CPU performance-per-dollar (a mac comparison)

[RichP]

A friend of mine today and I were discussing which mac offers the best CPU performance-per-dollar. It was started because we were discussing AppleTV as a cheapo machine for cluster setups and such.

Now, let me first disclaim things: This was ONLY based on CPU score, and price of the Apple machine in stock form. The CPU data was all from SPEC.ORG, using thier CINT2000 scoring. Macs are generally not listed, so I used an equivalent machine from Dell or IBM (which shouldnt matter all that much, they all share the same intel chipset and basic architecture)

The data is in "performance units per dollar" (Took the CPU score and divided it by the retail price)

Is the data bulletproof? far from it. but at least its some sort of comparison.

Here we go!:

Apple TV - 1Ghz Pentium M - 2.29
Mac Mini Core Duo 1.66Ghz - 2.32
MacPro Quad 2.00Ghz - 2.13
MacPro Quad 2.66Ghz - 2.20
MacPro Quad 3.00Ghz - 2.04

It appears the Mini is the best performance for dollar, with AppleTV a close second. Granted, this comparison has nothing to do with useability, memory, graphics, etc etc. But for building a cheapo render farm, or a "folding farm" the mini appears pretty strong.

Just figured I would share.

 

[zap2]

Interesting...also for the last Mac Pro...I assume you meant 3.0Ghz not 2.04Ghz...unless Apple stuck an unknow Mac Pro update in on my!

 

[panoz7]

I wouldn't have expected those results. I just did a few calculations and figured out the cost per GHz (Price / (Ghz * Cores)) of each of those:

Apple TV: $299 / Ghz (that one sure was easy )
Mac Mini 1.66: $180.42 / Ghz
Mac Mini 1.83: $218.31 / Ghz
Mac Pro 2.00: $265.12 / Ghz
Mac Pro 2.66: $227.44 / Ghz
Mac Pro 3.00: $268.25 / Ghz

In this case the 1.66 Ghz Mac Mini kills the others and the Apple TV is by far the worst. This doesn't correspond with your data very well. To me this means one of three things: either the Pentium M in the Apple TV performs better per Ghz then the Xeon in the Mac Pro (unlikely in my mind), the tests that spec.com uses aren't multi-core aware, or I'm an idiot and don't understand much about processors (most likely).

 

[Scarlet Fever]

Apple TV - 1Ghz Pentium M - 2.29
Mac Mini Core Duo 1.66Ghz - 2.32
MacPro Quad 2.0Ghz - 2.13
MacPro Quad 2.66Ghz - 2.20
MacPro Quad 2.04Ghz - 2.04

unless they have changed something, there was no 2.04GHz MacPro
EDIT; jeez you people are quick...

interesting data, though.

 

[FleurDuMal]

But can't different processor do different amounts of data per clock, meaning that this comparison is completely meaningless? For example, I'm guessing the processor in the Mac Pro can do a lot more per hz than the processor in the Mini

 

[RichP]

First..fixed the typo.

The most fair "comparison" is a CPU benchmark, as it should take into account what the CPU can do per clock cycle.

I had a hard time finding one site that just had lists of CPU benchmarks, would be interesting to see a more detailed analysis using other benchmarks.

The SPEC benchmarks list the number of cores and number of processors. Based on comparing machines, the scores take into account a dual-core (i.e. are substantially higher than the power of a similar single core) but dont show "double" the performance for a 2x2 arrangement. For the macpros, i doubled the CPU score to account for the quad core, which is, granted, is another "assumption"

About all of the CPU types we see, they are basically all the same architecture based off pentium M. (core duo, and core2duo, and xeon in the macpro, with different levels of tricks and whistles)

 

[siurpeeman]

i'll take a different approach.

i'll take the performance scores from macworld and divide that by the number of hundreds of dollars per system. for instance, macworld gave the 1.83 core duo mini a score of 165. at a cost of $799, the mac mini has a 20.65 performace per cost value. so for other systems...

mac mini 1.83 - 20.65
mac mini 1.67 - 25.38
macbook 1.83 - 15.29
macbook 2.0 (white) - 13.70
macbook pro 2.16 - 10.46
macbook pro 2.33 - 9.04
mac pro 2.66 - 11.96

i'm not going to do the whole lineup. it's easy to see that by the methods shown the cheapest systems will offer the most performance per cost. but it's important to note that such scores ignore the value of pricier systems. a mac mini cannot provide the portability of a macbook, and some people demand the performace of a mac pro regardless of cost. to many, a mac pro is worth a lot more than five or six minis. also, cost isn't a derivative of speed alone. you pay for luxuries like portability, built-in screens, options for expansion, more ram/hard disk space, discreet graphics, etc. to say that a mac mini has more performance per cost is misleading to say the least. the interesting thing, i think, is that the mini represents the worst value of the entire mac lineup.

 

[twoodcc]

yeah, i don't think this comparision means all that much. a computer is more than just a processor

 

[RichP]

yeah, i don't think this comparision means all that much. a computer is more than just a processor

This comparison noted all that, it was simply about CPU speed, which for some specific apps, is that that matters (number crunching)

I agree, the mini to me, represents one of the worse values for my user experience. but I needed a bunch of computers crunching math in an array, the mini, appears to be quite an option.

 

[Umbongo]

But can't different processor do different amounts of data per clock, meaning that this comparison is completely meaningless? For example, I'm guessing the processor in the Mac Pro can do a lot more per hz than the processor in the Mini

This is correct.

 

[localoid]

A friend of mine today and I were discussing which mac offers the best CPU performance-per-dollar. It was started because we were discussing AppleTV as a cheapo machine for cluster setups and such.
...

You might find this article an interesting read...

 

[localoid]

But can't different processor do different amounts of data per clock, meaning that this comparison is completely meaningless? For example, I'm guessing the processor in the Mac Pro can do a lot more per hz than the processor in the Mini

The follow is quoted from an rather interesting article that provides some insight into using clustered minis vs Intel Xeon dualies:

The Intel Dual Xeon Xserve looks promising as well in terms of raw cpu performance, but I’ve seen reports that it suffers from the same high power consumption as any other Dual Xeon system does: ~ 3A. However, this is TEN TIMES the power consumption of a MacMini, so for a high density web farm, this is not a better solution. Better to utilize that Intel Xserve as a database server to take advantage of it’s greater RAM capacity (32GB), increased threading (dual dual-core) and 64-bit capability.

 

[Pressure]

It's like saying any computer is better than no computer really heh.

 

[gnasher729]

...either the Pentium M in the Apple TV performs better per Ghz then the Xeon in the Mac Pro (unlikely in my mind)...

Actually, it will, at least at one GHz. Lets say you have memory that takes 100ns to read. At one GHz, that is 100 cycles, so that Pentium M would have to wait 100 cycles for memory reads. At three GHz, the same memory now takes 300 cycles to read.

That Pentium M has a slow harddisk coming with it. That will slow it down at times. However, the 3 GHz quad core MacPro would need a harddisk twelve times faster for the same percentage slow down, and it doesn't have that.

 

[twoodcc]

This comparison noted all that, it was simply about CPU speed, which for some specific apps, is that that matters (number crunching)

I agree, the mini to me, represents one of the worse values for my user experience. but I needed a bunch of computers crunching math in an array, the mini, appears to be quite an option.

yes, but as others have said, different processors can do more than others.....

 

[Umbongo]

This comparison noted all that, it was simply about CPU speed, which for some specific apps, is that that matters (number crunching)

I agree, the mini to me, represents one of the worse values for my user experience. but I needed a bunch of computers crunching math in an array, the mini, appears to be quite an option.

Why bother will Apple at all? Why on earth would you pay a large overhead for the form factor of the mini (actual design and mobile chipset and processor) when you could buy a proper tailored solution for less with better suited hardware.

 

[Chone]

A more accurate measurement would be the value of the whole computer by components and their OEM price, by those terms, a Mac Pro is the better value of them all, its impossible to build a Mac Pro equivalent computer for cheaper than 2500$, at least that was the case a few months ago.