Most powerful graphics-card under Cinebench OGL?

[peterburner]

Hello guys,

I want to improve my main MacPro 5.1 (12x2.93 GHz, 24 GB RAM, 10.9.4, ATI Radeon 5870, OWC Mercury Accelsior PCIe SSD) with a stronger graphics-card in terms of a higher Cinebench R15 OpenGL score. My current ATI 5870 scores 62 fps and I didn't find any card with a higher OGL-score for my MP 5.1/ OS X.

I do know the issues about Cinebench. But I have to rely on it since I am working daily with Cinema4D. So that is fixed. ;-)

The EVGA GTX 680 Mac, the Sapphire 7950 and especially all the overpriced Quadros simply do not get a score higher than my aging ATI 5870 - 62 fps.
So I did a lot of research on this. But I don't get any reliable results on which graphics-card (Mac- or flashed/ unflashed PC-cards) get the most high OpenGL-score under Cinebench R15.

I contacted Rob from bare feats, he supplied me with the score of the GTX 680 (weak 53 fps). Via ebay I asked Macvidcards (for 780/ 680) and Create_Pro (for Titan Black) for the Cinebench scores of the cards. But I did not get any useful answer, instead my wallet standing by... ;-)

So, could you please supply any relevant OGL scores, perhaps for a 12x2.93 Mac Pro? Which card scores the best in Cinebench R15 OGL for my Mac Pro 5.1?

Thanks a lot guys!
peter

Edit: the 7950 gets 68 fps - but to my mind thats not worth the pricey change of the card...

 

[MacVidCards]

CINEBENCH IS A JOKE FOR TESTING GPUS

They know it, Rob at Barefeats will confirm

It is 100% pointless and reason we didn't send you results.

Test is CPU bound and gives no insight into GPU performance

You have to understand that when I can place 20 different GPUs in my X5680 Mac Pro and get same result I KNOW IT IS A JOKE

They know it is a joke, they just don't care

 

[peterburner]

I know Cinebench relates the GPU-power to the single-core power. But I also know a 5870 scores higher than a 5770 or a 4870 or GT120. Thats one point I tried to make clear in my initial post: I do not want to start a discussion about Cinebench.
I just need some plain simple scores (e.g. as provided by cbscores.com). Thats all...

Thank you
peter

 

[666sheep]

I just need some plain simple scores (e.g. as provided by cbscores.com). Thats all...

So, according to my testing of various GPU/CPU combos you should choose Radeon 6870 (Barts XT Prototype on attached screenshot)

 

[peterburner]

Hey man,
thanks a lot.
I suppose this 6870 is a flashed PC-card?

 

[MacVidCards]

Hey man,
thanks a lot.
I suppose this 6870 is a flashed PC-card?

Perhaps you don't get the humor.

A 6870 beats a 7950....THINK ABOUT IT FOR A MINUTE

Here are the 2 important choices:

1. C4D is so poorly written that a 6870 trumps a 7950 (notice the 3 smiley faces and think) (implies that C4D has no relevance use of GPU
2. Cinebench is so poorly written that it can't tell the difference between a 6870 and 7950. (implies that Cinebench GPU has no relevance on GPU performance in C4D)

I have an email that I can't post here that clarifies that coder was REALLY PROUD he got new animation to work on old ap. (newer BMW skin on older car)

Think about that.

 

[peterburner]

I am wondering why you as a professional supplier get so furious at a potential customer who's is just asking for some scores to measure with?
So please relax and breathe.

I have to find a compareable measurement for the OpenGL-Performance in Cinema4D. A reasonable alternative to CB OGL test could be the fps of the OpenGL-scene at fullscreen-playback. Both of them cold be useful.

Rob provided me with the following:
Playback at fullscreen 2560x1440 (on 12-Core Mac Pro):

GTX 680 = 26.97 FPS
Radeon HD 7950 = 24.26 FPS

 

[lexR]

Hi Pete,

i think the thing that strikes a cord with some of the long term members here is that they dont like software that measures graphic performance but actually be heavily weighted on CPU performance, maybe if you give a reason as to why you want the Cinebench test done over other tests e.g. heaven or others, then it will have more relevance to them?

I will carry out a test when i get back later.

----------

I have to find a compareable measurement for the OpenGL-Performance in Cinema4D. A reasonable alternative to CB OGL test could be the fps of the OpenGL-scene at fullscreen-playback. Both of them cold be useful.

Rob provided me with the following:
Playback at fullscreen 2560x1440 (on 12-Core Mac Pro):

GTX 680 = 26.97 FPS
Radeon HD 7950 = 24.26 FPS

Sorry glance read that before i posted - apologies

 

[peterburner]

Hey Lex,

as mentioned, as a Cinema4D-artist I have to find a measure directly related to Cinema4D.

In case a 6870 beats a 7950 in Cinebench it might be the case that it also beats the 7950 in OGL-use of Cinema4D. So no matter if thats logical or not it should give a prediction of the cards performance under Cinema4D as Cinebench uses the same core as Cinema4D.

peter

 

[666sheep]

C4D seems poorly written.
It still does rely on OGL 2.1 (see requirements on Maxon website), hence IMO all these irrational scores we are getting. This, and fact that Cinebench (so C4D must be as well) is software OpenGL, i.e. CPU bound. As you can see on screenshot I posted before, top clocked Hex core CPUs "score" better in OpenGL than lower clocked CPUs with lower core count, with the same graphics card of course.

 

[sirio76]

C4D seems poorly written.
It still does rely on OGL 2.1 (see requirements on Maxon website), hence IMO all these irrational scores we are getting. This, and fact that Cinebench (so C4D must be as well) is software OpenGL, i.e. CPU bound.

Sorry but this is totally wrong.
C4D do use the latest OpenGL version, you are probably just looking at minimum requirement.
Cinebench(like C4D) do use hardware OpenGL, software shading is also available just in case your GPU do not work correctly but you have to set this in the preference panel(of course you get slower viewport performance).
The score are not irrational at all, is very well explained in this post from a user who do use C4D since many years, just like me(and unlike many people who have never used C4D but feels free to write about the software) :
https://forums.macrumors.com/posts/19410395/

Some data for the OP, I don't know if you are just interested in the oMP performance, anyway.. I'm running a nMP/8core D700, the OpenGL result is about 90 on OSX and 110 on Windows.
In real life.. I've just completed a huge work(about 50millions real polygons + 200millions polygons via render instances), I was able to navigate in the viewport @50/100FPS with just the real polygons and about 20/30FPS with all the render instances activated. Since C4D relay also on CPU for the viewport performance you will never get good result because of the slower single thread performance of the oMP.

 

[peterburner]

Sorry but this is totally wrong.
C4D do use the latest OpenGL version, you are probably just looking at minimum requirement.
Cinebench(like C4D) do use hardware OpenGL, software shading is also available just in case your GPU do not work correctly but you have to set this in the preference panel(of course you get slower viewport performance).
The score are not irrational at all, is very well explained in this post from a user who do use C4D since many years, just like me(and unlike many people who have never used C4D but feels free to write about the software) :
https://forums.macrumors.com/posts/19410395/

Thanks, totally agree. Of course C4D uses hardware OGL (since release 6?) as it can be selected as sirio76 said in the programs preferences.
Well, I do know how Cinebench works and I do know that it is only intended to give a reference on the OGL-performance of CINEMA4D and nothing else on your system. And that is by definition what I asked for in my initial post. Nothing else. ;-)

 

[Asgorath]

Sorry but this is totally wrong.
C4D do use the latest OpenGL version, you are probably just looking at minimum requirement.

Edit: My mistake, I enabled the GL version and sure enough C4D claims it's using the 4.1 core profile.

For what it's worth, I get 70.3 FPS with a Core i7-3770 (equivalent to the Late 2012 iMac) and a GTX 680. With Yosemite and a Core i7-4770 plus GTX 780 Ti I get more like 95-100 FPS. The biggest factor definitely seems to be the CPU speed, and the classic Mac Pros just have a really hard time keeping up these days.

 

[MacVidCards]

First round goes to...7970

This is on 3,1 running maxed out 3.2 Dual Quads

Note that literally EVERY Nvidia card I tested got same score.

So...why buy a GTX780 when a GT640 will match it frame per frame?

From lowly GT640 to GTX780, all within 1 fps

I'll take a whack on 12 core with a couple later but not expecting miracles.

I stand by my 100% meaningless comment.

I had always considered it worthless but got excited when they brought out new version.

But it seems just a new coat of paint/lipstick on a pig.

So either C4D is poorly written or this benchmark doesn't reflect it well.

 

[MacVidCards]

Cinebench 15 on 2nd fastest cMP 12 Core 5680

And sadly all of the 680s and 780s get stomped on by......a Tesla 2070...ie, a down clocked GTX480

Trust me, either C4D or Cinebench is poorly written, no other way to see it.

 

[peterburner]

Thanks for testing.
As mysterious as it seems the Cinebench performance reflects the Cinema4D performance as both is based on the same architecture.

@ Asgorath: your second result with the 780 Ti sounds great, but your tested machines are hackintoshes, right?

For now it seems that working with Cinema4D I have to stick with my Radeon 5870s.

Again, thanks for testing!
peter

 

[sirio76]

Trust me, either C4D or Cinebench is poorly written, no other way to see it.

You keep not understanding how C4D(and many other 3D packages) works. Your CPU can not feed the GPU fast enough with the 3D data, that's why you will never get better results no matter what GPU you use. It's not poorly written, it's just a how thing works in real life. Maybe you should start using 3D programs and understand how they works instead of making false claime on software you never used, this aren't game and all the gaming benchmark you are usually interested mean nothing here.

 

[peterburner]

Cinema4D is everything else but poorly written, trust me.
Interesting thing is that on Windows a stronger card also means a better OGL-result, just look at http://cbscores.com/index.php?sort=ogl&order=desc. So to my mind the graphics-card-drivers for Mac are poorly written or Mac OS X is the bottle-neck. Feels familiar somehow... :-/

 

[sirio76]

Cinema4D is everything else but poorly written, trust me.
Interesting thing is that on Windows a stronger card also means a better OGL-result, just look at http://cbscores.com/index.php?sort=ogl&order=desc. So to my mind the graphics-card-drivers for Mac are poorly written or Mac OS X is the bottle-neck. Feels familiar somehow... :-/

This is not totally accurate. OpenGL performance under Windows are 10/20%(or even less) better than under OSX(in most case that's what you get when you compare the same GPU and the same CPU running at same clock speed). The thing is that OSX results on CBscores are mostly coming from old CPUs @stock speed(=very low single core performance), while many Windows results are coming from newer CPUs heavily overclocked(even if that database is reporting stock speed>because CB15 do not detect OC speed) so the CPU is not the bottleneck like under OSX. For apple to apple comparison you should overclock your CPU under OSX, but that's another story.

 

[peterburner]

Na come on... ;-)

A difference like
GTX 780, 47 fps on 12core 3.33 MacPro (posted by MacVidCards)
GTX 780, 81 fps on 8core 3.51 Win PC (cbscores.com)
doesn't come from over clocking...

"same GPU and the same CPU":
Just run Cinebench on your machine under Mac OS X and then under Boocamp - you get totally different results almost like shown above!

 

[N19h7m4r3]

Na come on... ;-)

A difference like
GTX 780, 47 fps on 12core 3.33 MacPro (posted by MacVidCards)
GTX 780, 81 fps on 8core 3.51 Win PC (cbscores.com)
doesn't come from over clocking...

"same GPU and the same CPU":
Just run Cinebench on your machine under Mac OS X and then under Boocamp - you get totally different results almost like shown above!

Heres my nMP with D700's. Note how the CPU score under OSX is higher than Windows, but the OpenGL under Windows is better.

 

[peterburner]

I like the OGL-Scores of the nMP.
But on the CPU-side I love my nearly 1400 points.
To reach them with the nMP a few thousand bucks are on the loose (and still its only 12core... ;-/ )

 

[sirio76]

Na come on... ;-)

A difference like
GTX 780, 47 fps on 12core 3.33 MacPro (posted by MacVidCards)
GTX 780, 81 fps on 8core 3.51 Win PC (cbscores.com)
doesn't come from over clocking...

Apple vs orange, you are comparing different clock speed and different CPU architecture, also the PC results is most likely obtained on an OC CPU. Different CPU architecture do matter, 3ghz from a 7years old 8core are not the same as 3ghz from a modern IB 8core
My GPU score about 90 under OSX and about 110 under Windows, ther's a difference, but not huge.. if my viewport is spinning @27FPS under OSX instead of 30 under Windows, well.. I can live with that The new 8core have decent single thread performance(about 140CB points), can not compete with an OC haswell desktop CPU but is one of the best performance you can get from a Xeon.

----------

I like the OGL-Scores of the nMP.
But on the CPU-side I love my nearly 1400 points.
To reach them with the nMP a few thousand bucks are on the loose (and still its only 12core... ;-/ )

For me the time of dual CPU and local render power are over:
http://forum.vrayforc4d.com/threads/cheap-rendernodes-for-dr.14214/page-2
Almost 5000CB points for 4K $

 

[kennyman]

I like the OGL-Scores of the nMP.
But on the CPU-side I love my nearly 1400 points.
To reach them with the nMP a few thousand bucks are on the loose (and still its only 12core... ;-/ )

Your cinebench score is not a bad score at all, I get only 63.93 fps and this is with a 280X. SO even if you upgrade to the latest card or max out the CPU, it would not change the score dramatically. I use Maxon Virtualize from time to time and I hate it on MAC, it runs better on Windows with the same hardware. As others pointed out, Cinebench is a joke...., I am not even sure whether it shows the OpenGL performance of your hardware correctly lol

 

[sirio76]

You guys do not want to listen/understand.. CB is not a joke... is just that the old macpro single thread performance is lower than the new one and much lower than OC wintel machine. This is not a synthetic benchmark, is just how your GPU/CPU behave under normal everyday job and FWIW C4D viewport is just as fast as Max or Maya.
I know it's sad that after spending thousand of dollars in CPU/GPU/SSD upgrades the old machine is still slow in single thread operation(the vast mayority in everyday job) but that is how the things are.. Sorry. Ther's no need to denigrate a software just because your hardware do not shine in some tasks, the old machine is still a good performer