Quote:

Thanks Andreas, though I am hardly an authority on the subject, I just like the stuff, and I enjoy SIMD algorithm optimizations immensely. Not to mention, that no matter how much I read, I always end up thinking that I know nothing and have to read more. If I was an authority, I could perhaps influence Freescale to revive AltiVec and or enhance it -I did sent some email To Freescale with suggestions on AltiVec 2, a few years ago, didn't really get anything more than "thanks we'll look into it". :)

Quote:

Aha, thanks. I agree with you. So here is where Gerald was last seen in the wild:

[img]http://www.reuters.com/resources/r/?m=02&d=20090331&t=2&i=9520862&w=460&fh=&fw=&ll=&pl=&r=2009-03-31T231705Z_01_BTRE52U1SOQ00_RTROPTP_0_GERMANY[/img]

(Will inline images ever be back on MorphZone? Andre, why not?)

Quote:

Hardware development is mostly done by Pegatron -with the cooperation of bPlan-, and production is done solely by Pegatron but bPlan are very close to finishing Aura development on the EfikaMX, thus providing an high level abstraction to the devices/drivers of the device in an OS-agnostic way. Of course, an appropriate Linux kernel is also developed at the same time.

Quote:

Are you sure of any of this information, Andreas? At 65nm we were dealing with separate components. I haven't seen a tear-down of the newest XBOX360, but one chip would combine three different components.
Since the system has to operate the same as older XBOX360s this can't offer any architectural advantages, only lower cost of production.

That doesn't really strike me as an advantage to the user (unless Microsoft wants to pass on some of the savings). Its a little like Intel giving the latest Atoms an on die memory controller, but not changing the design to take advantage of it being on-die. There's no performance increase.

In a situation like that, I'd rather have separate components so I don't complicate or limit upgrade ability.

Quote:

Thanks for the clarification Konstantinos. So, what Genesi/bPlan/Pegatron are doing is making a new computer and custom operating system. Well, almost... But in many regards, it is. Once done, it's all a matter of convincing the world to write kernels for Aura.

The product is almost ready. Hey, I swear I red the word "August" there a moment ago...

>> I consider feanor to be an authority when it comes to SIMD computing.
>> So it's quite naturally for me to ask him on that.

> I am hardly an authority on the subject, I just like the stuff, and I enjoy SIMD algorithm
> optimizations immensely. Not to mention, that no matter how much I read, I always
> end up thinking that I know nothing and have to read more.

Alright, make that "an authority compared to most other MorphZone members including me" then ;-) Could you please have a look at my question over there?

> So here is where Gerald was last seen in the wild:
> http://www.reuters.com/resources/r/?m=02&d=20090331&t=2&i=9520862&w=460&fh=&fw=&ll=&pl=&r=2009-03-31T231705Z_01_BTRE52U1SOQ00_RTROPTP_0_GERMANY

Yes, AFAIK. Bigger version:

http://www.reuters.com/resources/r/?d=20090331&t=2&i=9520862&w=1024

> Hardware development is mostly done by Pegatron

That's what I had thought.

> with the cooperation of bPlan

Interesting, that's news to me.

> production is done solely by Pegatron but bPlan are very close to finishing
> Aura development on the EfikaMX

Again what I had thought.

Quote:

AFAIK, yes. Power7 -which is the only cpu currently supporting VSX- supports VMX as well. VSX is like an extention, or a superset as you put it. I don't know if that's guaranteed to be so according to the Power ISA, but at least on Power7 they implemented both. Though, I doubt that they would implement one without the other, to be honest.

Quote:

Let's just say that bPlan has lots of experience which proved valuable many times :)

Quote:

This sounds interesting. I still think there could be some potential.

> Are you sure of any of this information, Andreas?

I just quoted some things from the interwebs. How can one be really sure of anything written there? ;-)

> I haven't seen a tear-down of the newest XBOX360, but one chip would
> combine three different components.

Yes, that's how I read it too.

> Since the system has to operate the same as older XBOX360s this can't
> offer any architectural advantages, only lower cost of production.

Yes, no implications to the opposite. I was just referring to you stating that "all current processors offering this feature have horrible GPU performance". So I think this new Xenon version (or 'XCGPU' how it's called) seems to be a processor offering this feature while *not* having horrible GPU performance. But as I said, unfortunately not available outside Microsoft.

> In a situation like that, I'd rather have separate components so I don't
> complicate or limit upgrade ability.

I don't know if today there really would be a need to upgrade from an R500/R600 hybrid like the Xenos GPU yet.

No once console hardware is designed, its basic functionality is locked in until a completely new replacement is designed.
And with current consoles running all functions via a hypervisor, we're not likely to see complete access to functionality anyway (probably about the time they're ready to be replaced).
With regard to on-die consolidation, I rather meant that if it offered no benefit to a PC user, then it would present some disadvantages.
Again, with what AMD is proposing, I can see a clear advantage in performance rather than just a lowering of costs.
If PowerVR has advanced as much as you've pointed out, perhaps some of Intel's on-die projects will feature it.
Hopefully, we won't be forced in that direction with PPCs until an advantage to the user is present.

>> Does a CPU implementing VSX automatically support VMX/AltiVec,
>> i.e. is VSX a strict superset?

> AFAIK, yes. [...] VSX is like an extention, or a superset as you put it.

That would be the best thing I heard since reading first about VSX and would fully make up for the diverging focus from AltiVec/VMX.

> I don't know if that's guaranteed to be so according to the Power ISA,
> but at least on Power7 they implemented both.

Err, now that sounds like a contradiction to what you just wrote before. If POWER7 implements both VSX and AltiVec/VMX (separately), then it would mean that the latter is *not* automatically there by having the former implemented, i.e. a processor with VSX but *without* AltiVec/VMX compatibility would be technically possible (and conforming to Power ISA 2.06+).

> Though, I doubt that they would implement one without the other, to be honest.

I fear that if Power ISA 2.06+ allows to implement VSX without AltiVec/VMX compatibility then Freescale will do just that, if they adopt VSX at all that is ;-)

Quote:

Yes, you're right. The fact is that I'm confused myself. From the doc I posted earlier (page 4):

Quote:

It's apparent that VSX is an extention of VMX. I got confused because those few times I read about VSX, I missed -or they omitted- VMX mentioned. And until now -that is until you made me read more about it :) - I thought they were distinctive units.

[ Edited by feanor on 2010/7/7 1:00 ]

> No once console hardware is designed, its basic functionality is locked in
> until a completely new replacement is designed.

I know, but I'm not about the XBox360 or any other console hardware, but about the XCGPU in itself and the (theoretical, but most improbable) possibility to be able to use it in an own design which is not a console. But yes, since it's not available to anyone besides Microsoft anyway we can very well pretend this 'XCGPU' doesn't exist at all ;-)

> with current consoles running all functions via a hypervisor, we're not
> likely to see complete access to functionality anyway

Yes, I fully agree. See my statement there: https://morph.zone/modules/newbb_plus/viewtopic.php?topic_id=7186&forum=11#74250

> If PowerVR has advanced as much as you've pointed out, perhaps
> some of Intel's on-die projects will feature it.

Maybe, not unlikely. After all, Intel is a PowerVR licensee. Intel used the PowerVR MBX in XScale PXA27x support chipsets years ago, and they currently (since November 2008) use the PowerVR SGX 535 renamed as GMA 500 in Atom Z5xx (Silverthorne) support chipsets.

http://en.wikipedia.org/wiki/Intel_2700G
http://en.wikipedia.org/wiki/Intel_GMA#GMA_500
http://en.wikipedia.org/wiki/Poulsbo_(chipset)#Graphics_core

|||


[ Edited by Velcro_SP 20.04.2011 - 07:50 ]

> With the Linkbook and the Asia etc. the strategy is a little different. It consumes
> less wattage even when in full-operation. (At least that is the case with the 5121e
> and I think that is the case with the ARM systems.)

"The biggest downer is an amazingly short battery life, someting very surprising, given the supposed MPC5121e SOC."
https://morph.zone/modules/newbb_plus/viewtopic.php?topic_id=6726&forum=11&start=60#73290

@ Velcro

literally an e300/400 doesn't lead nowhere. And while there are the limebooks I don't see them marketed in a way I would expect a market success. Look, last week there was a 99 EUR offer at a supermarket chain (Netto) for a little netbook, probably MIPS driven and coming with WinCE. More or less crap probably, but also working. *If* the limebook was offered at a similar price, or say 200 EUR (at least significantly less than an Atom based netbook), then I would assume they could get sold in serious quantities. But they aren't. Look to the southafrican offer. It is a combined offer for the device and the mobile internet access. But the prices doesn't seem too attractive to me. For that price they easily could include a higher spec'ed device.
But as said, if they would sell those mini laptops for 150 EUR I think they would find their way to enough customers. But with the current price tag it is just too little power. An Atom based device is not really more expensive, the announced ARM devices are cheaper and more powerful. The Atom based devices have long lasting batteries now (often 6-9 hours).
With a 8610 you cannot compete pricewise with an Atom board, but it has some serious power. Hence I think it would be the best chip for a netbook.
The limebook comes with a too little battery. That's pretty unfortunate IMHO, because that way they don't use one of their advantages - the really low power uptake.

Anyway, my Eee costed me ~250 EUR (incl. 19% VAT) nearly two years ago - I think that's the minimum bar to jump over if you want to score.

It must meet at least one of teh following criteria:
a) cheaper
b) offer much longer battery time
c) more powerful

c is out of question with an e300/400MHz, b) unfortunately not implemented with the limebook, a) could probably be, but curently isn't.
So yet, the limebook pulled the bar and got more or less diqualified. And to keep that sport analogy: Maybe they overjumped the 2m bar, which is really impressive (I cannot!!). But it isn't enough to join the elite. And as long as you don't join the elite you are not earning money/fame and must try harder or accept that it is (serious) hobby/amateurism only. The competition is just very hard.

> last week there was a 99 EUR offer at a supermarket chain (Netto) for a
> little netbook, probably MIPS driven

It's ARM9 (VIA VT8505), similar to some of the Cherrypal Africa devices with VT8500 (i.e. those Africas that were not XBurst/MIPS).

>It's ARM9 (VIA VT8505), similar to some of the Cherrypal Africa devices with VT8500 (i.e. those Africas that were not XBurst/MIPS).

That particular processor is usually supported by a chipset that only offers 800 x 480 video resolution.

And the XCGPU you mentioned, Andreas, may be two separate dies under a common heatspreader.



[ Edited by Jim on 2010/7/7 20:30 ]

> That particular processor is usually supported by a chipset that only offers
> 800 x 480 video resolution.

I knew that this POS netbook only supports this low resolution, but I didn't know if that's due to the lowres display or the chipset. So thanks for this info. Anyway, I just wanted to tell Zylesea that it's ARM and not MIPS as he supposed.

> the XCGPU you mentioned, Andreas, may be two separate dies
> under a common heatspreader.

That may be, yes. I really don't know, just trying to interpret Wikipedia's "same package" and "integrated chip".

I thought it was MIPS (I just read the advert that flew into my letterbox and didn't paid more attention) , because when netbooks (the class of devices not the TM) were rather new, the all cheap offers by several Telcos (IIRC Phonehouse, Vodafone) were MIPS based.
Anyway, *I* don't thought that particular offer was too interesting, except for the fact that supermarkets offer 99 EUR netbooks today that were probably sold off quiet well.
I know enough ppl who are kind of addicted to supermarket special offers - I call that the Aldi addiction desease and if I ever happen to write a Psychology thesis I will chose that topic. Well, actually not too likely and if I eventually change to Psychology (could happen) I will rather end up with cognition/vision/motion .

Psychology? What are you crazy, Zylesea? Stick with a hard science, everyone I now in that field has issues.

And Andreas, POS (?!), that's the harshest term I've ever seen you use.

Quote:
Coming from low level neuroscience I am considering to switch to a group working on mental movement representation and embodyment. It is something embedded into a interdisciplinary institute (CITEC Bielefeld), but that particular group I am considering belongs to the faculty of psychology and sports. So, don't worry I'll probably not ending up on the couch (at least not because of *that*)...