Quote:

Very probably correct value is 1.9 MIPS/MHz/core - I have to search original source, but fact is, that G4 is faster on same frequency than G5. And G4 has 2.3 MIPS/MHz/core

Quote:

NXP QorlQ sheets for P5010, P5020, P5021 and P5040 (QP5020FS REV 6 and P50405021FS REV 4 ) says 3 DMIPS/MHz per core - i.e core e5500.
Of course, different document revisions can have different values, but a little.

e6500 core has 3.4 MIPS/MHz/core, but in single-thread mode, and 3.0 MIPS/MHz/thread ( 6.0 per core ) in dual-thread mode. This core is dual-thread capable and has SIMD (AltiVec), e5500 not. Unfortunately, we have no MorphOS computer with e6500 core yet.
For this numbers I don't know document nr, it is from NXP Community TechSupport. FactSheets for different e6500 CPUs gives slightly different values.



[ Edited by sailor 17.05.2023 - 13:20 ]

P.S.:
PowerPC architecture inside Amiga and AmigaNG
In Czech language, there still can be some erros, but majority of details is verified.

Some graphs are in the end.

[ Edited by sailor 17.05.2023 - 13:29 ]

>>> G5 ID 11,2: [...] 1.9 MIPS/MHz/core

>> This should be 2.9.

> Very probably correct value is 1.9 MIPS/MHz/core -
> I have to search original source

Please link the source if you find it. I'll spare you sources for 2.9 because the web is full of it ;-)

> fact is, that G4 is faster on same frequency than G5.

I doubt it as a general statement of fact. It is true for AltiVec/VMX code and for binaries using instructions the PPC970 lacks and has to emulate (for instance mcrxr which makes a PPC970-specific E-UAE binary necessary to make it faster on G5 than on G4), though.

https://morph.zone/modules/newbb_plus/viewtopic.php?forum=3&topic_id=8762&start=58
https://morph.zone/modules/newbb_plus/viewtopic.php?forum=3&topic_id=8762&start=61

>>> AmigaOne X5000: [...] 3.0 MIPS/MHz/core

>> Some newer Freescale/NXP docs say it's 3.3 (32-bit) and 3.4 (64-bit).

> NXP QorlQ sheets [...] says 3 DMIPS/MHz per core - i.e core e5500.

Yes, I know. That's why I was referring to newer docs (as recent as 2017, referenced in the linked comment of mine) that mention higher figures for e5500. That's not to say that one is more true than another, just that there's been inconsistent data around from Freescale/NXP.

> e6500 core has 3.4 MIPS/MHz/core, but in single-thread mode,
> and 3.0 MIPS/MHz/thread ( 6.0 per core ) in dual-thread mode.

As referenced in my linked comment, also the DMIPS figures for the e6500 have been inconsistently given by Freescale/NXP. Figures range from 5.4 to 7 dual-threaded, i.e. 2.7 to 3.5 per thread, and from 3.3 to 3.5 (non-SIMD) single-threaded.

> This core is dual-thread capable and has SIMD (AltiVec), e5500 not.

I'm well aware of this, of course.

Quote:

Great writeup with differences!

I do disagree a 2.5 GHz 11,2 Mac isn't powerful enough. For most things you're 100% right but:
1. For certain websites Wayfarer still is too slow.
2. For certain video files like 4k, even the mighty 2.7 has trouble.

The application set for MorphOS is becoming brilliant where you can use it for so much now, but when you need real cpu and bus grunt to get the job done even the 2.5 is too slow.

I would get a single processor 11,2 (2.0 or 2.3 GHz) and put a single 2.5 in it. Runs MUCH cooler and quieter with 0% performance hit.

> PowerPC architecture inside Amiga and AmigaNG
> [...] majority of details is verified. Some graphs are in the end.

I can see the 1.9 figure mentioned in your article and also the graphs, but no reference to any source.

I checked my "archives" and also not found direct original source. Respective I found notes with links to IBM pages, but links not works now and IBM is wayback-machine very unfriendly.
So I have only indirect comparisons like these:
BBN, Military+Aerospace Design
Second link has mosty Acces denied, but sometimes it can open.

Anyway it is not so much important. Important are real results.
Real speed of G5 970 CPU on the same frequency is lower than G4 and G3 and these two had 2.3 MIPS/MHz for sure.
Also all sources ( unfortunately there no original PA-Semi document ) says, PA6T-1682M has 2.2 MIPS/MHz/core, and in real life AmigaOne X1000 with 1.8 GHz is equal or faster than iMac 2.1 GHz.

So, I cannot say for sure what is exact value of G5 MIPS and I appologize if I misinformed somebody. But real-life comparisons says 1.9 is much closer to real number than 2.9.

If you wanted to take the time and found a real source from IBM, I will be be glad. And even I also like to be precise, I don't want to spend time on it.


[ Edited by sailor 18.05.2023 - 10:59 ]

Quote:

That would correspond with the fact that a 1,8 GHz Power Mac G5 running on one CPU (MorphOS) is roughly on par with PowerBook G4 @1,67 GHz.

So if I overclocked the G4 1.67 to 1.8GHz, it would be similarly efficient as the G5 2.1GHz in the iMac G5.

Quote:

Which is contradicted by the fact such a PowerBook did easily crush the X1000 in real world test back in the days.

-> Meaningless Instruction Per Second

There is also much more to performance then just the CPU running code in it's cache.

G4s can't make use of DDR(2) and even on old SDRAM they barely get to use the full performance
Pegasos G4 is even worse
X1000 doesn't do good as it should

G5 seems to be the only one with a proper RAM interface.

End result is that real performance of G5 (high clocked) G4 and X1000/5000 is so close with each having it's own minor strong point that performance just shouldn't be a deciding factor which to get.

Quote:

Yes something like that
We can do many real tests, but never will be absolutely fair.
- MorphOS is more efficient than AmigaOS 4
- AmigaOS 4 can use better graphics cards
- linux utilize all cores
- only common platform for all these computers is Debian 8 or similar distro with single-thread test. Newer distros have no support for exotics like Pegasos 2 and AmigaOne XE.
- if we use linux for testing, here is big disadvantage for Mac Mini and similar, since R100/R200 graphics drivers are poor.
- and has no sense made test with linux, as we want to use MorphOS

So best way is to use impressions and feeling with using of our computers.
And of course, time-to-time do some real benchmark, but we must have in mind, that benchmark is big simplification. Benchmark is simple answer for question which has no simple answer.


[ Edited by sailor 18.05.2023 - 13:27 ]

> Real speed of G5 970 CPU on the same frequency is lower than G4 and G3

As I wrote, for a fair comparison it's important that the binaries executed by the PPC970 do not contain instructions that the G5 lacks, else what you're actually measuring is how fast the G5 can emulate certain portions of the G3/G4. Ideally, binaries come in at least two fashions, one for pre-G5 (which could be two because of AltiVec) and one for G5. Except for E-UAE, I'm not aware of G5-specific binaries for MorphOS (and I doubt there are G5-specific code paths in the binaries). So considering that the very same MorphOS binaries do run on G2 (Efika) to G5 (as well as several Book III-E microarchitectures), what the G5 is actually running is generic PPC code not adapted to the peculiarities of the PPC970 instructions and microarchitecture, including missing instructions it has to emulate. As a result, I'm sure MorphOS programs on G5 could be even faster with binaries compiled for 32-bit PPC970.
I wrote about these things 3 years ago in another thread (comments #2 and #4).

> I cannot say for sure what is exact value of G5 MIPS

Just to make sure we're not talking past each other: What l'm referring to is not just MIPS (which is really meaningless) but Dhrystone 2.1 MIPS (DMIPS). Dhrystone 2.1 is a specific version of a specific benchmark algorithm.

> If you wanted to take the time and found a real source from IBM, I will be be glad.

At your service:

https://bunniestudios.com/blog/images/efuse3.pdf (p. 1, 6)
http://datasheet.digchip.com/205/205-00367-0-970.pdf (p. 13, 14)

> G5 seems to be the only one with a proper RAM interface.

According to RAGEMEM results, PA6T's memory bandwidth outperforms that of PPC970FX+CPC925. Would be interesting to know the RAGEMEM results of PPC970MP+CPC945. Can someone with PowerMac11,2 please do this (or link to results if already done)?

> So if I overclocked the G4 1.67 to 1.8GHz, it would be
> similarly efficient as the G5 2.1GHz in the iMac G5.

Efficiency means performance per watt, doesn't it? That opens quite another can of worms…

Quote:

In pure computing this would be true. But throughput of the g5 systems is way better. I.e. browsing and video replay on a G5 is quite better on my g5 than a G4, while things like jpg decoding is worse.

Comparing the myiMac G5 2.1 with my Powerbook (5.6) 1.67 shows that browsing on the iMac is a good tad smoother, and video replay works with higher resolutions w/o frame dropping, but decoding a jpg in showcase is faster on the Powerbook.

It's also noteworthy that (surprisingly) the SATA ssd on the G5 does not yield (significantly) higher speeds than the PATA/mSATA adadtper SSD package on the Powerbook.

In contrast to @sailor I think the iMac is a pretty nice system: It's not the fastest of all, but still conveniently fast while not bulky, relatively energy efficient and rather silent.