Disappointed with unexpectedly low clock speed

I ordered the basic model which states the CPU is "Intel Atom X5-E8000 2.00 Ghz" so imagine my surprise when my dmesg shows the actual clock speed is "CPU: Intel(R) Atom(TM) x5-E8000 CPU @ 1.04GHz"! Turns out that the number they list on their site is something called "burst speed" which is .... what exactly? That seems openly misleading to the customer, although it is partially my fault for not researching the processor (I didn't expect a company to present it so misleadingly).

I bought one of these expecting good performance. I write multi-threaded bioinformatics programs and was considering scaling up with some UDOOs, but it turns out the UDOO is actually slower than my raspberry pi 3!

Here's my rpi3 results from a standard run of one of my programs:

real 2m40.261s
user 10m2.740s
sys 0m0.330s

And here's the results from the same program, with the same arguments, compiled the same way on the UDOO:

real 3m11.864s
user 11m58.485s
sys 0m0.328s

(On both platforms, I compiled my program with -O3 -pthread on GCC)

Considering I could buy two or three rpi3s for the price of the UDOO X86 basic model, and the UDOO draws more current than the rpi3, I am pretty disappointed with this result. Live and learn, I guess.

 

Strange - I am getting much better performance on the UDOO compared to a Raspberry Pi 3. What OS are you using on the UDOO and where is that installed?

 

What if you do a sysbench -cpu on both systems? Do you get the same results?
Would be nice to see the difference for other Udoo X86 CPU's too.

$ sudo apt-get install sysbench

$ sysbench --test=cpu --cpu-max-prime=20000 --num-threads=4 run

 

Well, rpi3 is 1.2 GHz and UDOO X86 Basic is only 1.06 GHz. (Other models are 1.6 GHz, i.e. Advanced, Advanced Plus and Ultra)

 

The X86s are sold as 'up to 2.56´ GHz and 'up to 8GB RAM'. The amount of RAM is what you actually get, but the speed of the CPU is stretching the truth quite some part it seems.

From https://www.notebookcheck.net/Intel-Pentium-N3710-Laptop-Processor-Benchmarks-and-Specs.169233.0.html

 

Yes, indeed. Unfortunately their shop page is quite misleading and pretends that the cpu speed of the UDOO X86 Basic is 2 GHz.

I have seen other SBC companies, such as UP BOARD, use the "up to" weasel words, but the UDOO page literally does not say that! See the attached screenshot. "Stretching the truth" is generous

 

I'm using FreeBSD on the UDOO, and Raspbian Linux on the rpi3 (I have tested this program on both Linux and FreeBSD on a desktop PC at work and notice no difference when running this program - it is written in ANSI C and compiled with the newest version of GCC on both operating systems - so the different operating systems should be negligible). The OS is on an M.2 ssd on the UDOO, and an SD card on the rpi3.

edit: and in fact, the increase in processing time on the UDOO is roughly consistent with what one would expect when going from the 1.2 GHz rpi3 to the 1.0 GHz UDOO. Unfortunately I bought the UDOO thinking it was 2.0 GHz.

 

The homepage uses the word 'up to 2.56' GHz. "Up to" is in this case a weasel word, and as shown it suggests that, like the RAM, the CPU will be 2.56 GHz.

 

I started a new thread for CPU benchmarks: RPi 3 vs. UDOO X86 CPU benchmarks

 

I dont think you can compare an ARM cpu with an X86 CPU based on GHz at all!
Why is Udoo claiming that the Udoo X86 (Ultra, I suppose) is 10x faster that the Raspberry PI 3? Difference in GHz is not even 2x faster.
So a sysbench comparison would give more answers about comparing these boards. If we have sysbench results we can find reasons why for example one Udoo Basic has other results than an other one.
Sysbench results of the Raspberries can be found everywhere https://www.raspberrypistarterkits.com/comparison/raspberry-pi-3-pi-2-pi-b-benchmark-review/

 

Good action!

 

I by the way agree that the Udoo way of presenting CPU specifications is not as it should be.
It at least should mention normal and burst GHz values. Like the manufacturer Intel is doing.

 

Interestingly sysbench result is 10x faster as claimed -- post.

 

Now we are waiting for the sysbench test of the Udoo X86 Basic by @Ethin E.

 

Given that the results of speedtest seem to indicate that the 10x difference is not out of thin air the problem might be in the implementation, a bottleneck that makes code wait, on I/O, on result of another thread, and removes any speed difference.

 

My udoo mostly runs on the burst speed 2ghz-2.5ghz when I'm on Windows 10

 

I remember there being at least one post on this forum earlier saying that UDOO X86 is slower than RPi. Can't find it now but I think it was about some emulator which probably wasn't properly optimized for x86.

 

It is a little bit hidden in another topic: How to install Windows 7 or 10 onto a micro SD card?

 

I'm running it now and will post in the other thread.

However, I'd like to point out that this benchmark is useless to me, because I specifically wrote the program I use for my research, and optimized it to use POSIX threads for multithreading. To that end, my meaningful benchmark is running the program with a known configuration, and when I did that, the Rpi3 was faster by two minutes of cpu time.

 

Well after running the benchmark (see other results), it is clear that my UDOO can be faster than my Rpi3, but making my own programs faster would require some programming knowledge I don't have. My implementation utilizes POSIX threads, which ensures that all CPU cores are used. In fact, while the program is running, if I monitor it in top, the CPU usage is 400% (4 cores) on both my UDOO and Rpi3.

My own program performs a lot of mathematical computations, and the UDOO is generally configured to be faster (more minimal operating system, SSD instead of SD), so I am surprised by the sysbench results.