Macworld Mini Benchmarks compared with new iMacs

[FredT2]

I was looking at Macworld's benchmarks for the new iMacs, comparing them to the 2012 Mac Mini. The Mini model they use for comparison is the 2.6 i7 with fusion drive. All of the results look normal and reasonable until you get to HandBrake encode. In that test the top of the line iMac took 49 seconds while the Mini took 103 seconds, more than twice as long! Does anyone have an explanation for this difference?

[statenine]

The top of the line iMac runs a 3.4ghz i7, I am guessing that is the reason? Handbrake relies heavily on CPU so this would make sense to me. Maybe the better/dedicated GPU has something to do with it as well?

[FredT2]

Originally Posted by statenine

The top of the line iMac runs a 3.4ghz i7, I am guessing that is the reason? Handbrake relies heavily on CPU so this would make sense to me. Maybe the better/dedicated GPU has something to do with it as well?

The iMac CPU runs at 3.4 ghz, the Mac Mini at 2.6, about a 30% difference. I wondered about GPU, but I found that Handbrake doesn't use the GPU at all, so it's all processor.

[alexdd]

Originally Posted by FredT2

The iMac CPU runs at 3.4 ghz, the Mac Mini at 2.6, about a 30% difference.

Not true...My i7 2.6Ghz is 11900 vs 12.500 of the hi-end i7 iMacs..less than 10% difference and its 30% faster than the base iMac27''

[FredT2]

Originally Posted by alexdd

Not true...My i7 2.6Ghz is 11900 vs 12.500 of the hi-end i7 iMacs..less than 10% difference and its 30% faster than the base iMac27''

My point is that there is only a 30% difference in processor speed between the machine and that most benchmarks have differences even less than that, and yet Macworld has the Mini taking more than twice as long on the HandBrake test. Just wondering if there's an explanation for that, or perhaps a problem in the test.

[fa8362]

Originally Posted by FredT2

My point is that there is only a 30% difference in processor speed between the machine and that most benchmarks have differences even less than that, and yet Macworld has the Mini taking more than twice as long on the HandBrake test. Just wondering if there's an explanation for that, or perhaps a problem in the test.

Given that it's inconsistent with the other results, I'd say it's an obvious error.

[FredT2]

Originally Posted by fa8362

Given that it's inconsistent with the other results, I'd say it's an obvious error.

If it is, it's a consistent one. They previously tested a Mac Mini 2.3 i7 at 119. I guess I'll have to register at the Macworld site and see if I can get an answer there.

[jack92029]

I presume that we are talking about Geek benchmarks. What specifically does the Handbrake test examine?

More Ram and SSD have an effect on these test, as does doing the test immediately after boot up.

My best Geek benchmark was 10,955 @ 32Bit. The 64 bit seems to add 1000-1500 to the results. Are you comparing 32 or 64 bit tests?

[philipma1957]

HANDBRAKE is all about cpu. so handbrake on a 2.6 mac mini quad will run far slower then on a 3.4 imac quad. here is the passmark for a quad mac mini cpu



and an imac quad


http://www.passmark.com/cpubenchmark...0S+%40+3.10GHz


the imac is the i7 3770s the mini is the i7-3720m

[FredT2]

Originally Posted by philipma1957

HANDBRAKE is all about cpu. so handbrake on a 2.6 mac mini quad will run far slower then on a 3.4 imac quad. here is the passmark for a quad mac mini cpu



and an imac quad


http://www.passmark.com/cpubenchmark...0S+%40+3.10GHz


the imac is the i7 3770s the mini is the i7-3720m

I understand fully that a 2.6 Mini will run HandBrake more slowly than a 3.4 iMac. I don't understand, however, why it ran as slowly as it did in the Macworld test. If time to run a particular encode in HandBrake scales with processor speed, the the Mini should have completed the test in 64 seconds if the iMac took 49. I don't know what CPU Mark in the link you sent means, but in that chart the Mini cpu has a CPU mark of 8.557 while the 3.1 iMac cpu has a CPU mark of 9.53. If those represent relative performance, then the Mini should have completed the test in 56 seconds!

[philipma1957]

Originally Posted by FredT2

I understand fully that a 2.6 Mini will run HandBrake more slowly than a 3.4 iMac. I don't understand, however, why it ran as slowly as it did in the Macworld test. If time to run a particular encode in HandBrake scales with processor speed, the the Mini should have completed the test in 64 seconds if the iMac took 49. I don't know what CPU Mark in the link you sent means, but in that chart the Mini cpu has a CPU mark of 8.557 while the 3.1 iMac cpu has a CPU mark of 9.53. If those represent relative performance, then the Mini should have completed the test in 56 seconds!

yeah I had trouble linking to scores you found so I did not realize how short the test was.


the ratio of 9.53 to 8.557 for the imac's cpu to the mini's cpu means the imac should be 1.1137 faster then the mini. with a short test of 60 seconds or so the cache size matters


http://ark.intel.com/products/64891/...up-to-3_60-GHz


the cache on the mini is 6mb see above


http://ark.intel.com/products/65524/...up-to-3_90-GHz


the cache on the imac i7 is 8mb


so a short test of hand brake is bad testing. which your point that the test seemed wrong is completely correct. a 10 or 15 minute test would make the cache advantage much smaller. the cache advantage is 8 to 6 or 1.333 much higher then the speed advantage. you made a good 'catch' (pun intended) of the glaring difference in handbrake.

to illustrate look at a short aja test on a 320gb hdd allowing the hdd's test to be included

[FredT2]

Originally Posted by philipma1957

so a short test of hand brake is bad testing. which your point that the test seemed wrong is completely correct. a 10 or 15 minute test would make the cache advantage much smaller. the cache advantage is 8 to 6 or 1.333 much higher then the speed advantage. you made a good 'catch' (pun intended) of the glaring difference in handbrake.

Thank you so much for the information! I guess it's time to give the folks over at Macworld some feedback about their testing methodology.

[philipma1957]

your welcome I just posted an example above on including the cache for an hdd test look at it! It reads at 1978 MB/s LOL


on this post I ran a 8gb hdd test vs the 128MB test


still allowing the cache my scores are far more normal for a slow 5400rpm hdd 63 MB write and 31 MB read. same drive in both cases I allowed the cache to be used but the second test was huge 8gb so a 3mb cache on this mini (2010 model) does nothing much.

[tears2040]

Originally Posted by alexdd

Not true...My i7 2.6Ghz is 11900 vs 12.500 of the hi-end i7 iMacs..less than 10% difference and its 30% faster than the base iMac27''

Yeah and in ANYTHING regarding graphics it's about 75% slower.... I tested top of the line Mac Mini against my 2011 imac and in graphics the imac was about 3x faster.... Such a shame....

On the bright side though the new Mac Mini which will be released next year will have a better graphics card and more than likely I will upgrade to it. I actually like the design of the Mini but I need a better gpu.

[kobyh15]

What about thermal throttling? Surely the iMac quad is able to turbo boost higher and longer than the Mini quad.

[dasx]

Originally Posted by tears2040

On the bright side though the new Mac Mini which will be released next year will have a better graphics card

It's always nice to have a member of Apple's Mini Developer's Team in the forums.

[tears2040]

Originally Posted by dasx

It's always nice to have a member of Apple's Mini Developer's Team in the forums.

So are you saying the next Mac mini will not come with a Haswell HD 5000?

I mean you let me know, Genius Bar

[Mr. Retrofire]

Originally Posted by FredT2

I was looking at Macworld's benchmarks for the new iMacs, comparing them to the 2012 Mac Mini. The Mini model they use for comparison is the 2.6 i7 with fusion drive. All of the results look normal and reasonable until you get to HandBrake encode. In that test the top of the line iMac took 49 seconds while the Mini took 103 seconds, more than twice as long! Does anyone have an explanation for this difference?

The processor in the iMac has probably more transistors and better cooling (Turbo Boost 2.0), so that the CPU cores can do more in less time. The larger processors have probably also more transistors in the management units. The x264 H.264 encoder in HandBrake needs fast memory access, so that the larger L1, L2 and L3 processor caches can increase the encoding speed.

[Mojo1]

GPU performance requirements depends on what you plan to do with your Mac.

Here is an example of how a discrete GPU beats the pants off an integrated HD4000: http://www.barefeats.com/minivim.html.

[dasx]

Originally Posted by tears2040

So are you saying the next Mac mini will not come with a Haswell HD 5000?

I mean you let me know, Genius Bar

Did I say that? No no no… All I'm saying is I can't tell. Just as you CAN'T tell. You can guess/hope, that's all.

[ctyrider]

Originally Posted by dasx

Did I say that? No no no… All I'm saying is I can't tell. Just as you CAN'T tell. You can guess/hope, that's all.

Sure you can. If there IS another update of Mini, it will come with HD5000. This one is a given.

[dasx]

Originally Posted by ctyrider

Sure you can. If there IS another update of Mini, it will come with HD5000. This one is a given.

That's a guess too. You can't know it for sure so shouldn't really be saying guesses as if they were the truth.

Will the new Mini come with HD5000? Maybe. You don't know it though, do you?

[MatthewAMEL]

The Haswell GPU isn't called HD 5000. It's a whole new GPU line (GT1, GT2, GT3). GT1 (slowest) is supposedly twice as fast as the HD 4000.

Handbrake is CPU bound. My 2012 i7 Mini encodes the AppleTV 3 preset at almost exactly the same frame rates as my Late 2009 i7 iMac.

The Mini uses the Mobile Ivy Bridge 3720QM. The 27" iMac uses the Desktop Ivy Bridge 3770.

The Mobile Ivy Bridge can only hit 3.6GHz in SINGLE CORE operations. The Desktop version can run at 3.7GHz with all 4 cores.

The Mobile processors are also much more sensitive to thermal throttling. My Mini encodes the same frame rates as my iMac, but it runs at 96C vs 67C.

[ctyrider]

Originally Posted by dasx

That's a guess too. You can't know it for sure so shouldn't really be saying guesses as if they were the truth.

Will the new Mini come with HD5000? Maybe. You don't know it though, do you?

Dude, you seem to be logically challenged, so let me spell this out for you one more time. The next gen Mini will come with next gen Intel processors. The next gen Intel CPU architecture (Haswell) includes updated GPU, which is HD5000. You can keep playing silly semantic games and call this a "guess", but it's a fact based on the published Intel roadmaps, history of Mini updates, and just basic common sense.

[philipma1957]

Originally Posted by ctyrider

Dude, you seem to be logically challenged, so let me spell this out for you one more time. The next gen Mini will come with next gen Intel processors. The next gen Intel CPU architecture (Haswell) includes updated GPU, which is HD5000. You can keep playing silly semantic games and call this a "guess", but it's a fact based on the published Intel roadmaps, history of Mini updates, and just basic common sense.

my 2 cents is Haswell + HD5000 will be in the next mini ,

but apple is drifting towards arms that they built.

http://9to5mac.com/2012/09/15/apples...st-cpu-design/

my guess is 2013 brings us the haswell not a new arm, but it is a guess.