RISC OS London Show, Sat 24th October 2015

print A3 paper and 180gsm card and doesn’t cost the earth? That would be my ideal printer but I don’t think it exists.

Oki C831dn. Costs only half the earth (1500€ incl. VAT RRP), but prints paper weights up to 256g/m² and prints banners up to 1.3m long.

Wow, thanks for the mini reviews! Was fully expecting a two to seven month wait until any news got out ;-)

Not knowing the terminology (ATX, mini-ATX etc.)

Seems to be miniDTX. Interesting point is that, while smartphones and tablets are Cortex-A53, most ARM computers are still Cortex-A9 ones. So this board could be very useful for serious Linux users. Of course, a way to boot Linux from RISC OS would be a great thing (for Linux only users, RISC OS will then become a superbootloader). And dual head support for RISC OS (a la Geminus. Possibly as a paid package).

Adrian Lees was showing an updated Aemulor although it was not yet released. His plan was to make the next update a paid-for update ‘Aemulor Ultimate’ to be available for all platforms with the full functionality of Aemulor and Aemulor Pro so that this distinction would disappear.

Fantastic! IMHO, Aemulor and ADFFS are the two main ways to maintain a very good level of compatibility with older software (or non ARMv7 software).
The only real problem today with OMAP is sound support.

There is a Linux port for Titanium. If you look in their shop, you can see that it will be available with either RISC OS or Linux pre-installed.

Yes, but I was talking of RISC OS AND Linux (available for example with a !Linux bootloader).

Yes, but I was talking of RISC OS AND Linux (available for example with a !Linux bootloader).

Where the bootloader is RISC OS instead of U-Boot?

Mini-DTX is a bit odd of a choice, there’s not much in the way of chassis choices for it, without just sticking the board in a Micro-ATX chassis (which will work fine). (I’m aware of one actual Mini-DTX chassis in production – the NCASE M1 – but it’s not cheap at all. I’m going to use one for my next x86 PC build, though.)

In any case, I’m expecting this to be shockingly expensive – enough that buying one of the ARM server platforms may make sense, if you really want to play with SATA and PCIe on an ARM Linux system. If you just want a cheap 64-bit ARM platform, there’s one for $120-130 but you won’t get SATA or PCIe. And, Atom is very cost effective if you just want to run Linux and get some PCIe cards.

Also, the back panel layout is odd, using a mezzanine board for USB, instead of a dual stack DVI connector and getting it all on the board. And, that SD card location is just really strange, and will require removing the motherboard from the case to access it unless it’s in a Micro-ATX case.

Where the bootloader is RISC OS instead of U-Boot?

Yep, for example. A tool to load the kernel and define a few parameters (disc to be used).

Mini-DTX is a bit odd of a choice, there’s not much in the way of chassis choices for it, without just sticking the board in a Micro-ATX chassis (which will work fine)

I can find around 8 from my current web dealer. But, as you said, any ATX or MicroATX case will be OK.

enough that buying one of the ARM server platforms may make sense

We talk here of platforms that still cost several hundreds of euros.

If you just want a cheap 64-bit ARM platform, there’s one for $120-130 but you won’t get SATA or PCIe. And, Atom is very cost effective if you just want to run Linux and get some PCIe cards.

Yep, so basically, no SATA could be enough, x86 could be enough. And why not removing Linux too?
Then a Nokia under Windows Phone is a perfect alternative to a Linux+ARM+Sata platform :)

IMHO. For ARM Server Linux developments, Linux is not an option, ARM is not an option and Sata is not an option.

And, that SD card location is just really strange, and will require removing the motherboard from the case to access it unless it’s in a Micro-ATX case.

With Sata, why would you want to use it anyway?

Really, this platform is absolutely unique (for now) on this emerging market. Now, if we can have Geminus on it, I’ll probably buy one (for a developer, dual screen is [almost] not an option).

I can find around 8 from my current web dealer.

Do note that this isn’t just a 2-slot Mini-ITX case, as those may only have ~190 mm of vertical board clearance, and Mini-DTX needs 203 mm. It needs to truly be a Mini-DTX case (or a 3-slot Mini-ITX case like the M1, which was basically a case of “hey, let’s add a third slot area for Mini-DTX clearance or particularly small sub-Micro-ATX boards or to put things underneath a 2-slot GPU”). Links would be interesting, though, if you’ve got true Mini-DTX chassis options other than the M1.

We talk here of platforms that still cost several hundreds of euros.

And you think this board won’t cost a lot? It’s a low-volume board with very limited appeal outside of the RISC OS market, IMO.

Yep, so basically, no SATA could be enough, x86 could be enough. And why not removing Linux too?
Then a Nokia under Windows Phone is a perfect alternative to a Linux+ARM+Sata platform :)
IMHO. For ARM Server Linux developments, Linux is not an option, ARM is not an option and Sata is not an option.

Let’s not be facetious. It really depends on what you’re doing. If you want to play with ARMv8 (which the Titanium can’t do!), the cheap board will work. If you want to build a cost-effective power efficient Linux desktop, a Celeron N3150 will destroy the ARM platforms in the cost-effectiveness sense, while giving you SATA and much better PCIe than this thing. If you want to build a cost-effective power efficient Linux server, AFAIK Avoton (Atom C2000) ends up being cheaper than the ARM server boards with just as good RAM expansion, sometimes BETTER SATA, and PCIe.

With Sata, why would you want to use it anyway?

In a home environment, having an SD card reader is handy for digital camera download. (I mean, you could just hook one up to the front panel USB header, but still.) In a server environment, a popular way to do things nowadays, especially in *nix environments, is to boot the OS off of a small, cheap medium (SD card, USB key, that sort of thing), and store data on a RAID array (sometimes locally SATA/SAS-attached, sometimes a SAN).

Really, this platform is absolutely unique (for now) on this emerging market. Now, if we can have Geminus on it, I’ll probably buy one (for a developer, dual screen is [almost] not an option).

It’s unique, but it’s not that useful as a server IMO (not enough RAM, not 64-bit, not many cores), and ARM Linux on the desktop is not really a thing that’s taking off at all. (And, the PCIe on this board sucks for a lot of expansion, being x1 slots without even cutouts for x4/x8/x16 cards.) And, seems the Exynos 5 Dual’s TDP is around 8 W, with a SDP of 4 W. A Pentium N3700 or Celeron N3150 is at 6 W TDP, 4 W SDP, and I’d be wholly unsurprised if it’s faster. And, the SATA is better for SSDs, the PCIe is better for GPUs or whatever other expansion you want, etc., etc.

True DTX and Mini-DTX cases :
http://www.ldlc.com/informatique/pieces-informatique/boitier-pc/c4678/+fv285-7977,6214.html
34

Of course, you can also use any ATX or MicroATX case.

And you think this board won’t cost a lot? It’s a low-volume board with very limited appeal outside of the RISC OS market, IMO.

Of course it will. But I don’t expect it to cost the 5KE or more of current HDK for servers (sorry, I did want to say several thousands euros, not several hundreds). And here we have RISC OS. The only little problem is lack of RAM. 4 GB would have been better. Of course I expect easy dual boot and dual head support for RISC OS, with excellent stability. Something pricey (as the Iyonix) but fully usable (as the Iyonix) and a long term offer (as the Iyonix).

If you want to play with ARMv8

I don’t want ARMv8

cost-effective power efficient Linux desktop

I don’t want this either.
Server, Linux, ARM, Desktop, RISC OS.

In a server environment, a popular way to do things nowadays, especially in *nix environments, is to boot the OS off of a small, cheap medium

Source? I manage servers. They all boot on hard disc or network. For security reasons, SD card readers or USB ports are of course not present on production servers. Sometimes for firmware update.

ARM Linux on the desktop is not really a thing that’s taking off at all.

Tell that to Android users :)

seems the Exynos 5 Dual’s TDP is around 8 W, with a SDP of 4 W. A Pentium N3700 or Celeron N3150 is at 6 W TDP, 4 W SDP, and I’d be wholly unsurprised if it’s faster. And, the SATA is better for SSDs, the PCIe is better for GPUs or whatever other expansion you want, etc., etc.

And? I’m an ARM user, and a RISC OS user. I’m not here for an holly war. Else, I can go for it too.

It seems that Xeon E5v3 TDP is around 80 W. A Power8 with 165 W TDP will wholly unsurprised be much than twice faster. And the SAS is better for SSDs, the CAPI is better for GPUs or whatever other expansion you want, etc.

That’s not the point. The point is that true desktop/workstation/server ARM platforms are not very common (today we’re stuck with Cortex-A9 offers). Of course, we can use mobile processors for desktop use, but it’s suboptimal. We must often rely on USB to get things as Ethernet port, and I/O are very slow, because priority is given to the GPU. So every step in the right direction is good for me. Here the price will be probably better than the price of an ARM64 Server HDK (several k-Euros). There are things we won’t find on mobile HDK (Sata, native Ethernet, dual screen, PCI Express). So we can test more server things (than on mobile HDK) with less money (then with server HDK). If price is below 2KE, it’ll be OK for me. If it’s 2KE and 4 GB RAM, I see no real competitor for this today. But not tomorrow, with future low cost 96Boards Server.

Low cost x86 boards are not an argument. Of course, they are less pricey. Of course too, mainstream ARM motherboards are less pricey too.
My old computer, that is already paid, is even less pricey.

Source? I manage servers. They all boot on hard disc or network. For security reasons, SD card readers or USB ports are of course not present on production servers. Sometimes for firmware update.

It’s typically for virtualization hosts, as I understand. http://kb.vmware.com/selfservice/microsites/search.do?language=en_US&cmd=displayKC&externalId=2004784 talks about how it’s supported for ESXi, as an example.

Tell that to Android users :)

I said desktop, not pocket or tablet.

Low cost x86 boards are not an argument.

If you’ve already decided that your Linux platform must be ARM, then no, they’re not an argument. However, most people building real servers care about cost-effectiveness, not whether it’s ARM or not. And, the cost-effectiveness just isn’t there for ARM in the server role. There was a very narrow window and a fairly narrow niche in which it was (hosts where many relatively slow cores could work very well, and legacy binary x86 software wasn’t needed, to improve performance per watt (reduced power consumption and cooling costs, meaning a premium could be charged for the hardware and it still was cost effective), at the expense of performance per thread), but then Intel caught up to that market with the Atom C2000 and the Xeon D. There’s no legacy ARM software in the server market (please don’t count a Level 4 Econet server as “legacy ARM server software”), so there’s no need to pick ARM purely because it’s ARM, it needs to win on its merits.

This also counts for workstations – a Pentium N3700 board is going to be better in every way for the average Linux desktop user than the Titanium, so why would they buy the Titanium?

I’m not saying that the Titanium is a bad RISC OS board (it’s got some quirks for that application, sure, but it’s not bad), I’m saying it’s not going to be compelling as a Linux board, and that’ll probably even hold true in the ARM Linux space. Therefore, I wouldn’t expect it to succeed as a Linux board, and therefore it’s going to have the lack of economies of scale that come with being a RISC OS-centric product.

And here we have RISC OS. The only little problem is lack of RAM.

…and that RISC OS is hardly a server class operating system. You gave up on it yourself, did you not? ;-)

I don’t want ARMv8

It’s the up-coming thing. I think there will be interest in it because it may help ARM try to become more of a player in the server market traditionally held by x86 hardware.

For security reasons, SD card readers or USB ports are of course not present on production servers.

I think the reality of that statement depends a lot upon access control to the hardware itself. A server will need a method of interfacing (keyboard etc) for when remote connection is inoperable, and it will need a way to be able to run low level updates for when remote connection can’t do it (or has failed update). There is little difference from USB and SD as CD/DVD media or even, god help us, floppy discs. All can be vectors for external contamination; so the first level of defence is keeping unauthorised people out of the server room. The second level of defence is nobbling the hardware so it does nothing at all until an administrator mounts the device. You can put every epic virus and trojan in the world on a USB stick and shove it into the machine… and nothing will happen. The OS has been instructed to ignore it.

If an outsider can come and activate the device, the media itself doesn’t really matter. You’re screwed either way.

ARM Linux on the desktop is not really a thing that’s taking off at all.

Tell that to Android users :)

I’m not sure I would call a mobile phone or tablet a “desktop”. For a start, where are the windows? With very few exceptions, apps take over the screen. You can multitask, certainly, but you more or less cast your attention to one thing at a time and switch from one app to another. It is “multitasking” akin to the Psion 3a; and not like the desktop we know.

Perhaps, I might venture, Linux on the desktop might be more accepted if there were not several dozen distributions offering “the one true method” of running a window manager, and if the one that most people have heard of didn’t seemingly arbitrarily “change for the sake of change” almost as often as Firefox’s UI…

People, in general, want to use a computer. Not have to relearn a different bunch of methods and quirks at periodic intervals.

It seems that Xeon E5v3 TDP is around 80 W. A Power8 with 165 W TDP will wholly unsurprised be much than twice faster. And the SAS is better for SSDs, the CAPI is better for GPUs or whatever other expansion you want, etc.

Watts in? My P4 box is not used as often as it perhaps should be because it pulls around 100W without monitor. My EeePC pulls around – what was it – 65W (plus 30W for the monitor). My Pi? Pulls 7W (plus ~30W for the monitor).
I know which one I’d rather leave running – and indeed have. My Pi has been on since… about three weeks now. Whenever the last thunderstorm was. I can manage 7W in the background. It’s maybe a tenner a year like that. Even with the recent bizarre EDF bill backdated to 2012/2013.

We must often rely on USB to get things as Ethernet port,

Yes. And how can one offer gigabit ethernet with a USB port that does ~480mbps flat out? USB ethernet to USB storage, it it is a single USB channel to a hub? Well divide the bandwidth by two for that and you’re seeing maybe only twice the speed of bog-standard 100baseTX.
Example? The Beagleboard xM. USB leaves the OMAP as 8 bit data and some control signals and passes through a USB3320 PHY which converts it into traditional USB, which is then passed on to a LAN9514 four port USB plus Ethernet hub.
So you have four USB2 ports and a 100mbit ethernet port. All of which will be throttled by the link between the hub and the PHY which will have the capacity of one single USB2 port when running at maximum speed (assuming the OMAP can manage it)…

ARMv8

I’d avoid ARMv8 as there are fundamental flaws in the Virtualization extensions that make a Hypervisor both slow and allow the guest to become a higher privilege than the host, due to the botched exception model. ARMv8.1 is the one you want, finally a near industry standard Virtualization model.

Just need some tin! Next year possibly.

If you’ve already decided that your Linux platform must be ARM, then no, they’re not an argument.

We definitively don’t see it from the same perspective. I talk here of SERIOUS Linux users (not serious <> fun, but serious = pro). Core kernel devs, professional developers, etc. Here, there is no choice. The platform where most of Linux changes are made is ARM (because of Android), and the platform where most developments are made is ARM/Linux (because of embedded market). For a mainstream user, of course, a 2 kE board is a non sense, but believe me, for people who buy 5kE HDK, it’s something VERY interesting.

Mainstream desktop ARM platform will probably exists, but it’s not the Titanium. Probably more an Android 8 device or a Raspberry 5.

However, most people building real servers care about cost-effectiveness, not whether it’s ARM or not.

Once again, I do not talk of production servers, but of development. Today boards, without high speed Ethernet or Sata, did not give us a good view of performances and limits of future ARM boards. So we need HDK. They are very pricey. So we have two choice: stick with mobile chips, and expect some differences; use 32bit, with this card, or some modern ARM NAS (who are, in fact, good competitors to Titanium).

Of course, real ARM servers will be very different, with 48-96 cores setups, several gigahertz of speed and several hundreds or gigabyte of RAM. Microservers are NAS. Future ARM servers will be monsters… or will not be. But let’s face it: I not have 100 kE to buy one, and don’t have 1ME to buy a prototype. I even don’t have the 5-10kE needed to buy an hardware HDK.

There’s no legacy ARM software in the server market

False. ARM is the king of microservers. See low cost NAS: all ARM.

ARM is also (almost) the king of the biggest servers (along with Power and Mips). See 4G network and relays. Of course, I don’t count here clusters. Clusters are mostly x86 (Power too).

If servers were only PC, why IBM would still be in top5 without x86 offers? Anyway, that’s all of topic. My point is that Titanium is a very good board for RISC OS (as ARMX6), if good support. And a good board for new workloads testing under Linux, if correct price.

…and that RISC OS is hardly a server class operating system. You gave up on it yourself, did you not? ;-)

I don’t see RISC OS as a server OS. It’s just cool for me to see a board good for Linux and storage, with a developer compliant form, to be compatible with RISC OS.

to be compelling as a Linux board

To be useful and compelling is not the same thing.
5 kE HDK with a slow 4 core processor are not very compelling too, but very useful.

I think the reality of that statement depends a lot upon access control to the hardware itself.

OK OK. I mean accessible. We all use USB (or sometimes SD) for installation. It’s more reliable than CD-Rom. But we don’t boot (after installation) on it. Never. We have network and disks for this.

A server will need a method of interfacing (keyboard etc) for when remote connection is inoperable

Yep, Intel network management solutions :) Really I did not see my servers in person for years.

The only time we use USB it’s to put the first layer of OS (or virtualisation stack). From network or with a physical media (USB, CD-Rom, SD).

The second level of defence is nobbling the hardware so it does nothing at all until an administrator mounts the device.

More than this: these ports are only available at boot.

Just need some tin! Next year possibly.

Ahhhh. Back to topic. The same here.
Dual head will probably accelerate things :)

I’m not sure I would call a mobile phone or tablet a “desktop”.

No I can’t. But who cares? :) Desktop / mobile are just linked to technical limits. It’s client/server computing. And client is more and more Android, and less Windows. Of course, a lot of people use and will use desktop computers. But I know also a lot of people who ditch their computer completely for a mobile device.

The slow sales of PC prove that it’s not an epiphenomena. We (= the press) believed people just don’t renew their PC as fast as before. We know today that we were wrong: some don’t and won’t renew their PC.

ARM will probably never win the desktop war. But ARM is definitively the king of client-side and industrial computing.

Ahhhh. Back to topic. The same here.
Dual head will probably accelerate things :)

I think I’m actually way of topic, as the OP is about the London Show and I was referring to a machine with an ARMv8.1 SoC in it. Or more accurately ARMv8.1-R which hasn’t even been announced yet and may never exist.

ARMv8.1-A if and when a machine come out, would only be useful as a Type 1 Hypervisor in RISCOS terms, whereas ARMv8.1-R if it ever came out could run RISCOS with minor tweaks and run a Type 2 Hypervisor under that.

Having said all that, if an ARMv8-R based machine was to come out it could run RISCOS and a Type 2 Hypepervisor, but wouldn’t be particularly efficient when it comes to switching between the Hypervisor and Guest and would need the guest OS’s hardware vectors paravirtualizing – which kind of defeats the point of hardware virtualization in the first place! It’s no better than what ADFFS does today on StrongARM!!

Back on topic…

I did overhear a lot of conversations in my area about the general excitement at the show. Two things came up regularly:

1. R-Comp’s i.MX6 on an ultra-widescreen Dell display
2. Titanium

i.MX6 – Everyone was amazed at the resolution it was running at and I must admit it was a beauty to behold. Andrew would have to confirm the exact resolution it was running at, but it must have been getting close to 4K.

Titanium – The discussion here sadly wasn’t so positive, everyone that mentioned it raised doubts about the price (I don’t believe a price has been announced, so a lot of guessing going on.)

Sadly, I didn’t go to any of the lectures so can’t comment myself. Hopefully the videos will be up soon so I can see what I missed. There was a general feeling of excitement at the show though.

[ARMv8 things]

Agree on all points.

[Show things]

ARMX6 is really a cool platform, well balanced. I did not buy it because of two reasons: not enough money and stronger needs in term of processing power. Titanium is more closed to my business needs, as it’s a workstation for industrial developers, coupled with the desktop version of the Pyra (at home, I use a Pi, and it’s enough). I expect Elesar to provide code to use DSP or second core, drivers for dual head support, PCI Express extensions, etc. As it’s a long term product. I hope also to see some CDDVDBurn support.

Of course it’s not 64bit, but this is not important. ARM is not x86 and ARM 64bit is not ‘a better ARM 32bit’. They are different products for different markets. ARM 64bit will probably replace ARM 32bit completely one day… but IMHO not before 20 years :)

For me, Titanium is today the ultimate dev system for ARM 32bit (RISC OS or Linux), as ARMX6 is the ultimate system for desktop RISC OS use. People who works in electronic market will probably don’t find it very pricey (except if more than 2 kE). Of course some people will like to get it for 100 euros… and to get mainframes for 10.000 euros too.

I’m very pleased too with all the software announcements. Really a lot of things to buy now :)
Did we have news of the new strategy game?
This time I’m on topic :)

I cover a lot of embedded things on Silicon.fr (top5 IT B2B French media). So, of course, I made an article about the Titanium. http://www.silicon.fr/elesar-devoile-carte-mere-arm-32-bits-boitier-de-pc-129802.html

everyone that mentioned it [Titanium]raised doubts about the price (I don’t believe a price has been announced, so a lot of guessing going on)

The original price of a Risc PC (£1500) was mentioned as a likely price for a retail offering of the Titanium, cased. Bare circuit board was guessed at 1k€ leaving ARMX6 a good desktop machine for RISC OS taking price/performance balance into account.

i.MX6 – Everyone was amazed at the resolution it was running at

It was being demonstrated at 1920×1200, 2560×1440 and 3440×1440 on the R-Comp stand, the latter was the recommended maximum because 3840×2160 which was also feasible, was not such a good user experience when tried on a few monitors. Overall the 3440×1440, being demonstrated on a curved monitor, looked excellent but pricey (monitor > £500). Using a 3840×2160 display to show an eigen-value-adjusted 1920×1080 desktop apparently looked stunning but was probably a bit too expensive but at unadjusted 3840×2160 everything looked a bit small.

“The original price of a Risc PC (£1500)”: if that turns out to be correct, it’ll need to be pretty special to justify twice the price of R-Comp’s ARMX6, for example. I know it’s a small marketplace, and we should all support hardware dev, but when you can get a liquid-cooled, Haswell-E 8-core 4+GHz gaming PC with a NVIDIA GTX TitanZ graphics card which can probably run VRPC as fast or faster than any native hardware, plus you get a Wintel PC thrown in for free, for the same kind of money, you’ll need to be pretty dedicated to the platform to go with the Titanium IMHO.

Well it certainly makes my recent iMac purchase look like a bargain!
As for pegging it to Acorn’s price policy, how often did you hear
“I would buy a new RiscPC if I could afford one…”

4+GHz gaming PC with a NVIDIA GTX TitanZ graphics card which can probably run VRPC as fast or faster than any native hardware

not for everything. Also VRPC is limited to 128Mbytes of RAM and a 28Mbyte Wimp slot. Random file system writes are now faster on ARMX6 than VRPC 3GHz with a fast SSD drive on both.

I agree, though, that the ARMX6 price – performance ratio will meet most users’ needs.

“The original price of a Risc PC (£1500)” is far more than I would be prepared to pay.

I’ve already decided that my next computer is going to be a PiTop.

PiTop for me too, just waiting for them to become available

If the bare board is going to be €1000 (or £1000, which is even worse)… then my point about legitimate ARM server boards comes back into play, for using this with Linux.

http://www.ldlc.com/fiche/PB00191584.html €874.96 for an Applied Micro X-Gene board. Add GPU if you want something better than the ASPEED controller.