physically asymmetrical. The four main cores use high performance (G) transistors, which come in low and standard Vt flavors and operate at ~0.9V to reduce dynamic power consumption (low Vt transistors are faster, but much leakier). The 'companion' core and most of the SoC are implemented with low-leakage (LP) transistors that are only available in standard and high Vt and run at ~1.1V nominal. Generally the LP transistors have 10-100X lower leakage power, but operate about 2-3X slower. Tegra 2 was a compact the 49mm2 chip and Kal-El is estimated at 80mm2. The additional 3 extra cores, a faster GPU and more powerful video acceleration account for the bulk of the extra area, as the cache and memory interfaces are largely the same.

http://www.realworldtech.com/page.cfm?ArticleID=RWT100511155012

Some good articles there

 

[Ben90]

Here is a dualcore phone with less than half the cpu/gpu power of Tegra 3 running Need for Speed shift with the Hdmi out.

At the end of the video~ 2min 50 seconds.
http://www.youtube.com/watch?feature=player_embedded&v=qjlTRBZ0LhI

With a lot of input delay.

 

[tweakboy]

Now a typical smartphone has a 1ghz processor.

I still remember when like a 200mhz Pentium 3 copper processor was wicked fast. And actually, I do think that a 200 mhz pentium could actually run a web processor faster than your typical smartphone.

but as to the title, how do they compare?

Ya P1 200Mhz Pentium was fast because all you did on the computer was copy files and watch JPG on it, and listen to audio CD and watch a DVD using your ubber 2D card. lololol :wub::wub::wub:

Desktops are a whole new beast. Smartphones are a little beast too.

Smartphones will never have a OS like Vista or 7 ,, it will always be watered down version and custom built Phone OS or DROID or I5 mac etc. The 1Ghz processor is not comparable to the 1Ghz processor of 12 years ago. Its different technology. But if you wanna run a SuperPI on smartphone and one on computer,, the computer workstation will blow it out of the water,,, 1000x faster.

I can tell you the 1Ghz processor in the smart phone is much much much, faster then a 1Ghz from 12 years or so ago. The Die shrinkage and number of transistors.
One day the technology might come where you will have a full blown OS on your smartphone. I doubt that day is any where soon. gl

 

[smartpatrol]

Ya 200Mhz Pentium was fast because all you did on the computer was copy files and watch JPG on it, and listen to audio CD. lololol :wub::wub::wub:

Desktops are a whole new beast. Smartphones are a little beast too.

Smartphones will never have a OS like Vista or 7 ,, it will always be watered down version and custom built for DROID etc. The 1Ghz processor is not comparable to the 1Ghz processor of 12 years ago. Its different technology. But if you wanna run a SuperPI on smartphone and one on computer,, the computer workstation will blow it out of the water,,, 1000x faster.

One day the technology will come where you will have a full blown OS on your smartphone. I doubt that day is any where soon. 2020 maybe,, gl

2020?? No way. Windows 8 will be running on ARM SOCs within the next year. . . meaning we'll easily have smartphones with hardware capable of running a full-blown OS at this time next year. The only reason we won't have full-blown desktop OSes on our smartphones is because there's no need for that on a smartphone (yet).

And even then, I wouldn't be surprised to see smartphones running full desktop-like OSes in a couple years. E.G. a Windows 8 smartphone that can be docked and used as a full Windows 8 laptop.

 

[MrX8503]

Keep in mind the Iphone4S processor is running at 1Ghz.
So overclock it to 1.4ghz and it's faster than the Pentium3...

The iPhone4S is running at 800mhz and its Dual Core. The P3 I listed is only single core.

Is that single or multi core? I want to guess single core when comparing the 2500k and D525 cores. Theoretically a 2500k @ 1.5 GHz according to those calcs would be around 3500 points, which is only twice as fast as the D525, despite the already 2 core advantage.

The Atom is benched as a dual core.

1750 / 2 = 875
10,400 / 4 = 2,600 (+197%/core compared to Atom)

The 2500k@4.5ghz is 3x faster than Atom per core. Not accurate to compare processors this way because not everything is linear, but its the best we got to compare to a mobile SOC.

 

[Haserath]

Smartphones are only fast due to the software... I have noticed software optimization can really affect the experience.

My old Iphone 3GS has a lag free browser while scrolling in the browser while even my Droid 3 has stutter, since the 3gs has hardware acceleration and Android doesn't yet. Sometimes I even think the 3gs is able to load some pages faster over wifi with the new iOS5 update speeding it up again.
------
The Android/i OS's themselves are lightweight OS's compared to Windows by far; good luck running Windows with 512MB, or in the case of iOS on a 3gs 256MB, of Ram.

Btw, has anyone ever ran an atom dual core? These things are only slightly better than the pentium 4 clock for clock. In single threaded tasks the P4 creams the atom thanks to the higher frequency.
http://www.tomshardware.com/reviews/atom-d510-pentium-4-nettop,2649.html

I would say the atom is still too slow for a desktop, and a Tegra 3 will only be close to atom thanks to having 4 cores.

ARMv15 cores will help speed things up, especially if they are at 2.5ghz like Qualcomm is projecting them to be.

 

[IntelUser2000]

According to Geekbench...

My 2500k @4.5ghz = ~10,400 (+1577&#37

That seems little low. Searching Geekbench shows 3.3GHz base clock 2500K's scoring almost 10k.

http://www.overclockersclub.com/reviews/intel_corei7_2600k_and_corei5_2500k/7.htm

You could tell Geekbench doesn't exactly represent how CPUs perform as Sandy Bridge without Hyperthreading is way faster than Atom.

Conroe based Celeron E1200 outperforms the D510 by 50-80%: http://www.anandtech.com/bench/Product/110?vs=71

Sandy Bridge is yet another 80-120% faster than that Celeron, all comparing at similar frequency and number of cores.

 

[Blitzvogel]

The iPhone4S is running at 800mhz and its Dual Core. The P3 I listed is only single core.



The Atom is benched as a dual core.

1750 / 2 = 875
10,400 / 4 = 2,600 (+197%/core compared to Atom)

The 2500k@4.5ghz is 3x faster than Atom per core. Not accurate to compare processors this way because not everything is linear, but its the best we got to compare to a mobile SOC.

And my point, as clearly laid out by both you and me is that the overall Geekbench score to me isn't and should'nt be taken as a serious measurement of processor capability. The individual numbers matter, though, as they tell the real story than some simple synopsis of a single score, it's just as vague as 3DMark06 scores. Essentially, has not i-Series architecture on a basic level, shown to be a good 2.5x+ more capable per clock per core than Atom series in real world tests?

 

[Acanthus]

I don't think this would be accurate, but it should give you an idea.

According to Geekbench...

iPhone 4S = ~620
Pentium 3 @ 1.4ghz = ~740 (+19%)
Intel Atom D525 @ 1.80ghz = ~1750 (+182%)
My 2500k @4.5ghz = ~10,400 (+1577%)

So basically a top end smartphone today is about as fast as a low end Pentium 3 in its heyday. The next quadcores should be as fast as or faster than the Atom.

Just nitpicking, but 1.4GHz was the fastest pentium 3 ever made. It was the top bin Tualatin that defeated the early Wilamettes.

 

[Larries]

Someone posted a link to tegra 3's GPU performance in Mobiles section.

http://forums.anandtech.com/showthread.php?t=2207224

If that benchmark is anything that resembles true GPU power of tegra 3, the tegra 3 will be at least 25% slower than 543MP2...

 

[cbn]

"console quality graphics " perhaps at 960x540 smartphone resolution. Anything more implies making the jump to 128-bit memory bus which SoCs have not yet made. The potential performance of an ARM core is quickly outgrowing its external interfaces and until 3D stacked RAM comes out to widen the bus there will be big constraints on practical throughput compared to the 256-bit CPUs and GPU memory systems we have today.

ultramobile OSes have shown that a great user experience for most apps can run fine on a tiny CPU/RAM system but high res games are still uncharted territory for ARM SoCs. Compare that to a $100 PCIe GPU that gets over 100GB/sec and can barely run BF3 at 720p and you start to see huge thermal and I/O barriers standing in the way of energy-conscious phone SoCs achieving the level of 3D performance of consoles.

I would call a A4 Llano or PS3 the minimum of console quality graphics and these systems get about 10 and 20 GB/sec to RAM respectively compared to 2-5 GB/sec from a 32 or 64-bit wide SoC mch.

Great post.

I'll have to keep a watch out for wider buses on future SOCs.

 

[cbn]

Imagine the power draw of a laptop with a SSD,LCD screen, and a SOC chip with a 5th idle core with motherboard. Laptops have a huge battery compared to a phone.

The Qualcomm Mirasol displays are supposed to debut in 2012. They use very low power in transflective mode, but are still capable of color video playback. (If the tech works it should go a long way to helping your System vs SOC power budget tug-o-war.)

More info here---> http://www.technologyreview.com/computing/39135/?mod=chfeatured

E-Reader Display Shows Vibrant Color Video

Mirasol's reflective display is being tested by device manufacturers, and could appear on shelves next year.

Tuesday, November 15, 2011
By Tom Simonite

Even as the processing power and download speeds of mobile devices surge, one component still lags behind: the screen. LCD panels use significantly more power than any other component of a phone or tablet because of their need to pump out bright light to form an image.

The only practical alternative is e-ink, the technology used in the Amazon Kindle; it consumes orders of magnitude less power but sacrifices color and the ability to change images fast enough for video playback or smooth game play.

Now, after years of waiting, alternative technology that promises the best of both approaches is finally edging closer to commercialization. During a recent visit to mobile chipmaker Qualcomm's headquarters in San Diego, Technology Review tried out a full-color, 5.7-inch Android tablet with a display that offers rich colors under bright light, close to those of an LCD and not unlike the pages of a magazine. The prototype screen was also responsive enough for video playback and for a game of Angry Birds; it can deliver up to 30 frames per second.

Because the device's screen uses ambient light, like a printed page or e-ink display, the power consumption is a tenth or less of that of a comparable LCD, although the display also features a built-in light for use in the dark. Known as Mirasol, the technology was created by a startup company, Iridigm, acquired by Qualcomm in 2004.
Advertisement

"In the market today, you have the iPad at one end and things like e-ink at the other end. This is really meant to bridge both of those worlds," says Clarence Chui, who leads the group at Qualcomm developing the new technology. "It is extremely low power, full color, and can be looked at wherever you go."

The Mirasol display makes color in the same way as the wings of iridescent butterflies or peacock feathers—by being an imperfect mirror that tunes the color of incoming light before reflecting it back to the viewer.

In a Mirasol display, this is done by small cavities known as interferometric modulators, tens of microns across and a few hundred nanometers deep, beneath the display's glass surface. "It's the air gap between the back of that glass and a mirror membrane at the bottom of the modulator that sets the color," says Chui. Each modulator's mirror membrane can snap upward against the glass when a small voltage is applied, closing the cavity and displaying a black color to the viewer. Mirasol modulators are made using techniques similar to those used to pattern metals and deposit materials in computer chip manufacturing.