Quote:

GeekBench is funny: OSX is 2000 points slower than Windows in cpu hungry tasks it seems. E.g.: i7-3820QM 2700 MHz. That's the power of two G5/1.6Ghz, I guess there's something badly programmed (OSX or Geekbench, who knows). BTW iPad5 points like a dual G5/1.8ghz I would prefer to see real life tests instead of semi-synthethic tests. Let's say building one or various entire projects simultaneusly or rendering, scaling 100 pictures and applying filters,... I don't see many of these wunderbar ARM high-end desktop machines available anywhere, where do I plugin various SATA HDs? can I fit a modern PCIe gfx card?

Since ARM is so great can I have an ATX board that beats or even emulates last QuadG5? Since arm is so cheap, so powerful and so widespread it should be easy to find a dozen of boards that can be bought in almost all european countries, right?

[ Edited by Crumb 25.09.2012 - 00:51 ]

Quote:

Things like phones are locked but there are also an ever growing number of cheap dev boards which are not locked to anything. There's so many now I gave up trying to follow them.

Once the A15s arrive you'll see them appearing on them to. That'll give you the power of a G5 close to the price of an Efika.


Quote:

Cheap is an understatement. Look up "cloud stick" on eBay, ARM computer for £30!

Andreas_Wolf,
Quote:

Wow! I really worded that one badly.

What I meant was in relation to the 970 and the ARM CPUs you were comparing.

>> e6500 is certainly above Atom in performance.

> What I meant was in relation to the 970 and the ARM CPUs you were comparing.

I really don't want to hazard a guess here as to how e6500-based chips would compare to the PPC970 or to chips with Cortex-A9 cores or Krait cores or to the Apple A6 in Geekbench. And I guess we will never see Geekbench scores for e6500 anyway as there's no Linux/PPC version of it.

> iPad5 points like a dual G5/1.8ghz

Yes, minator already mentioned this in another thread.

> where do I plugin various SATA HDs?

I don't know about "various", but SATA isn't necessarily a problem any more:

https://morph.zone/modules/newbb_plus/viewtopic.php?forum=3&topic_id=7675&start=272

That was 4 months ago, so there's probably even more by now.

> can I fit a modern PCIe gfx card?

While there're various ARMv7 SoCs with integrated PCIe controller, I've yet to come across a board with ARMv7 SoC and PCIe slot.

> can I have an ATX board that beats or even emulates last QuadG5?

To quote myself from this thread:

"Today's fastest ARM CPUs are still slower than the PPC970 CPUs in most (i.e. 2.0+ GHz) G5 desktop machines."

> it should be easy to find a dozen of boards that can be
> bought in almost all european countries, right?

Right. You could try to start there:

http://en.wikipedia.org/wiki/List_of_single-board_computers#ARM_based

Code:

Most of these boards are pretty low-end and are fun to play with but these aren't a real replacement for desktop systems IMHO.

> these aren't a real replacement for desktop systems IMHO.

True. I guess that real desktop systems (i.e. what we understand by this term) will emerge with Cortex-A15 in 2013 at the earliest. It may even take as long as initiatives like nVidia's Project Denver (2014 or later) to bring ARM (ARMv8 in this case) really to the desktop. We'll see.

Considering the machines we are used to using, a Cortex-A9 might be adequate.

@Jim

Quote:

"Adequate", quite true, it will do what most people wants/needs to do with their computer (in fact, it *already does*, for millions of people around the world, albeit not in the shape of a "desktop"), but that's not what Crumb is talking about. I feel he wants workstation or "Gaming PC" performance, or thereabout at least. But it's not like his PPC machine that he is running MorphOS on comes close to Core i7 anyway, and a move to ARM would definitely be a step upwards from today, at least when/if that would happen (i.e. not today, or tomorrow, or this year, or...)


@Andreas

Quote:

Personally, I would bet on the latter. Not because Cortex-A15 per se wouldn't be enough, performance wise, but I think most chip/motherboard manufacturers will still aim for handhelds, smart-TV's, etc, with this core. And of course servers, since this is a pronounced target application for this core AFAIK, at least from ARM themselves. IIRC the nVidia president pronounced their "Denver" to be a "x86 killer" though, and I think they will make Denver based CPU's for all kinds of applications, including desktop and servers (and *I still* wouldn't be surprised if at least one of the upcoming console platforms will use this, most probable the new X-box, for several reasons).

I think we will see a fork in what we perceive as "desktop" in a close future. Most people (both professionally and at home) are definitely using over-performing computers for their computing tasks today. People are using core i5's and i7's mainly to type mails, browsing the web and office kind of applications. Expensive and power hungry CPU's, costing corporations and families thousands of dollars, completely in vain, when (as Jim said) a Cortex A9 could easily do the job they needs, for a fraction of the resources (both investment and running resources).

So with a "fork" I mean that I believe that the market for gaming PC's and workstations will continue its rapid hunt for performance and "marketing numbers" to advertise, while the "general use PC" will shrink in both PCB size (it won't need to be expandable, hence less slots), power consumption, performance and price. I think ARM will play a role here, and while I think Microsoft's agenda with ARM is first and foremost to meet Apple/Samsung/Google/etc in the handheld/tablet/smart-TV/whatever consumer markets, I also think they have more on their agenda with Windows 8 for ARM than just that, at least I think they want to open up some options. I think they are very familiar with nVidia's (and others as well) agenda with ARM, completely unknown to us of course, and I think they are "playing both sides" right now (both x86 and ARM) to see how things will fold. I think there is a bigger turmoil than ever in the industry right now, where paradigm shifts *definitely* can happen.

At least this is *my* opinion!

That said (again), I wouldn't mind MorphOS going x86, if it could be done in a smart way!

> Considering the machines we are used to using, a Cortex-A9 might be adequate.

I don't think a 1.5 GHz Cortex-A9 would be a match for the 1.5 GHz G4 in my Mac mini. A 1.5 GHz Krait maybe, but not Cortex-A9. Of course, this would also depend on whether MorphOS on such ARM CPU could make use of more than one CPU core. But then, if it could, there're also dual-G4 PowerMacs which are already supported by MorphOS with one core. So I think, for me to consider it as an adequate match (or even improvement), the ARM chip must be faster than my G4 on a per-core basis.

> Denver based CPU's [...] and *I still* wouldn't be surprised if at least one of the upcoming
> console platforms will use this, most probable the new X-box, for several reasons

I'm quite confident that Project Denver will be too late to both the next XBox party and the PS4 party.

At the moment if you want high end desktop performance you have no choice but Intel and AMD. I run a few apps that can easily max out even high end CPUs so I just got myself a high end laptop with a rather nice quad core i7.

The apps are Photo editing, Movie editing and Synth plug-ins.

One of those apps (photo editing*) is actually disc bound and I suspect movie editing* is too - even with an SSD. Only the synth plugin is purely CPU bound.

*In both cases this is exporting of the finished photos or movie. This will max out the processor but it will then spend more time accessing the disc.

ARM parts are mainly designed for mobile so are limited by the power available which tends to be very low. This however is changing. The A15 is designed to handle higher clocks and it'll be appearing in servers where you get wall socket power. Build them on a high speed process and they should be pretty potent. They wont be top end i7 level but they'll be a lot faster than today's ARM cores.

In the mean time I guess you'll have to do with the 2GHz dual A9 Rockchip just announced (shipping this year).

BTW what does desktop mean anyway? I've used a laptop as my main machine for a decade now. Slots? Who needs them?

> 2GHz dual A9 Rockchip just announced (shipping this year)

I hope this will be more real than the 2 GHz announced by Nufront (delivered as 1.5 GHz) ;-)

> what does desktop mean anyway? [...] Slots? Who needs them?

Crumb says he does, for PCIe graphics cards, and multiple SATA ports for HDDs :-)

@Andreas

Quote:

Smoke and mirrors...

I bet!

@minator

Quote:

That's actually what I mostly use my PC for.

Just the other day I had to use some "new" (for me) application (don't ask) that should render a quite large video file. It had some "use HW acceleration turbo" feature clicked in by default. I used that. And wondered why nothing happened. Then I restarted it without that check box, and it completed within a few seconds by CPU rendering...

However, I can't see myself using MorphOS for anything like this...?

takemehomegrandma,
Quote:

Not me.
Its still primarily web browsing, e-mail, and word processing.
So I guess I don't need Intel or AMD.

Andreas_Wolf,
Quote:


My personal benchmark for this would be 1080p video which even a slower A9 should be able to do.

>> I don't think a 1.5 GHz Cortex-A9 would be a match for the
>> 1.5 GHz G4 in my Mac mini.

> My personal benchmark for this would be 1080p video which
> even a slower A9 should be able to do.

Cortex-A9 at 1.5 Ghz or slower is definitely not able to do that. What is able to do that in many ARM chips is an auxiliary video decoding core that happens to reside in the same chip as the Cortex-A9. Utilizing such video decoding core under MorphOS would of course require a special MorphOS driver to be written for that particular core, not different from when you'd put a card with such video decoding chip into your PowerMac G4.

Except (of course) MorphOS doesn't support GPU assisted decoding (and I'm not using OSX).

> Except (of course) MorphOS doesn't support GPU assisted decoding

GPU accelerated video decoding is again another matter than video decoding done by dedicated video decoding cores. AFAIK, current devices with Cortex-A9 core(s) do it the latter way, not the former. MorphOS supports neither, and it wouldn't automatically support either solution if it was made to run on Cortex-A9.

Interesting. I had assumed they would use the GPU.
If the documentation for this function and existing Android or Linux code examples exists, why couldn't the guys use this functionality.

To me, this would be a great selling point on ISA transition.

>> GPU accelerated video decoding is again another matter than video
>> decoding done by dedicated video decoding cores. AFAIK, current
>> devices with Cortex-A9 core(s) do it the latter way, not the former.

> Interesting. I had assumed they would use the GPU.

I assume from the block diagrams of various chips with Cortex-A9 core(s) where it can be seen that the 2D/3D GPUs are distinct from the video decoding/encoding units:

http://www.freescale.com/files/graphic/block_diagram/IMX6Q_BD_IMG.jpg
http://www.samsung.com/global/business/semiconductor/file/product/Exynos4210-0-0.jpg
http://focus.ti.com/en/graphics/wtbu/OMAP_44x.jpg
http://www.stericsson.com/images/block_diagram/U9500.jpg
http://www.nufront.com/UploadFiles/image/%E6%97%A0%E6%A0%87%E9%A2%98.bmp

> If the documentation for this function and existing Android or Linux
> code examples exists, why couldn't the guys use this functionality.

If the premises you mention are met, then I think the MorphOS Team could provide support for the dedicated video decoding cores that are inside the ARM SoCs. It'd mainly be a matter of human resources I think, so a hypothetical MorphOS running on Cortex-A9 may (initially) very well lack support for several on-chip functionalities such as video decoding or 3D graphics.

Andreas_Wolf,
Quote:

Yes.
GPUs need more power than dedicated video blocks so that alone will get them used.

GPUs also aren't really suited for this sort of thing, they can accelerate some of it but not all.