|||

[ Edited by Velcro_SP 24.04.2011 - 10:25 ]

Quote:
MoC= Motherboard ona Chip.

If they still use the 5121 the Tricore is
1) The 603e @ 400 MHZ (which gets a bit less than 800 MIPS)
2) The gfx core
3) the 200 MHz AXE unit.

I guess because they don't want to say 400 MHz (which sounds too less today) and 800 MIPS (which isn't too impressive, too) they were just adding the theoretically achievable MIPS number of each subcore of the 5121 and put it to the overall number..
Cheap trick.

The 603e is quite fast at integer calculations, not so fast for floating point arithmetics and has no altivec unit.

http://trendsupdates.com/ces-2009-limepc-unveils-new-netbook-for-249/
http://www.umpcportal.com/2009/01/lime-note-and-lime-os-updated
http://www.liliputing.com/2009/01/limepc-199-freescale-powered-netbook-video.html
http://www.youtube.com/watch?v=rG_hS6pfv2w
http://www.youtube.com/watch?v=l1OBGl4csHU (apparently this)

...and as I assumed there are already people mistaking "1700MIPS" for a MIPS CPU.

terrible design

i.MX515

http://www.linuxdevices.com/news/NS8595694202.html
http://www.freescale.com/webapp/sps/site/prod_summary.jsp?code=i.MX515&tid=FSHBNR
http://www.liliputing.com/2009/01/freescale-powered-netbooks-could-hit-200-this-year.html

Quote:

Fantastic links as always, Andreas. I liked the "liliputing" one most, as it has a video.

Quote:

limepc-3.jpgYes, with the latest buzz about ARM from freescale, everyone (and themselves?) has forgotten about their PowerPC line. The fact that also LimePC has no mention to the processor family is quite shocking too. They mention "1700 MIPS Tri-core MoC", explicitly leaving out the brand and model. That's not normal, I'd say. If BBRV's statement about LimePC not having paid freescale (!) is true, I'd understand, though.
It seems we morphers are the only ones in the world that know that the LimePC uses the MPC5121e. I guess the very own freescale has forgotten it too, as all the buzz that accompanied the MPC512x family launch has completely vanished, and seems this new ARM based family is replacing it.
Well, at least, the MPC 5121e is shown in one screenshot of the article...

|||

[ Edited by Velcro_SP 24.04.2011 - 10:24 ]

> Don't know if PowerVR is the multimedia chip or the gfx chip.

The latter.

> Anyone know more about LimeOS? [...] Someone must have the info on
> LimeOS

http://projects.powerdeveloper.org/project/efika/109/entry/610
http://www.symphonyos.com
http://en.wikipedia.org/wiki/Symphony_OS
http://en.wikipedia.org/wiki/Mezzo_(desktop_environment)
http://www.limefree.org/download_2.shtml

Quote:

You're welcome, but it's not mine at all, it came in the article I referred to, which in turn came from the king-of-links Andreas Wolf.

Quote:

Perhaps you mean the "PowerVR MBX" core inside the MPC5121e. According to BBRV, "no public driver is available". The display unit ("DIU"), in the other hand, should pose no challenge, as it's almost only a frame buffer.
I guess it's all a matter of good relations with your supplier (Imagination Technologies and freescale, in this case). Which, according to BBRV again, can't be good at all... Anyway, I don't think any 3D acceleration is shown in that short video. Moreso, when the bundled software with the LimeBook is of general use, nothing exciting.
Actually, the only "exciting" feature of the LimeBook is that it has a PowerPC CPU. Count the number of people in the world that can get "excited" by that...

Quote:

You hit the nail: Hardware is only as good as the software it runs.

Quote:

And I bet we'll never know. Which is quite sad, mostly when there was once a plan for it to happen. Now, it seems freescale has forgotten about consumer/mobile chips based in PowerPC, seeing how aggressive they've just turned, with their shiny new i.MX515, and ARM based beast.

Quote:

Interesting. Were MorphOS team members involved in this discussion?

Quote:

There's been quite some talk about it, mostly in PowerDeveloper. It seems using all these internal devices of the MPC5121e would generate big bottlenecks. Heck, if everyone puts graphics engines outside of the CPU, in separate buses, since the Amiga chipset, it must be because of something. But I agree it all looked very fine on paper.

Quote:

Since ages, demos on the original Amiga chipset have shown that extremely written software can do things that look like magic but, in the end, merely bringing to the front a high resolution screen over the Workbench slowed all the system down. There's no more juice in the hardware, no matter how well you write programs. So, you can take my phrase above and swap terms: Software is only as good as the hardware it runs on. Doh!

Quote:

LimePC's corporate and technical information looks like a joke, moreso if you check their "community" section in "LimeFree.org" site.
I've only learnt (and not from LimePC themselves) that LimeOS is based on SymphonyOS, which is only yet another Ubuntu distribution with a semi-propietary desktop manager, with enphasis on constrained hardware, not a completely new, carefully written marvel like MorphOS is. Symphony's entry on Wikipedia directed me to a feature in Linux.com.

[ Edited by jcmarcos on 2009/1/12 15:15 ]

> with the latest buzz about ARM from freescale, everyone (and
> themselves?) has forgotten about their PowerPC line [...]
> I guess the very own freescale has forgotten it too, as all the
> buzz that accompanied the MPC512x family launch has completely
> vanished, and seems this new ARM based family is replacing it.

True, at least regarding *consumer devices*, but not surprising.

"The future as far as Freescale is *corporately* concerned for consumer devices is with i.Mx. The focus for the 5121e will be medical, industrial, home automation, etc."
http://www.powerdeveloper.org/forums/viewtopic.php?p=11568#11568

Quote:

I guess it would perform *very* similar as on an Efika. The e300 is the same. The AXE might have speed up some things, but since it isn't cache coherent I am not sure if you had to use the same tactics like on the BlizzardPPC/CyberstormPP with PowerUP, i.e. withe every context switsch the cache had to be discarded. Okay there is 128 KB SDRAM on the 5121 exclusively for the AXE, maybe that would have done well...
Anyway, leaving out the AXE you have a e300 and the PowerVR. The latter is not bad, but no powerhorse, too. I guess an Efika with Radeon 9250 performs better. The pci bus of the Efika might be some bottleneck, but I haven't recognized any 3D slow downs on the Efika.
IMHO the real advancement of th e5121 over the 5200 would have been SATA and USB2.0. The real advancement would have been the high integration rate and the very low price. The 5121 is similar priced as the 5200, but a 5121 design doesn't need all the additional goodies that a 5200 design requires.
But performancewise both should be very en par. Thus, we know how MorphOS performs on such a board. Not bad, but also not too great. I prefer my G3/600 much, much, much over the 5200. the days of a 400MHz e300 chip as netbook/notebook/desktop engine are over 8except for thre real low price market, but this Limepc thing is overpriced).
The 8610 is IMHO the 'hottest' ppc on the market.

> I guess it would perform *very* similar as on an Efika. The e300 is
> the same. [...] performancewise both should be very en par.

The MPC5121e has twice the cache (instruction as well as data) of the MPC5200B. Should bring little speedup.

> there is 128 KB SDRAM on the 5121 exclusively for the AXE

That's SRAM (Static RAM), not (S)DRAM ((Synchrone) Dynamic RAM).

Quote:

Erm... Actually, an external bus should allow for less bottleneck than using an internal device such as the PowerVR in the MPC5121e. That story about the Amiga's chipset bottleneck was supposed to be the explanation for this. This bottleneck was gone when we started using "alien" graphics processors, first on the original Zorro bus, and next in the "alien" PCI added-on buses.

Quote:

Indeed. Now, watch everything else go ARM in freescale, instead of PowerPC. What makes me nuts is knowing that every current generation games console uses some form of PowerPC! I remember my huge surprise when BBRV predicted this, three years ago.

Quote:

Again, Andreas, very on the spot. I just love the way you are able to correct all these details that pass unnoticed to some of us.
128 MB of static RAM just for the AXE coprocessor... Our beloved operating system can work entirely in half that size! By the way, I didn't realize the MPC5121e had such an amount of memory inside.

Quote:

Talking about being on spot: 128 *K*B not MB.

> Actually, an external bus should allow for less bottleneck than
> using an internal device such as the PowerVR in the MPC5121e.

Connected to the same chip, internal (i.e. on-chip) always beats external (i.e on-board or on-card).

> That story about the Amiga's chipset bottleneck was supposed to be
> the explanation for this. This bottleneck was gone when we started
> using "alien" graphics processors

That was due to the Amiga chipset becoming obsolete (so in fact you compare the chips rather than their connection to the system) and the way the Amiga chipset was connected to the rest of the system. Furthermore, it was on-board, not on-chip.

> I remember my huge surprise when BBRV predicted this, three years ago.

Predicted? Three years ago? I think they did nothing along this.

PS3 being PowerPC based became officially known in November 2004:
http://www.sony.net/SonyInfo/News/Press_Archive/200411/04-1129E/

XBox 360 being PowerPC based became officially known in May 2005:
http://web.archive.org/web/20050515013854/http://www.xbox.com/en-US/xbox360/factsheet.htm

Wii being PowerPC (G5 [sic!]) based had been rumoured at least since May 2005:
http://www.unika.com.cn/article/article.php/2572 (Chinese, Google translation)

...and officially confirmed (G3) in March 2006:
http://uk.wii.ign.com/articles/699/699118p1.html

So what exactly was there to predict and be surprised about 3 years ago?

Quote:

After seeing your links and its dates, obviously, the surprise in me came out of my ignorance.

Quote:

I guess I should dig a very deep hole, throw myself inside, and wait to come out when PowerPC rules the world... That statement from me, confusing kilobytes for megabytes is a hell of an embarrasing affair. Targhan, could you please reset my reputation level to zero?

|||

[ Edited by Velcro_SP 24.04.2011 - 10:24 ]

> It's not silly to think that a graphics co-processor might have
> 128 megs of RAM.

As was outlined numerous times already in this thread, the AXE is *not* the GFX core, but the PowerVR is.

> Since CherryPal and the LimeBook use the 5121E there's a
> two-for-one deal to be had in porting MorphOS to the 5121E

THTF/MTC have even more in the pipeline than just the LimeBook series, for instance the LimeBox/LimePC-X1 (which is in fact the CherryPal C114), the LimePC series, the LimeNote, the LimeMobile series, the LimeView, the LimePro (see linked YouTube video)... Just take a glance at limepc.com and iseeuon.com (Chinese).

|||


[ Edited by Velcro_SP 24.04.2011 - 10:22 ]

Quote:

Right. From what I've learnt, the "auxiliary execution unit" (AXE) is a 200 Mhz RISC coprocessor with a set of instructions typical for a digital signal processor (DSP), hence very suited for audio tasks.
In the MPC5121e, code is loaded into the AXE from the e300 core, and then execution is issued. One of the first things than come to mind is writing a program for the AXE that acts as some kind of "kernel" (more of an executive, or task scheduler), and that's already done, if I remember correctly.
This setup is somewhat similar to the PowerUP kernel in PowerPC accelerators for the Amiga, where there was the need to feed the alien processor with code, from the original processor running the main operating system.
Feel free to research...

[ Edited by jcmarcos on 2009/1/13 17:21 ]

[ Edited by jcmarcos on 2009/1/13 17:26 ]

> don't think I said "GFX core."

You said "graphics co-processor", which doesn't sound so different for me as the gfx core is a (on-chip) co-processor to the e300 core.
And before in this thread you made a distinction between "multimedia chip" and "gfx chip" (which btw must rather read "core" than "chip" in both cases).

> Are you saying that the AXE is a co-processor that has nothing at
> all to do w. graphics?

I rather let Freescale speak for me:

"AXE, a 32-bit RISC audio accelerator engine"

|||

[ Edited by Velcro_SP 24.04.2011 - 10:21 ]

> "A graphics co-processor" is not the same as "the graphics core,"
> for example some computers have had multiple graphics co-processors.

Still,

"It's not silly to think that a graphics co-processor might have 128 megs of RAM. Plenty of graphics cards have that much memory."

...doesn't make sense.
For what does a "graphics co-processor" that is not the gfx core need "128 megs of RAM"? Where are the "plenty of graphics cards" which have a "graphics co-processor" with "128 megs of RAM" that is not the gfx core?

|||

[ Edited by Velcro_SP 24.04.2011 - 10:20 ]