The Definitive riMac graphics benchmark thread

[fruitpunch.ben]

Hi all. I'm cross posting from #118, because it is getting out of hand:
All sorts of benchmarks are coming out as the new retina iMacs come in, but they are not standardized, and so it's impossible to compare. Here is what I have collected so far. If you can add to the information, please post, but do so with the same quality settings.
If you want any different benchmarks collected, ask and I'll add them to the list.
I will collect the results in this post.

EVERYONE with 2015 27" riMac: List your fps:
All tests at 1440p, list processor (i5/i7), RAM and card: (380, 390, 395, 395X)
1. Unigine Heaven running with:
- OpenGL/Direct X
- 2560x1440
- Quality Ultra
- Tessellation: Disabled
- Antialiasing: None
- Fullscreen
2. Ditto Cinebench

RESULTS: OS X/Open GL
UNIGINE:
1. Nvidia 680MX (2012): 21.4fps (7.7 - 47.6) i7 [Post #109]
2. Nvidia 780M (2013):
3. AMD 290 (2014):
4. AMD 290X (2014): 17.9fps (11 - 37.1) i5 [Post #3]
5. AMD 295X (2014): 21.7fps (13.3 - 44) i7 [Post #47]
6. AMD 380 (2015): 9.9fps (5 - 21.3) i5 [Post #12]
7. AMD 390 (2015): 15.7fps (9.8 - 31.7) i7 [Post #77]
8. AMD 395 (2015): 19.3fps (8.5 - 39.8) i5 [Post #154]
9. AMD 395X (2015): 20.7fps (9 - 42.9) i7 [Post #100]

CINEBENCH 15:
1. Nvidia 680MX (2012): 60.36fps (i7) [Post #110]
2. Nvidia 780M (2013): 75.08fps (i7) [Post #25]
3. AMD 290 (2014): 85fps (i5) [Mac4Ever review (French)]
4. AMD 290X (2014): 90.87fps (i5) [Post #63]
5. AMD 295X (2014): 104fps (i7) [Mac4Ever review (French)]
6. AMD 380 (2015): 78.7fps (i5) [Post #1]
7. AMD 390 (2015): 95.3fps (i5) [Mac4Ever review (French)]; 106.72 fps (i7) [Post #77]
8. AMD 395 (2015): 94.66fps (i5) [Post #95]
9. AMD 395X (2015): 107.04fps (i7) [Post #261]

Mac4ever:

Macotakara:

And huge thanks to barefeats for testing and @jerwin for assembling this into a table of gaming/pro apps benchmarks:

RESULTS: Windows/Direct X

 

[AnTaR3s]

Great post. I am sure this will be a lot of help for many people.

 

[jerwin]

cinebench 86.39 fps

 

[Alesc]

Late 2013 iMac

 

[twilexia]

Fantastic post and compilation!!!

 

[kødskjold]

I don't know if I should be impressed with the 680MX or disappointed in the 395X, on the UNIGINE test. I've been waiting for this 27" riMac update for the past 5 months. So far, I'm rather underwhelmed.

 

[Laai]

I don't know if I should be impressed with the 680MX or disappointed in the 395X, on the UNIGINE test. I've been waiting for this 27" riMac update for the past 5 months. So far, I'm rather underwhelmed.

Wait up. Sadly, DFX tests need some time. Newer chip is not always faster than the older one on every test. Additionally, as time goes on and newer drivers come out for current chips, games are optimized more and more to take advantage of it.

On the other hand, I do not understand why Apple does not put desktop class DFX card in this machine. The space is there, if you custom design cooling the card can fit in current design. I also do not understand, why Apple does not upgrade the cooling fan to a bigger one, THE SPACE FOR IT IS THERE. On the fan part, there are only upsides: bigger fan=lower RPM for same airflow and silent machine, when the fan spins up, much more air is pushed through and you can cool the iMac much better.

 

[AnTaR3s]

Wait up. Sadly, DFX tests need some time. Newer chip is not always faster than the older one on every test. Additionally, as time goes on and newer drivers come out for current chips, games are optimized more and more to take advantage of it.

On the other hand, I do not understand why Apple does not put desktop class DFX card in this machine. The space is there, if you custom design cooling the card can fit in current design. I also do not understand, why Apple does not upgrade the cooling fan to a bigger one, THE SPACE FOR IT IS THERE. On the fan part, there are only upsides: bigger fan=lower RPM for same airflow and silent machine, when the fan spins up, much more air is pushed through and you can cool the iMac much better.

We will never understand, they will never tell. I have made my peace with that a long time ago, though its frustrating from time to time...the solution for will come hopefully with Thunderbolt 3 and external GPUs. Then they will never ever sell me an iMac again.

 

[kødskjold]

hopefully with Thunderbolt 3 and external GPUs. Then they will never ever sell me an iMac again.

If they would make a respectable mac-mini update (TB3, quad-core), I would totally go this route. There are some brilliant desktop GPUs and monitors out there. I made Newegg wishlist with a self-build at the same pricepoint as the iMac. The boss graphics and G-SYNC/Freesync are tempting. But I want it on OSX!

 

[phi7]

Could someone with a 395 or 395x please do a Blender Cycles rendering benchmark?
My iMac will arrive in 3 weeks and I am quite curious how the new GPU performs. The test is not very complicated and should only take a few minutes.

You need the latest version of Blender (2.76) which should support AMD GPUs -> http://www.blender.org
The speed test is done using the BMW 1M Test scene by MikePan -> Download BMW27.blend
Just double click the scene after you've installed Blender and go to "render" -> "render image" (or just hit F12)
After the scene has finished rendering you should get the time on the upper left part of the screen

You can also check whether or not the GPU is detected
("File" -> "User Preferences" -> "System" Tab -> Compute Device (lower left))

 

[762999]

Could someone with a 395 or 395x please do a Blender Cycles rendering benchmark?
My iMac will arrive in 3 weeks and I am quite curious how the new GPU performs. The test is not very complicated and should only take a few minutes.

You need the latest version of Blender (2.76) which should support AMD GPUs -> http://www.blender.org
The speed test is done using the BMW 1M Test scene by MikePan -> Download BMW27.blend
Just double click the scene after you've installed Blender and go to "render" -> "render image" (or just hit F12)
After the scene has finished rendering you should get the time on the upper left part of the screen

You can also check whether or not the GPU is detected
("File" -> "User Preferences" -> "System" Tab -> Compute Device (lower left))

by default my test was in CPU mode, it's a good idea that people validate as you indicated. thanks for the good find

 

[vetruvian]

Hi all. I'm cross posting from #118, because it is getting out of hand:
All sorts of benchmarks are coming out as the new retina iMacs come in, but they are not standardized, and so it's impossible to compare. Here is what I have collected so far. If you can add to the information, please post, but do so with the same quality settings.
If you want any different benchmarks collected, ask and I'll add them to the list.
I will collect the results in this post.

EVERYONE with 2015 27" riMac: List your fps:
All tests at 1440p, list processor (i5/i7), RAM and card: (380, 390, 395, 395X)
1. Unigine Heaven running with:
- OpenGL/Direct X
- 2560x1440
- Quality Ultra
- Tessellation: Disabled
- Antialiasing: None
- Fullscreen
2. Ditto Cinebench

RESULTS: OS X/Open GL
UNIGINE:
1. Nvidia 680MX (2012): 21.4fps (7.7 - 47.6) i7 [Post #109]
2. AMD 290 (2014):
3. AMD 290X (2014):
4. AMD 295X (2014):
5. AMD 380 (2015):
6. AMD 390 (2015):
7. AMD 395 (2015):
8. AMD 395X (2015): 20.7fps (9 - 42.9) i7 [Post #100]

CINEBENCH 15:
1. Nvidia 680MX (2012): 60.36fps (i7) [Post #110]
2. AMD 290 (2014): 85fps (i5) [Mac4Ever review (French)]
3. AMD 290X (2014): 90.87fps (i5) [Post #63]
4. AMD 295X (2014): 104fps (i7) [Mac4Ever review (French)]
5. AMD 380 (2015):
6. AMD 390 (2015): 95.3fps (i5) [Mac4Ever review (French)]
7. AMD 395 (2015): 94.66fps (i5) [Post #95]
8. AMD 395X (2015): 102.87fps (i7) [Post #44]

RESULTS: Windows/Direct X

Could someone explain why the 680MX beats the 395X on Unigine but gets smoked by the 395X in Cinebench? I'm guessing its an Open GL thing?

 

[SBruv]

Could someone explain why the 680MX beats the 395X on Unigine but gets smoked by the 395X in Cinebench? I'm guessing its an Open GL thing?

I, too, have been wondering about this. It gives me hope… yet makes me sad.

 

[mjohansen]

I, too, have been wondering about this. It gives me hope… yet makes me sad.

Could it have anything to do with driver support?

 

[vetruvian]

Could it have anything to do with driver support?

Could be. I hope someone can run them in bootcamp. Looks like OP has a Windows 10 section prepped.

 

[762999]

Could someone explain why the 680MX beats the 395X on Unigine but gets smoked by the 395X in Cinebench? I'm guessing its an Open GL thing?

it's been debated before, for me, Unigine is much closer to a gaming experience than Cinebench could be. For computational testing, I would rely more on a Blender test than Cinebench.

 

[TechZeke]

Could it have anything to do with driver support?

That would be my guess. AMD isn't known for the best driver support in-general, and I doubt these being "new" cards is helping especially since all these cards except for the M395X have only appeared in these iMacs.

 

[mjohansen]

Does Apple provide updated drivers together with updates to OSX or how does that usually work?

 

[TechZeke]

Could someone explain why the 680MX beats the 395X on Unigine but gets smoked by the 395X in Cinebench? I'm guessing its an Open GL thing?

It makes no sense, since in the thread that this thread spinned off from, the Unigine benchmark put the M395 at 15% below a 970M(both done by ninja2000?) at the same settings. It was even closer in actual gaming benchmarks on sniper elite 3 and CoH 2, with the M395 getting within 1-3 fps on both games. A 970M should smoke the 680MX, so I'd expect the M395X to be even faster the 970M and destroy the 680MX. Ninja2000's benchmarks were the most useful from that thread, since he owned both of the machines and tested at the same settings.

 

[Nirurin]

I would also like to see the 780m from 2013 added to the list (as it's the newest non-AMD card, plus the one I actually have haha). I have seen someone in post #4 already put up a benchmark to go on the list.

Besides, the 780m also had 4gb of ram, so it might be a bit more on-par with certain benchmarks.

I've just been considering selling mine, as it still has a year of applecare left, and upgrading to a new 5k imac... but I'm not sure it's going to be worth it.

 

[Alesc]

I would also like to see the 780m from 2013 added to the list (as it's the newest non-AMD card, plus the one I actually have haha). I have seen someone in post #4 already put up a benchmark to go on the list.

Besides, the 780m also had 4gb of ram, so it might be a bit more on-par with certain benchmarks.

I've just been considering selling mine, as it still has a year of applecare left, and upgrading to a new 5k imac... but I'm not sure it's going to be worth it.

Can you try to bench yours? I'm surprised by the results of mine...

 

[Nirurin]

Can you try to bench yours? I'm surprised by the results of mine...

Yeh probably if the programs are free, I've not actually tried before.

I'll try and get the benchmark things for OSX in a minute, and also do it on bootcamp tonight. I'll keep you updated

 

[SBruv]

It makes no sense, since in the thread that this thread spinned off from, the Unigine benchmark put the M395 at 15% below a 970M(both done by ninja2000?) at the same settings. It was even closer in actual gaming benchmarks on sniper elite 3 and CoH 2, with the M395 getting within 1-3 fps on both games. A 970M should smoke the 680MX, so I'd expect the M395X to be even faster the 970M and destroy the 680MX. Ninja2000's benchmarks were the most useful from that thread, since he owned both of the machines and tested at the same settings.

Well, going by the info gathered so far, it looks like the M395X doesn't destroy anything, sadly. My i7/395X is coming tomorrow, so I'll be running some actual gaming benchmarks in Windows. I fear it's basically going to be on par with the 680MX, if not slightly slower. :-/

 

[Fl0r!an]

Could someone explain why the 680MX beats the 395X on Unigine but gets smoked by the 395X in Cinebench? I'm guessing its an Open GL thing?

Cinebench OpenGL benchmark isn't meaningful at all since it's CPU limited on most machines.
OS X Nvidia drivers tend to have a higher CPU overhead which explains the huge performance drop.

Btw, Nvidia benchmarks should be done with latest web drivers, Kepler has received big performance gains in many applications.

 

[Nirurin]

Here is OSX Cinebench 15.

Imac late 2013, i7, 780m 4gb, with 24gb of ram. Done on OSX.

I ran the test 3 times, and got 77, 73 and then 75. So I stuck with the 75. I did have a bunch of browser windows open in the background still because I'm downloading Unigine still, so I dunno if that effected anything. I'll try it again later and repost if it is any different.

Still, the 780 is pretty much halfway between the 680 and the 295, which is where you'd expect it to be.