https://www.youtube.com/watch?v=6SqYMU81l8Y
PowerMac G5 - Is it Obsolete?
The 8-Bit Guy

His conclusion is sobering.

Some people like driving girly cars like the Prius, some of us like driving Toyota Tundra trucks. I'm the latter. He can keep his piddly mini while I use my G5

My only problem with the G5 is the thermal issue(s).

Quote:
Did you replace the thermal paste? Worked for me.

Quote:

I have one of the Quad G5 models. I don't notice the heat output, they do run Linux, and MorphOS support hasn't even been announced for them.
It has been announced for the X5000 which is more power efficient (and is also supported by Linux).

Not too sobering.
Plus we have future X64 support to look forward to.
Things actually look pretty good.

Is the 2006/7 dual core intel mac mini really faster than a quad core G5?

Is the 2006/7 dual core intel mac mini really faster than a quad core G5?

Quote:
The levels of peformance appear to be quite similar at least: Benchmarks

The G5 seems to have a lead in most benchmarks. Both systems should support four threads.
Fairly comparable, but a PCIe G5 would have better expandability (even if it is only PCIe 1.0).

We have the added benefit of the support of newer graphic cards.
As the drivers for those improve we should have better video performance than the Mini.
Also, if Posideon were to be upgraded, we might be able to support USB 3.0 cards with the G5.
And of course, those options would be possible for the X5000 as well.

So, again, we don't come off too bad in a comparison like this.

Maybe not, but that is 10 year old hardware. How does it fare against a modern processor (i5 or i7)?

Quote:

I don't know about the rest of you, but the i7, i5, and i3 only seem like slight improvements over the older Core2 series.
I have an i7 laptop, but frankly its core count and hyperthreading capabilities make it closer to an i3 desktop.
I have an older Core2 Quad system that performs better, and with a discrete graphics card it dusts the Intel gpu built into the laptop.

As cpus go, when it was produced the G5 was quite a competitive processor.
It hasn't aged too badly, except that its a real power hog, and it could use some improvement in its multimedia/Altivec instructions.
Its got a really nice floating point unit.

> the G5 [...] could use some improvement in its multimedia/Altivec instructions.

Anything missing in G5's AltiVec instructions compared to G4?

Quote:

No, they should be identical.
They only seem dated in comparison to continuing development in X64 instructions.
And its obviously better to have them than not (as in the e5500 core).

I actually had to struggle to come up with any real deficiencies for the G5.

In fact, when you compare it to the highest power draw AMD cpus, it still doesn't look that bad.

> They only seem dated in comparison to continuing development in X64 instructions.

Yes, Intel's AVX-512, for instance, stomps AltiVec into the ground.

Quote:

Difficult to say as there is no established test for it. It's clear a current I7 is way above a G5, but not dimensions from it. From the different sources I read I assume a current 4 core i7 should be about 4 to maybe 10 times faster than a top G5 quad. IMHO still a surprising good result for the G5. But I may be wrong with my assumption (thought it would be a bigger difference).

The problem of MorphOS is that with its single core usage most of this processing power just does not get used. And you cannot easiliy benefit from software optimizations done by the gazillions of developers for x64 (like the well known JS or JS JIT examples that impact our small comunity pretty hard). And ppc is not really progressing any more.

My bottom line: Old ppc systems are still surprising powerful and it's sad this architecture goes down the drain. But we cannot change it and should go our way toward x64.

Quote:

Here is a stupid question regarding use of other cores, which I think has been asked before, but I don't remember if it was ever answered, and if answered, what the answer was. Since MorphOS currently can only make use of one core of multi CPU and/or multi-core CPU systems, would it be technically possible to run a virtual machine client on MorphOS, which ran on one or if the quad core G5 was ever supported, ran on the other 3 cores? Is it technically possible to write a virtual machine client for running Linux PPC, for example, on the 1 or 3 cores that MorphOS is not using, and either switch to a screen showing the desktop of the VM Linux machine to run Linux software, or if easier/better, install a 2nd video card in one of the PCIe slots, to run a 2nd monitor, where the VM Linux desktop could be displayed full time, with a single mouse and keyboard and shared clipboard?

No doubt it would be more worthwhile to focus on completing MorphOS for x64, instead of spending any programming resources on this VM client for our current version of MorphOS, but I am just curious if it is even possible to do on our current PPC version of MorphOS? If it is not, should we perhaps request such a feature to be incorporated into the new x64 version of MorphOS, as this would solve the initial problem of a lack of native software to run, by allowing us to run a VM Linux distro, accessible from the x64 MorphOS desktop, so we could run almost any available x86/x64 Linux software, until native x64 MorphOS versions can be ported, or created from scratch. One of the concerns from current PPC MorphOS users, is that there will be almost zero software to run, once the x64 version of MorphOS is released.

I am not familiar with how AmigaOS4 is able to run some Linux software, but I assume it is running some kind of X-Windows compatible GUI wrapper, which I have no idea of how it would work, or if that route is a good or bad option for running more available software. Since there does not seem to be much discussion regarding AmigaOS4.x's ability to run some Linux software using that method, I must assume that it is NOT a great solution for making AmigaOS4.x compatible hardware run more software.

> you cannot easiliy benefit from software optimizations done by the gazillions
> of developers for x64 (like the well known JS or JS JIT examples that impact
> our small comunity pretty hard).

This is only true for a small part of the "software optimizations done by the gazillions of developers for x64", even if some of them have relatively big impact for us. Fortunately, the vast majority of software optimizations is done in scalar, endian-agnostic code written in high-level language.

> ppc is not really progressing any more.

It is, just not in an area that would benefit a desktop operating system.

http://morph.zone/modules/newbb_plus/viewtopic.php?forum=3&topic_id=9463&start=87

> would it be technically possible to run a virtual machine client on MorphOS, which
> ran on one or if the quad core G5 was ever supported, ran on the other 3 cores?

I don't know if this is possible in general, but if it is (in ASMP style like iFusion running on OS3/WarpOS), then such VM would be restricted to running on one core. So if you have more than two cores like in the quad-G5 or the X5000/40, you would have to run more VMs to use more cores. This is because, while the OS running inside the VM may be SMP-capable, the VM itself would still be a MorphOS program and thus bound to the known limitations.

> I am not familiar with how AmigaOS4 is able to run some Linux software,
> but I assume it is running some kind of X-Windows compatible GUI wrapper

It's X-Windows (X11), but not a GUI wrapper. The programs run with their genuine GUI.

> which I have no idea of how it would work, or if that route is a good or
> bad option for running more available software. Since there does not seem
> to be much discussion regarding AmigaOS4.x's ability to run some Linux
> software using that method, I must assume that it is NOT a great solution
> for making AmigaOS4.x compatible hardware run more software.

Your assumption regarding AmiCygnix-like solutions is apparently shared by the MorphOS team:

http://morph.zone/modules/newbb_plus/viewtopic.php?topic_id=8220&forum=3&start=9
http://morph.zone/modules/newbb_plus/viewtopic.php?topic_id=7785&forum=9&start=23
http://morph.zone/modules/newbb_plus/viewtopic.php?topic_id=7785&forum=9&start=31
http://morph.zone/modules/newbb_plus/viewtopic.php?topic_id=7785&forum=9&start=32
http://morph.zone/modules/newbb_plus/viewtopic.php?topic_id=7785&forum=9&start=48

Quote:

Sounds about right, two (or more) virtually machines, and the VM software would have to have enough intelligence to manage core assignments and other house keeping tasks in a multiple VM setup.
That last part sounds like a job for some kind of hypervisor.

> the VM software would have to have enough intelligence to manage
> core assignments and other house keeping tasks in a multiple VM setup.

In my scenario, core assignment would be managed by an ASMP-enhanced MorphOS. The two (or more) VMs each run on its own core and aren't even aware of each other.

> That last part sounds like a job for some kind of hypervisor.

Yes, implemented at OS level in my scenario.

Quote:

Considering that most supported systems don't have hardware hypervisor, that makes sense

How likely will it be, that we will be able to run other OSes in a virtual machine, while running MorphOS for x64? Edit: I didn't word that very well, so let me try again. How likely is it that either the MorphOS Dev. Team members, or outside 3rd party software developers who create or port software for/to MorphOS, will create a VM program that allows us to run Linux, Android, Windows, MacOSX, or any other OS in a virtual environment running on top of MorphOS for x64?

I know that this is an answer that only the members of the MorphOS Dev. Team can answer for sure, and they might not be able to answer it right away, unless they already know how much work it is to create a VM program for the new MorphOS, but perhaps some of the rest of you already know how much work it is to create a VM capable program, and predict if one will be possible, from the small group of developers that still create software for MorphOS.



[ Edited by amigadave 26.08.2016 - 18:43 ]

Quote:

I think it is very likely.
Am really rather cluelless about virtualization, but isn't the missing ability to access the MMU the show stopper for virtual maschines on todays MorphOS. MorphOS x64 should provide proper MMU access. I mean, the objective of MorphOS x64 with a compability breach is to provide the missing modern features of current MorphOS.