On the subject of benchmarks. Sysmon now include socres for an X5000 + R7 250X in the benchmarks section.

Scores for Ragemen.

Sysmon1.jpg


The CPU speed of the X5000 is nearly are third faster.

The X5000 has faster caches, especially L2 but gets oblitrated by the X1000 when it comes reading and writing to memory, apart frrom the "tricky write", whatever that actually means.

Reading from VRAM is over 3 times faster on the X1000 but when writing the X5000 is over 3 times faster. I understand that faster writes to VRAM are what matters and when Amigakit posted the Cow3D results the X5000 put everythiing else to shame.

Scores for SDL Bench.

Sysmon2.jpg

My X1000 has an R9 280 vs the R7 250X in the reference X5000 setup.

X5000 wins in most tests althought the X1000 streaks ahead in the hardware rectfil test at 320X240. Not sure why this is.

> Scores for Ragemen.
> Sysmon1.jpg

RAGEMEM (sans MIPS and VIDEO BUS tests) comparison between my 2.3 GHz PowerMac G5 (=1.00), the X1000 (first column) and the X5000 (second column):

L1
0.59 | 0.65 : READ32
0.84 | 0.93 : READ64
0.78 | 0.87 : WRITE32
0.78 | 0.87 : WRITE64
L2
0.30 | 0.39 : READ32
0.35 | 0.53 : READ64
0.30 | 0.60 : WRITE32
0.24 | 0.53 : WRITE64
RAM
1.04 | 0.27 : READ32
1.42 | 0.48 : READ64
1.97 | 1.12 (0.72*) : WRITE32
2.26 | 1.03 (0.66*) : WRITE64
0.91 | 6.09 : WRITE (Tricky)

*Edit: a more recent report on X5000 RAGEMEM results gives way lower score (about one third less) for RAM WRITE32 and RAM WRITE64, decreasing the relative scores from 1.12 to 0.72 and from 1.03 to 0.66.

> The CPU speed of the X5000 is nearly are third faster.

"Benchmarking MIPS is NEVER relevant. From a benchmark to another, the loop used is not the same, giving different results."
http://os4depot.net/index.php?function=showfile&file=utility/benchmark/ragemem.lha

[ Edited by Andreas_Wolf 26.12.2017 - 12:32 ]

Quote:

How do you run Ragemem on PowerMac? Through OS4emu?

> How do you run Ragemem on PowerMac? Through OS4emu?

Yes, like I did with my Mac mini 6 years ago :-)

Results for PowerMac DP 2,7 GHz:

---> L1 <---
READ32: 13843 MB/Sec
READ64: 19109 MB/Sec
WRITE32: 10274 MB/Sec
WRITE64: 20476 MB/Sec
---> L2 <---
READ32: 12895 MB/Sec
READ64: 17083 MB/Sec
WRITE32: 9872 MB/Sec
WRITE64: 19651 MB/Sec
---> RAM <---
READ32: 3030 MB/Sec
READ64: 3059 MB/Sec
WRITE32: 1468 MB/Sec
WRITE64: 1580 MB/Sec
WRITE: 432 MB/Sec (Tricky)

Quote:

Those figures would probably be better on a DDR2 equipped Quad 2.5 GHz PowerMac, but we don't support those.
And I'm not sure we should.
It would discourage people from buying X5000 systems.

> It would discourage people from buying X5000 systems.

...unless they also want to run OS4 ;-)

Quote:

If PCIe Quad G5 are faster and a better value for the money, why not I don't see what kind of hardware someone uses MorphOS on helping the morphOS Team one way or the other

Quote:

The only way buying a MorphOS license for an X5000 helps the MorphOS Dev. Team, is that it makes them feel a little better about their decision to port MorphOS to such an expensive computer. If the team decides to later support the Quad Core G5 PowerMac, it will be even more energy inefficient, using 4 cores to run a single core capable OS, on a computer that uses so much energy to run those 4 G5 CPU's. At least the X5000 is more energy efficient. That and the ability to use newer PCIe video cards are the strongest selling points for the X5000.

I'm happy to see that my X1000 is not drastically slower than the X5000, but the appeal of using one computer to run both AmigaOS4.x and MorphOS3.10 does make me consider selling my X1000 so I can buy an X5000.

I think I will still wait until Hyperion releases a version of AmigaOS4.x that implements some kind of multiple CPU/Core support, and also wait for the quad core X5000 to be released for sale, so there is no hurry to sell my X1000, as it may take another 10+ years before Hyperion finally releases AmigaOS4.2 with multi-core support of some kind. Actually, I have my doubts that Hyperion will ever release AmigaOS4.2 that actually has all the features promised for that version. Even with all the assistance that A-Eon has provided, trying to inject some new life into the AmigaOS4.x community, Hyperion continues to provide zero evidence of improving their ability to produce updates at a reasonable pace.

@amigadave

A G5 "Quad" only has 2 2core CPUs

Supporting the Quad would also mean supporting Single-CPU-DualCore G5s which are not liquid cooled and might be a bit less ineffective than the current 2-CPU-SingleCore G5s.

Supporting any given PCIe in a G5 is just a matter of providing the init code (won't be able to show OF or use OSX)


Your X1000 allready has 2cores and might in reallife allready be faster then an X5000 (regardless wether the OS uses 1 or 2 cores)

> Your X1000 [...] might in reallife allready be faster then an X5000

Are you referring to AltiVec here? Or are there any non-SIMD reallife benchmarks showing the X1000 being faster than the X5000 per core?

Quote:

I think there are very few benchmarks that will show the X1000 to be faster than the X5000, but I do like the fact that the PA6T has Altivec support, and I will probably stick with my X1000 after I have more time to consider the pros and cons of each computer. The only reason I am even considering a X5000, is because of the future MorphOS support. I know that it makes no sense for the MorphOS Dev. Team to support the X1000, and I would not ask them to waste time that could be better used to complete the new x64 version of MorphOS, but that does not stop me from wishing that MorphOS could run on my X1000.

Who knows, maybe with enough influence from A-Eon, AmigaOS4.x will eventually become more bearable to use (but I have no faith in Hyperion completing AmigaOS4.2 as advertised).

As I have previously demonstrated that I have no previous experience with, and know very little about virtual machines, and exactly how SMP works, I will make a statement, or question about the X5000 that may be stupid, but here goes.

Since AmigaOS4.x and MorphOS currently only run on one CPU core, is it feasible at all to think that it could be possible for someone, or some group of programmers (surely outside of Hyperion Entertainment and the MorphOS Dev. Team members) could devise a way to run MorphOS on one of the X5000's CPU cores, while simultaneously running AmigaOS4.1FE in a virtual machine, on the other X5000 CPU core, possibly using separate video cards and monitors, and able to share a portion of memory, for clipboard sharing, but set aside separate memory for the rest of each system, to avoid problems of one OS over writing memory being used by the other OS.

I'm thinking that using a virtual machine this way would be like having a bridgeboard to run the 2nd OS, but instead of a physical bridgeboard, the 2nd currently unused CPU core, would be tasked with running the 2nd OS, on the virtual machine. It would be best, if just one keyboard and mouse were needed, and could be switched back and forth with a special key press combination, or the switching was automatic and the mouse could be dragged from one monitor to the other, which would trigger the switch for both the keyboard and the mouse.

That would be so much better than dual booting, and if successful, perhaps it could also be extended to allow running Linux, or MacOSX in virtual machines on the 2nd core, instead of AmigaOS4.1FE.

I suppose that running AmigaOS4.1FE for Classic Amigas through EUAE on MorphOS, would give similar results, but the performance would be much less, and the limitations of OS4.1 through EUAE make it less of a good option. I'm guessing that AmigaOS4.1FE would be easier to get running on a virtual machine (if what I am describing even qualifies as a virtual machine?), instead of the other way around, starting the X5000 with AmigaOS4.1FE and then running a virtual machine to startup MorphOS3.10

Possible yes, feasable no.

Writing such a VM would be a gigantic task even if you had the support of the guest OS provider to write drivers to it.

Without it you would need to "emulate" an X5000 inside it so you could use the X5000 version of the guest OS.

Another way might be a port of UAE with "passthrough" PPC-EMU running the Amiga version of the guest OS.

Still wouldn't touch the lack of accces to the 2nd core on the host OS.

While I doubt feasability as well, the Hypervisor technology integrated into the qoriq processors offer a Base for a quite elegant solution. A lot of things for virtualization is done in Silicon here, could ease things up quite significantly.

[ Editiert durch Zylesea 06.01.2017 - 10:07 ]

> run MorphOS on one of the X5000's CPU cores, while simultaneously running
> AmigaOS4.1FE in a virtual machine, on the other X5000 CPU core, possibly
> using separate video cards and monitors, and able to share a portion of memory,
> for clipboard sharing, but set aside separate memory for the rest of each system,
> to avoid problems of one OS over writing memory being used by the other OS.

Why devise a VM when the e5500 core has this capability built in? See:

http://www.nxp.com/files/32bit/doc/white_paper/EMBEDDED_HYPERVISOR.pdf (refers to the e500mc core but is valid for e5500 and e6500 as well)
http://www.linux-kvm.org/page/E500_virtual_CPU_specification (also for e5500 and e6500)
http://morph.zone/modules/newbb_plus/viewtopic.php?forum=16&topic_id=10057&start=31
http://www.google.com/search?q=site:nxp.com+%22embedded+hypervisor%22

> perhaps it could also be extended to allow running Linux, or MacOSX in virtual machines
> on the 2nd core, instead of AmigaOS4.1FE.

Using the e5500's embedded hypervisor (Topaz), running Linux in parallel would be no problem. MacOSX wouldn't be possible, though, as it's not compatible with the e5500 core on supervisor level (and lacks drivers for the QorIQ P5's on-chip controllers).

> I'm guessing that AmigaOS4.1FE would be easier to get running on a virtual machine
> [...], instead of the other way around, starting the X5000 with AmigaOS4.1FE and then
> running a virtual machine to startup MorphOS3.10

Why do you think this would be easier than the other way round?

Quote:

Because, as I stated in the original post with this question, it would most likely be done by programmers who are not part of the MorphOS Dev. Team, and Hyperion Entertainment, and MorphOS is currently tied to hardware, while AmigaOS4.1FE is not tied to a specific machine, though it is tied to a specific model of machines. I guess I should have clarified that statement by saying the full versions of both, and not the time limited version of MorphOS.

[ Edited by amigadave 05.01.2017 - 16:08 ]

The only thing that is more tied to the HW in that case would be the code reading the MAC address from the onboard NIC.

It's like saying being 1st in a marathon is harder cos you have to run through that ribbon at the finish.

>>> I'm guessing that AmigaOS4.1FE would be easier to get running on a virtual machine
>>> [...], instead of the other way around, starting the X5000 with AmigaOS4.1FE and then
>>> running a virtual machine to startup MorphOS3.10

>> Why do you think this would be easier than the other way round?

> Because [...] MorphOS is currently tied to hardware

In a VM, where also the NIC is virtualized, a MAC address has to be conceived anyway. Why not simply use the MAC address the particular MorphOS keyfile is tied to?

Quote:
Moreover the PCIe-G5s are a much less pain in the ass Linux-wise - no AGP-Card problems any longer... Also max. 16GiB DDR2 instead of 8GiB DDR. And the thing I liked best about my G5 11,2 was that to my surprise it was much more quiet than my G5 7,3 even without changing the thermal paste!

Quote:

I did not know that in a virtual machine that the NIC MAC address was also "conceived anyway". In that case, I suppose that both systems are essentially equal in difficulty.

Quote:

Quote:

Everyone (including AeonKit) knew from the beginning about the PA6T's situation, but bought into it nevertheless, with open eyes.

Quote:

Technically, you really only need one platform in order to write/test Radeon HD drivers.

Quote:

What's a CFE bug?

Quote:

Since the similarities between the Nemo and Cyrus are (AFAIK) virtually non-existent, it could very well be that the X5000 is a new experiment all over again...

> Freescale has [...] 20% higher clock

It's 2.0/1.8-1 = 11%.

> 30% cheaper

It's about 15%.

> Nemo is more expandable board

Yes, more RAM slots, SATA/PATA ports and USB ports.

> What's a CFE bug?

http://morph.zone/modules/newbb_plus/viewtopic.php?forum=3&topic_id=11372&start=211