> A very interesting document.

Indeed. Thanks again for pointing me to it.

> Why do you think page 14 has these details?

Let me clarify first how I read that roadmap chart: the abscissa has three columns, with the left one representing the past (PowerQUICC and e600), the middle one the present (QorIQ P) and the right one the future (QorIQ T). The ordinate has the performance level as well as wattage going from low to high with products at corresponding performance/wattage level put at the same height.

> The QorIQ – P5 P5020 and P5010 listed as
> "Increase FP Perf
> Next Gen process
> Security plus
> AltiVec"
>
> And the QorIQ – P4 including the P4080 and P4040 listed as
> "Next Gen Core
> Increase FP Perf
> Security plus
> AltiVec"

As I read it, those details refer to QorIQ T5 and T4, not to P5 and P4. As I said before, T series is supposed to switch to a 28 nm process node (from the P series's 45 nm) and to have AltiVec. And apparently, Freescale also intends to increase the floating point performance in T series compared to P series.

> plus
> "e600 +Soc"

That's in the "past" column according to my reading, put slightly below QorIQ P4/T4 in performance and wattage.

> The P5 were in initially supposed to not implement AltiVec.

Yes, and the roadmap chart only says that QorIQ T2 to T5 (i.e. 28 nm e5500 and e500mc cores) will get AltiVec.

> The P4s do not currently feature AltiVec and are e500mc based.

Correct.

> It looks like Freescale intends to move the P4 to an e600 core

That's absolutely not what I read from that chart. It's rather that Freescale is painting the QorIQ P4 (and P5) chips as replacements for the MPC86xx ("e600 + Soc") chips. I had addressed this observation in the past already:

https://morph.zone/modules/newbb_plus/viewtopic.php?topic_id=7001&forum=3&post_id=76991#76991

> its hard to tell from this document which processors will feature AltiVec.

Yes, to my mind it's not clear yet whether future QorIQ P2 to P5 (i.e. 45 nm e5500 and e500mc cores) chips, to be released before first QorIQ T chips, are supposed to get AltiVec or not.

Thanks, whoever created that page made it fairly hard to decipher. Your explanation seems spot on.
I wonder if we're going to see T5010 and T5020 processors along with the T5040?

[ Edited by Jim on 2011/1/29 21:48 ]

> I wonder if we're going to see T5010 and T5020 processors along with the T5040?

T5020 is actually there, that's why I listed it. See HTML version.

This is going to sound foolish. Which page is the t5020 mentioned on. I can only find the T5040.

And now back to the thread topic......... for about $1,100 you can finally pre-order a new SAM460ex motherboard. Not a complete system, just the motherboard without any RAM or graphics card (onboard graphics capable with up to 64mb shared RAM)

I can't wait to see some detailed performance comparisons between the SAM460ex @1.15GHz and a G4 PowerMac @1.42GHz, or G4 MacMini @1.5GHz (currently fastest stock PPC Mac hardware that is fully supported by MorphOS)

You forgot to mention the PCI-e slot with only 4 lanes and a processor without AltiVec instructions. Oh and the memory benchmarks I've seen are no faster than my PC133 equipped Quicksilver.
Overall, there's not much difference between this board and the P2020 Freescale board Andreas and I discussed (except the Freescale board is only $595 with a hard drive pre-loaded with Linux). And we dismissed the Freescale offering because it had many of the faults present in the SAM460EX.

I'd love to see some new MorphOS hardware, but Andreas' posts have convinced me that we need to consider the P5010, P5020, and T5020.
If we could base something on one of those and charge SAM level prices for a board competitive with the X1000 I think we'd have a winner.

> Which page is the t5020 mentioned on. I can only find the T5040.

The T5020 is mentioned on the same page and line the T5040 is. As I said, you can see it in the HTML version I linked to. In the original PDF file the T5020 is simply cut off due to being awkwardly placed.

> just the motherboard without any RAM or graphics card

...but with a boxed version of OS4 ;-)

> onboard graphics capable with up to 64mb shared RAM

Do you know for sure the SM502's VRAM is being shared with main RAM on the Sam460ex? According to the SM502 databook either configuration is possible (with dedicated VRAM being the faster one, of course). But I'm not sure which one ACube went for on the Sam460ex.

ftp://ftp.siliconmotion.com.tw/databooks/SM502_MMCC_Databook_V1.00.pdf (page 17)

> the memory benchmarks I've seen are no faster than my
> PC133 equipped Quicksilver.

I take it you refer to the RAGEMEM benchmark results ("RAM" section). Could you please provide your Quicksilver results here?

> Overall, there's not much difference between this board and the
> P2020 Freescale board Andreas and I discussed (except the
> Freescale board is only $595 with a hard drive pre-loaded with Linux).

Now you're being unfair towards the Sam460ex. In my opinion it has some features which surely make for an advantage over Freescale's P2020RDB from an end-user's point of view:

* one PCIe x4 with x16 connector + one PCIe x1 + one PCI (vs. two PCIe x1)
* six USB2 + one USB1.1 (vs. one USB2)
* SATA (vs. none)
* on-board audio (vs. none)

Besides, it seems you're once more confusing the $595 P2020RDB and the $3,500 P2020DS, the latter being the one that comes with an integrated HDD.

> we dismissed the Freescale offering because it had many
> of the faults present in the SAM460EX.

The $595 P2020RDB is to be dismissed because it's in no way suited for the desktop market (see comparison to the Sam460ex above), and as I told you already in this very thread, the $3,500 P2020DS is to be dismissed because it's just way too expensive for our purposes. Moreover, anything lacking a proper FPU (like the e500v2 core in the P2020) should be dismissed for desktop computing purposes.
None of these points apply to the Sam460ex (the price argument at least not to the same extent), and I never said or implied I'd dismiss "the Freescale offering because it had many of the faults present in the SAM460EX".

New hardware to run MorphOS on is not that important, unless it is faster, and has better features than the currently supported G4 machines, and it can be bought for about half the cost of the SAM460ex.

I don't believe that ANY hardware to run MorphOS on is worth a purchase price beyond $500 to $600, until MorphOS has matured a bit further and there is 3 to 4 times more good, productive software that runs natively on MorphOS. But that is just my opinion (which I think is shared by some users and maybe some of the MorphOS Dev. Team members). I just can't see that too many people would be willing to spend more than $500 to $600 for any computer system if it can only be used to run MorphOS. Maybe it would make sense for MorphOS users that like Linux PPC as their main system and they would be willing to spend more than $500 to $600 for a combined MorphOS/Linux system.

Optimizing MorphOS to run better and faster on the currently supported hardware, adding more features to MorphOS and completing the work already started to port MorphOS to some G4 PowerBooks, G5 PowerMacs and G5 iMacs, until there is a cost effective, better alternative, are great decisions that the MorphOS Development Team have made, and several of the best reasons to choose MorphOS as the best Amiga compatible OS at this time and for the foreseeable near term future. What we need more than any newer, or faster hardware right now is more and better software (hence the thread about donating hardware to new MorphOS developers).

Sorry to be repeating myself, as I am sure I have written these opinions before.

Edit: I am not sure about the shared VRAM vs dedicated VRAM being used on the SAM460ex that can be pre-ordered from AmigaKit, or Vesalia.

[ Edited by amigadave on 2011/1/30 10:32 ]

Cool! I see that now. Freescale needs some better graphic designers.

[ Edited by Jim on 2011/1/30 18:33 ]

Quote:

Yes, I've repeatedly confused those two boards even though the prices for each are listed in the original document you quoted.

http://www.freescale.com/files/32bit/doc/brochure/PWRARCHQIQSG.pdf

But the SAM460EX still only offers a PCIe X16 slot with four lane which would slow down many video cards.

And the processor is fairly slow, even if overclocked by .15Ghz..

As I listed in the CPU/memtest thread, my current 7447A processor produces these numbers:

Write 568 MB/sec. Verify 387 MB/sec

This is with PC133 Memory.

These figures were quoted by Piru on Amiga.org.
SAM460 AMCC460 1.167GHz
Code:
---> RAM <---
READ32: 311 MB/Sec
READ64: 310 MB/Sec
WRITE32: 521 MB/Sec
WRITE64: 521 MB/Sec
WRITE: 1251 MB/Sec (Tricky)

That would seem to indicate that the SAM460EX is slower than some PC133 equipped Powermacs.

And of course there is the lack of AltiVec(although we face that with the first e5500 cored products).

Finally, Piru's comparative benchmarks of distributed.net tasks .

OGR
Sam 460 1.0GHz: 10,124,948 nodes/sec
Pegasos 2 G4 1.0GHz: 20,844,783 nodes/sec
Mac Mini G4 1.5GHz: 31,267,175 nodes/sec

RC5-72
Sam 460 1.0GHz: 3,286,052 keys/sec
Pegasos2 G4 1.0GHz: 10,678,428 keys/sec
Mac Mini G4 1.5GHz: 15,701,333 keys/sec

Obviously these are greatly skewed by the presence or lack of AltiVec, but they're still pretty dramatic.

Dave has a point, it offers us no clear advantage over what we already have,. In fact its probably less powerful.


So if a second hand Mac can wipe the floor with this unit, why consider it?

> the SAM460EX still only offers a PCIe X16 slot with four lane

An x4 slot is still better than P2020RDB's x1 slot and even P2020DS's x2 slot.

> the processor is fairly slow

It has a proper FPU, at least ;-)

> As I listed in the CPU/memtest thread, my current 7447A processor
> produces these numbers: Write 568 MB/sec. Verify 387 MB/sec

Thanks. I just compared MorphOS's memtest and RAGEMEM on my machine and have found that RAGEMEM reports an up to 8% better performance for RAM read (write is the same).

> These figures were quoted by Piru on Amiga.org.

...from a November's posting of mine from this very thread, btw :-)

> That would seem to indicate that the SAM460EX is slower than some
> PC133 equipped Powermacs.

Yes, indeed. But we shouldn't forget that such memory benchmark apparently is still CPU bound in a way, as your memtest result for read on your 133 MHz FSB PowerMac is better than mine on my 167 MHz FSB Mac mini.

> of course there is the lack of AltiVec

Same on P2020RDB and P2020DS.

> Piru's comparative benchmarks of distributed.net tasks .

Btw, those Mac mini results are quoted from a December's posting of mine from this very thread, in which I also quoted those Sam460ex results and which is much more extensive by listing all the different dnetc cores' results for both the Sam460ex and my Mac mini.

> So if a second hand Mac can wipe the floor with this unit, why consider it?

Don't ask me. I don't say I'd consider it or the MorphOS Team should consider it. On the other hand I know that there're people who for whatever reason object against purchasing second hand hardware and thus prefer purchasing slower new hardware over faster old hardware.

@amigadave

I think your post nails it pretty much. Nevertheless I'd see a chance for a low performane ultra low cost board. Think about a US$ 50 board with some RAM, inbuild gfx, usb and ethernet connectors and an sd card slot - all based on a 5125. Something to toy around. Pretty much like the very cheap ARM boards. If seomething like that would exist, a specially prized MorphOS version for these ultra low end boards would come in handy (say 50 EUR).
But beside from such a ultra low cost/low end thing new hardware would need to offer a significant improvement over existig kit and your price range of about 500-600 (US$-EUR) is a realistic one.

I could see an ARM board for about $150.
Andreas, myself, and many others have been discussing ARM for a while.

Bur I keep coming back to the PPCs that Andreas has mentioned because they look like good candidates for a less complicated upgrade (if they can be obtained at even remotely affordable prices)

I've been thinking about ARM as a way to set up a low cost low power draw Ubuntu system. But I don't want to lose the ability to run PPC software.

> I could see an ARM board for about $150.

I can see one for $125:

http://search.digikey.com/scripts/DkSearch/dksus.dll?keywords=296-23428-ND

;-)

Quote:

I wasn't speaking of ARM boards (I know about them and discussed ARM boards and MorphOS since years here already ...). I am talking about a potential ultra cheap ppc board that might get some tinkering momentum just *like* the popular cheap ARM boards. But weaker and cheaper.
50 US$ for 400MHz, inbuild gfx, 256 (even better: 512) MB RAM, usb, eth, sd slot - all on a few square cm. With the 5125 this is pretty possible.
Cheap stuff for your hobby, that's the way forward. Not some exclusive überexpensive elitist stuff that again isn't up to date but eventually stuck in some midclass level.

Anyway, with the upcoming QorIQs (I advocate them here since a quite long time) PPC may have a very last chance, but this is quite another story and I don't see any producer picking them up for a GP board. *If* there would be a significant tinker scene already, demand for a GP QorIQ Txxxx board could be imaginable though. Hence my thoughts about an ultra cheap ppc board which I wrote about here and elsewhere several times already...

[ Editiert durch Zylesea an 2011/1/31 1:52 ]

Update:

Quote:

ACube now report 7% to 13% better results:

http://www.amiga.org/forums/showpost.php?p=610580

Correction:

>> what do you think about the PowerEN?

> In the context of mobile (read: netbooks/notebooks) or desktop
> computing I don't find it that much interesting because it seems
> to lack in on-chip peripheral controllers: there's apparently just
> (as much as four) 10GbE and only two PCIe (Gen2) lanes, nothing more.

Better information on PowerEN's PCIe:

"PowerEN PCIe support includes three Gen2 ports, which can be used as root complex (8x and 1x) or end-point (16x) in the PCIe tree."
http://www.power.org/resources/downloads/PowerEN_Dev_Cloud.pdf (page 3)

Quote:

Wow! Almost up to my PC133 figures. Actually, considering the speed of the processor these figures aren't that bad and they're likely to improve.

>> ACube now report 7% to 13% better results:
>> http://www.amiga.org/forums/showpost.php?p=610580

> Wow!

I'm surprised to see you enthused now and here. You already commented on these updated figures half an hour after they were published ;-)

http://www.amiga.org/forums/showpost.php?p=610592

> these figures aren't that bad

...and "surprisingly slow" at the same time according to you ;-)

Its all relative. Just shows what you get when you team fairly good memory with a relatively lame processor (especially when you push it just to get another 15%).

Once I realized they were over clocking the processor it occurred to me that they were probably not using an ideal ratio to derive their memory clock (that they are actually running memory slower than it could be).

From the new figures you linked, I'd say they've figured out a way to up the memory clock (or make some other change - like lowering latency).

There may even be some additional tweaks they could make.

Although, now that I think about it, I wonder how reliable this system will be if they're resorting to these tricks?

Enthusiastic or skeptical? I don't know Andreas, I keep coming back to those processor benchmarks. The 460 is still a pretty bad product.

> Once I realized they were over clocking the processor it occurred to me that
> they were probably not using an ideal ratio to derive their memory clock (that
> they are actually running memory slower than it could be). From the new
> figures you linked, I'd say they've figured out a way to up the memory clock

"On the Sam460ex the DDR is overclocked as the CPU."

That means both the CPU and the memory bus run overclocked by 15%. If I'm not mistaken this would make for a memory clock of 230 MHz. Whether that's really the reason for the reported improvement I can't say. The old figures had been reported at a CPU clock of 1.167 GHz (i.e. overclocked by 16.7%), which presumably translates to a memory clock of either 233 MHz (5:1 ratio) or 212 MHz (5.5:1 ratio, which would fit your theory of increased memory clock).

> or make some other change - like lowering latency

This would be the case where they kept the 5:1 ratio, i.e. actually lowered the memory clock from 233 to 230 MHz.

> There may even be some additional tweaks they could make.

Yes, that's what is said:

"Next U-Boot versions for our board will include modifications on the DDR config code to achive higher speed transfers."

> I wonder how reliable this system will be if they're resorting to these tricks?

"On the Sam440ep models, the DDR runs [...] overclocked on the 733 Mhz model (147 Mhz)"

Apparently, the overclocking of the Sam460ex is similar to the 733 MHz Sam440ep-flex where both the CPU and the memory bus run overclocked by 10%. So far, I didn't hear about reliability problems on this machine.

Thanks Andreas,
That clarifies some things that occure to me when I read your posts.

It was obvious that they were most likely just using a frequency boost to obtain the over clock. You don't get an odd gain like 16.7% from a multiplier change. And since the memory clock would be affected by this it seemed likely that initially they changed the ratio that memory was based on to prevent memory instabilities.

Your ability to ferret out all these details continues make make this topic interesting.

One tends to think of Acube boards as rather low tech systems based on relatively slow processors, but the company apparently pushes these parts harder than you would expect for a company building boards intended for both the hobbyist and industrial markets.

[ Edited by Jim on 2011/2/2 17:15 ]

> You don't get an odd gain like 16.7% from a multiplier change.

Generally speaking, you get a 16.7% frequency gain by changing the multiplier from 6 to 7. But the multiplier of the PPC460EX is set to 5 by default, so yes, this couldn't be the case here.

> since the memory clock would be affected by this it seemed likely that initially they
> changed the ratio that memory was based on to prevent memory instabilities.

Let me summarize how I understand you and me perceive how ACube may have proceeded with this:

1. CPU: 1000 MHz, bus: 200 MHz, multiplier: 5 (stock default)
2. CPU: 1167 MHz, bus: 233 MHz, multiplier: 5 (created memory instabilities)
3. CPU: 1167 MHz, bus: 212 MHz, multiplier: 5.5 (non-ideal ratio)
4. CPU: 1150 MHz, bus: 230 MHz, multiplier: 5 (production model)

Any objections? :-)

> the company apparently pushes these parts harder than you would expect for a
> company building boards intended for both the hobbyist and industrial markets.

I think ACube does the pushing solely for the Amiga market mainly for the psychological effect of the numbers. Pegasos II G4 has so far been the fastest hardware for OS4 and shines with a CPU clock of 1.0 GHz. To create the impression of providing a hardware with a faster CPU they simply had to go above 1.0 GHz for marketing reasons alone, knowing that there're still people who directly translate clock speed to performance even with diverse CPU types. Now with the overclocking, the Sam460ex delivers as much DMIPS as the Pegasos II G4 ;-)