Chip maker Marvell, which acquired the Xscale ARM RISC chip business from Intel in 2006, is now officially gunning for the chip giant in the market for low-powered, cloudy infrastructure servers. If there is such a market. At the ARM Techcon 2010 in Intel's hometown of Santa Clara, California, Marvell unveiled its quad-core …
What we really need is...
1) Some motherboards sporting this or similar chips
2) Those motherboards to be in *TX (ideally mini-ITX or micro-ATX) format so that we can use standard chassies
3) Those motherboards to have DIMM sockets for memory rather than have the memory soldered on so that we can use cheaply available memory in them rather than be stuck with whatever the manufacturer decides to do
4) Most importantly, we need such motherboards to be available in single-figure quantities at prices at least in the similar ball-park to comparable Atom boards
Back in the days around the CHRP debacle, a fair-sized group of Linux enthusiasts was crying out for someone to sell them a PowerPC motherboard that was relatively consumer-priced and decently compatible with mass-market PC parts. I'm not sure whether they were denied out of self-interest (neither Apple nor IBM can have been keen on the idea of competition from white boxes) or mostly just incompetence (it *is* the PowerPC Consortium we're talking about), but history suggests that the PowerPC architecture probably wasn't really in a position to blow off that influential support base.
Build it in an Intel-compatible pinout?
How easy would it be to build it in a Xeon-compatible pinout and stick it in a readily available Xeon-based box (and stick a Linux on it)? Just as an experiment, obviously.
This could get interesting if done properly (not sure if hampering it with a Xeon pinout counts as proper, but...).
Has anyone tried that intel swap since Cyrix?
Intel wasn't too fond of that game.
There are many reasons why we have so many pin patterns today.
I think the high-level of integration between chip and chipset make this a bit of a non-starter. It would mean using the same proprietary connections as a Xeon and, thus, having to license them from Intel. Plus, one can assume that the actual ARM modules are quite a bit smaller than the Xeon's with their masses of cache.
As TPM says, real world benchmarks with comparative loads - web transactions with nginx and Postgres are likely to get tongues wagging. Can you build a data centre out of them without a cooling system?
I love this.
Now I want it in a mini-ITX form factor board with some basic desktop bits and bobbins. Then I want a Gingerbread installer for it.
My new desktop has arrived.
... already exist, but they are either outrageously expensive (aimed at development and prototyping of systems):
or are not available in small quantities at a same price:
They still fail the requirements I mentioned above WRT cost and basic features such as DIMM expandable memory.
Close enough? No Gigabit Ethernet or SATA (or USB3), and no memory expandability, but it does run Ubuntu Linux, Android, and Chrome OS.
That is fantastic. :D NEW CHRISTMAS PROJECT GET.
If they build it...
...what will it run?
Whilst I've seen arm builds of linux kernels, I have no idea if these are designed for server usage (I've seen builds for PDA's). Are there any cloud ready OSs out there that can exploit this CPU, or will server houses be required to recompile for this new target?
I do hope that this does succeed, but I can't help feeling that they've invented the infernal combustion engine before they've developed a fuel. I'm sure the knowledgeable reg readership will not hesitate to bring me upto speed however.
Complete Linux distros exist for ARM architectures sporting all the packages available in the x86 ports. So as far as software goes, ARM will do everything x86 will if you use Linux.
Power's performance isn't as great as the theoretical figures show. PowerPC (a very small sibling of full fat Power) has atrocious performance compared to x86 (G5 barely kept up with a Pentium 3 5 years it's senior before Apple rightfully canned it in favour of x86). Power's performance per Watt is also pretty dire, not to mention that it's overall power consumption makes it unsuitable for reasonably sized servers (things that fit into 1U).
There is already am ARM port of Debian (and Fedora).
Don't underestimate the leverage MS and Intel and apply. Intel are smart enough to recognize that a push from non-x86 side is way bigger threat than AMD's x86 products ever were or are going to be. Google about what happened to the Asus' ARM based eee that mysteriously got canceled after being shown at a show and generated a huge interest. Lenovo Skylight went a similar way. The closest we have now is the somewhat disappointing Toshiba AC100, which is late and not as good as the other two canceled machines I mentioned, but I await with great interest details on how to get a full fat Linux distro running on it instead of the wholly unsuitable Android they ship it with. As for economies of scale, volume of ARM variants already dwarfs the volume of x86 variants (ARM is in most mobile phones, and other embedded systems), so that isn't the only cost factor.
Ubuntu already has pretty complete support for a variety of ARM systems, so getting Linux support shouldn't be a big problem. The most obvious stumbling-blocks for ARM as a server processor are lack of hardware virtualisation and 32-bit addressing, I would guess. Both are addressed by the Cortex-A15 design, but we won't see any A15s on sale for years yet.
Hardware virtualization is something which is essential in x86 world, specifically for microsoft (since there you are trying to install an OS from a DVD, not one that you built yourself) and is sure handy for installing linux from a standard coaster too. But you shouldn't need it on an ARM server. Without hardware virtualization support, the OS can 'virtualize' storage over the network in several ways (ATAoe in addition to normal network shares like nfs) and can also 'virtualize' whatever display/console technology you need over the network (for a server, that might as well be just sshd, but could be vncserver too). This doesn't support running more that one virt, machine per CPU, but who needs to do that when the CPU is so small and cheap? It you want a more 'physical' virtual disk, you could make a chip that connects to the PCIe and emulates a SATA controller, tunnelling the traffic over some fabric to the actual disk (like that weird 256xatom box that was announced a month or two ago). The whole Wintel development history has been a story of new hardware conforming to existing software (including, e.g. every modern display adapter can emulate all the VGA/EGA/CGA cruft from the 80's) -- and to some extent we've all been trained to expect that and see it as normal -- but the way it's done elsewhere is you allow for a bit of system coding so the software can adapt to new hardware. It will result in a more efficient system, compared to hardware virtualization.
If the server market is ripe for attack by non-x86 archs, then why aren't we hearing more about POWER these days? Is it because x86's grip on the market is actually secure? Or because the opening that exists is only for low-power architectures like ARM? Or because IBM is too fat and happy from fleecing high-end scientific-cluster builders and the mainframe faithful to want to risk a serious bid for the server mainstream? (If so, is IBM likely to change its mind if it sees ARM making serious inroads?) Or is POWER actually making big gains lately, and I just haven't heard about it? Not being sarcastic, I'm genuinely curious...
Power7 owns the Unix market
IBM sells Mainframes, Power and System X. Customers buy the right systems for the right workload and the price corresponds to the value. We don't use mainframe for our printer servers and we don't use x86 for our expensive software licenses. It's because of Larry's pricing we put all their software on Power.
RE: Power7 owns the Unix market
Whilst I'm amused by Ms Park's idea that "Power7 owns the Unix market", the rest of the post is accurate. Why would you expect an enterprise CPU like Power to be competing in the sub-Xeon space? Us customers usually buy whatever will do the job with the least hassle at the lowest pricepoint, and for a lot of stuff below the enterprise level, x64 is currently the only real choice. It's a bit like asking Porsche why they haven't cornered the small hatchback market, it's just not where Porsche plays.
I hope they can keep running with this
What I fear is that none of the established mobo manufacturers will play with this for fear of upsetting their Wintel masters and it will be relegated to specialty uses and appliance style products.
I want what has been listed above. TX form factor boards that take standard memory SIMMS, and provide all the usual PCI, USB and networking goodies.
Then I want a Debian port to install on it.
RE: I hope they can keep running with this
"What I fear is that none of the established mobo manufacturers will play with this...." Well, Goat Toe Jam, that's similar to the worries people had when AMD first started talking about making a competitor to Intel's Xeon. If a small competitor puts something together that starts to make money then the bigger players will get interested. Intel saw this coming which is why they went for the Atom route as a defensive move, but if the new ARM chips can be cheap, reliable and in good supply then they'll probably do very well in non-Windoze appliances, which could give the economies of scale to attract the server vendors into making low-end servers in quantity. The "scared of their Wintel masters" hyperbole doesn't hold - all the top vendors, even Dell, have worked with Linux and non-Intel options.
It's not hype, Matt. It's a matter of fact, Matt.
"The "scared of their Wintel masters" hyperbole doesn't hold - all the top vendors, even Dell, have worked with Linux and non-Intel options."
You've already forgotten the evidence from Dell+Intel in the US justice system which shows you just how much Intel were willing to pay to motivate Dell to ensure Dell volume products stayed Intel? Fraudulent accounting, hundreds of millions of dollars of kickback in a single quarter, and so on?
Ever tried buying Linux-ready volume product from Dell? There's a reason folk talk about the "Windows tax", a reason other folks understand even if you don't. Microsoft have just been lucky so far; lots of folk know why the manufacturer and retailer adverts all say "xxx recommends Windows 7 Braindead Bees Knees Edition" somewhere in the ad, even if you don't understand.
RE: It's not hype, Matt. It's a matter of fact, Matt.
The mere fact that Intel and M$ have to (alledgedly) pay bribes to the vendors to keep them inline just shows they are not "in fear of their masters" but happilly taking Intel's and M$'s cash. If either was the "master" they would simply threaten to withdraw their product, not pay extra to keep the vendor's sweet. Would you like me to explain the definition of "master", in very small words, just so you can understand it?
"....Ever tried buying Linux-ready volume product from Dell?...." Actually, yes I have, but we're probably a lot bigger customer than you so it's probably not surprising that Dell were a lot more helpful with us.
get us the real details man
Timothy Prickett Morgan, great ....Now go and find out the real Details Man.
Get SOMEONE, Anyone with access to these and related SOC/chips and give us the real details, and (seriously) run x264 on them to give people some real life data throughput for these quad 1.6 chips Neon SIMD etc
its true that much of the existing Arm v7 Cortex A8/A9 x264 NEON 128bit SIMD assembly needs writing yet to give at least a 30% further speed boost, but at least we have existing data to then compare against an existing 800Mhz and other Arm Cortex A8/NEON
2010-08-05 08:20:04 < Dark_Shikari> if you care about optimizing for ARM though, the first thing you should do is finish up the NEON....
v7 Arm cortex A8/Neon 128bit SIMD Freescale i.MX51
Such things... #
Posted Wednesday 10th November 2010 10:06 GMT
... already exist, but they are either outrageously expensive (aimed at development and prototyping of systems):
or are not available in small quantities at a same price:"
Gordan, your not looking hard enough mate, theres more than a single Arm vendor you know, and there already is a low yield (as in the price can come down a LOT with scale) v7 Arm cortex A8/Neon 128bit SIMD Freescale i.MX51running at 800MHz, available in single quality's.
that's the Genesi Efika single core A8 in both box ($200) and laptop ($300) form for developers AND enthusiast/Mod Builder alike right now for instance.
see the youtube video's
"OpenVG hardware acceleration by Freescale"
OpenVG overlayed on top of OpenGL 2 seems like a very nice flash scripting replacement If someone writes the code to use them in a generic video app using FFmpeg etc.
"Microsoft Word on ARM Powered Laptop using Genesi and Citrix solutions"
very cool for a cheap developer/enthusiast kit on the face of it, that ORA boot firmware sounds interesting as you dont need an expansive/expensive BSP
they are actually the one's getting and initially supporting for the long time now the actual full Arm Linux porting of different distro's too as they did with the PPC linux before that, Plus OC there's already QNX RTP6 (the RIM OS now) ported to it for long time now , AROS and google OS along side the usual full Linux distro's.
***Gordan, your not looking hard enough mate, theres more than a single Arm vendor you know, and there already is a low yield (as in the price can come down a LOT with scale) v7 Arm cortex A8/Neon 128bit SIMD Freescale i.MX51running at 800MHz, available in single quality's.
that's the Genesi Efika single core A8 in both box ($200) and laptop ($300) form for developers AND enthusiast/Mod Builder alike right now for instance.***
I saw those before. Thanks for reminding me. :)
Shame they don't have a DIMM slot for memory expansion. 512MB isn't really enough when Firefox and Chrome have a basic memory footprint of 100MB - depressing as that may be.
Where's PPC, they dropped the ball and ran away
Where's POWER? #
Posted Wednesday 10th November 2010 10:07 GMT
If the server market is ripe for attack by non-x86 archs, then why aren't we hearing more about POWER these days?"
Where's PPC, simple, they dropped the ball, they and the PPC (powerPC) licensee's actually had and ran away from the PPC SOC market place, there was a market to be had, and Genesi tried to cater for that market in a limited way with their Original "PPC Efika", but when the actual PPC SOC's produced kept coming back with Hardware error after error on each new chip revision , Genesi finally gave up trying to help them make a cheap generic PPC SOC board (£99) and moved over to Freescale's i.MX51 v7 Arm cortex A8/Neon 128bit SIMDat 800MHz and made the Efika MX PCB