Mac Mini Price Significantly Less than DIY PC

OSViews ^ | Posted Jan 14, 2005 - 01:08 PM | Harry Rider

[Swordmaker]

The Mac mini has received rave reviews from some of Apple's harshest critics for its low price. However, some of them are still implying that a comparably equipped PC can be had for slightly less. Harry Rider submitted the following editorial to osOpinion / osViews which continues his series on comparing Macs to equally-equipped DIY PCs. This time his comparison deal's with Apple's new Mac mini.

--

Sometime last year, I wrote an editorial comparing the price of Apple's new iMac G5 with a comparably equipped PC to see which of the two was less expensive. Apple has a reputation for selling "premium-priced" hardware and I've found that not to be true as Apple's prices are in fact less expensive. To the surprise of many, the iMac turned out to be nearly $250 less than an equally equipped DIY PC.

Apple's incorrect premium-priced reputation was best summarized by osViews's editor's prelude to my editorial: "The reason for [the false reputation] stems from the fact that Apple doesn't allow you to build your own computer from commodity parts. But that doesn't make PCs less expensive... though it does make the Mac less configurable."

Apple sells a limited number of computer configurations. As a result, the company has managed to secure larger margins than average... though not because their computers require you to pay more, but because they require you to buy more. In essence, they didn't offer a low-end machine thus allowing you to buy less and get less as you can with the large assortment of PC suppliers as well as the opportunity to build your own.

With Apple's introduction of the Mac mini early this week, the company has finally started competing in the low-end computer market. This caused me to ask the same question as I did before... "How does Apple's latest computer stack up against a comparably equipped PC?"

Same as before, I started by building a PC at Dell's web site to match the specs listed on Apple's Mac mini. The problem with this strategy is that Dell doesn't offer an equivalent system. You can buy a config with less and pay less, or you can buy one with more and pay more. So to make the comparison fair, it had to be made against a DIY PC.

So how does Apple's new computer stack up??

Here are the specs of the lower-end Mac mini:

1.25GHz PowerPC G4
256MB of PC2700 (333MHz) DDR SRAM
ATI Radeon 9200 with 32MB of DDR SDRAM
40GB Ultra ATA
Slot-loading Combo Drive (DVD-ROM/CD-RW)
One FireWire 400 port; two USB 2.0 ports; DVI output; VGA output (adapter included)
Built-in 10/100BASE-T Ethernet and 56K V.92 modem
Mac OS X version 10.3 “Panther
iPhoto, iMovie, iDVD and GarageBand), AppleWorks, Quicken 2005 for Mac, Nanosaur 2, Marble Blast Gold

Price: $499

--

Here are the specs of a comparably equipped DIY PC

Intel Pentium 4 2Ghz CPU................$145
P4S800 (SiS 648FX) Series P4 Motherboard................$63
256MB of PC2700 (333MHz) DDR SRAM................$51
ATI Radeon 9200................$90
40GB Ultra ATA................$56
DVD-ROM/CD-RW................$60
Case/PSU................$40
Windows XP Professional................$199
Adobe Photoshop album (Compares with iPhoto)................$50
Windows Movie Maker (Compares to iMovie)................$0
Ulead DVD MovieFactory (Compares with iDVD)................$40
Fruity Loops (Compares with Garage Band)................$80
Microsoft Works (Compares with Apple Works................$50
Quicken 2004................$60
Nanosaur 2................$15
Marble Blast Gold................$15

Hardware Price: $505
Hardware and Software price: $1,022

In my last editorial comparing the iMac and a DIY PC, I read reviews of my article from other sources redoing my comparison but leaving out software and then bringing down the cost of the PC as a result. This seems to be a common practice amongst PC hardware review sites and I think it's a mistake. Software is an area where Apple ads value to their hardware. I think their mistake can be attributed to the fact that most bundled software on PCs is shareware or junkware, so factoring in equivalents to Apple's iLife is so easily overlooked.

Regardless, to appease those that only look at hardware, I ran the figures both ways. The Apple hardware comes out less in both scenarios. The price comparison reiterates that Apple doesn't charge the supposed "premium" that is often equated with their systems. Can we finally do away with the fallacy that implies that Apple charges a premium for their computers?

[general_re]

Yeah, frequency means everything.

Hey, that's the second time you've done that - I never said that, and there's little point in continuing this if you plan on ignoring what I say. Now, if the G4 is sooooo very smokingly hot, it ought to be relatively simple for you to show me some real-world results, not abstract discussions of why it should be fast. Or fess up and admit that there aren't any such results. Your choice.

[HAL9000]

When you get your Mac mini, please keep us posted on how it goes.

Here is an item that should be included in the comparison of the $399 Dell vs. the $499 Mac: The Dell Dimension 2400 has a 90-day warranty. The Mac mini has a one-year warranty.

[HAL9000]

What matters is raw crunching power, and the G4 ain't got it where it counts. Never has.

The scalar unit of the G4 is a bit slower than comparable x86-based computers, but it has a respectable level of performance. I doubt that the typical user would perceive a major difference in scalar speed.

However, the G4's vector processing unit - which functions as a "supercomputer" - can easily outperform a lot of x86 computers if the software is designed correctly.

Applications that are designed to use vector processing hardware are a lot more common on Mac computers than Windows PCs. The Mac OS X operating system uses vector processing extensively for things like image processing, audio processing, etc. As far as I know, the Windows operating system doesn't use vector processing hardware much at all, and most of the PC apps that do use it are games.

In comparing the number crunching capabilites of the systems, it should also be noted that Apple's "Quartz Extreme" handles a lot of the screen drawing logic on the graphics chip instead of the G4. Windows won't be offloading it's screen imaging tasks to the GPU until Longhorn. It will be interesting to see how well Longhorn's advanced imaging system will perform on your current hardware.

[N3WBI3]

At least he is providing something (theory or not) all you have done is barked that the g4 would get smoked...

[Sloth]

A thoroughly dishonest comparison.

[television is just wrong]

I bought one. It is great. The only thing is that there is no airport card or antenna in it yet.

Not til at least 3/21

[general_re]

However, the G4's vector processing unit...can easily outperform a lot of x86 computers if the software is designed correctly.

And if the problem you're attacking lends itself to vectorization - lots of things don't.

Windows won't be offloading it's screen imaging tasks to the GPU until Longhorn.

Ummm, what? All GUI-based operating systems have taken advantage of graphics coprocessors for literally years now with stuff like hardware blitters and so forth to accelerate windowing operations. QE throws some additional compositing and 3d acceleration into the windowing system, but this is hardly revolutionary - it's certainly the next evolutionary step, but it's not exactly on a par with inventing the wheel or some such.

As far as I know, the Windows operating system doesn't use vector processing hardware much at all, and most of the PC apps that do use it are games.

"As far as you know" being, apparently, not very. Never heard of SSE/SSE2/SSE3? You do know that those are vector processing instruction sets for the x86 family, supported in hardware since the PIII, right? Heck, the first rudimentary vector processing on x86 has been around since the MMX/Pentium-166 days, back when the altivec-less G3 was the top of the line on the Apple side. Personally I regularly use audio and video encoding software that leverages the vector processing capability of x86, so I'm not sure where you're getting your info from.

[general_re]

[N3WBI3]

hmm things I dont see in there..

the g4 1.25 chip, or the amd chip..

Here however I do see it stacking up pretty good against much higher clock processors $ for $..

[antiRepublicrat]

Now, if the G4 is sooooo very smokingly hot, it ought to be relatively simple for you to show me some real-world results, not abstract discussions of why it should be fas

I don't know recent test on the G4 except Apple, and of course you'll discount that out of hand. But here's one of a G4 500 MHz vs. a PIII 1 GHz. Remember, the PIII is faster per-clock than the P4 of the time, and the latest G4 is a couple generations ahead of that old 500 MHz model.

So, we have a 500 MHz chip barely getting beaten by a chip clocked twice as fast. That's pretty clear.

[general_re]

Even if you boost the score by 25% to reflect the higher clock rate, it's still nothing to write home about. And the rating you link to is based on...what, exactly?

[antiRepublicrat]

it should also be noted that Apple's "Quartz Extreme" handles a lot of the screen drawing logic on the graphics chip instead of the G4. Windows won't be offloading it's screen imaging tasks to the GPU until Longhorn.

It should also be noted that not only does OS X now offload screen compositing to the graphics card, but as of Tiger (coming in a couple months) it will also offload a lot of video and graphics editing filters off to the graphics card. That will seriously skew many G5 benchmarks in favor of the Mac.

[antiRepublicrat]

"As far as you know" being, apparently, not very. Never heard of SSE/SSE2/SSE3? You do know that those are vector processing instruction sets for the x86 family, supported in hardware since the PIII, right?

You seem to have forgotten my little architectural lesson too. SSE on Intel shares duty with the floating point unit, so you're hosed if you want to do floating point and SSE at the same time, and this double-duty design doesn't make for very good SIMD in the first place. The G4/G5 have dedicated multi-part SIMD units separate from the floating-point. SIMD on the P4 is a joke.

Funny thing is, none of the non-SIMD apps I use require much processor power anyway, yet the the ones that do need power use SIMD extensively.

[general_re]

I seem to recall that my suggestion was an Athlon.

[dennisw]

bttt

[general_re]

SSE on Intel shares duty with the floating point unit, so you're hosed if you want to do floating point and SSE at the same time

No. That's no longer true - since the advent of SSE2, virtually all floating-point work is done through the SSE2 extensions rather than the x87 FPU. It's not an either-or thing, and hasn't been since SSE2 was introduced along with the P4.

[antiRepublicrat]

Hey, that's the second time you've done that - I never said that

You are correct. You said "what matters is raw speed," correctly taken to mean clock speed. The P4 has "raw speed," but that's all it has. Going electricity, I'd put it as the P4 gives a lot of voltage, but very little amperage. AMD and IBM/Freescale prefer to load up on amps instead.

[general_re]

"Raw speed" in the sense of more work done per unit of time. By that measure, there's nothing to apologize for. Unless maybe you're Apple pursuing the "megahertz myth" for all it's worth.

[antiRepublicrat]

I seem to recall that my suggestion was an Athlon.

AMD is always a good suggestion. Like the PPC, it does more work per clock than a Pentium 4/Xeon.

[antiRepublicrat]

That's no longer true - since the advent of SSE2, virtually all floating-point work is done through the SSE2 extensions rather than the x87 FPU. It's not an either-or thing, and hasn't been since SSE2 was introduced along with the P4.

"Virtually all." Sorry to disappoint you, but the P4 has one FP/SIMD pipeline, with all calculations being done by one FP/SIMD unit, the other unit used only for memory operations. The G4 has one FP unit with a dedicated pipeline and four vector processor with four units (permute, two integers and FP) that can operate on two instructions per cycle, and it has its own dedicated pipeline. That is why FP/SIMD is the PPC's strong point and why a lower-clocked PPC will beat a higher-clocked P4.

BTW, the Pentium's limited number of registers, a throwback to the very old days, also hurts it.