@iggy

Quote:

We now rely on multiple cores and other trickery, but clock speed are only moderately faster.

Single core performance rises, but you are right it is slow since 2006/2007 (unlike late 1990s/early 2000s).

Eg.:

In 2002 my father bought Pentium 4 1.4 GHz (513 SpecInt2000 = cca 4 SpecInt2006)
In 2007 we bought Core 2 Quad Q6600 2.4 GHz (2500 SpecInt2000 = 18 SpecInt2006
In 2011 my brother bought Core i5-2500K 3.3 GHz I use right now (40 SpecInt2006)

Best single core SpecInt2006 score has today Core i7-5960X 4.5 GHz (82.5 SpecInt2006), but this certainly is not for home users. Core i5-4690K used in 1000 EUR desktop computers scores 60 SpecInt2006.

@tlosm

Quote:

tlosm wrote: @cdimauro There is a lent on the top of the image click on it and will explode in big screen .... more attention please ... look like you have an eagle eye only where you like

Your poor english doesn't help. Can you speak Italian and ask clearly what you want? Because I already took a look at the first two images of your site, and I found nothing interesting.
Quote:

... i ask you how much thread had that machine for core ?

Directly from "my" site:

Intel® Core™2 Extreme Processor X7900
Intel® Xeon® Processor 5130

So, both processors are dual core, and without hyperthreading.

Now do you want to continue to make a comparison with your quad core machine?
Quote:

dont try to play like is your usual ... for change the cards on table ... bla bla bla

ROFL. Don't worry! I don't need it, and I've already replied to your non-sense comment before.

I reported a lot of data coming from respectable and professional sites (even Mac site!), and using real-world applications.

I don't care about tests made by an unknown which spend his time at generating videos and numbers which have no practical use, except auto-glorify himself.

If you have something to say against the data that I reported, you're free to do it.


@pavlor

Quote:

pavlor wrote: @cdimauro Quote: It's very well know that Quake 3 is an SMP application: Well, yes Mac port has SMP by default. I must admit you got me... this time. However, this doesn´t changes my point about Quake 3: both CPU and GFX speed are benchmarked.

Yes, and if the graphic card is the same, then you're testing the CPU, right?

As I already reported, looking only at the G5s you can see how the dual core / SMP is working. That's because the graphic card is the same, of course.

If you would have read the article, you should have seen that this apply as well to the PC which I used for comparison. In fact:

"All machines were tested with 1GB of RAM and the ATI Radeon 9800 Pro graphics card; the Mac version of the graphics card has a maximum of 128MB of RAM, while the high end for PCs is 256MB. Most of the PCs used dual, RAID-striped hard drives; the Apple systems did not. We retested the Alienware Aurora with the 128MB Radeon 9800 Pro card and without RAID for more-direct comparison with the G5 systems."

So, the comparison is valid, and you can see that a single core x86 is able to outperform the dual core G5.
Quote:

Quote: However, what kind of optimizations are you talking about? We don´t know what options were used during compilation. Different builds could give variable results (eg. look at ffmpeg being slower and slower with newer releases).

Well, those were closed source programs released by well known professional companies. I expect that they used the best setting for compiling their product.

Since there's no way to know what you ask, we have no chance other than to take the binaries as they are, and the coming data too.


@iggy

Quote:

iggy wrote: @cdimauro Quote: Threads and cores are the same, right? Glad to see that you know how ridiculous these comparisons can get. Hard enough to compare similar architectures (like my Core2quad system vs an i5, which btw is still underwhelming) let alone across ISAs.

The only thing which makes sense is comparing the real products which people is using. If you have to make a render, for example, it matters if a certain platform can reduce the rendering time, because you get sooner the result that you need.

That's why it's completely non-sense running synthetic benchmarks.
Quote:

The thing that worries me is the a single G5 core doesn't compare that badly still.

No, and it can be used even today. It depends entirely on what you, as a user, need to do with it.
Quote:

We now rely on multiple cores and other trickery, but clock speed are only moderately faster. BTW - I still want my 10GHz P4 intel. You said you'd scale it up for us.

If you believe in god, you have to ask him, and not me: I'm not able to change the physic's laws. Only to accept them, as they are.

Intel, like all other companies, found that with the 90nm it wasn't easier anymore to scale with the clock frequency, as it was continuously made up to that time.

That's why, after that, chips stopped scaling so well in frequency, and companies had to focus on integrating multiple cores to get more performance (for algorithms that can be scaled in such way).
Quote:

The point in that little bit of flippery being that I wouldn't poke at the bear (IBM). If they hadn't relied on off the shelf parts for the PC, Intel would not be in the position they are now.

IBM sold its fabs. The only thing that it can make is to continue with the pure research in such field, and sell the results, if they are good.

It's out of question that it can compete with other companies which have their own foundries.

@cdimauro

Quote:

So, the comparison is valid, and you can see that a single core x86 is able to outperform the dual core G5.

As GFX card drivers play major role in this benchmark, you can´t directly compare per core performance of G5 and "x86".

Take Gallium3D vs proprietary Linux drivers vs proprietary Windows drivers as example - proprietary drivers may be from the very same company, still give different results on the same hardware.

@pavlor: Apple is not Linux, and graphic drivers came/come directly from the graphic vendor (Ati, in this case).

@cdimauro

Quote:

Apple is not Linux

Of course not...

Quote:

and graphic drivers came/come directly from the graphic vendor (Ati, in this case).

If you think so...

@pavlor: it's no me thinking so. Graphic drivers for Apple machines came directly from Ati. No question here.

You can ask if the o.s. influences the results, and the answer is clearly yes.

Unfortunately, Windows version for PowerPCs is not available, and you can't test the G5 using the same o.s..

Anyway, Apple users had no other chance to use their G5 machines as they were, getting those results.

The same applies to Amiga users using the same cards: they get results with their systems, as they are.

@cdimauro

Quote:

they get results with their systems, as they are.

But you directly compared CPU vs CPU speed... which - as you now admit - aren´t directly comparable with Quake 3 benchmark.

@pavlor

osx drivers are made by ati and nvdia ... i dont know for intelhd.
same is for linux on x86 ... different is on linux ppc where are used different drivers no cuda no catalysts ... full of endianess and with really poor performances.

_________________
I love Amiga and new hope by AmigaNG
A 500 + ; CDTV; CD32;
PowerMac G5 Quad 8GB,SSD,SSHD,7800gtx,Radeon R5 230 2GB;
MacBook Pro Retina I7 2.3ghz;
#nomorea-eoninmyhome

@tlosm

G5 are fast CPUs and to some point relevant even today.

They have "snappines' which so many much faster x86/64
computers miss due to Windows torturing the poor things.
I think problem for Apple was that they went RISC to gain
more speed at less $$$ but PCs got cheaper too fast so
the plan failed.PPC 604 did good , G4 too but after that Intel
dominated market (even though Athlon 1 Ghz is much faster
than any P3 including Tualatin).

@all

Could PPC emulation on X64 be used as a stepping stone to full porting to X64 architecture?

Just curious if that could be possible.

_________________
1 AmigaOne G4XE (OS4 Pre-Release Update4)
Minimig
Sam440ep + OS4.1FE
Sam460cr + OS4.1FE

@1Mouse

Not the one to answer but MacOSX PPC (say 10.4) can only
run PPC (:O) software for older Macs in fully emulated Mac OS 9
(only one supported and it already broke a lot of compatibility to
System 7 which had most titles).

It actually boots whole OS9 in a window much like UAE so I think
Apple thought so.

@1Mouse

Quote:

Could PPC emulation on X64 be used as a stepping stone to full porting to X64 architecture?

You can run eg. QEMU/MorphOSPPC on MorphOS NG/X64 and retain so compatibility. However, proper integration like Petunia/Trance is much harder task.

Rather than doing an all out port over to X64, was think emulation could be used and then over a period of time make more and more elements X64 native.

Personally I think the PPC range of processors are expensive and therefore cause the systems they are built around to be expensive.

ps. I have allegiance to any particular OS or any particular hardware.

Last edited by 1Mouse on 12-Dec-2015 at 11:27 AM.

_________________
1 AmigaOne G4XE (OS4 Pre-Release Update4)
Minimig
Sam440ep + OS4.1FE
Sam460cr + OS4.1FE

@1Mouse

Quote:

Personally I think the PPC range of processors are expensive

Eg. AVNet lists T1024 (two e5500 cores, 1.4 GHz) for 60 USD...

Manufacturing of PowerPC boards is expensive, not used SoCs.

Quote:

Rather than doing an all out port over to X64, was think emulation could be used and then over a period of time make more and more elements X64 native.

I fear this way would be too hard in comparison to simple recompile and UAE-like solution for compatibility.

@pavlor

Quote:

pavlor wrote: @cdimauro Quote: they get results with their systems, as they are. But you directly compared CPU vs CPU speed... which - as you now admit - aren´t directly comparable with Quake 3 benchmark.

If you can run Quake 3 without an o.s. you're free to show how to do it, so we can see the crude numbers coming from G5 and x86 processors.

In the meanwhile real people ran real software which used the underneath o.s., and got the results which were shown...


@1Mouse

Quote:

1Mouse wrote: Rather than doing an all out port over to X64, was think emulation could be used and then over a period of time make more and more elements X64 native.

No: PowerPC is big-endian, whereas x64 is little-endian.

You can port the kernel first, and then the rest of code, but an hybrid isn't possible (unless you can use something like VAMOS or Emumiga, which can help a lot, but they are just prototypes).
Quote:

Personally I think the PPC range of processors are expensive and therefore cause the systems they are built around to be expensive.

Expensive (as platform) and underpowered.

@cdimauro

XAMOS is just for running AMOS programs on windows.

_________________
http://lifeofliveforit.blogspot.no/
Facebook::LiveForIt Software for AmigaOS

@NutsAboutAmiga: what about copperlists? Are they supported?

@cdimauro

Quote:

If you can run Quake 3 without an o.s. you're free to show how to do it, so we can see the crude numbers coming from G5 and x86 processors.

You see Quake 3 is not well suited for benchmarking of raw CPU speed among various CPU and OS combinations.

@pavlor

Quote:

pavlor wrote: @cdimauro Quote: If you can run Quake 3 without an o.s. you're free to show how to do it, so we can see the crude numbers coming from G5 and x86 processors. You see Quake 3 is not well suited for benchmarking of raw CPU speed among various CPU and OS combinations.

People plays games too, and I think that they want to see some pragmatical result even on this field.

You also like to report Doom results to compare platforms performances, even using emulators, and... emulators inside emulators...

@cdimauro

Quote:

People plays games too, and I think that they want to see some pragmatical result even on this field.

Of course. Quake 3 is (or was in 2003) ideal for benchmarking of 3D games performance on many different computer configurations.

Quote:

You also like to report Doom results to compare platforms performances, even using emulators, and... emulators inside emulators...

And in nearly every of my Doom related posts I add caution that Doom benchmarks both GFX and CPU.