Mac/PC showdown: and the winner is...

Posted by: Joe Petrik on 15 October 2004

Circumstance has allowed me an opportunity to try an interesting comparison -- well, interesting to me; you may be bored to tears -- putting a current G5 Mac up against a current Xeon-based Dell workstation on exactly the same task.

The comparison is interesting since both machines happen by chance to be the same purchase price (within a few dollars) after education discount, so the comparison is really about how much computing power you get for $x with a Mac vs a name-brand PC. The Mac is my home machine, bought in early August 2004 and the PC is my work machine, bought in early September 2004.

The two computers in question are spec'd as follows:

Entry-level Mac G5
* Dual 1.8 GHz 970fx processors
* 2 GB of RAM
* 160 GB SATA HD
* OS X (10.3.5)

Dell 670 Precision Workstation
* Single 3.2 GHz Xeon processor
* 2 GB of RAM
* 160 GB SATA HD
* Windows XP (SP2)

Both machines have the current "CS" version of Adobe Photoshop installed. The task was to render a 24,000 x 18,000 pixel image (which works out to be 1.2 GB) by rasterizing a PDF file, then apply a filter to that image.

Both computers had their Photoshop memory preference set to 75% of available RAM, so both had enough memory available to render a 1.2 GB image with at least a 100 MB to spare. No other applications, except OSs, were running at the time.

Here are the results for time taken to render a 1.2 GB image file:
Mac -- 10:24 (~10.4 minutes)
PC -- 3:32 (~3.5 minutes)
PC comes out about 3 times faster

Here are the results for time taken to apply a filter:
Mac -- 4:16 (~4.25 minutes)
PC -- 2:40 (~2.7 minutes)
PC comes out about 1.5 times faster

Obviously, this isn't the be-all and end-all of tests, but on at least one application I use regularly a decent PC workstation has a substantial speed edge over a current G5 Mac. Easy of use and the fun factor are clearly and decisively in the Mac's favour -- at least for me -- but that comes at a price that many will not want to pay.

The upshot is that I prefer Macs and will continue to recommend them, but I may back off from boasting how blindingly fast the new G5s are.

Joe

Posted on: 19 October 2004 by Alex S.

I had him down as a Pixel.

Posted on: 19 October 2004 by garyi

More of a microbe really.

I was just running a test on my mac.

In Photoshop, make a new RGB white bg 8x10 300ppi document. Open Add Noise and set to Uniform, Monochromatic, 400% and click OK. Then open Radial Blur, set to Amount 100, Method Zoom, Quality Best but do not hit OK yet.

As you hit OK, start timing. Continue timing until the filter finishes, then stop.

I got this off another website.

I got 4 minutes 20, mine is a 1.6 though, lowest spec possible, lol.

Posted on: 19 October 2004 by Joe Petrik

Gary,

quote:
I got 4 minutes 20, mine is a 1.6 though

I just ran the test on my Xeon machine at work and it came in at 1:43, which is ~2.5 times faster than a single 1.6 GHz G5.

I'll try it at home tonight with my dual 1.8 if I get a chance.

Joe

Posted on: 19 October 2004 by garyi

Joe that would be fair enough to my simple brain in that its twice as powerful.

Posted on: 19 October 2004 by sideshowbob

http://homokaasu.org/rasterbator/

-- Ian

Posted on: 19 October 2004 by Derek Wright

Did Gary's test it took

1 minute 9 seconds

Using a dual 2.0 ghz with 1.5gb of memory

The cooling fans were getting up speed during the process

Derek

<< >>

Posted on: 19 October 2004 by garyi

Derek I feel slightly deflated.

Posted on: 19 October 2004 by sideshowbob

Macs have never been speed demons. That's why they're rarely used for complex computing tasks. Graphic designers would be more productive using PC hardware, but they have style reasons for their preference (and, I suppose, Mac colour and font management has historically set an industy standard, although that's largely down to Adobe rather than Apple).

-- Ian

Posted on: 19 October 2004 by Dan M

6min 9sec on my iBook (700 MHz G3 384 MB)

Waaaaay faster than I would have expected from iGary's numbers, which has me thinking --

In System Preferences / Energy Saver / Optimize Energy Settings what is set? On my iBook, when plugged in (Power Adapter), Options has my 'Processor Performance' set to 'Highest'. How's yours set?

Dan

Posted on: 19 October 2004 by Dan M

Take a look at Mac's G5 PDF -- the figure on page 28 seems to suggest a dual 1.8GHz G5 handily beats a dual Xeon (>3GHz). So what gives?

Dan

Posted on: 19 October 2004 by sideshowbob

quote:
So what gives?

Marketing.

-- Ian

Posted on: 19 October 2004 by matthewr

"which has me thinking"

Had me thinking maybe some people are doing their 8x10s as cms and some as inches

All modern computers, of whatever ilk, are insanely fast and, outside of specialist applications like high end video editing, you can get several times as much as you need for about $1000.

The only people who worry about it are the sort who obsess about which sort of computer *other* people should buy. They need to get some perspective.

Matthew

Posted on: 19 October 2004 by woody

I've often heard this thing about Macs (and MacOS) being more reliable than PCs...so what's that all about?

-- woody

Posted on: 19 October 2004 by Joe Petrik

Weird. Despite it coming out way behind the 3.2 GHz Dell Xeon when rasterizing an image, my dual 1.8 GHz came out ahead of the Xeon on Gary's test -- 1 minute and 11 seconds (for the dual 1.8 G5) vs 1 minute 43 seconds (for the 3.2 GHz Xeon). So, on this test, the Mac is ~45% faster.

_____________________________________

Matthew,

quote:
Had me thinking maybe some people are doing their 8x10s as cms and some as inches.... The only people who worry about it are the sort who obsess about which sort of computer *other* people should buy.

Your laptop didn't fare so well, did it?

Joe

Posted on: 19 October 2004 by Dan M

quote:
Originally posted by matthewr:
"which has me thinking"

All modern computers, of whatever ilk, are insanely fast and, outside of specialist applications like high end video editing, you can get several times as much as you need for about $1000.

The only people who worry about it are the sort who obsess about which sort of computer *other* people should buy. They need to get some perspective.

Matthew,

Well in my defense, I often run my dual Xeon Dell work machine full whack on both processors for 30 days, using simulation code written in F90. A 4x speed up would get me done in about a week. So, I've very interested in this since we got just got in the top end G5 and I'll be seeing if I need to trade in my 3 year old dual 1.6GHz Dell. FWIW, my 'other' work machine is a clustered Symmetric MultiProcessing system with 1,600 POWER4 processors with a 1.3-GHz clock cycle (each can perform up to four floating-point operations per cycle, so peak of 8.32 TFLOPs), though its hard to get more than a few of the 32-way nodes to yourself.

cheers,

Dan

Posted on: 19 October 2004 by Joe Petrik

Dan,

quote:
FWIW, my 'other' work machine is a clustered Symmetric MultiProcessing system with 1,600 POWER4 processors with a 1.3-GHz clock cycle (each can perform up to four floating-point operations per cycle, so peak of 8.32 TFLOPs), though its hard to get more than a few of the 32-way nodes to yourself.

What kind of pron needs that much processing power?

Joe

Posted on: 19 October 2004 by sonofcolin

Joe,

Do you have the photoshop G5 plugin installed? If not, this could explain your findings.

In reality, you should see little difference in performance between the 2 systems you've described (except of course reliability).

Maybe the G5 ran your actions twice and you just didn't notice! Check your action scripts and make sure they're identical!

Posted on: 20 October 2004 by Stephen Bennett

quote:
Originally posted by Joe Petrik:

What kind of pron needs _that_ much processing power?

Joe

Me! I'm making music and movies.

Stephen

PS Who you calling a pron? I'm a geek.

Posted on: 20 October 2004 by Stephen Bennett

quote:
Originally posted by sonofcolin:
Joe,

Maybe the G5 ran your actions twice and you just didn't notice! Check your action scripts and make sure they're identical!

And check the CPU monitor! You owe it to us and posterity....

Stephen

Posted on: 20 October 2004 by matthewr

Joe said "Your laptop didn't fare so well, did it?"

I would need more information before I could do a test -- the inches/cm thing for sure, plus which version of PS, how much RAM, etc. My laptop has a Hyper Threading (a sort of psuedo multi-processing) P4 which post-dates my current version of PS (6.0) -- plus one might reasonable expect to get different numbers from 6, 7, and CS, and so on.

My laptop should produce something around G4 Powerbook numbers I would imagine. My desktop is much faster (at least at what i do) and would be more interesting as it's currently running half-cocked unless I install 64-bit XP and there is a 64-bit version of PS.

Dan M "I often run my dual Xeon Dell work machine full whack on both processors for 30 days, using simulation code written in F90"

I think that qualifies as a specialist application -- finding out which is faster would require specfici research and ideally custom benchmarking.

"Well in my defense"

I was making a general point (that performance doesn't really matter) rather than having a go at you.

Matthew

PS Joe I just bought a new monitor based on Mitsubishi's new Diamondtron U2 tube -- frankly it roxxxxxx and CRT is not dead. A 22" was 1/3 less than my 19" cost 4 years ago as well.

Posted on: 20 October 2004 by garyi

Matthew just select 8x10 in the drop down menu, no one said this was a serious teast just a good indicator.

I think if anything it proves that benchmarkers are making money for old rope, you can make it say anything, if JOe had opened this thread with my test the whole perspective would be skewd

Posted on: 20 October 2004 by matthewr

Gary -- But 8x10 what? I have a choice between inches and cms (and points, and picas, and pixels, etc). Incheas vs cm is the difference between a 3.2M and 20.6M image...

Plus the version of PS has to be the same for it to be remotely meaningful.

Matthew

Posted on: 20 October 2004 by Joe Petrik

Sonofcolin,

quote:
Do you have the photoshop G5 plugin installed?

Yes. Incidentally, both machines (the Xeon-based Dell and my G5 Mac) have Photoshop CS with the latest plug-ins installed.

_______________________

Stephen

quote:
I'm making music and movies.

Well, good pron needs a decent soundtrack. ;-)

_______________________

Matthew,

quote:
My desktop is much faster (at least at what i do) and would be more interesting as it's currently running half-cocked unless I install 64-bit XP and there is a 64-bit version of PS.

Did you buy one of these...

quote:
I just bought a new monitor based on Mitsubishi's new Diamondtron U2 tube -- frankly it roxxxxxx and CRT is not dead. A 22" was 1/3 less than my 19" cost 4 years ago as well.

I was rather impressed with my previous Mitsubishi monitor, too. I also rather like my Apple LCD monitor, but a Mitsubishi CRT is much, much better value.

Joe

Posted on: 20 October 2004 by matthewr

Joe,

The new CRT is now ahead of my state-of-the-art S-IPS based laptop panel, and what I had assumed was stellar performance from the panel was the gradual demise of my old Mitsubishi.

The panel is still good but the CRT is fantastic and for £400, compared to approx. £900 for a 1600x1200 there really is no contest.

Matthew

Posted on: 20 October 2004 by Dan M

Joe asked:What kind of pron...

I think you answered your own question:

In short, rasterizing is a way of resizing a file to the exact size, resolution and colour space you want (say, a 3x5-foot poster at 300 dpi in RGB space) without loss of quality

Matthew -- I knew you were not having a go at me.

I also admit the best way to see if the G5 would give me a decent boost is to suck it and see, i.e. profile the same code on both machines - which I will do. However, with all the hype around the G5 (and that it uses the same chip as my 'other' machine) I was hoping for 'blow the doors off' performance. I'm currently waiting for a library for the mac so can't do the test at the moment.

cheers

Dan