In search for info on PPC cpus

[AmigaMac]

Post the said benchmark IF it proves your case.

I don't have to prove any case.  You're the one whining about software app availability/access for these mags, not me!

[Hammer]

AmigaMac wrote:

Secondly, it doesn’t cover dual Opteron 246 @2.0 on VIA K8T800..

We live not in a perfect world.  How often do they update the G4/G5 benches... see my point?

Opteron is targeted mostly at the server market anyways.

Note that, NVIDIA's nForce3 Pro**(1-way) and nForce3 250*(2-way, 4+4) targets workstations/high performance gaming machines. Also holds true for VIA's Socket 940 motherboards for  2-way 4+1 and 1-way  motherboards.

*soon to be released.

Athlon FX 51 is identical to Opteron 148 in very respects, except in name. Such regime is similar to Athlon XP and Athlon MP renaming schemes.

One should take note of Opteron’s suffix, its companion Socket 940 motherboard and targeted price in relation to market targeting.  

[Hammer]

AmigaMac wrote:

Post the said benchmark IF it proves your case.

I don't have to prove any case.  You're the one whining about software app availability/access for these mags, not me!

That’s what they said in their limitations page.
The availability and “sameness” of the applications for both platforms takes precedence.

[Hammer]

AmigaMac wrote:
At least give the PS benches:

http://www.geocities.com/sw_perf/PSBench.html

I would like to add that the Barefeet.com's link has PS7 benchmarks (IF one follows the embedded links).

[AmigaMac]

That’s what they said in their limitations page.

How could there be limitations running benchmarks between a Mac running Final Cut Pro versus a PC running Premiere?  The only limitation here is poor execution from journalism itself by either magazine.

I would like to add that the Barefeet.com's link has PS7 benchmarks (IF one follows the embedded links).

I followed the links as well as that I keep up with Barefeats other benchmarks as well.

[Hammer]

I followed the links as well as that I keep up with Barefeats other benchmarks as well.

Why did you state  "At least give the PS benches:"?

How could there be limitations running benchmarks between a Mac running Final Cut Pro versus a PC running Premiere?

IF one should apply the same logic i.e. why not run UT2003 on Mac and QuakeIII on the PC since they are both FPS?

The issue is about running the same/similar code base in both of the platforms. Both are running similar code on legacy mode.

 The limitation here is poor execution from journalism itself by either magazine.

Should one bring in Newtek’s Video Toaster for Windows?

What about the following video editing suites;
1. Matrox Rt2500 Video Editing Suite.
2. Sonic Foundry Vega 4.0.
3. Pinnacle Studio (what ever version or edition)
4. Ulead's Media Studio 7 and Video Studio 7
5. Cyberlink PowerDirector
6.  'etc'.

Why not have an all-out expensive Video editing shootout?  

IF Apple ever ports their MacOS X iTune software to Win32 (i.e. not using Music Match) then the issue is completely different.    

[AmigaMac]

Why did you state "At least give the PS benches:"?

Because CineBench was your focus.

IF one should apply the same logic i.e. why not run UT2003 on Mac and QuakeIII on the PC since they are both FPS?

The issue is about running the same/similar code base in both of the platforms. Both are running similar code on legacy mode.

Not always... I personally like seeing competing applications benchmark as well (games excluded).  Like fxPaint versus Photoshop versus Gimp, etc...

Should one bring in Newtek’s Video Toaster for Windows?

What about the following video editing suites;
1. Matrox Rt2500 Video Editing Suite.
2. Sonic Foundry Vega 4.0.
3. Pinnacle Studio (what ever version or edition)
4. Ulead's Media Studio 7 and Video Studio 7
5. Cyberlink PowerDirector
6. 'etc'.

Why not have an all-out expensive Video editing shootout?

Now that's what I'm talkin' about  :-P

IF Apple ever ports their MacOS X iTune software to Win32 (i.e. not using Music Match) then the issue is completely different.

Where have you been, Apple already ported iTunes to Windows 2k/XP...

http://www.apple.com/itunes/

[Rodney]

AmigaMac wrote:

Careful with theoretical performance claims when practical performance indicates otherwise.

It's a matter of opinion... practical and/or theoretical.

Are you saying performance is a matter of opinion? In that case, everything is. Its more of a matter of what applications you use. The PPC is different to x86 and is faster or seems to do things faster in some cases but in others the x86 blows it away. Such as, compiling was said to work faster on pentiums than PPCs in a test once. I think PPCs may be better at floating point? or is the other way around, not sure... anyway the point is, their both good and bad at different things.

[Hammer]

Because CineBench was your focus.

The rest of the information is included in the link.

"http://www.geocities.com/sw_perf/PSBench.html" benchmarks can not be use to compare with other sites i.e. they have their own point scoring regime (reference to an 1000 MHz TBird Athlon running Photoshop 6.01 and PS5Bench) i.e. Cinebench 2003’s CPU Rendering values are usually compatible with other site’s Cinebench 2003’s CPU Rendering values.

Secondly, there's minor issue with the non-discloser of the type of chipset/motherboards in K7 Athlon and Pentium 4s based solution.

For encoding benchmarks, one could go for open source encoding tools.

Not always... I personally like seeing competing applications benchmark as well (games excluded). Like fxPaint versus Photoshop versus Gimp, etc...

That would be very hard thing to do since it involves some subjective comparisons (e.g. How hard to do certain things) and the resulting image quality issues.

Such reviews are more about the applications than the processors. Care to start up a DV magazine?

Where have you been, Apple already ported iTunes to Windows 2k/XP...

Not at the time of MacWorld’s review…

[Waccoon]

I find it hard to believe anyone would want an ad campaign entitled, "World's Fastest Personal Computer".  Nobody can hold that title for more than a few minutes, so it's an awful lot of trouble to try it.  :-)

It bugs me that people are squabbling over a few minor marks.  Who cares if the performance is off by a few percent?  People were slamming AMD a few years ago because the P4 was beating the Athlons.  "Ooh, in this benchmark, the P4 is a whopping 10% faster!  What an embarrasment!  AMD has completely lost it!"  Yeah, never mind the price spread.  I didn't want to spend twice as much money for a sliver of performance, which is why I got a 2.4Ghz P4 instead of a 3Ghz.  (If I had known my stability issues were because of bad memory, I would never have replaced my 2600+, even though it was very noisy).

Just because something is faster, doesn't mean it's better.  Besides, how expandable is the G5 tower?  To me, a tower is NOT a tower if it only offers one drive bay and three PCI-X slots.  This isn't so much an x86 vs PPC issue as it is a PC vs Mac issue.

As for why I slam the AmigaOne if performance isn't important, well, the price/performance margin of the AmigaOne is just *TOO* different from a typical PC.  There are limits.

[iamaboringperson]

I have two books about programming for the PowerPC CPU's.

I found both of them new for $2, and there was a huge pile of the books for that price!

I bought 2, but I should have bought more to sell on ebay

[CodeSmith]

@Hammer

IBM
G5 @1.6 GHz delivers 222 points (1Ghz FSB).
G5 @1.8 GHz delivers 251 points (1Ghz FSB).

AMD
K7-XP @1.6 Ghz delivers 207 points (266Mhz FSB).
K7-XP @2.2 GHz delivers 265 points (400Mhz FSB).
K8-FX @2.2 GHz delivers 305 points (@ Core speed).

Intel**
P4-C @3.2 GHz delivers 380 points (800Mhz FSB).

That's closer to my what my wallet can take :-) Looking at the 1.6GHz mark (most affordable and simplest to compare, 3.2GHz = 2x 1.6GHz), it looks like the differences are actually pretty minimal:

G5:        222 points
K7-XP: 207 points
P4-C:    190 points

Intel actually looks a bit sucky, saved as usual by sheer clock rate (I don't think you can actually buy P4Cs at 1.6), but other than that they're all within less than 10% of each other (except G5 vs P4C, that's closer to 15%).  Maybe a few FPS more or less in Quake 3 (or a couple more minutes in a 3D render), but for all intents and purposes fluff.  That's actually kindof good, it means that though the G5 is not the CPU for "the fastest desktop computer" (that was not going to last long anyway), PPC is at least a credible contender.  If you care about just benchmarks, the PPC does have a (slim) lead over the Intel system.  You didn't quote prices, but based on what I normally see, the AMD is probably your best bang for buck.  But we all already knew that :-)

[Varthall]

Thanks for all your infos and links, especially the arstechnica articles were highly informative. Reading the G4 vs K7 comparison I had the impression that G4s are more attractive to coders for its design elegance...
The performance depends on many factors, and it's difficult to really say which is faster and how much it is, althought it seems that G4s are more and more lagging behind P4s and Athlons, but this depends on what applications are used. So, I'll say to linux users that PPCs are roughly equivalent to x86, more expensive but with less heating being the most evident differences. I don't know if the MacOs emulation possibility will be seen as a attractive bonus for them, thought.
An important issue is that with the A1 and the Pegasos, now its finally possible to build ppc computers without having to buy complete Macs. Also, it's nice that there's a cpu in the desktop market without the x86 isa legacy.

Varthall

[Hammer]

CodeSmith wrote:
@Hammer

IBM
G5 @1.6 GHz delivers 222 points (1Ghz FSB).
G5 @1.8 GHz delivers 251 points (1Ghz FSB).

AMD
K7-XP @1.6 Ghz delivers 207 points (266Mhz FSB).
K7-XP @2.2 GHz delivers 265 points (400Mhz FSB).
K8-FX @2.2 GHz delivers 305 points (@ Core speed).

Intel**
P4-C @3.2 GHz delivers 380 points (800Mhz FSB).

That's closer to my what my wallet can take :-) Looking at the 1.6GHz mark (most affordable and simplest to compare, 3.2GHz = 2x 1.6GHz), it looks like the differences are actually pretty minimal:

G5:        222 points
K7-XP: 207 points
P4-C:    190 points

Intel actually looks a bit sucky, saved as usual by sheer clock rate (I don't think you can actually buy P4Cs at 1.6), but other than that they're all within less than 10% of each other (except G5 vs P4C, that's closer to 15%).  

For X86 @ 1.3GHz comparisons refer to
http://www.cpuid.com/PentiumM/index.php

It’s a review on Pentium M(400FSB, Banias) vs Athlon XP (266FSB, Palomino**) vs Pentium IV (Northwood) vs Duron (Morgan) vs VIA C3 (Nehemiah) vs Celeron(Tualatin) in regards to their efficiencies. All X86 cores has been set to 1.3Ghz clock speed.

**Limitation of the review i.e. Athlon's other cores was not reviewed e.g. Barton (with 400FSB + L2 512KB), T-Bred-B/Thoron (with 333FSB/400FSB), Athlon 64(Claw-Hammer) and Athlon FX (Sledge Hammer).

As for the prices, refer www.pricewatch.com (one of many) for pricing indications…

@Varthall
From the date Arstechnica's G4 vs K7 comparison was written, the X86 world has moved on to newer X86 cores.  The cpuid.com link also includes Pentium M's thermal(watts) information.

For PowerPC G4's thermal(watts), refer to
http://www.geek.com/procspec/apple/g4.htm

[Hammer]

K7-XP @2.2 GHz delivers 265 points (400Mhz FSB).

Addendum;
 
One of my test Athlon XP box @2.08Ghz/nForce2 400 Ultra(GA-7N400Pro2) with DDR400 FSB scored 263 points in Cinebench 2003.

[minator]

For X86 @ 1.3GHz comparisons refer to
http://www.cpuid.com/PentiumM/index.php

Interesting, The Pentium-M turns out well as expected but I know the C3 sucked but it sucks really badly in some of those tests.

--

The modern G4s (7447, 7457) should come in similar to the 1.3GHz Pentium-M but with 1/3 the power consumption.  The top x86s will be around 2X faster but up to 10X the power consumption.

Don't go by power consumption figures for Mac G4s because they are not the same CPUs.  They only used the top binned chips which can clock faster and use considerbly more power.

Now find someone who in general usage (i.e. not Photoshop) who actually notices the difference between 1GHz and 2GHz.

--

As for benchmarks, look at what they are comparing, many CPU benchmarks which are comparing using different programs are not just comparing CPUs.

If you want to measure the CPU performance only you have to use the same OS, same application and same compiler.  Very few if any CPU comparisions do this.


--

Note: I don't generally give advice to people selling competitors products but this is about CPUs...