@billt
Quote:

billt wrote: The PA Semi in X1000 is a G5 class chip, not G4.

The G thing is an Apple thing. If one refers to that terminology the PA6T is a mobile G5. That is if the G5 key feature is 64 bit. The PA6T is 64 bit as is the 970 family by ibm, but low power end doesn't reach that computing power. The G4 (as primarily known from the 744x) is 32 bit only, but the core is rather on par with the PA6T core, the Altivec implementation exceeding teh one of teh PA6T. .
But the G terminology is leading nowhere. The 744x has its strengthes and weaknesses as has the PA6T or the 970xx.
Benchmarks are providing a better impression than any label. And they indicate that the PA6T is not bad, but honestly expectaions were much higher.

_________________
My programs: via.bckrs.de
MorphOS user since V0.4 (2001)

@Zylesea

"expectaions were much higher"

Some might have such expetations, some not.
Anyway, PA6T performs in between 970FX and 970MP, higher than G4.

But we need the SMP AOS for it.

_________________
- KimmoK
// For freedom, for honor, for AMIGA
//
// Thing that I should find more time for: CC64 - 64bit Community Computer?

@KimmoK

Quote:

But we need the SMP AOS for it.

Did Hyperion Entertainment ever promise SMP support or did they promise to use 2nd core in PA6T in the future?

_________________
Amiga Developer
Amiga 500, Efika, Mac Mini and PowerBook

@itix

The official word is that modern PPC SMP technologies (including emerging Freescale AMP generation) will be used fully, starting from AOS4.2.
(it might be from4.2.531b forward but anyway, neither initial AOS4.2SMP release year is announced and x1000 users are getting the upgrade for free)

It was confirmed by HyperionMP little while ago.
(official word from Hyperion.)
"The multicore support design for AmigaOS allows support of all of these features over time starting in AmigaOS 4.2."

Plans are plans. Let's see... ... ...

In the meanwhile I would love to see the "multicore support design document" where can I sign NDA for that???

btw. repost of interesting article about AMP series.

Last edited by KimmoK on 10-May-2012 at 07:40 AM.
Last edited by KimmoK on 10-May-2012 at 07:37 AM.
Last edited by KimmoK on 10-May-2012 at 07:34 AM.
Last edited by KimmoK on 10-May-2012 at 07:34 AM.
Last edited by KimmoK on 10-May-2012 at 07:27 AM.

_________________
- KimmoK
// For freedom, for honor, for AMIGA
//
// Thing that I should find more time for: CC64 - 64bit Community Computer?

@OldFart

Quote:

When that market has moved to full integration of CPU and GPU, there's little left for some vendors to produce high perporming graphics chips anymore. Hm, time to invent/develope our own AGA+++++ chip?

Maybe there's little left for others to produce new graphic chip. I would like to see our own AGA +++ chip. There is such possibility. I know an old ex amigan capable of creating interesting graphic chip. In fact, this dude is able to create graphic chip capable of Full HD real time ray tracing. I have some images on my desk and, believe me, i'm impressed of the quality achieved.

The problem is that for a project such this, many resource are needed (time and money). I've spread some e-mail here and there talking about this project but, as usual, i've got no-reply.

Could the Amiga be again a reference machine in the Graphic sector? Maybe, but as usual....

_________________
retired

@KimmoK

HyperionMP didnt confirm SMP.

_________________
Amiga Developer
Amiga 500, Efika, Mac Mini and PowerBook

@itix

Ok, he did not say Symmetric.
Anyway, the use of multithreading and multiprocessing (multiple CPUs, more than 2GB RAM, etc...) is being done (somehow).

From the OpenGL article:
"Finally, multi-threading and multi-CPU support is an important factor. If you are aiming at making a system an integral part of the OS, then support for multithreaded operation is as much a requirement as support of multiple CPU's."

UPDATE:
More SMP related stuff here:
http://amigaworld.net/modules/newbb/viewtopic.php?topic_id=34171&forum=33&start=40#627520
"The foundation for SMP support was put in place, a clear picture exists what needs to be done to accomplish it and how."

So, to me SMP is said/announced officially. But how exactly, does it twist the meaning of symmetric, etc, is to be seen.

Last edited by KimmoK on 11-May-2012 at 07:07 AM.
Last edited by KimmoK on 11-May-2012 at 07:05 AM.

_________________
- KimmoK
// For freedom, for honor, for AMIGA
//
// Thing that I should find more time for: CC64 - 64bit Community Computer?

@KimmoK

Amiga have always supported multithreaded operations in sense gfx operations could run on a background.

_________________
Amiga Developer
Amiga 500, Efika, Mac Mini and PowerBook

Quote:

Zylesea wrote: @olegil Quote: olegil wrote: @olegil Actually, I have a question. What sort of application is it we need more oomph for now. I think a PA6T or G4 1.67 GHz is close to be convenient. From my experience with the G4 I still desire more power for: - Full HD video replay (the PA6T may do some full hd codecs, but probably not everything an dwill be on its limit, the G4 is definietly not able to handle full hd in a convenient way) - webbrowsing with heavy computing stuff and/or embedded video in high resolutions - I recognized that much of my Hollywood stuff is rather cpu demanding. - complex pdfs are also not scrolling smoothly. - high resolution compressed pics (read: digi cams with high resolutions) take a while to display (the 8 MP jpgs from my digicam take about 0.35-0.8s to load on my 1.5 GHz G4 with Showgirls. That's of course still quick, but not instant. My older pics ( 2 MP) take about 0.06-0.12s to load, that's about the time of the double click,. That's instant, it feels real time) With CPU power it is easy: the more the better. But all in all I think reaching i7 level is not the most pressing issue, but a bit more juice than the G4 or PA6T provides would be very handy for rather basic everyday stuff. A G5 2.7 GHz would proabably be fine cpu power wise - unfortunately it's a bit energy hungry and neither OS4 nor MorphOS supports it anyway...

Very good analysis, as far as I can tell. One thing I'm wondering, though: Are your G4 tests using Altivec? I'm fairly certain it's possible to use multiprocessing to speed up all of your tasks there. pictures and video is typically rather easy, both jpeg (including j2k) and mpeg (1/2/4, including h.264) divide the picture into smaller blocks as part of the codec processes. The PDF problem could be harder, depending on what exactly is causing the problem. Using evince in Linux, I typically have to wait an excruciating time when scrolling fast in large files. This is on a 3.06GHz Core2Duo. Seems it's rendering everything from the top and I have to wait for the whole file to be rendered before I can see the bits at the end.

_________________
This weeks pet peeve:
Using "voltage" instead of "potential", which leads to inventing new words like "amperage" instead of "current" (I, measured in A) or possible "charge" (amperehours, Ah or Coulomb, C). Sometimes I don't even know what people mean.

@KimmoK

Quote:

KimmoK wrote: Some might have such expetations, some not. Anyway, PA6T performs in between 970FX and 970MP, higher than G4. .

Nope----

All benchmarks we've seen sofar suggest that the P6T preforms like a slightly lower clocked dual-G4 setup with superb RAM-throughput.....

I'm just not bored enough to install Linux on that dual 1.8GHz QuickSilver to run those tests....

_________________
- We don't need good ideas, we haven't run out on bad ones yet
- blame Canada

@Zylesea

One thing it would be very nice to know is how much of a bottle neck is the processing power of the E600 core and how much is the memory interface of the 744x. I would have really liked to try out an 8641(D), with the direct DDR2 interface. Unfortunately, the only numbers I have are for code running from the cache.

_________________
This weeks pet peeve:
Using "voltage" instead of "potential", which leads to inventing new words like "amperage" instead of "current" (I, measured in A) or possible "charge" (amperehours, Ah or Coulomb, C). Sometimes I don't even know what people mean.

@olegil

The jpg decoding times are from Showgirls that AFAIK utilizes Altivec.

And about the bottleneck issue of the G4: Somewhere on youtube there is this video from a 8610 based board (NEC redtail) where full hd replay is demoed. The 8610 clocks up to 1330 MHz and has a rather little L2 cache (256 kb), nevertheless it is able to do full hd (albeit the system is probably highly optimized and we don't know the used codec). But it's a thing a 744x based system is not able to deliver.
Because of this I still assume that the PA6T will be able to deliver full hd video replay (the mem controller is really good). Currently MPlayer using only one core and current AOS4.1 is not able to _decode_ the stream and hence is the limit, but I guess there's some room for optimization.

_________________
My programs: via.bckrs.de
MorphOS user since V0.4 (2001)

@Zylesea

Didn't know the 744x can't do full HD. Agreed this is a bandwidth issue so the PA6T should be able to do that if the 8610 can.

_________________
This weeks pet peeve:
Using "voltage" instead of "potential", which leads to inventing new words like "amperage" instead of "current" (I, measured in A) or possible "charge" (amperehours, Ah or Coulomb, C). Sometimes I don't even know what people mean.

@olegil

Board: Freescale MCP8610 dev board
CPU: 8610 @ 1066 Mhz
RAM: DDR2 533 Mhz - 512 MB
OS: Linux Debian 6 stable
Kernel: old 2.6.xx

Blender 2.49 takes 8:58.25 seconds (running the test from a shell)

Hope this helps

_________________
Massimiliano Tretene, ACube Systems, Soft3

An other dev board in Acube's hands ...

Petrol.

@m3x

Quote:

Board: Freescale MCP8610 dev board CPU: 8610 @ 1066 Mhz RAM: DDR2 533 Mhz - 512 MB OS: Linux Debian 6 stable Kernel: old 2.6.xx Blender 2.49 takes 8:58.25 seconds (running the test from a shell)

you're asking for trouble, max?

Last edited by AmigaBlitter on 10-May-2012 at 04:49 PM.

_________________
retired

@m3x

what is this all about?

@m3x

Nice

Amiga One X1000 results from this thread :

6:54 min with Wazp3D/os4.1
Linux 1 core 7 min
Linux 2 cores 3:26 min


Sam440ep-flex 800mhz = 17:33:27

http://amigaworld.net/modules/newbb/viewtopic.php?mode=viewtopic&topic_id=35671&forum=34&start=60&viewmode=flat&order=0#664189

Last edited by Spectre660 on 10-May-2012 at 08:06 PM.

_________________
Sam460ex : Radeon Rx550 Single slot Video Card : SIL3112 SATA card

@m3x


mpc8610

@amitv

They could win 700 euros if board design goes open source?

Nice to see some next SAM

_________________
Future Acube and MOS supporter, fi di good, nothing fi di unprofessionals. Learn it harder way!