mersenneforum.org

mersenneforum.org (https://www.mersenneforum.org/index.php)
-   Hardware (https://www.mersenneforum.org/forumdisplay.php?f=9)
-   -   This year's small computer (https://www.mersenneforum.org/showthread.php?t=19481)

fivemack 2014-07-12 16:33

This year's small computer
 
After a certain degree of fuss (the only EU supplier who had it in stock was SECO in Italy, and more than three months elapsed between order and delivery), my nVidia Jetson TK1 board arrived yesterday. Cost £178.49 including delivery.

Take it out of the box, plug in an Ethernet cable, turn it on, ask your DHCP server which address it most recently handed out, and you can ssh into it without trouble; building GMP6 required explicitly giving armv7l in the system triplet (it was auto-detected as 'neon' which the rest of the configure didn't understand), but after about an hour it's running gmp-ecm.

Not especially fast:

[code]
echo "(10^71-1)/9" | ecm -c 1 1e6
[/code]

takes 8.5 seconds on one CPU (also 8.5 seconds on each CPU if you run four in parallel), so five times slower than one core of i7/4930K, but twice as fast as the ODROID Cortex-A9 that I had previously. Comparing with a more realistic competitor, one core of Avoton C2750 takes 4.7 seconds; obviously 64-bit arithmetic is really useful for ECM!

There is a quite capable GPU, but I haven't found a build of the nVidia tools for the Ubuntu-14.04 that is installed on the board; so I have it doing polynomial selection without involving the GPU.

Haven't got a spare power meter to work out how much electricity it uses yet; there is a fan, but it's pretty quiet.

I imagine next year's small computer will be an AMD Seattle, with 64-bit Cortex-A57 processors, and I wouldn't be amazed if that was more competitive with the Avoton; but nor would I be amazed if Intel brought out a 3GHz 16-core 14nm lots-of-Atoms SoC.

For a more serious problem (2340_736),
[code]
echo 18140989185283655973469449579704944039975362878755465578374551246406222559010534117329360835005049754874622232679931513368003677417541836205899311159219761141488825676850446673 | gmpecm/ecm-6.4.4/ecm -v -c 1 4e8

Step 1 took 7107196ms
Step 2 took 1387825ms
[/code]

xilman 2014-07-12 18:26

[QUOTE=fivemack;377949]After a certain degree of fuss (the only EU supplier who had it in stock was SECO in Italy, and more than three months elapsed between order and delivery), my nVidia Jetson TK1 board arrived yesterday. Cost £178.49 including delivery.[/QUOTE]
Playing with small computers myself.

Four Parallellas fitted with 16-core Epiphany co-processors arrived from Adapteva about 3 weeks ago. Not yet had chance to play with them, partly for lack of connecting gubbins (now largely sorted) but mostly because of Real Life™ issues. The four credit card machines, four US-pronged power supplies (one of the connecting gubbins problems here in the UK), four SD cards pre-loaded with Ubuntu, shipping, sundry baksheesh and a "free" T-shirt cost a total of £463.46 --- about $790 at the spot rate today.

Also got hold of a DEO-Nano FPGA dev kit at about 15% of the above cost. Cute little thing, also CC sized, with no processor at all --- unless you want to build one yourself. A 32-bit RISC processor is freely available and clocks in at 100 BogoMIPS --- roughly five times as fast as my first Linux box.

henryzz 2014-07-12 19:43

I had a 32-bit binary around so I gave it a whirl:
[code]
GMP-ECM 6.2.3 [powered by GMP 4.2.1_MPIR_1.1.1] [ECM]
(10^71-1)/9
Input number is (10^71-1)/9 (71 digits)
Using B1=1000000, B2=1045563762, polynomial Dickson(6), sigma=3673851391
Step 1 took 2714ms
Step 2 took 3370ms[/code]64-bit
[code]GMP-ECM 6.4.2 [configured with MPIR 2.5.1] [ECM]
(10^71-1)/9
Input number is (10^71-1)/9 (71 digits)
Using B1=1000000, B2=1045563762, polynomial Dickson(6), sigma=1986288319
Step 1 took 2153ms
Step 2 took 2184ms[/code][code]GMP-ECM 7.0-dev [configured with MPIR 2.6.0, --enable-openmp] [ECM]
(10^71-1)/9
Input number is (10^71-1)/9 (71 digits)
Using B1=1000000, B2=1045563762, polynomial Dickson(6), sigma=1:12988648
Step 1 took 1779ms
Step 2 took 2168ms
[/code]
Around a 1.5x speedup for 64-bit but it is also an older ecm and mpir. This is on a Q6600

Xyzzy 2014-07-12 19:59

[url]http://arstechnica.com/information-technology/2014/07/hummingboard-looks-like-a-raspberry-pi-but-packs-in-more-power/[/url]

fivemack 2014-09-30 00:06

In a moment of lunacy, I ran linear algebra on the Jetson board.

The right options seem to be la_block=8192 la_superblock=98304 (the board has 32k-per-core L1 caches and a shared 2M L2 cache over four cores; la_superblock=196608 is about the same speed, 393216 is a good deal slower, la_block=16384 is a lot slower).

It takes just under ten hours on four threads for a 1.98M matrix, compared to just over two hours for four threads on i7/4770. Not bad for a machine that fits under an iPad Mini.

Batalov 2014-09-30 00:46

That reminds me of the second most famous quote from Jasper Beardly.
No, not the "That's a paddlin'!" quote.
[QUOTE]Jasper: No. I'm here for Microwave Cookery. No, wait. [pause] Coping with Linear algebra on a Microwave CPU.[/QUOTE]

henryzz 2014-09-30 10:55

That's quite nice. Have you had a chance to measure power yet?

fivemack 2014-09-30 12:48

Power metering is mostly telling me that USB3 external hard drives are much more serious power hogs that I'd anticipated: 25-30W while spinning up, 8W at idle. Will switch around the drives and the devboard tonight and see if I can get some better numbers.

fivemack 2014-09-30 19:20

Sitting at a command prompt: 3.5W

Running one thread of ECM: 7.0W

Running four threads of ECM: up to 13.7W in stage-1, up to 14.2W in stage-2

Reading the cycles for msieve (from a USB3 drive whose power is not being monitored): 4.0W

Making the matrix: 5.5W

Constructing packed rows: 3.8W

Actual four-thread LA phase: 13.0W

henryzz 2014-09-30 21:18

That looks like it is possibly slightly more power efficient than a modern pc(quite a bit better than my Q6600).
It is a shame it doesn't have enough memory for large LA jobs or it would be a good machine to leave on a 6 month job.

Xyzzy 2014-10-08 02:31

[url]http://recode.net/2014/10/07/kano-lets-kids-build-their-own-computers-and-learn-to-code/[/url]

chappy 2014-10-11 20:01

[url]https://www.kickstarter.com/projects/atomcomputer/neutron-a-full-blown-windows-pc-in-the-palm-of-you[/url]

Xyzzy 2014-11-01 18:39

We are intrigued with the idea of using a smaller computer for our main computer.

In particular, the smaller footprint, reduced operating noise and electrical power will be appreciated. (Unfortunately the VESA mount on our display is blocked by the pedestal, so we will not be able to mount the [URL="http://www.intel.com/content/www/us/en/nuc/overview.html"]NUC[/URL] to the back of our display.)

We toyed with the idea of getting a laptop, but we already have a decent display, laptops are expensive and they are generally difficult to set up with Linux, unless you run Linux in a virtual machine.

We expect the new parts to arrive next Thursday.

We need to think of a solution for playing sound. Maybe we will test a Bluetooth speaker, assuming the Bluetooth works with Linux. We currently use [URL="http://www.centos.org/"]CentOS 7[/URL] for our operating system.

New parts for system:
[URL]http://www.newegg.com/Product/Product.aspx?Item=N82E16820147316[/URL]
[URL]http://www.newegg.com/Product/Product.aspx?Item=N82E16820148730[/URL]
[URL]http://www.newegg.com/Product/Product.aspx?Item=N82E16833106244[/URL]
[URL]http://www.newegg.com/Product/Product.aspx?Item=N82E16856102052[/URL]

Existing parts for system:
[URL]http://www.newegg.com/Product/Product.aspx?Item=N82E16823816001[/URL]
[URL]http://www.newegg.com/Product/Product.aspx?Item=N82E16824260111[/URL]
[URL]http://www.newegg.com/Product/Product.aspx?Item=N82E16825122015[/URL]
[URL]http://www.newegg.com/Product/Product.aspx?Item=N82E16826816006[/URL]
[URL]http://www.newegg.com/Product/Product.aspx?Item=N82E16833124490[/URL]
[URL]http://www.newegg.com/Product/Product.aspx?Item=N82E16842101419[/URL]

Old computer:
[URL]http://www.newegg.com/Product/Product.aspx?Item=N82E16811139007[/URL]
[URL]http://www.newegg.com/Product/Product.aspx?Item=N82E16813131821[/URL]
[URL]http://www.newegg.com/Product/Product.aspx?Item=N82E16817139015[/URL]
[URL]http://www.newegg.com/Product/Product.aspx?Item=N82E16819116501[/URL]
[URL]http://www.newegg.com/Product/Product.aspx?Item=N82E16820147096[/URL] ×2
[URL]http://www.newegg.com/Product/Product.aspx?Item=N82E16820167122[/URL]
[URL]http://www.newegg.com/Product/Product.aspx?Item=N82E16835608019[/URL]

Batalov 2014-11-01 19:53

So I guess, Mike, you read the Neutron kickstarter's webpage comments?

Amazing, simply amazing. When I first read that page, I thought, "wow! $750 'early bird'... $850... that's steep, but they build the board etc..." but they [I]don't[/I], they just assemble a NUC, SSD and wireless adapter into a shiny plastic box. Of course, Mike can do it all better than them and save ~$400.

Xyzzy 2014-11-01 21:08

[QUOTE=Batalov;386668]So I guess, Mike, you read the Neutron kickstarter's webpage comments?[/QUOTE]:ttu:

Xyzzy 2014-11-06 21:07

3 Attachment(s)
We just "assembled" the NUC. It took under 20 minutes. Part of that time was used to find a suitable screwdriver!

We haven't installed an operating system, messed with the BIOS or tested the memory yet. But, here are some (very poorly taken) pictures!

VictordeHolland 2014-11-06 23:35

[QUOTE=Xyzzy;387041]We just "assembled" the NUC. It took under 20 minutes. Part of that time was used to find a suitable screwdriver!

We haven't installed an operating system, messed with the BIOS or tested the memory yet. But, here are some (very poorly taken) pictures![/QUOTE]
Nice!

P.S.
Your big case looks like it could use some de-dusting :P .

Xyzzy 2014-11-07 03:22

2 Attachment(s)
The BIOS on this thing is crazy cool, and it even includes comprehensive overclocking options! It is totally GUI with mouse support. Every line or option has detailed mouseover information. (We turned off the overclocking stuff.)

We didn't alter the BIOS settings too much. Basically we turned off UEFI booting and adjusted the slowest fan speed from 40% to 0%.

In real-life use the fan rarely spins up at all. If it does spin up, it stops as soon as the load is removed. We can use the fan sound to determine if Firefox tabs are behaving badly!

If we hit a crazy web page full of animated crap the fan spools up pretty quickly. (We also adjusted the maximum fan speed to 100%.)

We might mess with the "reaction time" of the fan because right now it spins up immediately when under load. We think the heatsink can easily absorb transient spikes. We think this adjustment is the "damping" factor.

This page is one that will turn the fans to "meltdown imminent" mode easily: [URL]http://ie.microsoft.com/testdrive/Performance/FishIETank/Default.html[/URL]

Our overall experience is silent, which is really nice. Generally the entire system draws 32 watts, most of which is used by our relatively large LCD.

Both WiFi and Bluetooth work in Linux with no fancy setup. They just work! (Seriously, this is amazing!)

For our casual work, which is easily 99% of what we do, we are unable to tell a difference between the NUC and our monster computer. The monster computer is scheduled for a rebuild (and cleaning!) and it will be repurposed into something very interesting. (Details to follow!)

During the memory testing procedure, when there were no power management options in effect, the CPU capped out at around 62°C. It took a little over an hour to run through one pass of the memory tester. The memory tester is SMP aware, which must be a new development since we have never seen it before.

We installed Ubuntu 14.04 first because for some weird reason, CentOS 7 (and RHEL 7) will not install to a blank SSD. We have a habit of "secure-erasing" our SSD when we install a new system. But, we had to let Ubuntu partition everything and run through the complete install and then install CentOS on top of all that. What is really weird is the partitioner in CentOS knows to destroy the old partition table and create a new one. It just cannot create one from scratch.

The cardboard box the NUC ships in has a photocell built in, and when you open the box it plays the "Intel chime" very loudly. We would have preferred they omit this "feature" and sell the NUC for a few dollars less, but for other people it might be a "real cool feature". It was definitely loud and unexpected!

We have attached more (poorly taken) pictures of the memory tester and of our new, streamlined desktop "experience". Unfortunately, we had to turn our UPS (wattmeter display) to the side, but that gives us lots of room. The iPad sure came in handy for diagnosing things during the install!

Batalov 2014-11-07 03:28

[QUOTE=Xyzzy;387062]... The monster computer is scheduled for a rebuild (and cleaning!) and it will be repurposed into something very interesting. (Details to follow!)[/QUOTE]
Turn it into a 24 x 7 real life fish tank [STRIKE]with [URL="http://ie.microsoft.com/testdrive/Performance/FishIETank/Default.html"]the same fish[/URL][/STRIKE] with life fish (Fish2 through Fish21)! The forum will finally get snappy.

Xyzzy 2014-11-07 03:57

1 Attachment(s)
Here is a picture of the wattmeter when the display has gone to sleep.

Note that the NUC is idle, not suspended or hibernated!

Xyzzy 2014-11-07 04:04

1 Attachment(s)
[QUOTE=Batalov;387065]Turn it into a 24 x 7 real life fish tank [STRIKE]with [URL="http://ie.microsoft.com/testdrive/Performance/FishIETank/Default.html"]the same fish[/URL][/STRIKE] with life fish (Fish2 through Fish21)! The forum will finally get snappy.[/QUOTE]We posted that the monster computer would be repurposed, not reporpoised!

:razz:

LaurV 2014-11-07 04:16

[QUOTE=Xyzzy;387062].[/QUOTE]
:tu:

Batalov 2014-11-07 04:33

[QUOTE=Xyzzy;387071]We posted that the monster computer would be repurposed, not reporpoised!

:razz:[/QUOTE]
Oy vey, reporpoised, I feel back in New York, ...your idea is indeed even worse than mine. And mine was bad!

Xyzzy 2014-11-10 18:18

[URL]http://arstechnica.com/information-technology/2014/11/20-model-a-is-the-cheapest-raspberry-pi-ever/[/URL]

[QUOTE]After releasing a revised version of the Raspberry Pi Model B (dubbed “B+") in July 2014, the low-cost computer project has produced its cheapest model yet, which will retail for just $20—$5 less than the previous Model A.[/QUOTE]

Xyzzy 2014-11-10 18:22

[YOUTUBE]Jq5nrHz9I94[/YOUTUBE]

kladner 2014-11-10 19:17

Beautiful!

EDIT: At first I thought it was a Lego design for parking garages!

Xyzzy 2014-11-10 19:23

More info: [url]http://www.southampton.ac.uk/~sjc/raspberrypi/[/url]

Xyzzy 2014-11-13 04:32

[QUOTE=Xyzzy;386656]We need to think of a solution for playing sound. Maybe we will test a Bluetooth speaker, assuming the Bluetooth works with Linux.[/QUOTE]Amazingly, the Bluetooth sound works very well!

We found a pile of clearance Bluetooth speakers at the PX today and we chose [URL="http://www.cnet.com/products/monster-clarityhd-micro/"]this one[/URL] for $50 marked down from $130!

We don't know if we will ever use the portability that it offers, but as a standalone speaker it greatly exceeds our expectations.

We ended up disabling the onboard sound and microphone in the BIOS to simplify things.

Xyzzy 2014-11-13 18:47

[QUOTE=Batalov;386668]…into a shiny plastic box.[/QUOTE]Besides the fact that their acrylic solution is not shielded to reduce RFI, we are concerned that they are "missing the boat" WRT cooling. Our NUC uses its substantial metal case as a heat sink. Even after being idle for over an hour the case is always warm to the touch. Under stress it probably could double as a coffee cup heater!

FWIW, playing around further with the fan profile has allowed us to significantly reduce how often the fan kicks in. It now is very slow to kick in, kick off and change speeds. In the end, the fan runs longer but only when the box is stressed a lot more than just scrolling a graphics-heavy page. We could have underclocked the CPU to achieve the same results, or maybe even a quieter result overall, but it is nice to have a snappy experience.

ewmayer 2014-11-13 21:59

[QUOTE=Xyzzy;387566]Besides the fact that their acrylic solution is not shielded to reduce RFI, we are concerned that they are "missing the boat" WRT cooling. Our NUC uses its substantial metal case as a heat sink. Even after being idle for over an hour the case is always warm to the touch. Under stress it probably could double as a coffee cup heater![/QUOTE]

And if you use a smooth flat-bottom coffee mug and add a thin film of water between case and mug you get enhanced cooling of the NUC and your coffee warmed. Win-win!

Xyzzy 2015-01-08 18:40

[url]http://arstechnica.com/gadgets/2015/01/hands-on-with-intels-new-mini-desktops-faster-smaller-more-expandable/[/url]

Xyzzy 2015-01-30 16:12

[QUOTE=Batalov;386668]So I guess, Mike, you read the Neutron kickstarter's webpage comments?[/QUOTE]They still have not shipped. (!)

[url]https://www.kickstarter.com/projects/atomcomputer/neutron-a-full-blown-windows-pc-in-the-palm-of-you/comments[/url]

Batalov 2015-01-30 18:58

Well, there are some projects that give all kickstarter brand a bad rep.

I was amused to learn that ze 'Maurice et Patapon' animated series was [URL="https://www.touscoprod.com/fr/project/produce?id=91"]funded via a kickstarter-like channel[/URL]. That was (maybe) money well spent... or not.
[YOUTUBE]CMbCDnAyhUw[/YOUTUBE]
[QUOTE]Maurice: En somme notre rôle consiste à tenir compagnie à des cons, gâteux, asociaux, seule et dépressifs.
Patapon ("Le Psy[I]CHAT[/I]re"): Oui.
Maurice: On est des assistantes sociales.
Patapon: Plutôt psychanalystes. On leur coûte un max et on ce fou de leur gueule.
Maurice: Voilà c'est ça que je veux dire.
___________________(translation)________________

Maurice (dog): Basically, our role is to keep company with idiots, senile, asocial, loners and depressive types.
Patapon (cat): Yes.
Maurice: We are social workers.
Patapon: Psychoanalysts, rather. Charge them at max and let them shoot their mouths.
Maurice: Yep, that's what I mean.[/QUOTE]

Xyzzy 2015-02-02 16:49

[url]http://techcrunch.com/2015/02/02/raspberry-pi-2/[/url]

pinhodecarlos 2015-02-02 22:03

[QUOTE=Xyzzy;394242][URL]http://techcrunch.com/2015/02/02/raspberry-pi-2/[/URL][/QUOTE]

How to run BOINC on PI:

[url]http://www.raspberrypi.org/downloads/[/url]

[url]http://burdeview.blogspot.co.uk/p/raspberry-pi-boinc-project-ive-created.html[/url]

Mark Rose 2015-02-02 22:40

[QUOTE=pinhodecarlos;394264]How to run BOINC on PI:[/QUOTE]

Answer: slowly.

Batalov 2015-02-02 22:55

[QUOTE=pinhodecarlos;394264]How to run BOINC on PI:

[url]http://www.raspberrypi.org/downloads/[/url]
[/QUOTE]
The PIDORA download (Fedora Remix, they say) may not be acceptable to some users.
In some countries, it could be [URL="http://www.garant.ru/news/481391/"]even considered a criminal offense[/URL] :max: :davieddy:

jasong 2015-02-03 08:13

[QUOTE=Batalov;394269]The PIDORA download (Fedora Remix, they say) may not be acceptable to some users.
In some countries, it could be [URL="http://www.garant.ru/news/481391/"]even considered a criminal offense[/URL] :max: :davieddy:[/QUOTE]
The only reason I know that's a Russian site is the ru in the web address. Keep me from seeing the web address and tell me Eqyptian or Iranian, I'll be none the wiser. Not saying the 3 countries have anything in common, just babbling about how stupid I am.

Batalov 2015-02-04 02:52

Let's assume that you already learned how to use google search.
Next step is [URL="https://translate.google.com/?hl=en"]google translate[/URL].
For practice, try it on that site, and then you can try it on [URL="http://lenta.ru/news/2013/05/23/pidora/"]this one[/URL]
Was it that [URL="https://translate.google.com/translate?hl=en&sl=ru&tl=en&u=http%3A%2F%2Fwww.garant.ru%2Fnews%2F481391%2F"]hard[/URL]?

Xyzzy 2015-02-04 18:33

[url]http://arstechnica.com/gadgets/2015/02/mips-creator-ci20-released-interesting-though-not-very-useful/[/url]

xilman 2015-02-04 19:18

[QUOTE=Xyzzy;394442][url]http://arstechnica.com/gadgets/2015/02/mips-creator-ci20-released-interesting-though-not-very-useful/[/url][/QUOTE]Intriguing. I was a big fan of MIPS systems back in the day and still have several stashed away in the loft.

A significant part of the RSA-129 project ran on MIPS machines. The main relation collecting machine at MIT was a DECserver 240, the same model as two in my collection.

The nostalgia market is still there but whether that is enough to keep the architecture alive is rather doubtful.

Xyzzy 2015-02-05 18:19

[url]http://arstechnica.com/gadgets/2015/02/review-intels-broadwell-mini-pc-is-a-next-generation-ultrabook-in-a-box/[/url]

Mark Rose 2015-02-05 19:33

[QUOTE=Xyzzy;394569][url]http://arstechnica.com/gadgets/2015/02/review-intels-broadwell-mini-pc-is-a-next-generation-ultrabook-in-a-box/[/url][/QUOTE]

Is it sad how the things whose size I could related to the most in the picture at the top of the article were USB ports?

chalsall 2015-02-05 20:29

[QUOTE=Mark Rose;394572]Is it sad how the things whose size I could related to the most in the picture at the top of the article were USB ports?[/QUOTE]

LOL... Might it be possible the large USB ports were provisioned because they provided legacy support? (And, besides, the extra surface area required would help with conductive / radiative heat exchange...)

fivemack 2015-02-09 16:01

At last, Cortex-A53 on a devboard
 
OK, it's back-ordered for eight weeks and only appears to have US suppliers, but

[url]https://www.96boards.org/products/hikey/[/url]

has an eight-core 1.2GHz A53 devboard for $129.

1GB memory, 4GB eMMC, and networking would have to be over either wifi or a USB2 dongle, but it is an eight-core ARMv8 board.

[url]www.96boards.org[/url] promise a 'enterprise edition', which definitely has SATA and hopefully has gigabit ethernet, with 'first products 2015Q2'

henryzz 2015-02-10 19:03

[QUOTE=fivemack;395015]OK, it's back-ordered for eight weeks and only appears to have US suppliers, but

[url]https://www.96boards.org/products/hikey/[/url]

has an eight-core 1.2GHz A53 devboard for $129.

1GB memory, 4GB eMMC, and networking would have to be over either wifi or a USB2 dongle, but it is an eight-core ARMv8 board.

[url]www.96boards.org[/url] promise a 'enterprise edition', which definitely has SATA and hopefully has gigabit ethernet, with 'first products 2015Q2'[/QUOTE][url]http://en.wikipedia.org/wiki/Comparison_of_ARMv8-A_cores[/url] suggests that the a53 is half the speed of a a57.

fivemack 2015-02-11 10:45

[quote]
[url]http://en.wikipedia.org/wiki/Compari..._ARMv8-A_cores[/url] suggests that the a53 is half the speed of a a57.
[/quote]

True; [url]http://www.anandtech.com/show/8718/the-samsung-galaxy-note-4-exynos-review/4[/url] has some real benchmarks (admittedly for 32-bit code) with A53 scoring 597 on specint2000.gcc and A57 scoring 1549. Also a real measurement of A57 at 1W/core at 1500MHz and A53 at 100mW/core at 900MHz.

On the other hand, in most circumstances a system with a slower chip that you can buy is preferable to one with a faster chip that you can't.

VictordeHolland 2015-02-11 17:13

[QUOTE=fivemack;395159][URL="http://www.anandtech.com/show/8718/the-samsung-galaxy-note-4-exynos-review/4"]http://www.anandtech.com/show/8718/the-samsung-galaxy-note-4-exynos-review/[/URL] [/QUOTE]
That is a nice in depth article also elaborating on the differences in manufacturing processes (Samsung vs. TSMC) and 28nm vs. 20nm.
That A57 is power hungry >1300Mhz! No way that a phone will sustain those frequencies for more than a few seconds.

fivemack 2015-02-11 23:29

Whilst [url]https://www.youtube.com/watch?v=w6dK62_B3lI[/url] is a video, it does have a bit more information about the 96boards project; for example it confirms that the enterprise spec includes Ethernet (and explains why the consumer spec doesn't).

I still don't know if it's realistic to hope for affordable ThunderX boards (48 2.5GHz v8 cores, quad-channel DDR3) by the end of the year, but I have more hope now.

VictordeHolland 2015-02-19 00:44

Qualcomm Snapdragons are not very likely to come to small computer boards, but at least it looks like Cortex A72 (successor of the A57) is coming at the end of this year. Looks like the Snapdragon 618 and 620 could become interesting SOCs with 2160p30 HEVC (H.265) en/decode options.[STRIKE] I just hope Qualcomm can get enough 14/16nm FINFET resources by then so they are not produced by TSMC 'ancient' 28nm HPm.[/STRIKE]
[URL]http://www.anandtech.com/show/8990/qualcomm-announces-new-soc-lineup-modems[/URL]

[edit]
According to Ars Technica their are going to use 28nm :-(
[quote]The chips, both of which will use a 28nm manufacturing process, will begin shipping in consumer devices in the second half of this year.[/quote][URL]http://arstechnica.com/gadgets/2015/02/qualcomms-new-snapdragon-chips-bring-high-end-features-to-midrange-socs/2/[/URL]

Xyzzy 2015-04-06 15:05

[url]http://en.yibada.com/articles/24755/20150406/intel-opens-compute-stick-pre-orders-amazon-retailers-starts-110.htm[/url]

Dubslow 2015-04-06 21:31

[QUOTE=Xyzzy;399470][url]http://en.yibada.com/articles/24755/20150406/intel-opens-compute-stick-pre-orders-amazon-retailers-starts-110.htm[/url][/QUOTE]

That's the first thing in a while from the electronics industries that has excited me. It's even cheaper without Windows, that's a first as well.

ldesnogu 2015-04-07 06:26

[QUOTE=Dubslow;399502]That's the first thing in a while from the electronics industries that has excited me. It's even cheaper without Windows, that's a first as well.[/QUOTE]
The non-Windows version comes with less RAM and disk space: 1GB/8GB vs 2GB/32GB. I'll wait for Chromebit, but I'm biased :smile:

Dubslow 2015-04-07 06:55

[QUOTE=ldesnogu;399529]The non-Windows version comes with less RAM and disk space: 1GB/8GB vs 2GB/32GB.[/QUOTE]

....................

xilman 2015-05-07 19:39

A [URL="https://www.kickstarter.com/projects/1598272670/chip-the-worlds-first-9-computer"]Kickstarter project[/URL] has fired up to produce a 1GHz ARM cpu + gpu, 512M RAM and 4GB storage on a 4cm x 6cm board at a price of $9. Over here that translates to £6 plus shipping and baksheesh. Shipping is likely to cost markedly more than the hardware :sad:

chalsall 2015-05-07 20:15

[QUOTE=xilman;401931]Shipping is likely to cost markedly more than the hardware :sad:[/QUOTE]

Atoms vs. bits...

Perhaps the manufacturer will outsource for those "far away"? (Very likely if those who "pledge" state this as a condition.)

Xyzzy 2015-05-15 17:20

[url]http://arstechnica.com/gadgets/2015/05/asrock-packs-cherry-trail-and-usb-type-c-into-a-fanless-mini-desktop/[/url]

fivemack 2015-06-09 21:07

On keeping things cold
 
I have a Xeon D-1540 machine now. It's in a nice small case - [url]https://twitter.com/fivemack/status/608362123924774912[/url] - it's nice and quiet, and it overheats if I try running more than a trivial load.

Looking at it with a thermal camera, the hotter bits are the AST2400 lights-out-management chip, the i350 dual gigabit PHY, and two components marked VITEC PR72-221 right next to the CPU which are 45-amp 220-nanohenry inductors.

Balancing a spare 92mm case fan on top of the heat sink helps a bit - the temperature converges to 60C or so with four threads sieving, but with sixteen threads active it rises at about 0.2C/second.

I have an EC3838M fan that I bought from Maplin, which is a 38mm cube that runs at 8000rpm, is advertised as 11 cubic feet per minute, and does manage to keep the temperature static at 68C with forty threads running, but it's just balanced on the heat sink and I can't see how to attach something like it more firmly. Picture of the contraption at [url]http://www.chiark.greenend.org.uk/~twomack/server-photo.JPG[/url]

Any advice? Lots of companies offer big slow fans, but this feels like a situation where I want a fairly small (the heat sink is about 60mm on a side) super-fast fan.

fivemack 2015-06-09 21:36

Xeon D-1540 benchmarks
 
echo "(10^71-1)/9" | ecm -c 1 1e6

1.220/0.740 seconds for one thread
1.276/0.876 seconds for each of eight threads running simultaneously

Comparable job on a C2750 was 2.766/1.925 seconds, so this machine is about 2.5 times the speed.

On i7/4770, 0.740/0.536 for one thread and 0.828/0.564 for each of four; so this machine is about 60% the speed, which seems not unreasonable since it's running at 2.4GHz and the i7/4770 at 3.6 or so.

kladner 2015-06-09 22:03

I realize that you are in the UK, but perhaps you could use [URL="http://www.frozencpu.com/products/8858/fan-545/Sunon_60mm_x_38mm_Extreme_Fan_-_3-Pin_PMD1206PMB1-A.html?tl=g36c15s927"]this site[/URL] to track down models to look for.

LaurV 2015-06-10 03:09

[QUOTE=fivemack;403766]I have a Xeon D-1540 machine now.[/QUOTE]
Wow! You have a freaking big foot! :shock:

fivemack 2015-06-10 20:33

This may not in fact have been Supermicro's best motherboard. 30 watts at idle, 83 watts running sixteen threads of msieve.

Do I have just to expect that level of idle power consumption as a side effect of having 32GB of DDR4? I am surprised that gigabit-ethernet PHYs still take noticeable power.

chris2be8 2015-06-11 15:30

Can you change the heat sink? One with fan mountings or a built in fan would help. Or a water cooling system (LaurV posted pictures of his recently).

Chris

PS. The server photo is at [url]http://www.chiark.greenend.org.uk/~twomack/server-photo.JPG[/url] ([url]http://chiark.greenend.org.uk/~twomack/server-photo.JPG[/url] gets an error message).

fivemack 2015-06-12 00:21

I've asked Boston Micro, who sold me the motherboard, to try to figure out the Supermicro part number for the heat-sink-with-fan that they use on the version of the board that has 10Gbit Ethernet, and see if they can order one of those as a replacement for the current heat sink.

FCBGA 1667 is a 37.5-mm-square package, which sounds as if it's probably a standard size (and has a integrated heat spreader, which makes me much less worried about breaking it while changing the heatsink); looking at the picture of the motherboard in the manual, and scaling by the length of the DIMM slots, the heat sink is 6cm square and the mounting seems to be with screw-holes with the centre-to-centre distance being 54mm; is that a standard heat-sink size and fixture?

fivemack 2015-06-14 09:24

Xeon-D linear algebra timings
 
This is on a 4.8M matrix from a C152

[code]
Threads | million dimensions per day | total runtime | efficiency
1 3.2 36:34 (100)
2 5.6 20:43 88%
3 8 0.14:37 83%
4 10.2 11:26 80%
5 11.8 09:51 74%
6 13.5 08:38 71%
7 15.0 07:46 67%
8 16.2 07:11 64%
[/code]

I'm quite pleased with the efficiency figures, since I've got eight cores sharing a single memory controller, even if it is a two-channel DDR4-2133 memory controller

For comparison, on an i7/4790K with dual-channel DDR3-1866
[code]
1 4.7 24:44 (100)
2 7.8 14:52 83%
3 10.4 11:12 74%
4 11.9 09:47 63%
[/code]

[code]
for u in 1 2 3 4; do taskset -c 1-$u /home/nfsworld/msieve-svn/trunk/msieve -v -ncr -t $u & p=$!; sleep 600; kill $p; sleep 20; done
[/code]

pinhodecarlos 2015-06-14 09:49

How did you calculate the efficiency? (edit: got your numbers, you determine it proportional to 1 core running )
Do you have an energy meter? Will the consumption be linear as you increase the number of cores? I don't think so....

fivemack 2015-06-14 10:00

I do have an energy meter - but it's in an outbuilding and I don't have a webcam rigged up to point at the energy meter yet, so I've got point measurements and couldn't very easily record energy use at each core count.

A little further up the thread I mentioned that I was seeing 30W at idle and about 83W flat-out.

For the aid of anyone else reading this: 'efficiency 73% on 5 CPUs' means that the dimensions-per-hour figure for five CPUs is 73% of five times the figure on one CPU.

pinhodecarlos 2015-06-14 10:04

[QUOTE=fivemack;404048]

For the aid of anyone else reading this: 'efficiency 73% on 5 CPUs' means that the dimensions-per-hour figure for five CPUs is 73% of five times the figure on one CPU.[/QUOTE]

I was updating my post while you were answering because I did understand how you calculated your "efficiency".

masser 2015-06-18 00:51

I'm considering this build:

[URL="http://pcpartpicker.com/p/JW3Z3C"]http://pcpartpicker.com/p/JW3Z3C[/URL]

I already have the PSU and RAM from previous combo purchases. Anyone here have experience with AMD's AM1 line?

VictordeHolland 2015-06-27 14:46

Anandtech reviewed a Xeon-D on a SuperMicro board and got an idle power reading of 31W, very close to that of fivemack. [url]http://www.anandtech.com/show/9185/intel-xeon-d-review-performance-per-watt-server-soc-champion[/url]

VictordeHolland 2015-06-30 11:50

2 Attachment(s)
AnandTech tested the Samsung Exynos 7240 (including detailed power consumption). Produced on Samsung 14nm FinFet it needs less power to operate at a higher frequency than the Samsung Exynos E5433 (on 20nm planar).
5433: 4x A57 cores @1.9GHz => 7.39W
7240: 4x A57 cores @1.9GHz => 4.12W
7240: 4x A57 cores @2.1GHz => 5.49W

See also attached images and anandtech article:
[URL]http://www.anandtech.com/show/9330/exynos-7420-deep-dive/5[/URL]

I hope Hardkernel is going to use it in a new Odroid board :P .

ldesnogu 2015-06-30 13:39

[QUOTE=VictordeHolland;405031](...) Samsung Exynos 7240 (...)

I hope Hardkernel is going to use it in a new Odroid board :P .[/QUOTE]
Not this year [URL="http://forum.odroid.com/viewtopic.php?f=54&t=13062"]it seems[/URL] :sad:

VictordeHolland 2015-07-04 11:34

[QUOTE=ldesnogu;405037]Not this year [URL="http://forum.odroid.com/viewtopic.php?f=54&t=13062"]it seems[/URL] :sad:[/QUOTE]

I'm not really suprised, it would seem Samsung is reserving the 14nm high-end production capabilities for it's own products (and maybe Apple).

It are interesting times with Samsung, Qualcomm and Intel battling it out on the high-end low power mobile market. At the moment Samsung seems to have the best smartphone SOC, but that could change when Qualcomm releases their own ARM core implementation. Or maybe Intel have an ace up their sleeve...

debrouxl 2015-07-14 13:51

Some interesting computing equipment based on 64-bit ARM SoCs is supposed to become available for both rental (e.g. [url]http://www.design-reuse.com/news/37844/ovh-armv8-public-cloud-cavium-thunderx-processor.html[/url]) and purchase (e.g. [url]http://www.cavium.com/newsevents-GIGABYTE-announces-384-Core-2U-server-powered-by-Cavium-ThunderX-ARMv8-processors.html[/url]) this quarter.

ldesnogu 2015-07-21 07:43

Cavium ThunderX seems to be gaining traction even in HPC in the form of CPU + NVIDIA Tesla: [url]http://insidehpc.com/2015/07/e4-computer-engineering-accelerates-arm-with-gpus/[/url]

Xyzzy 2015-08-25 00:49

[url]http://arstechnica.com/gadgets/2015/08/intel-introduces-its-smallest-socketed-form-factor-yet-the-55/[/url]

Xyzzy 2015-09-02 14:33

[url]http://arstechnica.com/gadgets/2015/09/acer-unveils-the-revo-build-a-tiny-modular-stackable-pc-starting-at-180/[/url]

debrouxl 2015-09-02 17:11

More raw power than the RPi 2, only $15: [url]http://linuxgizmos.com/15-dollar-orange-pi-pc-hacker-sbc-packs-quad-core-soc/[/url] .

Xyzzy 2015-11-11 15:32

[url]http://arstechnica.com/gadgets/2015/11/nvidias-jetson-tx1-dev-board-is-a-mobile-supercomputer-for-machine-learning/[/url]

fivemack 2015-11-12 11:56

I never managed to get anything compiled for the single Kepler core on my TK1, which makes me a little more wary of getting a TX1. The pricing is a bit beyond trivial ($599 - I'd need the base-board since interfacing with the 400-pin custom board-to-board connector is well beyond my skills), but 4xA57 is a perceptible amount of computation and 4GB an acceptable amount of memory.

An interesting question is how deeply nVidia are committed to making the baseboard usable with multiple module generations; whether I can upgrade in 2020 to a TV1 with 16xA77 and a single Volta core just by buying a new module.

Xyzzy 2015-11-26 14:48

[URL]http://arstechnica.co.uk/gadgets/2015/11/pi-zero-a-full-raspberry-pi-for-just-4/[/URL]

[QUOTE]Have you ever wanted a Raspberry Pi but thought that £16/$25—roughly three packs of cigarettes—was a little exorbitant for a complete computer capable of running Linux? Well, I have good news: the Raspberry Pi Foundation has just released the Pi Zero, a single-board computer that costs just £4/$5.[/QUOTE]

paulunderwood 2015-11-26 16:34

[url]http://www.bbc.co.uk/news/technology-34922561[/url]

Another write-up on the Pi Zero.

Can one buy micro-usb keyboards and mice? :smile:

How do you network it?

paulunderwood 2015-11-26 16:53

I seems you can buy a cheap micro-usb hub with ethernet and say 3 usb ports.

How is this board powered? :smile:

VictordeHolland 2015-11-26 22:11

[QUOTE=paulunderwood;417328]I seems you can buy a cheap micro-usb hub with ethernet and say 3 usb ports.

How is this board powered? :smile:[/QUOTE]
Micro USB (5V), but I don't believe a cable or adapter is included (else I would just buy it for the adapter :) ).

VictordeHolland 2016-01-06 21:00

HiSilicon Kirin 950 in the Huawei Mate 8:
[URL]http://www.anandtech.com/show/9878/the-huawei-mate-8-review/3[/URL]

HiSilicon Kirin 950 key features:
4x 2.3GHz A72 cores
4x 1.8GHz A53 cores
CCI-400 interconnect
2x32bit LPDDR4 @1333MHz memory controller

In terms of architectural improvements of the A72 over the A57, they are mostly in memory and floating point performance. Little to none in integer, if you cancel out the 200MHz frequency benefit of the A72 in the 950 compared to the A57 in the Samsung 7420.

3734mW with 4x A72 @2.3GHz in Kirin 950
compared to
5486mW with 4x A57 @2.1GHz in Samsungs 7420

Some of the improved power efficiency may be from the improved ARM architecture, the rest is probably largely due to the (newer/better) TSMC 16nm FinFET+ manufacturing node (also called 16FF+) compared to Samsungs 14nm. Keep in mind these 16/14nm nodes are more marketing names, as they are dimensionally more similar to 20nm with FinFET. Intel's 14nm FinFET is still miles ahead.

LaurV 2016-01-07 02:16

[QUOTE=VictordeHolland;421417]HiSilicon Kirin[/QUOTE]
Hi VictordeHolland!

Xyzzy 2016-01-07 16:27

[url]http://arstechnica.com/gadgets/2016/01/intels-new-atom-and-core-m-compute-sticks-get-faster-and-look-better/[/url]

VictordeHolland 2016-01-08 08:03

[QUOTE=Xyzzy;421468][URL]http://arstechnica.com/gadgets/2016/01/intels-new-atom-and-core-m-compute-sticks-get-faster-and-look-better/[/URL][/QUOTE]
Core M3 and M5, Yeay!
*Kuch* price *kuch*.

Xyzzy 2016-01-19 20:09

Skylake NUCs: [URL]http://www.intel.com/content/www/us/en/nuc/products-overview.html#newestnucs[/URL]

i5 versus i3: [url]http://ark.intel.com/compare/91160,88180[/url]

Note DDR4 and 32GiB capacity!

:mike:

masser 2016-02-06 02:24

[url]http://blog.codinghorror.com/the-scooter-computer/[/url]

Xyzzy 2016-02-13 15:40

[url]http://www.theinquirer.net/inquirer/news/2446966/canonical-releases-snappy-ubuntu-core-for-the-intel-nuc-iot-starter-kit[/url]

[QUOTE]The announcement explained: "We focused on the Intel NUC for its relatively low cost point for a starter platform (around $150) and broad availability (you can even find them on Amazon).[/QUOTE]

VictordeHolland 2016-02-13 18:48

Odroid U2
 
I bought an Hardkernel Odroid U2 some time ago. I wanted to use it to play media from a PC/external harddrive. But after a firmware update of my blu-ray player it was rendered useless as the blu-ray player could play files with H264/x264 codec from an external harddrive directly.

Hardkernel Odroid U2 Specs:
[URL]http://www.hardkernel.com/main/products/prdt_info.php?g_code=G135341370451[/URL]
Samsung Exynos4412 Prime
Process node: 32nm HKMG
CPU: 4x ARM Cortex-A9 @1.7GHz, 1MB shared L2 cache
Mem: 2GB LPDDR2-880
GPU: ARM Mali-400MP4 @440MHz
LAN: 100mbit
OS: Ubuntu 14.04 or Android 4.4.4
5V 2A adapter (so it shouldn't consume more than 10W)

It is collecting dust for some time now, but I noticed the MLucas and ECM for ARM threads in the past weeks.
[URL]http://mersenneforum.org/showthread.php?t=20846[/URL]
[URL]http://mersenneforum.org/showthread.php?t=20614[/URL]

So I might try to get those working. And if I get lucky post some benchmark results here ;). From the schematics it looks like it only has a single VFPv3 unit for (fast) floating point calculations, so it's probably going to be :mike::mike::mike:.

xilman 2016-02-13 18:59

[QUOTE=VictordeHolland;426232]I bought an Hardkernel Odroid U2 some time ago. I wanted to use it to play media from a PC/external harddrive. But after a firmware update of my blu-ray player it was rendered useless as the blu-ray player could play files with H264/x264 codec from an external harddrive directly.

Hardkernel Odroid U2 Specs:
[URL]http://www.hardkernel.com/main/products/prdt_info.php?g_code=G135341370451[/URL]
Samsung Exynos4412 Prime
Process node: 32nm HKMG
CPU: 4x ARM Cortex-A9 @1.7GHz, 1MB shared L2 cache
Mem: 2GB LPDDR2-880
GPU: ARM Mali-400MP4 @440MHz
LAN: 100mbit
OS: Ubuntu 14.04 or Android 4.4.4
5V 2A adapter (so it shouldn't consume more than 10W)

It is collecting dust for some time now, but I noticed the MLucas and ECM for ARM threads in the past weeks.
[URL]http://mersenneforum.org/showthread.php?t=20846[/URL]
[URL]http://mersenneforum.org/showthread.php?t=20614[/URL]

So I might try to get those working. And if I get lucky post some benchmark results here ;). From the schematics it looks like it only has a single VFPv3 unit for (fast) floating point calculations, so it's probably going to be :mike::mike::mike:.[/QUOTE]It would make a fine ECMNET client. My Parallella systems did sterling work on the (forthcoming) HCN extension tables.


Paul

VictordeHolland 2016-02-16 21:29

4 Attachment(s)
Thanks to the Debian package I got Mlucas working (sort of) on the Odroid-U2 with Ubuntu 14.04.
Some minor problems with running the selftest. Some radices would result in an ERR_ASSERT or threadpool_init failed error, this happened with the -s tiny (FFT 88K), small (FFT 768K) and m (FFT 3584K) tests.
But if I do:
[code]
mlucas -s s -fftlen 768[/code]Then 768K and 3584K work without spitting out errors (see also the screenshots)

Here some preliminary results (FFT size in K, followed by ms/ilter, errors, radices)
[code]
128 msec/iter = 9.73 ROE[avg,max] = [0.244838170, 0.312500000] radices = 16 16 16 16 0 0 0 0 0 0
144 msec/iter = 12.18 ROE[avg,max] = [0.233816964, 0.281250000] radices = 36 8 16 16 0 0 0 0 0 0
160 msec/iter = 13.03 ROE[avg,max] = [0.242075893, 0.312500000] radices = 20 16 16 16 0 0 0 0 0 0
176 msec/iter = 15.92 ROE[avg,max] = [0.282393973, 0.375000000] radices = 44 8 16 16 0 0 0 0 0 0
192 msec/iter = 14.96 ROE[avg,max] = [0.223990304, 0.250000000] radices = 24 16 16 16 0 0 0 0 0 0
208 msec/iter = 18.59 ROE[avg,max] = [0.264927455, 0.312500000] radices = 52 8 16 16 0 0 0 0 0 0
224 msec/iter = 18.92 ROE[avg,max] = [0.249218750, 0.312500000] radices = 56 8 16 16 0 0 0 0 0 0
240 msec/iter = 20.66 ROE[avg,max] = [0.236710031, 0.281250000] radices = 60 8 16 16 0 0 0 0 0 0
256 msec/iter = 21.03 ROE[avg,max] = [0.281250000, 0.281250000] radices = 32 16 16 16 0 0 0 0 0 0
288 msec/iter = 25.00 ROE[avg,max] = [0.229432896, 0.312500000] radices = 36 16 16 16 0 0 0 0 0 0
320 msec/iter = 27.60 ROE[avg,max] = [0.254017857, 0.312500000] radices = 40 16 16 16 0 0 0 0 0 0
352 msec/iter = 32.23 ROE[avg,max] = [0.290244838, 0.406250000] radices = 44 16 16 16 0 0 0 0 0 0
384 msec/iter = 32.07 ROE[avg,max] = [0.222670201, 0.281250000] radices = 24 16 32 16 0 0 0 0 0 0
416 msec/iter = 37.82 ROE[avg,max] = [0.250251116, 0.312500000] radices = 52 16 16 16 0 0 0 0 0 0
448 msec/iter = 39.09 ROE[avg,max] = [0.237262835, 0.281250000] radices = 56 16 16 16 0 0 0 0 0 0
480 msec/iter = 42.05 ROE[avg,max] = [0.230156599, 0.281250000] radices = 60 16 16 16 0 0 0 0 0 0
512 msec/iter = 43.43 ROE[avg,max] = [0.375000000, 0.375000000] radices = 32 16 32 16 0 0 0 0 0 0
576 msec/iter = 51.38 ROE[avg,max] = [0.230461775, 0.281250000] radices = 144 8 16 16 0 0 0 0 0 0
640 msec/iter = 57.27 ROE[avg,max] = [0.269419643, 0.312500000] radices = 160 8 16 16 0 0 0 0 0 0
704 msec/iter = 65.28 ROE[avg,max] = [0.296651786, 0.390625000] radices = 176 8 16 16 0 0 0 0 0 0
768 msec/iter = 70.51 ROE[avg,max] = [0.230664062, 0.312500000] radices = 192 8 16 16 0 0 0 0 0 0
[/code]And the more relevant for the current wavefront and comparison with 'real' desktops:
[code]
1024 msec/iter = 121.70 ROE[avg,max] = [0.298214286, 0.312500000] radices = 128 16 16 16 0 0 0 0 0 0
1152 msec/iter = 142.69 ROE[avg,max] = [0.225310407, 0.250000000] radices = 144 16 16 16 0 0 0 0 0 0
1280 msec/iter = 161.44 ROE[avg,max] = [0.251618304, 0.312500000] radices = 160 16 16 16 0 0 0 0 0 0
1408 msec/iter = 185.52 ROE[avg,max] = [0.297056362, 0.375000000] radices = 176 16 16 16 0 0 0 0 0 0
1536 msec/iter = 195.56 ROE[avg,max] = [0.234742955, 0.312500000] radices = 192 16 16 16 0 0 0 0 0 0
1664 msec/iter = 208.36 ROE[avg,max] = [0.254631696, 0.312500000] radices = 208 16 16 16 0 0 0 0 0 0
1792 msec/iter = 222.32 ROE[avg,max] = [0.234012277, 0.250000000] radices = 224 16 16 16 0 0 0 0 0 0
1920 msec/iter = 243.65 ROE[avg,max] = [0.235016741, 0.281250000] radices = 240 16 16 16 0 0 0 0 0 0
2048 msec/iter = 255.25 ROE[avg,max] = [0.310714286, 0.312500000] radices = 256 16 16 16 0 0 0 0 0 0
2304 msec/iter = 297.26 ROE[avg,max] = [0.228341239, 0.281250000] radices = 288 16 16 16 0 0 0 0 0 0
2560 msec/iter = 339.70 ROE[avg,max] = [0.256682478, 0.312500000] radices = 160 16 16 32 0 0 0 0 0 0
2816 msec/iter = 384.56 ROE[avg,max] = [0.296219308, 0.375000000] radices = 176 16 16 32 0 0 0 0 0 0
3072 msec/iter = 413.85 ROE[avg,max] = [0.239704241, 0.281250000] radices = 192 16 16 32 0 0 0 0 0 0
3584 msec/iter = 370.28 ROE[avg,max] = [0.231487165, 0.281250000] radices = 224 16 16 32 0 0 0 0 0 0
4096 msec/iter = 455.10 ROE[avg,max] = [0.282142857, 0.312500000] radices = 128 16 32 32 0 0 0 0 0 0
[/code]Timings are almost linearly increasing with FFT size, so the Odroid-U2 doesn't seem bandwidth limited. The Cortex-A9 cores are just not very powerful. I don't know if it is using the VFP3 extension/unit to accelerate floating point calc. How can I check if it is being used? At the start of MLucas it says:
[code]
Mlucas 14.1

http://hogranch.com/mayer/README.html

INFO: testing qfloat routines...
CPU Family = ARM Embedded ABI, OS = Linux, 32-bit Version, compiled with Gnu C [or other compatible], Version 4.8.2.
INFO: Using inline-macro form of MUL_LOHI64.
INFO: MLUCAS_PATH is set to ""
INFO: using 53-bit-significand form of floating-double rounding constant for scalar-mode DNINT emulation.
INFO: testing IMUL routines...
INFO: System has 4 available processor cores.
INFO: testing FFT radix tables...
Mlucas command line options:
[/code]For comparison purposes a Core2Duo E7400 @2.8GHz with DDR2-800 (looking at the 2thread timings it is probably single-channel) running the latest mprime.
[code]
Intel(R) Core(TM)2 Duo CPU E7400 @ 2.80GHz
CPU speed: 2800.89 MHz, 2 cores
CPU features: Prefetch, SSE, SSE2, SSE4
L1 cache size: 32 KB
L2 cache size: 3 MB
L1 cache line size: 64 bytes
L2 cache line size: 64 bytes
TLBS: 256
Prime95 64-bit version 28.7, RdtscTiming=1
Best time for 1024K FFT length: 16.134 ms., avg: 16.250 ms.
Best time for 1280K FFT length: 20.692 ms., avg: 27.239 ms.
Best time for 1536K FFT length: 25.937 ms., avg: 26.121 ms.
Best time for 1792K FFT length: 30.732 ms., avg: 31.735 ms.
Best time for 2048K FFT length: 34.630 ms., avg: 35.848 ms.
Best time for 2560K FFT length: 42.596 ms., avg: 44.089 ms.
Best time for 3072K FFT length: 52.945 ms., avg: 53.301 ms.
Best time for 3584K FFT length: 66.552 ms., avg: 82.680 ms.
Best time for 4096K FFT length: 70.515 ms., avg: 73.019 ms.
Best time for 5120K FFT length: 87.307 ms., avg: 94.365 ms.
Best time for 6144K FFT length: 108.814 ms., avg: 122.418 ms.
Best time for 7168K FFT length: 131.005 ms., avg: 149.828 ms.
Best time for 8192K FFT length: 144.657 ms., avg: 156.096 ms.
Timing FFTs using 2 threads.
Best time for 1024K FFT length: 15.166 ms., avg: 15.980 ms.
Best time for 1280K FFT length: 17.386 ms., avg: 18.087 ms.
Best time for 1536K FFT length: 21.690 ms., avg: 22.476 ms.
Best time for 1792K FFT length: 25.742 ms., avg: 26.861 ms.
Best time for 2048K FFT length: 28.558 ms., avg: 33.612 ms.
Best time for 2560K FFT length: 36.754 ms., avg: 39.345 ms.
Best time for 3072K FFT length: 46.804 ms., avg: 48.510 ms.
Best time for 3584K FFT length: 82.085 ms., avg: 83.430 ms.
Best time for 4096K FFT length: 60.806 ms., avg: 64.677 ms.
Best time for 5120K FFT length: 76.952 ms., avg: 78.853 ms.
Best time for 6144K FFT length: 90.912 ms., avg: 91.943 ms.
Best time for 7168K FFT length: 112.518 ms., avg: 123.265 ms.
Best time for 8192K FFT length: 123.618 ms., avg: 132.638 ms.[/code]1024K 121.70ms vs. 16.25ms (=~13% of a single Core2Duo core)
2048K 255.25ms vs. 35.85ms (=~14%)
4096K 455.10ms vs. 73.02ms (=~16%)

Conclusion: it's slow, but we already knew that.
But, it uses[U] very little[/U] power. When idle on desktop my wallmeter reads 0W, which actually means it is using <1W and it can't get an accurate reading.
When running Mlucas on a random low DC exponent, let's say M35218831 (it's already DCed and I'm not going to finish it, as it would just take too long ;) )
[code]M35218831: using FFT length 1920K = 1966080 8-byte floats.
this gives an average 17.913223775227866 bits per digit
Using complex FFT radices 240 16 16 16[/code]It's using 6.2W with very little variance and barely getting warm to the touch.

For webbrowsing and Libreoffice it is not a bad little thingy. Watching Youtube in 480p is ok, but it struggles with 720p. Setting Chromium to use x264 instead of VP9 for video's makes it slightly more fluent.

[edit]
Next objective is to try to get GMP-ECM working.

ldesnogu 2016-02-16 21:48

[QUOTE=VictordeHolland;426575]For comparison purposes a Core2Duo E7400 @2.8GHz with DDR2-800 (looking at the 2thread timings it is probably single-channel) running the latest mprime.
[/QUOTE]
Wouldn't the comparison be more fair if you also used Mlucas on your C2D?

VictordeHolland 2016-02-16 23:11

[QUOTE=ldesnogu;426576]Wouldn't the comparison be more fair if you also used Mlucas on your C2D?[/QUOTE]
Yes it would, but it wouldn't change the outcome by much. The E7400_Ubuntu machine is not running at the moment (it uses too much electricity and I haven't got the space for it). I could have compared it with a i5 2500k or a i7 3770k both running P95 under Win7. Both of them have AVX, thus that would make the difference even larger. I don't think there is more than single digit % difference in performance between Mprime and MLucas on pre-AVX machines.

Running GMP-ECM on the Odroid-U2 probably makes more sense.

henryzz 2016-02-16 23:37

Considering that it is asm vs c code it isn't that bad.
That single channel memory is crippling it though. The memory on Core 2s is bad but I haven't seen it that limited on a dual core before only quads.


All times are UTC. The time now is 02:58.

Powered by vBulletin® Version 3.8.11
Copyright ©2000 - 2021, Jelsoft Enterprises Ltd.