2013 27" i7 vs 2012 27" i7

[CrAkD]

So the mac pros pricing has finally pushed me to the iMac I wanted a 6 core but at $4k I'm priced out.

I have a nehalem hackintosh that's been running strong for years well I saved and waited for the right powerhouse mac. I'm tired of not being able to update software and ready for a new mac.

I want to get a second Thunderbolt Display so a refurbished 2012 would be welcome. But I have had a sata 3 ssd in my hackintosh for years so I'm worried the 2012 hard drive options may be a little slower. I'm also not sure about graphics I don't play many games but I'm planning on keeping the iMac for quite a few years like I did my nehalem.

Is fusion worth it or should I hold out for straight ssd?

Machine will be used to run protools, convert blurays/DVDs and light video editing.

Do you think I should save money on the 2012 or will the hard drive speed of the 2013 really future proof me?

 

[joema2]

So the mac pros pricing has finally pushed me to the iMac I wanted a 6 core but at $4k I'm priced out...Is fusion worth it or should I hold out for straight ssd?...Machine will be used to run protools, convert blurays/DVDs and light video editing...Do you think I should save money on the 2012 or will the hard drive speed of the 2013 really future proof me?

I have both 2012 and 2013 iMac 27. The 2012 is 3.4Ghz i7, 3TB Fusion, 16GB, GTX-675, and 2013 is 3.5Ghz i7, 3TB Fusion, 32GB, GTX-780M. Below are some basic benchmarks I ran:

Lightroom 5.2 import 102 raw 23 megapixel stills and build 1:1 previews (min:sec):

2012 iMac 27: 4:55
2013 iMac 27: 2:19

Lightroom 5.2 export 102 raw 23 megapixel stills to .jpg at 90% quality (min:sec):

2012 iMac 27: 5:09
2013 iMac 27: 4:19

64-bit GeekBench 3 benchmark, single core:

2012 iMac 27: 3548
2013 iMac 27: 3955

64-bit GeekBench 3 benchmark, multi-core:

2012 iMac 27: 13,728
2013 iMac 27: 14,839

QuickBench disk benchmark (large test, no caching, 5 pass average):

2012 iMac 27 average read: 419.7 MB/sec
2013 iMac 27 average read: 657 MB/sec

2012 iMac 27 average write: 326.0 MB/sec
2013 iMac 27 average write: 329.9 MB/sec

Re SSD vs Fusion, a large SSD is expensive, and most external drives aren't as fast as Fusion. The only external drive I have that's as fast is an 8TB Pegasus R4 in RAID 5.

In general I think Fusion is a good idea -- it gives a lot of storage at a moderate price (relative to SSD) and is very fast.

Re the hard drive speedup of the 2013 with PCIe interface vs 2012, as you can see above it makes a bigger difference on some things than others. If you're reading from cache, it's very fast. If you're writing, not as much improvement. Whether you would perceive a major difference vs 2012 would depend on your workload characteristics.

 

[CrAkD]

I have both 2012 and 2013 iMac 27. The 2012 is 3.4Ghz i7, 3TB Fusion, 16GB, GTX-675, and 2013 is 3.5Ghz i7, 3TB Fusion, 32GB, GTX-780M. Below are some basic benchmarks I ran:

Lightroom 5.2 import 102 raw 23 megapixel stills and build 1:1 previews (min:sec):

2012 iMac 27: 4:55
2013 iMac 27: 2:19

Lightroom 5.2 export 102 raw 23 megapixel stills to .jpg at 90% quality (min:sec):

2012 iMac 27: 5:09
2013 iMac 27: 4:19

64-bit GeekBench 3 benchmark, single core:

2012 iMac 27: 3548
2013 iMac 27: 3955

64-bit GeekBench 3 benchmark, multi-core:

2012 iMac 27: 13,728
2013 iMac 27: 14,839

QuickBench disk benchmark (large test, no caching, 5 pass average):

2012 iMac 27 average read: 419.7 MB/sec
2013 iMac 27 average read: 657 MB/sec

2012 iMac 27 average write: 326.0 MB/sec
2013 iMac 27 average write: 329.9 MB/sec

Re SSD vs Fusion, a large SSD is expensive, and most external drives aren't as fast as Fusion. The only external drive I have that's as fast is an 8TB Pegasus R4 in RAID 5.

In general I think Fusion is a good idea -- it gives a lot of storage at a moderate price (relative to SSD) and is very fast.

Re the hard drive speedup of the 2013 with PCIe interface vs 2012, as you can see above it makes a bigger difference on some things than others. If you're reading from cache, it's very fast. If you're writing, not as much improvement. Whether you would perceive a major difference vs 2012 would depend on your workload characteristics.

Thank you that was very helpful....however it makes me want to save more or wait till the 2013s show up in the refurb store haha

 

[WilliamG]

That huge discrepancy between Lightroom RAW imports might be more to do with what spinning hard drive, too. I can't think of anything else that would cause such a massive speed difference, especially since the pcie portion isn't a factor, and Haswell is a tiny speed increase.

 

[Outrigger]

Thank you that was very helpful....however it makes me want to save more or wait till the 2013s show up in the refurb store haha

Also consider the 2012's 16GB of RAM vs. 2013's 32GB. That makes a big difference.

 

[joema2]

That huge discrepancy between Lightroom RAW imports might be more to do with what spinning hard drive, too. I can't think of anything else that would cause such a massive speed difference, especially since the pcie portion isn't a factor, and Haswell is a tiny speed increase.

I repeated that test twice, once after a reboot and before starting Lightroom. This implies it's not RAM caching. As you said, the amount of speedup is far greater than the Haswell CPU improvement, and it's not GPU-bound. That only leaves disk I/O, and the only major improvement in the disk subsystem was the PCIe interface.

True, a faster disk interface normally has no major advantage over a slower one if testing on a single rotating drive, and the slower interface is already far faster than that drive. However the 128GB SSD built into the Fusion drive could have easily cached that, and unlike RAM you can't clear it by rebooting.

Accessing SSD via PCIe can be much faster than SATA, and it's possible the SSD portion of the Fusion drive was a factor in this one test.

When I get time I'll re-run it with the files on my Pegasus R4, then with the files on a slower external drive. The CPU & GPU elements will not change, so the performance changes significantly that implies its I/O bound. The total data size isn't that big -- just 2.67GB across 102 files, so normally you wouldn't expect an I/O bottleneck, but maybe LR is doing lots of small reads. I'll look at the cumulative I/O count in Activity Monitor.

 

[WilliamG]

Also consider the 2012's 16GB of RAM vs. 2013's 32GB. That makes a big difference.

Yep, true.

To the OP, if you can get a 2012 cheaper than the 2013, no question get the 2012. The differences are truly negligible in 99% of tasks. Big file reads will be faster on the 2013, and so will big file transfers over 802.11ac. That's it.

Speaking of which, has anyone sourced an 802.11ac card yet? I want to stick one in my 2012 when I open it up to swap out the hard drive and SSD for larger ones.

 

[Bear]

Thank you that was very helpful....however it makes me want to save more or wait till the 2013s show up in the refurb store haha

Waiting for them to hit the refurb store is not necessarily a bad idea. Save some money and maybe get a better system.

 

[ekiro]

Is it worth selling my 2012 27-inch iMac at a $600-1000 loss to buy the next year's model?

It's at a loss because I forgot to pay off the 6mo-no-interest card on time... kind screwed me for another $300...

 

[Bear]

Is it worth selling my 2012 27-inch iMac at a $600-1000 loss to buy the next year's model?

It's at a loss because I forgot to pay off the 6mo-no-interest card on time... kind screwed me for another $300...

Most likely not, unless you are going from a base 27" to a CTO model that fits your needs better.

 

[CrAkD]

Waiting for them to hit the refurb store is not necessarily a bad idea. Save some money and maybe get a better system.

yeah maybe I will do that as a happy medium. I can get a 2012 almost maxes for $2k or under the new one is like $2700 so thats a thunderbolt display right there.

 

[WilliamG]

I repeated that test twice, once after a reboot and before starting Lightroom. This implies it's not RAM caching. As you said, the amount of speedup is far greater than the Haswell CPU improvement, and it's not GPU-bound. That only leaves disk I/O, and the only major improvement in the disk subsystem was the PCIe interface.

True, a faster disk interface normally has no major advantage over a slower one if testing on a single rotating drive, and the slower interface is already far faster than that drive. However the 128GB SSD built into the Fusion drive could have easily cached that, and unlike RAM you can't clear it by rebooting.

Accessing SSD via PCIe can be much faster than SATA, and it's possible the SSD portion of the Fusion drive was a factor in this one test.

When I get time I'll re-run it with the files on my Pegasus R4, then with the files on a slower external drive. The CPU & GPU elements will not change, so the performance changes significantly that implies its I/O bound. The total data size isn't that big -- just 2.67GB across 102 files, so normally you wouldn't expect an I/O bottleneck, but maybe LR is doing lots of small reads. I'll look at the cumulative I/O count in Activity Monitor.

I'd like to know what you see. Thanks for the reply.

 

[ekiro]

Most likely not, unless you are going from a base 27" to a CTO model that fits your needs better.

I have the 2012 27-inch model all maxed out (3.4GHz i7, 32GB, 128GB + 3TB Fusion). Totaled around $2.8k. I did the Bill Me Later thing for no payments for 6 months, interest free, but stupid me missed the opportunity to pay it off at the 6th month. So my new total investment is around $3.1K. I think I can sell it for $2.2K for sure. Lowest $2K. It would only then cost me an additional $600 out of pocket to get the 2013 iMac maxed out. You think my numbers are pretty accurate?

(I don't buy the RAM from Apple.)

 

[Crunch]

You can always break up the Fusion drive, if you want to have pure SSD performance, but unfortunately Apple still only supplies you with 128 GB Flash on both the 1TB and 3 TB versions. Still, 128GB SSDs, especially the new PCIe ones from Samsung, are no longer slower than their larger capacity counterparts. My sweet spot is the 240/256GB category.

I am just SERIOUSLY peeved that it doesn't include Thunderbolt 2. That's just ridiculous when less than a month later, every entry-level 13" MacBook Pro w/ Retina Display now supports the 20Gbps Thunderbolt 2, which is required for 4K displays. @#$%^&*()

Mine might go back in favor of my usual Retina MacBook Pro + Thunderbolt Display combo.

 

[Bear]

I have the 2012 27-inch model all maxed out (3.4GHz i7, 32GB, 128GB + 3TB Fusion). Totaled around $2.8k. I did the Bill Me Later thing for no payments for 6 months, interest free, but stupid me missed the opportunity to pay it off at the 6th month. So my new total investment is around $3.1K. I think I can sell it for $2.2K for sure. Lowest $2K. It would only then cost me an additional $600 out of pocket to get the 2013 iMac maxed out. You think my numbers are pretty accurate?

(I don't buy the RAM from Apple.)

Very not worth upgrading.. only a few percent performance difference. Wait another year or three.

 

[etichi]

Also consider the 2012's 16GB of RAM vs. 2013's 32GB. That makes a big difference.

You forgot an important caveat, which is only when the machine has broken the 16G threshold will the 32G machine perform better. If the machine doesn't do this then there is 0 difference.

 

[RodPinto]

I did that and have no regrets... Faster wi-fi is great and double the RAM on the gpu too...

Is it worth selling my 2012 27-inch iMac at a $600-1000 loss to buy the next year's model?

It's at a loss because I forgot to pay off the 6mo-no-interest card on time... kind screwed me for another $300...

 

[ekiro]

I did that and have no regrets... Faster wi-fi is great and double the RAM on the gpu too...

I have my iMac wired. Lets face it, it sits on a desk all day. The wireless is just to make it look cleaner with one cable instead of 2. But if you have anything plugged into the IO it defeats the purpose. I use a white cat6 cable to match the white power cable and use zip ties to merge and count our them nicely. I do some gaming on this machine and require the best speeds and latency, so wire is the way to go with that.

 

[joema2]

...I do some gaming on this machine and require the best speeds and latency, so wire is the way to go with that.

I also formerly thought so. However attached are Speedtest.net results from my 2013 iMac 27 using 802.11AC wireless (121 mbps) vs my wife's 2012 iMac 27 using wired 100Base-T ethernet (95 mbps) to the same Asus RT-AC66R router. This test was repeated multiple times with the same results. The ISP is Comcast Xfinity 105.

 

[ekiro]

I also formerly thought so. However attached are Speedtest.net results from my 2013 iMac 27 using 802.11AC wireless (121 mbps) vs my wife's 2012 iMac 27 using wired 100Base-T ethernet (95 mbps) to the same Asus RT-AC66R router. This test was repeated multiple times with the same results. The ISP is Comcast Xfinity 105.

Every encounter any interference with your Wifi signal?

 

[joema2]

Every encounter any interference with your Wifi signal?

So far, no. My 2013 iMac is within 10 feet of the Asus RT-AC66R router, so that probably helps.

At the other end of my house, my wife's 2012 iMac with 100Base-T wired ethernet gets better performance than wireless, about 92 mbps wired vs about 60 mbps wireless. However her iMac doesn't have 802.11ac, only 802.11n, so maybe the newer wireless protocol would be better.

So what you said is still often correct. Wired is more reliable over more situations than wireless. However wireless is continuously getting better, and in some cases it's better than wired.

 

[joema2]

I repeated that test twice, once after a reboot and before starting Lightroom....
When I get time I'll re-run it with the files on my Pegasus R4..

OK I re-ran the Lightroom import test and cannot reproduce that. I don't know how that fast 2:19 time was achieved.

On 10.9.0, importing from a 3TB Fusion drive, using the same version of Lightroom 5.2 and the same files I get 4:44 for import and 4:29 for export.

Activity Monitor shows it's mostly CPU-bound, not disk-bound. On import it's about 8 megabytes/sec read, 1 megabyte/sec write with CPU at about 70% across four cores.

On export it's about 4 megabyte/sec writes, with similar CPU. So this type of activity is more CPU than disk bound, and no matter how fast or slow the disk, it shouldn't make much difference.

I did some further disk copy tests comparing the 3TB Fusion Drive and the 8TB Promise Pegasus R4 in RAID 5. Each of these repeated twice, and the average taken:

3TB Fusion, disk-to-disk copy of 143 files (5.7 TB): 203 megabytes/sec

3TB Fusion-to-R4: 514 megabytes/sec

R4-to-R4: 216 megabytes/sec

R4-to-3TB Fusion: 279 megabytes/sec

Comments: due to iMac's 32GB RAM and 128GB SSD cache on the 3TB Fusion Drive, it's very hard to get a consistent number not affected by caching. If your workload falls into the cache it is greatly accelerated.

I don't know why the copy from 3TB Fusion to Pegasus R4 was so fast. I repeated the test many times. Typically RAID 5 isn't fast for writes, but this was consistently very fast. The R4 must be doing some kind of write-behind optimization.

In general the Fusion drive is very fast and doesn't require management.

The Promise Pegasus R4 is also very fast and with RAID 5 can tolerate a single drive failure without losing data.

 

[joema2]

One more item: for a Fusion Drive, the app iStat Menus shows what disk I/O goes to the SSD part vs rotating part.