ARM Cortex-A9 PandaBoard ES Benchmarks

Phoronix: ARM Cortex-A9 PandaBoard ES Benchmarks

The performance of the dual-core ARM Cortex-A9 OMAP4460 configuration as found on the PandaBoard ES is quite commendable and in this Phoronix review the dual-core 1.2GHz ARM system with PowerVR SGX540 graphics is being compared to several Intel Atom, Pentium M, and Core Duo configurations running Ubuntu Linux throughout. To spice things up, the pre-production OLPC XO-1.75 was also thrown into the testing mix with its single-core ARMv7 800MHz Sheeva processor.

http://www.phoronix.com/vr.php?view=16888

Not too bad for an ARM cpu, power consumption would also be nice. The Core duo still owned it in raw preformance, let alone c2d, 1st gen I series and sandy bridge.


I wait for the day my phone can be a renderslave to my pc.

Were these all multithreaded benchmarks, then? I missed if you mentioned or not. And more important (IMO) what's the performance per watt?

How about rounding articles off with things like power consumption, likely availability of the processor in netbooks/tablets/phones etc. and a conclusion?

I bought the original PandaBoard with the aim of using it as a HTPC/home entertainment system. Apart from the PowerVR GPU, the Panda also packs hardware video acceleration from TI (called the ducati engine). The Ubuntu 11.10 PPA now has a working set of libraries (libdce etc) and firmware to play HD video. As far as I can tell only gstreamer supports libdce currently, so no ffmpeg based apps will work. Totem works ok with the occasional crash.

Great to see some ARM vs x86 benchmarks.

So the slowest of the slowest single-core Intel Atom processor is in most cases still faster than a dual-core ARM processor. Sort of what I expected but nice to see some benchmark results confirm this.

First; because this is ARM it might help to clue us in on the processors structure a little bit. That is what it has execution unit wise. There is so much variability here that one needs to be reminded. In the intel world it is pretty easy to remember that say AVX is on Sandy Bridge and Bulldozer. With ARM it is hard to remember who implemented what.

Second; it would pay to explain your negativity towards the GPU. I can imagine it has to do with its closed nature, but one shouldn't have to guess.

Third; you need to include an AMD solution in the mix. Specifically a Zacate based board. We already know that Core is several times faster, what we need to understand is how processors targeting the same general market fare.

Fourth; with ARM all over the place (compared to ATOM) performance wise it might be interesting to try to determine the impact of compiler performance on the numbers. That is if at all possible build with GCC and CLang. If compiler maturity on ARM is an issue this might highlight where one compiler is weak with respect to the other. Or maybe the ARM architecture just has inherent weaknesses relative to ATOM. Which goes back to detailing exactly what is included on chip with this specific ARM implementation.

In any event keep up the good work. It is actually a bit of a shock that ARM is now delivering this level of performance. With larger caches and 2GHz clocks A9 would likely beat ATOM at far lower powers.

Firstly,

Thanks for the article. However can you please please please put in power metrics next time? It's one thing to show how ARM stands vs x86 in terms of performance but for my own curiosity the benches are begging for performance per watt.

If the ARM cpu can do 25% of the performance at a 100th of the power that's a big win in my eyes (well for certain scenarios of course).

Cheers.

Quote:

---

Originally Posted by bradmccormac

Firstly,

Thanks for the article. However can you please please please put in power metrics next time? It's one thing to show how ARM stands vs x86 in terms of performance but for my own curiosity the benches are begging for performance per watt.

If the ARM cpu can do 25% of the performance at a 100th of the power that's a big win in my eyes (well for certain scenarios of course).

Cheers.

---

I explained in the article why I didn't do power tests...

Quote:

---

Originally Posted by Michael

I explained in the article why I didn't do power tests...

---

after doing phoronix for all these years you should have learned that people just look at the pictures :p:p:p:p

Quote:

---

Originally Posted by Michael

I explained in the article why I didn't do power tests...

---

(As well as why you don't like the graphics situation)

It is possible to circumvent the issue with some laptops--at least on Thinkpads, you can set the BIOS to default to external display only.
With some other laptops, you can switch the video to an external monitor. It still wouldn't be a perfect comparison, but it beats no numbers.

What would be a fair comparison are the various small x86 boards (mainly VIA and Intel, perhaps Intel/ION)--they'll be cheaper (newegg has $60-80 without ION, ~ $120+ for the ION boards) and roughly similar hardware.

Quote:

---

Originally Posted by Ibidem

It is possible to circumvent the issue with some laptops--at least on Thinkpads, you can set the BIOS to default to external display only.
With some other laptops, you can switch the video to an external monitor. It still wouldn't be a perfect comparison, but it beats no numbers.

---

Exactly.

This was a great review, but it's of limited relevance without power consumption figures.

Quote:

---

Originally Posted by Ibidem

What would be a fair comparison are the various small x86 boards (mainly VIA and Intel, perhaps Intel/ION)--they'll be cheaper (newegg has $60-80 without ION, ~ $120+ for the ION boards) and roughly similar hardware.

---

And when there's so many non-Premium AdBlock'ers, those systems come from where?

Quote:

---

Originally Posted by aragon

This was a great review, but it's of limited relevance without power consumption figures.

---

I've been scouring the PandaBoard site and the FAQ suggests that this board spikes over 1.2A under heavy load, and their AC power supply suggestion is for 4A max current handling. These currents are at 5V.

So peak power consumption of around 10W would be my guess, FWIW.

Quote:

---

Originally Posted by Michael

And when there's so many non-Premium AdBlock'ers, those systems come from where?

---

Do you honestly think that you would have much revenue from the people that use AdBlock if they didn't use it? Revenue comes from clicks, not views. The people that use AdBlock wouldn't click, so you would get no revenue from them regardless.

Quote:

---

Originally Posted by Shining Arcanine

Do you honestly think that you would have much revenue from the people that use AdBlock if they didn't use it? Revenue comes from clicks, not views. The people that use AdBlock wouldn't click, so you would get no revenue from them regardless.

---

You're incorrect, Phoronix is paid on an impression basis.

armel or armhf?

Something that should be noted (that I didn't see in the article) is that the OLPC XO 1.75 is running a soft-float ABI now, which will be replaced with a hard-float ABI hopefully soon. So, all the SciMark benchmarks are not representative of the hardware, since they're not performed using hardware floating point instructions.

I'm eagerly awaiting arm boards with sata, pcie, and standard ddr3 slots :)
probably not things they care to target yet ...

usefull review, people ask for arm to be benchmarked often in forums around the web

Quote:

---

Originally Posted by Michael

And when there's so many non-Premium AdBlock'ers, those systems come from where?

---

Less beer, more work worth reading. You'd have more readers if you put more effort into rebuilding your reputation with the general public. Try linking one of your articles on any tech site with Linux users, it's often considered a laughing stock.

If your rep was better you'd have more readers and more readers is more clout when hitting up companies for review samples, more samples means more articles, more articles means more page views. Rinse and repeat.

I like many others have allergies, the most relevant here are to overzealous advertising and to bad journalism. I'll put up with one, but never both.

At least you've stopped trying to review PSUs, You have neither the hardware nor the knowledge to do that justice. Just wish allot of other sites would stop cluttering Google with PSU reviews that say little more then "well it has wires and it turns on..."

Quote:

---

Originally Posted by Soul_keeper

I'm eagerly awaiting arm boards with sata, pcie, and standard ddr3 slots :)
probably not things they care to target yet ...

usefull review, people ask for arm to be benchmarked often in forums around the web

---

Yeah, I'm wanting to see what those fabled quad 2.5Ghz ARM SoCs can do, especially compared to an AMD Z-01(5.9w dual 1Ghz w/ 276Mhz 80SP GPU, DDR3 1066) but there have been a few ARM servers http://www.linuxfordevices.com/c/a/N...stems-R1801e-/ --|-- http://www.geek.com/articles/chips/c...rver-20110314/ --|-- http://www.theregister.co.uk/2011/10...e_server_chip/ HP's Moonshot etc.

Yes, I know nobody with AMD will read this
Side note, AMD guys, where is all the hardware based on the Z-01, C-60 and E-450? These refined versions of the Bobcat series APUs should be taking the low end market by storm by now since they where launched months ago! Furthermore, why can't I find a high end gaming laptop thats all AMD? For years the best AMD mobile CPUs where paired with no better then an HD*670 GPU, why is there never even an all AMD desktop replacement laptop with a top end mobile GPU like the HD6990m?

Quote:

---

Originally Posted by Kivada

Less beer, more work worth reading. You'd have more readers if you put more effort into rebuilding your reputation with the general public. Try linking one of your articles on any tech site with Linux users, it's often considered a laughing stock.

If your rep was better you'd have more readers and more readers is more clout when hitting up companies for review samples, more samples means more articles, more articles means more page views. Rinse and repeat.

I like many others have allergies, the most relevant here are to overzealous advertising and to bad journalism. I'll put up with one, but never both.

At least you've stopped trying to review PSUs, You have neither the hardware nor the knowledge to do that justice. Just wish allot of other sites would stop cluttering Google with PSU reviews that say little more then "well it has wires and it turns on..."

---

I'm not complaining at all about traffic levels at all considering they trump most all other Linux news sites, I'm complaining about AdBlock users.

Quote:

---

Originally Posted by RussDill

armel or armhf?

---

Yes, please specify.

Or do a comparison between armel and armhf.

Quote:

---

Originally Posted by Michael

And when there's so many non-Premium AdBlock'ers, those systems come from where?

---

OK, I guess I'll comment one or two more lines out of /etc/hosts (currently 4803 lines and 143,888 bytes):o
(Before I got a hosts file, phoronix was almost unusable without switching to xlinks2--once I clicked a link to phoronix, waited ~5 minutes, went and ate dinner, came back, and waited a few more minutes for it to finish loading :rolleyes: --in case you're wondering why it takes so long, the CPU is an Atom N270 that spends most of the time at 800 MHz, in an Aspire One)

Michael,

Thank you for the interesting article, its great to have some ARM benchmarks! I'd love to see more ARM coverage on Phoronix. Another system you could benchmark would be the Beagleboard and of course the Raspberry Pi (when its released).

BTW.. contgrats on making the frontpage of Slashdot on this article!!

Quote:

---

Originally Posted by Michael

I explained in the article why I didn't do power tests...

---

Seems you did :-)
I was too focused on numbers . Sorry :-(

Quote:

---

the x264 benchmark, in the stock configuration

---

Well, that does mean it uses hand-optimized assembler code on x86, but not on ARM, right?

Also, it's understandable that power consumption tests are quite difficult with such different hardware, but then maybe such conclusions should be a bit more cautious:

Quote:

---

While this dual-core ARM Cortex-A9 platform was generally not competitive with the Intel Atom hardware[...]

---

I think it is very competitive: If it uses less than half the power of the Atom then just throw two more ARM cores in and there you have the competitive CPU, at least for multithreaded use cases.

Quote:

---

Originally Posted by Michael

I'm not complaining at all about traffic levels at all considering they trump most all other Linux news sites, I'm complaining about AdBlock users.

---

Yet when any of us Adblockers tells you what you can do to get us to whitelist Phoronix you brush us side to do nothing but whine that we're somehow costing you obscene amounts of money, which isn't even remotely possible, since you're apparently spending more on beer every week then what us adblockers would generate in a month.

As others have stated, if you truly feel you're being robbed, become a paid subscription only site, I'm sure it'll work just great for you...

Quote:

---

Originally Posted by thofke

Yes, please specify.

Or do a comparison between armel and armhf.

---

tests were done with the stock Ubuntu 11.10, as such it's armel. agreed that armhf comparison would be nice.

Thanks Michael, it's great to have some good ARM numbers.

Quote:

---

Originally Posted by ChrisXY

I think it is very competitive: If it uses less than half the power of the Atom then just throw two more ARM cores in and there you have the competitive CPU, at least for multithreaded use cases.

---

Not in perf/$. For 189$ you can get an E-450.

I should add, it's very likely TI subsidizes that, so the real price would be more than 189$. They are known to have done so in the past with their dev boards.

Quote:

---

Originally Posted by curaga

Thanks Michael, it's great to have some good ARM numbers.



Not in perf/$. For 189$ you can get an E-450.

---

I agree. It's not competitive not from a price/performance nor from a price/features point of view. The Raspberry Pi (US$25 to $35) looks much more interesting especially if they can get it to run XBMC.

diff --git a/arch/arm/mach-omap2/id.c b/arch/arm/mach-omap2/id.c
--- a/arch/arm/mach-omap2/id.c
+++ b/arch/arm/mach-omap2/id.c
@@ -231,7 +231,7 @@ static void __init omap4_check_features(void)
case 1:
default:
/* Standard device */
- omap_features |= OMAP4_HAS_MPU_1_2GHZ;
+ omap_features |= OMAP4_HAS_MPU_1_5GHZ;
break;
}
}

The above is a patch that will run the 4460 at 1.5ghz, replacing the 1.2ghz state. Pretty easy to see what I did is just fudge a board id check.. I put a GPU memory heatsink on the thing just for safety... I tried to drop the mv down to 1350 instead of 1375... but did it quick, and probably missed quite a bit of loose ends.. and thus that was fail..

Thermal throttling is pretty annoying when compiling a kernel.. ./kern/drivers/staging/thermal_framework/governor/omap_die_governor.c has the goods for the whens and whats of that.. The actual temp sensor (sysfs) readout is quite different from what the thermal throttling setup repots as values (in kmsg when certain temp reached).. so... thats a bit odd.. temp1_input seems to be more realistic.

Currently, this setup is painful.. Lots of glitchy things happening with X when using the latest pvr crap.. unity is... unity.. but even worse when such tomfoolery is going on.. For anyone who is setting their PB-ES up now, save yourself the time and pain -- and move the rootfs off of the sd card and onto nfs, usb-stick, usbhdd.. whatever.. just not the SD.. I don't care if its class 20000x mega-super.. it sucks.. morse code has more iops..

Michael,

Could you please also run a comparison of an ARM softfloat vs. hardfloat userlands? The standard softfloat solution has a huge performance overhead for EVERY single floating point instruction, which unnecessarily disadvantages the processor in comparison to x86. Also, various FPU options may be of interest (neon, vfp).

Thanks!

erm.. I'm probably doing something wrong/off.. but for comparison's sake..

7-Zip 9.20 Copyright (c) 1999-2010 Igor Pavlov 2010-11-18
p7zip Version 9.20 (locale=C.UTF-8,Utf16=on,HugeFiles=on,2 CPUs)

RAM size: 641 MB, # CPU hardware threads: 2
RAM usage: 425 MB, # Benchmark threads: 2

Dict Compressing | Decompressing
Speed Usage R/U Rating | Speed Usage R/U Rating
KB/s % MIPS MIPS | KB/s % MIPS MIPS

22: 1097 139 767 1067 | 25756 199 1166 2325
23: 1105 154 730 1126 | 25525 200 1170 2337
24: 1095 146 808 1177 | 25146 199 1171 2333

and.. i forgot to turn off a swap partition I had enabled on the sd card.. so I ran it again..

7-Zip 9.20 Copyright (c) 1999-2010 Igor Pavlov 2010-11-18
p7zip Version 9.20 (locale=C.UTF-8,Utf16=on,HugeFiles=on,2 CPUs)

RAM size: 641 MB, # CPU hardware threads: 2
RAM usage: 425 MB, # Benchmark threads: 2

Dict Compressing | Decompressing
Speed Usage R/U Rating | Speed Usage R/U Rating
KB/s % MIPS MIPS | KB/s % MIPS MIPS

22: 1112 153 708 1081 | 25799 200 1165 2329
23: 1106 154 732 1127 | 25574 200 1172 2341
24: 1102 156 757 1185 | 24896 197 1170 2310

Hey Michael,

Love the ARM articles! Would love for you to go more in-depth on the ARM architecture. Some ideas would include:

- Soft Float vs VFP.
- NEON vs VFP
- All of the above vs x86 (Atom)
- Power consumption of mainboard/cpu only (arm vs atom).

I know some of the benchmarks may be a bit tricky, especially considering Ubuntu 11.10 lacks good arm support; however, maybe with the latest GCC and kernel, some of these tests may become viable. Knowing what the performance limitations of the ARM core vs Atom and the power consumption of the ARM dev board vs. an Atom (ie N510) would be very interesting!

The phoronix got 1154 DMIPS out of their pandaboard with their dhrystone benchmark (DMIPS=dhrystone score/1575) . ARM got 2075 DMIPS out of their 830mhz single core. These benchmarks are also discussed in realworld tech forums, and someone there got 2.5 times phoronix result on the 7zip benchmark on equivalent 1.2Ghz dual core cortex A9. (exynos based). So maybe the operating system didn't get all the drivers right and runs it at fraction of real clockspeed.

Quote:

---

Originally Posted by josmala

The phoronix got 1154 DMIPS out of their pandaboard with their dhrystone benchmark (DMIPS=dhrystone score/1575) . ARM got 2075 DMIPS out of their 830mhz single core. These benchmarks are also discussed in realworld tech forums, and someone there got 2.5 times phoronix result on the 7zip benchmark on equivalent 1.2Ghz dual core cortex A9. (exynos based). So maybe the operating system didn't get all the drivers right and runs it at fraction of real clockspeed.

---

Well, fwiw.. the kernel included in the 11.11 linaro 'distro' had my pandaboard-es running at a smoking 920mhz.. I don't know if this is the case with the latest bundled kernels - as that inspired me to get the source and attain the proper clock speed in any way I could... so, its possible something like that is going on - that and running everything off of the sd card is awful (from a usability standpoint) - I wouldn't be surprised to impact benchmarks in ways you wouldn't think of..

Quote:

---

Originally Posted by gururise

- NEON vs VFP

---

NEON isn't a replacement for VFP as it doesn't support IEEE754. In particular no double precisions and just one rouding mode. ARMv8 ISA will have proper IEEE754 SIMD instructions.

BTW, here are my benchmark results from gentoo (hardfp) running on origenboard (dual-core ARM Cortex-A9 @1.2GHz): http://openbenchmarking.org/result/1...AR-1112277AR91