Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
It's the "score". Higher is better. How much vram does the GTX-660 have? The m395X shows some clear advantages in a great many of those tests.
The GTX-660 has 2GB VRAM. It is an older but still pretty good midrange card in a 4Ghz Windows PC. I think it pulls up to 250 watts, more than the entire iMac, so it's pretty good the "mobile" M395X is competitive.
 
...I am leaning towards returning mine and getting the M395X. I'm truly curious about how VRAM affects performance in FCPX (and maybe Premiere Pro if you want to test that?)...


Thanks for running the Textures and Film Noir tests. My updated numbers and yours are attached. Interestingly in these two tests the M395X was actually slower than the 780m, whereas it is generally faster. I rebooted, recreated the FCP X test library and re-ran those on both machines six times each to be sure, and took the average. The M395X was a little faster than the M395.

If you could run the Heaven and Valley benchmarks that would be good. Also Cinebench is easy to run. Sorry if you've already done these; I can't remember. See below links.

You might also want to re-run the Tessmark tests because your numbers were so far off. If they are consistent that doesn't mean there's anything deficient about the M395 -- it does very well on most tests. The M395X is usefully faster but not hugely faster in most cases. There are always odd results which is why running multiple tests are important.

It's probably a good practice to temporarily disable Time Machine and Spotlight when doing benchmarks. They can throw off results. Time Machine can be disabled from the menu bar, Open Time Machine Preferences. Spotlight can be disabled in System Preferences>Spotlight>Privacy, then press "+" to add your hard drives to this exclusion list. Remember to undo these steps after benchmarking.

Heaven: https://unigine.com/products/benchmarks/heaven/
Valley: https://unigine.com/products/benchmarks/valley/
Cinebench: http://www.maxon.net/products/cinebench/overview.html
 

Attachments

  • FCPX_FX_Bench_780mvsM395X.jpg
    FCPX_FX_Bench_780mvsM395X.jpg
    177.1 KB · Views: 560
  • FCP_FX_Bench_M395XvsM395.jpg
    FCP_FX_Bench_M395XvsM395.jpg
    107.3 KB · Views: 507
...I'm truly curious about how VRAM affects performance in FCPX (and maybe Premiere Pro if you want to test that?)...


Unfortunately I can't easily test Premiere. I have CS5 on my Windows machine but would be too hard to install it on my Mac.

It occurred to me we should test LuxMark. It is widely used and easy to run. Unfortunately the 3.0 and 3.1 versions crash on OS X, so the 2.1 version is the last that runs error free. My results are on the attached spreadsheet. Procedure:

(1) Download LuxMark 2.1 for OS X: http://www.luxrender.net/wiki/LuxMark_v2
(2) Select Scene
(3) It will either auto start on OpenCL GPU only, or you select it from Mode menu
(4) Write down the result, select next scene and repeat
 

Attachments

  • iMac27Benchmarks.jpg
    iMac27Benchmarks.jpg
    723.4 KB · Views: 499
Unfortunately I can't easily test Premiere. I have CS5 on my Windows machine but would be too hard to install it on my Mac.

It occurred to me we should test LuxMark. It is widely used and easy to run. Unfortunately the 3.0 and 3.1 versions crash on OS X, so the 2.1 version is the last that runs error free. My results are on the attached spreadsheet. Procedure:

(1) Download LuxMark 2.1 for OS X: http://www.luxrender.net/wiki/LuxMark_v2
(2) Select Scene
(3) It will either auto start on OpenCL GPU only, or you select it from Mode menu
(4) Write down the result, select next scene and repeat

Sounds great. I will run this when I get back.

Those 119 fps figures for the m395 look suspicious.

Here are the results for my 290x, a significantly less powerful card with 2 GB ram

View attachment 598360 View attachment 598361 View attachment 598362 View attachment 598363 View attachment 598364 View attachment 598365 View attachment 598366

Looks like a bug.

Yeah, I think my numbers look a little off. For some reason I had many tests that hit this odd 7198/119 fps barrier (and none of them went past it). This makes me wonder if having only 16GB RAM was a bottleneck. Jerwin how much RAM were you running with the m290x?
 
...I think my numbers look a little off. For some reason I had many tests that hit this odd 7198/119 fps barrier (and none of them went past it). This makes me wonder if having only 16GB RAM was a bottleneck...

I have seen several cases where during this stress testing that FCP X and/or OS X become "deranged" and performances really drops. I can't tell if it's the OS, the app or if the FCP library has become damaged. If you see suspiciously low numbers during benchmarking, close the test library, create a new one, reboot the system and re-import the test clip.
 
Sounds great. I will run this when I get back.



Yeah, I think my numbers look a little off. For some reason I had many tests that hit this odd 7198/119 fps barrier (and none of them went past it). This makes me wonder if having only 16GB RAM was a bottleneck. Jerwin how much RAM were you running with the m290x?
Twenty Four gigabytes. It essentially means that I have a large cache (and I suppose memory leaks won't be as dangerous)
Gputest doesn't use all that much main memory-- perhaps 50 megabytes. It may consume very large amounts of vram.
 
...Unfortunately the 3.0 and 3.1 versions crash on OS X, so the 2.1 version is the last that runs error free...

To be more accurate, only the "Lobby" test of LuxMark 3.1 crashes -- the LuxBall HDR and Neumann tests run OK.

You can't cross-compare numbers between 2.1, 3.0 and 3.1, so having a complete result from ver. 2.1 is useful. However 3.1 is the current version so it's probably best to also run that for the two tests that work. The 3.1 version is at the bottom of this page: http://www.luxrender.net/wiki/LuxMark#Binaries

My numbers for LuxMark 3.1 are included in the attached updated table.
 

Attachments

  • iMac27Benchmarks.jpg
    iMac27Benchmarks.jpg
    741.1 KB · Views: 452
  • Like
Reactions: sebban
This was a great thread, especially that last couple of pages when the video editing folks chimed in. This was very helpful. I ended up ordering a 5K i7 m395 256GB 8GBRAM (upgrading to 32GB). I think that is the sweet spot as far as having a machine that will last a while/future proof/price. I'll be updating (and adding to) my current set-up that consist of a early 2013 maxed out rMBP and 2006 2.66GHz MP. All used for different stages/levels of the video production process. I plan on using the new iMac for mainly editing/color-grading/effects to exporting files in all the various formats. While the rMBP will still be used for liter editing tasks/graphics/titles/web. And the macpro... well, it's been demoted to capturing/archiving older format video/audio for a documentary I'll be working on for the next year. I love these machines. They last and seem to always find a way to still do certain tasks in my life. Thanks again.
 
  • Like
Reactions: rdav
I ended up with a 27" i5 2TB with 8GB ram (will go to 24gb if needed) and a M395 but I am using it for audio production and studio work. I also believe the m395 is really the best of all choices and this article helped me out making decision.

http://barefeats.com/imac5k15.html

It was really the only place I could see all of the cards against each other in any meaningful way. So far 16 tracks of 24 bit at 96khz is not very taxing at all but no anomalies, jitters, or artifact noticed. Quite a nice step up from my older P4 XP based system ($30K new when purchased).

AZ
 
I ended up with a 27" i5 2TB with 8GB ram (will go to 24gb if needed) and a M395 but I am using it for audio production and studio work. I also believe the m395 is really the best of all choices and this article helped me out making decision.

Congrats, that's the exact build I have its a great configuration. Enjoy your new baby :)
 
Great thread! Mainly read the last 5 pages, so sorry if this was mentioned previously. Are there any energy consumption/thermal comparisons of the 395 vs 395x, under light use of the GPU (i.e. not doing any graphics work)?
 
Now that's it been out for a while can someone answer it in the most basic terms.
How big of a difference can we expect?
From what I've gathered- on this thread, the biggest jump is from the 380 to 390 (40%?), but the 390 uses a dated chip...so is it worth jumping to the next base model?

$1,949 m390 $2,129 +2TB HD
$2,149 m395
$2,379 m395x
The biggest jump is from the m395 to m395x- from m390 with 2tb to upgrade the card is only $20.

I know iMacs are not gaming machines- cheaper to build your own- way cheaper- but I use it for other things.
http://www.notebookcheck.net/Computer-Games-on-Laptop-Graphics-Cards.13849.0.html

*I purchased the m390 but then cancelled because I did not want to regret buying it....the whole "I should've just coughed up the extra $200"
 
Now that's it been out for a while can someone answer it in the most basic terms.
How big of a difference can we expect?
From what I've gathered- on this thread, the biggest jump is from the 380 to 390 (40%?), but the 390 uses a dated chip...so is it worth jumping to the next base model?

$1,949 m390 $2,129 +2TB HD
$2,149 m395
$2,379 m395x
The biggest jump is from the m395 to m395x- from m390 with 2tb to upgrade the card is only $20.

I know iMacs are not gaming machines- cheaper to build your own- way cheaper- but I use it for other things.
http://www.notebookcheck.net/Computer-Games-on-Laptop-Graphics-Cards.13849.0.html

*I purchased the m390 but then cancelled because I did not want to regret buying it....the whole "I should've just coughed up the extra $200"

Yes, M390 is AMD Pitcairn GPU, an older GCN 1.0 generation chip. M395 is new Tonga, GCN 1.2 which uses memory bandwidth more efficiently and allows a bit higher resolutions or better texture quality and has also DSP included (if Apple is ever starting to use it). At least in Windows you should be able to use it. M390 --> M395 gives around 10-20% speed boost.

4GB of video memory is useful, if you play Windows games. For Mac, it's less useful, but some pro apps like video editing can use it. M395 --> M395X gives another 10% plus memory advantage that can be a big thing on Windows gaming and some openCL intensive apps.

I have no experience on boot camp on 5k iMac, but what I've read there is no full support for screen resolution. It will work up to 4k.
 
Last edited:
I have no experience on boot camp on 5k iMac, but what I've read there is no full support for screen resolution. It will work up to 4k.

Apple provided 5K drivers for bootcamp some time ago. You're no longer relegated to 4K.
 
They don't care. "Let it boil at 100C" they say.. so it does. :apple:

i5 3.3GHz consumes 26W less than i7 4.0GHz when at 100% utilization..

Wouldn't be surprised, if Apple has counted that these chips can do 100C 37 months.. and then die. Just to survive over the Apple care. Then we can sell theses suckers next computer and there will be less machines in the second hand market.
It seems like the overheating issue of the Core i7 is not as big of a deal as being reported.

I have played some Elder Scrolls Online at Max settings at 2.5K and machine (4GHz i7 395X) isn't overheating or getting into the triple digits.

Here is a interesting video.

That said I have my Rampage Extreme motherboard arriving in about a week. I'm not going to be doing much gaming on the iMac. That said having seen the screen I will be upgrading my monitors for my main machine.
 
It seems like the overheating issue of the Core i7 is not as big of a deal as being reported.

I have played some Elder Scrolls Online at Max settings at 2.5K and machine (4GHz i7 395X) isn't overheating or getting into the triple digits.

Yes, 2015 version has an improved heat dispatching solution. I haven't heard what is the actual difference to previous model. Fixit tore down only 4k 21.5" model in 2015.
 
Yes, M390 is AMD Pitcairn GPU, an older GCN 1.0 generation chip. M395 is new Tonga, GCN 1.2 which uses memory bandwidth more efficiently and allows a bit higher resolutions or better texture quality and has also DSP included (if Apple is ever starting to use it). At least in Windows you should be able to use it. M390 --> M395 gives around 10-20% speed boost.

4GB of video memory is useful, if you play Windows games. For Mac, it's less useful, but some pro apps like video editing can use it. M395 --> M395X gives another 10% plus memory advantage that can be a big thing on Windows gaming and some openCL intensive apps.

I have no experience on boot camp on 5k iMac, but what I've read there is no full support for screen resolution. It will work up to 4k.

Oh thanks.
I'd love to go for the m395x but the price is just way too high.
I can justify the m390 to m395 jump- faster CPU/larger HDD- but the m395x does not offer anything else...it's just a +$200 GPU bump.
 
Just left the Apple Store and more confused now. I went in with a memory card with a 12 minute video clip from my drone. It was recorded in 1080p.

I downloaded the file to all three trims of the 27 inch iMac. I opened up three windows in Safari. Two streaming video in 1080P and one streaming in 4K.

I imported the video file to iMovie, added the same filter and same adjustments, and exported the file to the desktop.

The lowest spec iMac was the fastest. The mid spec iMac was behind by about 10 seconds. The top spec iMac was behind by 30sec.

Export settings were the same. 1080P, Pro setting, and compression set for best quality.

M380 took 7:30
M390 took 7:40
M395 took 8:00

Shouldn't it be the other way around?

I have a good deal on the late 2014 3.5ghz i5 with the m290x. I am a photographer who is incorporating more video work as I progress through the year.

The difference between the three current video cards is throwing me for a loop and making me wonder if the old M290x might be my best choice money wise because from what I understand everybody is expecting an upgrade this coming November anyways, nvidia?
 
Got home and did the same test with my old mbp. It's a '10 2.66GHz core i7 with GeForce GT 330M 512mb.

Anybody care to guess how long it took that to do the same test?
 
Got home and did the same test with my old mbp. It's a '10 2.66GHz core i7 with GeForce GT 330M 512mb.

Anybody care to guess how long it took that to do the same test?
Make an article. I'd be interested to read it. I have my doubts but it would be uninteresting read.
 
Register on MacRumors! This sidebar will go away, and you'll see fewer ads.