GTX 680 (2 or 4MB VR) and X-Plane 10

[dpny]

That PC will be more powerful for gaming than the Mac Pro and it will be a lot cheaper.

I don't know what you need your Mac to do, but take a look at the top of the line Mac Mini. With an i7 it benches faster than the base Mac Pro. You could, conceivably, get a Mini, build a PC, take the 680 from your current Mac Pro and put it in the new PC and have a Mac and a gaming PC for less than the cost of a new Mac Pro.

Depending on what happens with Apple's pro hardware this year, I am considering something like that.

 

[MacsRgr8]

I don't expect a Mac Pro at WWDC but maybe Apple will give us a clue as to the direction they are going, or not.

Would be nice to find out if and what they'll be doing to the Mac Pro. It would be a first if Tim were to say anything at all, though.

Pretty hard to be an Apple lover and an X-Plane fan at the same time, eh..?

 

[oldmanmacdummi]

As promised a report after upgrading Mac Pro 4.1:
Changes:
ATI Radeon 4870 → GTX 680 2 GB
8 GB Ram → 24 GB (3x8)
x-plane on SSD

First impression after quicktest was a frustation.
I used the same settings as before and the FPS were slightly higher.
But after setting texture resulution to very high and later extreme
and HDR to 2xSSAA+FXAA I was very surprised that the framrate did not drop
with a much better picture.
It is also possible to set the CLAOD DETAIL to 100%.
Fazit: Not the FPS raised but the picture quality is much better.
I'am greatly satisfied.
oldidumm

 

[Topper]

My 2008 Mac Pro, GTX 680, & X-Plane

My GTX 680 in my 2008 Mac Pro is nice but I wish it would do better.
I keep thinking that my CPU is a bottleneck for X-Plane.
It’s the old story where the GPU is great but it is being hindered by an outdated CPU.

On the 10th, Barefeats threw an EVGA GeForce GTX 680 Mac Edition card into a 2010, 2009, and 2008 Mac Pro.
They had received reports that GPUs run slower in the 2008 Mac Pro as compared to the 2010 Mac Pro.
Barefeats benchtested the three computers with Motion, Cineman 4D, Resolve, and Inigine’s Valley.
In each case the GTX 680 ran much, much faster in the 2010 Mac Pro than it did in the 2008 Mac Pro.
In the Motion - Render Ram benchtest, the GTX 680 ran twice as fast in the 2010 Mac Pro than it did in the 2008 Mac Pro (36.9 fps compared to 17.9 fps).

I swear that one of these days I am going to have an outstanding computer with top-of-line processors, motherboard, and video card.

 

[MacsRgr8]

My GTX 680 in my 2008 Mac Pro is nice but I wish it would do better.
I keep thinking that my CPU is a bottleneck for X-Plane.
It’s the old story where the GPU is great but it is being hindered by an outdated CPU.

Yep, same goes for the 7950, be it less "dramatic".
The more I read about the 7950 (Sapphire Mac Edition), the more I am happy with it inside my Mac Pro 2008.

CPU, PCIe bus speed, RAM speed, etc. are the limiting factor on this ol' Mac Pro with these cards.
And, to be honest, I am happy with that. Finally we have grfx cards which put a 5 year old CPU to shame.

X-Plane 10 is definitely held back by other hardware limitations besides the 7950 or 680 (Mac editions) on a Mac Pro early 2008.

I am so, so, so hoping for a new Mac Pro. Hopefully we will know @ coming WWDC.

If not:
Latest,greatest and fattest iMac seems okay... EXECPT the 2 GB VRAM limit on the 680MX!!!

Maybe, I'll try to get a refurbished 3.33 GHz hex-core 2012 Mac Pro...

 

[Topper]

Latest,greatest and fattest iMac seems okay... EXECPT the 2 GB VRAM limit on the 680MX!!!
Maybe, I'll try to get a refurbished 3.33 GHz hex-core 2012 Mac Pro...

For your sake, I hope you get a new Mac Pro or a new xMac.
iMac: a desktop computer with a mobile gpu. Just shoot me now!

 

[MacsRgr8]

For your sake, I hope you get a new Mac Pro or a new xMac.
iMac: a desktop computer with a mobile gpu. Just shoot me now!

Thanks!

I would prefer an xMac regarding X-Plane 10: Better an i7 than a Xeon. Just as long as you are able to get the fattest AMD grfx card inside... (as I will be using OS X).

 

[dpny]

For your sake, I hope you get a new Mac Pro or a new xMac.
iMac: a desktop computer with a mobile gpu. Just shoot me now!

The 680 MX in the top of the line iMac benches just a little slower than the 7950. It's a very good card.

----------

Thanks!

I would prefer an xMac regarding X-Plane 10: Better an i7 than a Xeon. Just as long as you are able to get the fattest AMD grfx card inside... (as I will be using OS X).

The i7 is overkill for X-Plane 10: it won't make use of the extra threads. You're better off with a faster i5.

 

[Topper]

The 680 MX in the top of the line iMac benches just a little slower than the 7950. It's a very good card.

Yes, the 680MX iMac gets very respectable results.
I personally would never own one but to each their own.

The i7 is overkill for X-Plane 10: it won't make use of the extra threads. You're better off with a faster i5.

You're listening to the x-plane.org gurus. And I have great respect for AndyGoldstein. In almost all cases they are talking more about the best bang for your buck machine rather than the ultimate X-Plane computer.
In the same threads that you are referring to there are i7 users who believe that they are getting noticeably better results.
I don't know who is right but I am tired of people telling me that my GPU isn't powerful enough or that my CPU is hindering the performance of my GPU.
I'll be getting an OC'd i7-3970X computer with a GTX Titan. People will say I spent too much money but I can live with that.

 

[GermanyChris]

Yes, the 680MX iMac gets very respectable results.
I personally would never own one but to each their own.



You're listening to the x-plane.org gurus. And I have great respect for AndyGoldstein. In almost all cases they are talking more about the best bang for your buck machine rather than the ultimate X-Plane computer.
In the same threads that you are referring to there are i7 users who believe that they are getting noticeably better results.
I don't know who is right but I am tired of people telling me that my GPU isn't powerful enough or that my CPU is hindering the performance of my GPU.
I'll be getting an OC'd i7-3970X computer with a GTX Titan. People will say I spent too much money but I can live with that.

How many threads is key here. If xplane only uses 2-4 threads then there will be no additional benefit of the 6 core. There is a reason most gamers stick with the i5 and clock the heck out of it most games don't use 4 threads let alone 8 or 12 hyper threads..

related get the 3930k you can OC it the same and spend half the money.

 

[Topper]

related get the 3930k you can OC it same and spend half the money.

I've been told that before so I guess I had better listen (and save). Thank you.

 

[MacsRgr8]

Core i5 or i7, whatever.
Overkill is the best word out there.... I wish my hardware was overkill for X-Plane 10.

The GTX 680 MX in the iMac might be similar to a 7950, but:
a) only 2 GB VRAM (yes, it does matter in X-Plane 10)
b) bad OS X drivers.

 

[Topper]

Core i5 or i7, whatever.
Overkill is the best word out there.... I wish my hardware was overkill for X-Plane 10.

The GTX 680 MX in the iMac might be similar to a 7950, but:
a) only 2 GB VRAM (yes, it does matter in X-Plane 10)
b) bad OS X drivers.

I agree with you on every aspect.
I also agree with GermanyChris that the 3930k (which is a Core i7) can be nicely overclocked and is very powerful.

 

[Asgorath]

My GTX 680 in my 2008 Mac Pro is nice but I wish it would do better.
I keep thinking that my CPU is a bottleneck for X-Plane.
It’s the old story where the GPU is great but it is being hindered by an outdated CPU.

On the 10th, Barefeats threw an EVGA GeForce GTX 680 Mac Edition card into a 2010, 2009, and 2008 Mac Pro.
They had received reports that GPUs run slower in the 2008 Mac Pro as compared to the 2010 Mac Pro.
Barefeats benchtested the three computers with Motion, Cineman 4D, Resolve, and Inigine’s Valley.
In each case the GTX 680 ran much, much faster in the 2010 Mac Pro than it did in the 2008 Mac Pro.
In the Motion - Render Ram benchtest, the GTX 680 ran twice as fast in the 2010 Mac Pro than it did in the 2008 Mac Pro (36.9 fps compared to 17.9 fps).

I swear that one of these days I am going to have an outstanding computer with top-of-line processors, motherboard, and video card.

I was really glad to see this article, as it does a fantastic job of showing just how out-of-date the 2008 Mac Pro is compared with modern technology. The system itself is now 5 years old, and at the time of release the CPU technology was already a year or two old (since Xeons always lag behind the consumer CPU models). The Nehalem-based 2009 Mac Pro was a really big step up over the 2008 model, and folks that still have a 2008 or earlier Mac Pro have to realize that their CPUs simply cannot keep up with a modern high-end GPU like the GTX 680. It's also a shame there are still so many folks posting that the NVIDIA drivers suck, since all the recent Barefeats benchmarks suggest otherwise, but I guess that will never change.

 

[dpny]

It's also a shame there are still so many folks posting that the NVIDIA drivers suck, since all the recent Barefeats benchmarks suggest otherwise, but I guess that will never change.

Simply switching to Windows on Boot Camp, and thus using the Windows nVidia drivers instead of the OS X nVidia drivers--immediately doubles the fps on my machine. It's blatantly obvious to anyone who has done what I've done that OS X nVidia drivers are terrible when compared to the Windows drivers. Want instancing on your nVidia card? You need to be in Windows.

Once again, Barefeats X-Plane tests are horrible, as they test none of the things which kill frame rate in X-Plane. It's akin to testing gaming cards by measuring which one runs screen savers the best.

 

[Asgorath]

Simply switching to Windows on Boot Camp, and thus using the Windows nVidia drivers instead of the OS X nVidia drivers--immediately doubles the fps on my machine. It's blatantly obvious to anyone who has done what I've done that OS X nVidia drivers are terrible when compared to the Windows drivers. Want instancing on your nVidia card? You need to be in Windows.

Once again, Barefeats X-Plane tests are horrible, as they test none of the things which kill frame rate in X-Plane. It's akin to testing gaming cards by measuring which one runs screen savers the best.

Okay, so you're talking about one specific app: X-Plane. Right? The Barefeats tests for games and apps outside of X-Plane still stand, and suggest that the drivers don't completely suck in general. Just look at the scaling in Unigine Valley compared with the 7950 -- the AMD card basically gets no faster going from a MacPro3,1 to a MacPro5,1 while the NVIDIA card improves by 33%.

I'm not an X-Plane user and thus am not seeing the issues you're talking about, my basic point is that I think your assessment is a little unfair based on the performance of the NVIDIA driver in every app outside of X-Plane. I play at lot of WoW and SC2, and my GTX 680 absolutely flies at 2560x1600 with everything cranked up.

 

[MacVidCards]

Simply switching to Windows on Boot Camp, and thus using the Windows nVidia drivers instead of the OS X nVidia drivers--immediately doubles the fps on my machine.

Screenshots or it didn't happen.

I don't think Nvidia has done a bad job at all. And I have the feeling that Apple doesn't make it easy anyway. OSX doesn't support latest OpenGl, not Nvidia's fault, that's Apple.

 

[Asgorath]

And, as I've mentioned before in other threads, in my experience a lot of the massive (esp 2x) perf deltas come down to differences in the driver multithreading implementation.

Compare these posts:

http://store.steampowered.com/news/4211/
http://blogs.valvesoftware.com/linux/faster-zombies/

for example. Many games have a config switch to enable/disable the driver multithreading, and I can count on one hand the number of games I've seen a substantial improvement from enabling it on the Mac side, while it usually gives the expected ~2x perf improvement on Windows. YMMV.

 

[dpny]

Okay, so you're talking about one specific app: X-Plane. Right?

Seeing how this thread is specifically about X-Plane: yes. Didn't that that was hard to understand.

As I've said repeatedly, the X-Plane devs have confirmed the current OS X nVidia drivers do not support instancing under OS X. As X-Plane has an object budget of something like 25,000 per frame, the lack of instancing is the difference between a usable simulator and a slide show.

The relevant line from my Windows install of X-Plane:

This video card is: DX10 or 11 - With instancing (found glMapBufferRange)

The relevant line from my OS X install of X-Plane:

This video card is: DX10 - No instancing (found glMapBufferRange)

Same computer, different operating systems.

As above, the X-Plane devs are aware of this, and have been working with Apple/nVidia to get the driver up to snuff. However, there is no way to know if that will ever happen. Until it does, OS X is not a viable platform for X-Plane. I have no idea about any other games, as the only other game I play is Mass Effect 3, which runs fine in Parallels.

----------

Screenshots or it didn't happen.

Go and read the X-Plane.org hardware forums, or read the lead X-Plane graphic programmer's blog. He knows more about OpenGL programming than just about anyone in the biz and has a lot of posts about the differences between Windows and OS X drivers.

This issues has been done to death in the X-Plane world. There's lots of info out there for you to find.

 

[MacVidCards]

Go and read the X-Plane.org hardware forums, or read the lead X-Plane graphic programmer's blog. He knows more about OpenGL programming than just about anyone in the biz and has a lot of posts about the differences between Windows and OS X drivers.

This issues has been done to death in the X-Plane world. There's lots of info out there for you to find.

Since YOU were the one claiming that "Simply switching to Windows on Boot Camp, and thus using the Windows nVidia drivers instead of the OS X nVidia drivers--immediately doubles the fps on my machine." I was hoping that YOU could provide screenshots showing this. I just did a search and read the blog, didn't see anything about it running literally twice as fast in Windows.

This sounds a lot like the guy who puts the noisy oversized muffler on his car and claims it's faster because he can tell. Shouldn't be very hard to prove if you are so adamant that it is common knowledge. If you could take the time to post it here, I would be happy to forward to the Nvidia Mac team.

 

[dpny]

Since YOU were the one claiming that "Simply switching to Windows on Boot Camp, and thus using the Windows nVidia drivers instead of the OS X nVidia drivers--immediately doubles the fps on my machine." I was hoping that YOU could provide screenshots showing this. I just did a search and read the blog, didn't see anything about it running literally twice as fast in Windows.

X-Plane's built in fps tests. Same machine--3.33 GHz Xeon, 16 GB RAM, 4 GB SC 670--booted into OS X and Windows.

OS X tests:
Test 1: FRAMERATE TEST: time=92.9, frames=4752, fps=51.14 GPU LOAD: time=92.9, wait=4.1, load=4.4%

Test 2: FRAMERATE TEST: time=93.2, frames=3299, fps=35.40 GPU LOAD: time=93.2, wait=2.9, load=3.1%

Test 3: FRAMERATE TEST: time=96.1, frames=1991, fps=20.72 GPU LOAD: time=96.1, wait=1.7, load=1.8%

Test 4: FRAMERATE TEST: time=99.4, frames=1896, fps=19.08 GPU LOAD: time=99.4, wait=1.6, load=1.6%

Test 5: FRAMERATE TEST: time=92.8, frames=5658, fps=60.99 GPU LOAD: time=92.8, wait=5.3, load=5.7%

Windows tests: 

Test 1: FRAMERATE TEST: time=92.8, frames=7013, fps=75.54 GPU LOAD: time=92.8, wait=1.9, load=2.0%

Test 2: FRAMERATE TEST: time=93.1, frames=3950, fps=42.45 GPU LOAD: time=93.1, wait=1.0, load=1.0%

Test 4: FRAMERATE TEST: time=96.4, frames=2449, fps=25.40 GPU LOAD: time=96.4, wait=1.2, load=1.2%

Test 5: FRAMERATE TEST: time=92.8, frames=8826, fps=95.11 GPU LOAD: time=92.8, wait=1.5, load=1.6%

Unfortunately, test #3 wouldn't run in my Boot Camp install. As you can see, big increases, but not double. As this is X-Plane's built in test, it only tests performance with respect to things the sim can control: object density, view distance, weather, etc. The plane used is one of the default planes stripped down as much as possible, with an almost non-functional cockpit.

If I load up a third party plane with a complicated, 3D cockpit, like the Mig-29 or T-38, the difference between OS X and Windows becomes enormous. Those two planes are essentially unusable in OS X, as my frame rate will be around 15 fps just sitting in the plane on the tarmac. If I take off and get into some clear weather it might rise all the way to 19 or 20. Now, if I load the same plane on Windows, I will see 24 to 25 fps on the tarmac, and better than 30 fps while flying. That's the doubled frame rate I'm talking about.

I'm not pulling this stuff out of my ass. I've spent a year, and several hundred dollars, trying to find a way to make X-Plane run well on what should be a powerful Mac. I'm not alone: read the X-Plane.org forums and you'll see a lot of posts with people trying to find a way to make their Mac Pros work as well as Windows machines, and no success. No matter what I've done, it's all come down the same issue: there's no getting around the fact that nVidia's OS X drivers lag considerably behind their Windows drivers.

If you could take the time to post it here, I would be happy to forward to the Nvidia Mac team.

Go ahead, but they already know. The X-Plane devs have a good working relationship with both nVidia and AMD. There has been incremental improvement; the drivers for 10.8.3 were about 15% faster than those for 10.8.2. But it's not enough.

 

[oldmanmacdummi]

@dpny
Will you please tell us, which Mac OS you used for this test?
And did you change anything in the rendering settings?
An what is the reason for testing it only with a simple plane?

 

[Asgorath]

No matter what I've done, it's all come down the same issue: there's no getting around the fact that nVidia's OS X drivers lag considerably behind their Windows drivers.

The point that many people seem to forget is that Apple writes the OpenGL framework that sits between the app and the NVIDIA driver. On Windows, NVIDIA implements the entire OpenGL driver themselves. Perhaps some of these differences in performance are due to the different driver models between the two OSes?

Out of curiosity, what do the AMD results look like for Mac and Windows? Are you suggesting that AMD gets the same level of perf as NVIDIA's Windows driver in both cases?

 

[derbothaus]

Simply switching to Windows on Boot Camp, and thus using the Windows nVidia drivers instead of the OS X nVidia drivers--immediately doubles the fps on my machine. It's blatantly obvious to anyone who has done what I've done that OS X nVidia drivers are terrible when compared to the Windows drivers. Want instancing on your nVidia card? You need to be in Windows.

I get your logic but it isn't that simple. It isn't just Nvidia. It is more a problem of Apples OpenGL implementation than the drivers. And even more so the fact that these Mac games are usually lousy ports. My 5870 doubled in performance going from OpenGL to DirectX in Win7 just the same as Nvidia cards. 75FPF in TF2 and 150+ in Windows. It is simply put, Windows gaming vs OS X. Cards brands don't matter as much. A little. But not enough to blame one over the other. Blame the game community on OS X.

 

[Asgorath]

If I load up a third party plane with a complicated, 3D cockpit, like the Mig-29 or T-38, the difference between OS X and Windows becomes enormous.

The FPS test options are failing for me with this error:

0:00:-0.000 E/SYS: +-------------------------------------------------------------------------------
0:00:-0.000 E/SYS: | Trying to run a plane that is not default or net!
0:00:-0.000 E/SYS: |
0:00:-0.000 E/SYS: |
0:00:-0.000 E/SYS: |
0:00:-0.000 E/SYS: | (fm_start.cpp:184)
0:00:-0.000 E/SYS: +-------------------------------------------------------------------------------

so I went and bought the MiG-29 model to see what's going on. With default options and just sitting on the runway, I'm seeing 10.9 FPS with all the rendering options cranked up while running at 1024x768 (to ensure that the test is limited by the CPU/driver). If I enable HW instancing, that nearly doubles to 19.9. Note that this is on a 2.66GHz MacPro4,1 so not the fastest CPU around, and assuming that I'm reading the FPS counter correctly.

The really strange thing is that Apple's OpenGL Profiler tool doesn't even show glDrawElementsInstanced being called. The vast majority of time in both cases is in glDrawElements, and that's only accounting for less than 20% of the total app time in both cases.

So, in short, I'm having a hard time understanding how this is a "crappy NVIDIA driver" causing a perf difference between Mac OS X and Windows. The OpenGL Profiler is also showing that the app is not taking advantage of the driver multithreading like it would on Windows, so once again I suspect that the delta is due to the driver running in a separate thread on Windows and in the same thread as the app under OS X, which can result in a doubling of framerates on Windows in ideal circumstances.

The math also works out so that if I'm at 20 FPS on OS X with about 20-30% of time inside the driver, and you move that time to a separate thread, then the framerate should go up by 20-30% which results in 24-25 FPS. Of course, I'm not sure what the framerates are on my system under Windows and whether that math would be correct.