[zerostar]

Dual G5 vs Dual Xeon

I ran the following PSBench tests on my two main work machines. This includes 21 Photoshop Tests run 3 times and the average score is given.

I consider the G5 Powermac to be a workstation level machine, as the G5 was around $4500 fully setup and the HP xw8000 was around $4300 fully setup.

I pulled RAM (to even the machines out) and used the same 300+MB/sec RAID to make the tests as fair as possible.

Specifications are as follows

Apple PowerMac G5
Dual 2.5GHz G5 PPC Processors
2.5GB DDR-3200 RAM
Ati Radeon 9800 Pro 256MB Video
Internal 250GB SATA Drive
Mac OS X 10.4.1 All Updates Applies
Adobe Photoshop CS 1

HP xw8000 Workstation
Dual IntelXeon 3.2GHz Processors
2.5GB DDR 2700 RAM
Matrox Parhelia 256MB Video
2x80GB SCSI Ultra Drives
Windows XP Pro w/all Updates
Adobe Photoshop CS 1

TEST RESULTS (Winner of each is in bold)

All times are in seconds, shorter times are better.

90 Deg. Clockwise Rotation
G5 - 0.2
Xeon 0.2
Tie

9 Deg. Clockwise Rotation
G5 - 1.5
Xeon - 1.5
Tie

.9 Deg. Clockwise Rotation
G5 - 1.5
Xeon 3.7

1 Pixel Blur
G5 - 0.6
Xeon - 2.9

3.7 Pixel Blur
G5 - 0.9
Xeon - 4.8

85 Pixel Blur
G5 - 1.3
Xeon - 3.8

50% 1pix Unsharp Mask
G5 - 0.5
Xeon 1.5

50% 3.7pix Unsharp Mask
G5 - 1.1
Xeon 3.4

50% 10pix Unsharp Mask
G5 - 1.3
Xeon - 3.5

Despeckle Filter
G5 - 0.4
Xeon - 1.6

RGB to CMYK Filter
G5 2.7
Xeon - 3.6

60% Reduction
G5 - 0.4
Xeon - 0.6

Lens Flare
G5 - 2.7
Xeon - 4.9

Color Halftone
G5 - 7.7
Xeon - 7.6

NTSC Color Filter
G5 - 1.7
Xeon - 2.5

Accented Edges
G5 - 47.3
Xeon - 27.6

Pointillize Filter
G5 - 16.2
Xeon - 13.7

Watercolor Filter
G5 - 41.9
Xeon - 34.9

Polar Coordinates
G5 - 1.8
Xeon - 3.3

Radial Blur
G5 16.3
Xeon 16.7

Lighting Effects
G5 - 3.0
Xeon - 2.0

As you can see the G5 won in 16 of the 21 tests and 2 were a tie. Now I don't know how a Pentium 4 M compares to a Xeon or how well a newer 3.6 Xeon wold compare in these tests, but this is interesting to me none the less with the current Apple situation.

What does all this mean? I didn't get any work done so-far today.

[Apple Pro Underwear]

It also is not indicative of how these tests will be in 2008 when the G5 finally reaches 3Ghz and the Intel processors reach _____Ghz.

[zerostar]

Originally Posted by Apple Pro Underwear

It also is not indicative of how these tests will be in 2008 when the G5 finally reaches 3Ghz and the Intel processors reach _____Ghz.

Of course not, but who know where either will be by then? I would still put my money on Intel to lead the pack at that point.

I would LOVE to see Mac OS X Intel benchmarks on a similar machine, to see what difference the OS made, if any in these scores.

[Dalhectar]

The P4 architecture maxs out somewhere in the 4-5 GHz range. The reason Intel is going Dual Core and adopting 64-bit is because they are running out of P4 headroom. It's also the reason they are going to adpot a Pentium M based core to the desktop. But look at these benchmarks comparing a PPentium M to a Dual Core P4 and AMD64... it gets trounced...

http://www.gamepc.com/labs/print_content.asp?id=pm915
...hint scroll down to the Photoshop benchmarks and other creative/media encoding stuff.
http://www.anandtech.com/printarticle.aspx?i=2382
...look at MCC Winstone 2004 benchmarks

If you look at the type of apps Professional Mac users would use, the Pentium M in its current form is a step backwards. Between now and 2006-07 some things will happen, such as more optimizations for the Pentium M core, Pentium M going Dual core, and perhaps much higher clocked Pentium M (imagine what you can clock a Pentium M to if you applied the cooling mechanism of a PowerMac G5 to it). So perhaps once applied in a Apple Tower in 06-07 a Pentium M based (Yonah core) system could trounce a P4 based Xeon, G5, AMD64, and whatever else comes down in any application. The timeframe Jobs puts out gives time for Intel and Apple to switch platforms.

[zerostar]

Looks like a Pentium 4 EE would be a better choice for now? Especially if they hit 4 and 5 GHz, can the P4 EE do Dual Processor Configs?

[turtle777]

I guess we really don't know what Intel got up its sleeves. Since Steve would now both, the PPC and Intel roadmaps fairly well, I assume that Intel looks much more promising. They alluded to that, but I don't expect to see hard facts anytime soon, since the details of these roadmaps are probably confidential...

-t

[budster101]

Steve was probably pitting IBM against Intel for a few years now. Holding the carrot out for Intel, and when IBM said "uncle", or rather said, in not so many terms, lata' to Steve, he was ready to give Intel the go ahead for production. You don't promise machines in one year and not have production in the planning for a while.

[OreoCookie]

Intel will not hit 4 or 5 GHz anytime soon.

The Pentium 4 is dead. Dead. The next one will be based on the Pentium M which currently maxes out at 2.13 GHz (if my memory serves me right), but outperforms many 3+ GHz Pentium 4s.

[Apple Pro Underwear]

Originally Posted by zerostar

Of course not, but who know where either will be by then?

Who knows?

Steve does.

that's why it's INTEL or bust.

[budster101]

Pentium 4 runs tooooo hot. Fugettuhboudit.

[olePigeon]

I'm a little saddened. IBM's seeding Cell developer kits now with Linux and will be holding a "How to Program for the Cell" at LinuxTag 2005.

Sigh. I would've loved to see Apple jump onto the Cell architecture instead of Intel. Intel just seems so... dead end.

[budster101]

Right about the Cell technology. I'm a bit put off by that as well. Weird.

[Apple Pro Underwear]

Originally Posted by olePigeon

I'm a little saddened. IBM's seeding Cell developer kits now with Linux and will be holding a "How to Program for the Cell" at LinuxTag 2005.

Sigh. I would've loved to see Apple jump onto the Cell architecture instead of Intel. Intel just seems so... dead end.

just from my mac forum knowledge...

dont you think Apple would love to as well? dont you think apple scoured every scenario before coming to this conclusion. just from common sense, i would think that the Cell is not the answer and it does not solve the problems Apple has and thats why they need to move to Intel.

the truth probably is that the Cell technology is "dead end" (in terms of Apple products).

[TampaDeveloper]

Originally Posted by zerostar

TEST RESULTS (Winner of each is in bold)

All times are in seconds, shorter times are better.

Color Halftone
G5 - 7.7
Xeon - 7.6

As you can see the G5 won in 16 of the 21 tests and 2 were a tie. Now I don't know how a Pentium 4 M compares to a Xeon or how well a newer 3.6 Xeon wold compare in these tests, but this is interesting to me none the less with the current Apple situation.

What does all this mean? I didn't get any work done so-far today.

Ok. First you need to recount... that brings it down to 15. Next, I'm pretty sure that 7.7 is greater than 7.6, so you can unbold the G5.

So G5 won in 14 of 21 tests, P4 won in 5 test, and 2 were a tie.

NOW... how much of this is related to the video card?

http://graphics.tomshardware.com/graphic/20031229/

Card Chip DX Generation Chipclock Memoryclock Buswidth Pipes
Radeon 9800 Pro 256 ATI R350 DX 9 380 700 256-bit 8x1
Parhelia 128 Parhelia 512 DX 8.1 220 500 256-bit 4x4

(128 has same specs other than memory as 256) see www.matrox.com

Ok. So we have the G5 using a video card with 1.32x the chip clock and 1.4x the memory clock.


I'm pretty sure that Steve Jobs and the management at Apple didn't just toss a coin to come to the conclusion that they needed to abandon PPC. It sucks, yes. But you can bet your ass that they have some pretty substantial evidence that x86 Mac is quicker than PPC Mac.

[olePigeon]

Originally Posted by TampaDeveloper

Card Chip DX Generation Chipclock Memoryclock Buswidth Pipes
Radeon 9800 Pro 256 ATI R350 DX 9 380 700 256-bit 8x1
Parhelia 128 Parhelia 512 DX 8.1 220 500 256-bit 4x4

I think those are mostly AltiVec, not video card enhanced.