Mac pro 2013 gaming performance

[Beta Particle]

Here's dual D700's beating a Titan in the latest 3Dmark

The Titan has been surpassed by the GTX 780Ti now, and will be replaced with the Titan Black Edition next month. (essentially a 780Ti with 6GB RAM)

But that's two cards against one. Crossfire may produce high framerate numbers, but due to frame pacing issues, it is not a smooth experience.

 

[N19h7m4r3]

The Titan has been surpassed by the GTX 780Ti now, and will be replaced with the Titan Black Edition next month. (essentially a 780Ti with 6GB RAM)

But that's two cards against one. Crossfire may produce high framerate numbers, but due to frame pacing issues, it is not a smooth experience.

Entirely depends on the game, and drivers at the time. One cannot make a blanket statement like that.

I've run multi and single GPU setups, and my last Xfire experience was extremely pleasant. Further tests are needed, and at the moment MP's are in very short supply with people that can sit down and just analyse this for us.

Single and Dual also mean nothing, it's the end result that does. Over all performance.

 

[iEisberg]

http://www.barefeats.com/tube07.html

and

http://www.barefeats.com/tube06.html

Are they going to add a D700 comparison as well?

 

[Sinx2oic]

Are they going to add a D700 comparison as well?

Yep barefeats is waiting for the 8 core with D700's to test - said it's coming sometime in Feb

 

[energyboy]

I maxed everything out on a fully loaded Mac Pro and games run perfect on it.

You need to watch the first minute of this:

http://www.youtube.com/watch?v=JBN5MNpYWlY

 

[MikaMac]

For those interested in gaming via bootcamp with the nMP I may have some slightly disappointing news.

From my tests with a couple of games (12 core, D700), it does not appear that you can simply turn on CF in the driver and have all games benefit automatically. I just got the machine today and still plan to test a bunch of games, but with a couple I've tried so far I'm getting identical FPS with and without Crossfire enabled in the FirePro control center, which would indicate its not working in those games at least. Further, the FPS I'm getting (observed via Fraps) isn't spectacular, it's very close to a PC gaming rig I built a couple years back which is rocking a single GTX 670, leading further evidence that the results I'm getting are from a single D700.

I've seen other benchmarks though (diablo, wow) where people are clearly getting different results with/without CF enabled, so it may depend on the game. Next on my list is BF4 and Skyrim.

Edit: Tried it out, and BF4 is obviously attempting to use both cards - there's a clear FPS difference with it enabled vs. not when running through an in-engine cinematic (a fairly controlled environment), but it also doesn't run particularly great either way. Performance is a little choppy/eratic under crossfire, and nowhere near double performance. This doesn't surprise me as BF4 on my PC has had a number of issues that require bleeding edge drivers on consumer cards, custom tailored for the game. I'll try more games tomorrow.

 

[VirtualRain]

For those interested in gaming via bootcamp with the nMP I may have some slightly disappointing news.

From my tests with a couple of games (12 core, D700), it does not appear that you can simply turn on CF in the driver and have all games benefit automatically. I just got the machine today and still plan to test a bunch of games, but with a couple I've tried so far I'm getting identical FPS with and without Crossfire enabled in the FirePro control center, which would indicate its not working in those games at least. Further, the FPS I'm getting (observed via Fraps) isn't spectacular, it's very close to a PC gaming rig I built a couple years back which is rocking a single GTX 670, leading further evidence that the results I'm getting are from a single D700.

I've seen other benchmarks though (diablo, wow) where people are clearly getting different results with/without CF enabled, so it may depend on the game. Next on my list is BF4 and Skyrim.

Edit: Tried it out, and BF4 is obviously attempting to use both cards - there's a clear FPS difference with it enabled vs. not when running through an in-engine cinematic (a fairly controlled environment), but it also doesn't run particularly great either way. Performance is a little choppy/eratic under crossfire, and nowhere near double performance. This doesn't surprise me as BF4 on my PC has had a number of issues that require bleeding edge drivers on consumer cards, custom tailored for the game. I'll try more games tomorrow.

Which drivers are you using? Bootcamp Drivers supplied by Apple? Latest AMD Radeon Catalyst drivers? Catalyst FirePro drivers? It might be worth trying all of these, but at the very least, the latest AMD Radeon drivers.

 

[kfscoll]

There's an option in the Catalyst Control Panel to force CrossFireX even if there's not specific game support for it. If you're using the Boot Camp drivers you probably need to at least try AMD's drivers to make sure you can activate CrossFireX at all times.

Regardless, what you're seeing (CrossFireX only working some of the time) isn't an issue that's peculiar to the nMP; it's sometimes just the way SLI and CrossFire are. I bet once a couple more iterations of D700-specific drivers are released things will improve in this regard.

 

[Beta Particle]

Entirely depends on the game, and drivers at the time. One cannot make a blanket statement like that.

It has nothing to do with the game - Crossfire does not have proper frame pacing, as AMD did not even consider it to be an issue until very recently. It should be fixed on the latest GPUs (R9 290/X) but the ones in the Mac Pro are a generation out of date.

Driver support for Crossfire is game dependent, and AMD is often late in releasing driver updates in time for the latest games, when Nvidia almost always has beta drivers out before the games launch.

Single and Dual also mean nothing, it's the end result that does. Over all performance.

No, having the same synthetic performance on a single card, is far better than having it on dual GPUs. (or worse, triple or quad GPUs)

There is additional latency as you add cards, and additional problems that introduce stuttering even when the reported framerate is high, due to frame pacing issues, and not all games/applications can take advantage of dual GPUs, so that halves your performance right there. (assuming the game scales perfectly - which most do not)

 

[Sinx2oic]

For those interested in gaming via bootcamp with the nMP I may have some slightly disappointing news.

From my tests with a couple of games (12 core, D700), it does not appear that you can simply turn on CF in the driver and have all games benefit automatically. I just got the machine today and still plan to test a bunch of games, but with a couple I've tried so far I'm getting identical FPS with and without Crossfire enabled in the FirePro control center, which would indicate its not working in those games at least. Further, the FPS I'm getting (observed via Fraps) isn't spectacular, it's very close to a PC gaming rig I built a couple years back which is rocking a single GTX 670, leading further evidence that the results I'm getting are from a single D700.

I've seen other benchmarks though (diablo, wow) where people are clearly getting different results with/without CF enabled, so it may depend on the game. Next on my list is BF4 and Skyrim.

Edit: Tried it out, and BF4 is obviously attempting to use both cards - there's a clear FPS difference with it enabled vs. not when running through an in-engine cinematic (a fairly controlled environment), but it also doesn't run particularly great either way. Performance is a little choppy/eratic under crossfire, and nowhere near double performance. This doesn't surprise me as BF4 on my PC has had a number of issues that require bleeding edge drivers on consumer cards, custom tailored for the game. I'll try more games tomorrow.

Thanks for testing games, will be interesting to see how different drivers effect it. Are you running 2560x1440? I know the D700's won't be cutting edge but still hoping they might beat my 680.

 

[MikaMac]

Which drivers are you using? Bootcamp Drivers supplied by Apple? Latest AMD Radeon Catalyst drivers? Catalyst FirePro drivers? It might be worth trying all of these, but at the very least, the latest AMD Radeon drivers.

The boot camp drivers supplied by apple. I'll certainly play around with other options.

 

[MikaMac]

Hey, some here's some positive news from testing BF4 some more.

In the apple-supplied "FirePro Control Center" you can set application specific options, including manually choosing an application specific profile for managing crossfire. The list of apps with custom profiles is pretty large and covers most games, however it didn't have a profile for BF4, which might have explained why it ran kinda poorly.

However, there IS a profile for Battlefield 3. On a whim, I forced my BF4 exe to use the AMD-supplied BF3 profile, fully expecting it to not work at all. Surprise surprise, it worked perfectly! In the in-engine cinematic I'd been testing I saw smooth, stutter-free performance at roughly double the FPS I'd seen previously, which for those who like numbers was about 80 FPS on ultra, 1900x1200 (the native res of my display, or I'd test higher).

 

[antonis]

Hey, some here's some positive news from testing BF4 some more.

In the apple-supplied "FirePro Control Center" you can set application specific options, including manually choosing an application specific profile for managing crossfire. The list of apps with custom profiles is pretty large and covers most games, however it didn't have a profile for BF4, which might have explained why it ran kinda poorly.

However, there IS a profile for Battlefield 3. On a whim, I forced my BF4 exe to use the AMD-supplied BF3 profile, fully expecting it to not work at all. Surprise surprise, it worked perfectly! In the in-engine cinematic I'd been testing I saw smooth, stutter-free performance at roughly double the FPS I'd seen previously, which for those who like numbers was about 80 FPS on ultra, 1900x1200 (the native res of my display, or I'd test higher).

Yup, now you're talking. That's a very nice performance.

 

[Anim]

On a whim, I forced my BF4 exe to use the AMD-supplied BF3 profile, fully expecting it to not work at all. Surprise surprise, it worked perfectly! In the in-engine cinematic I'd been testing I saw smooth, stutter-free performance at roughly double the FPS I'd seen previously, which for those who like numbers was about 80 FPS on ultra, 1900x1200 (the native res of my display, or I'd test higher).

Great find! Thanks for sharing that tweak.

 

[Beta Particle]

I saw smooth, stutter-free performance at roughly double the FPS I'd seen previously, which for those who like numbers was about 80 FPS on ultra, 1900x1200 (the native res of my display, or I'd test higher).

Unless you're still using a CRT, no monitor can sync to 80fps, so it cannot have been "smooth, stutter-free performance".

 

[Anim]

Unless you're still using a CRT, no monitor can sync to 80fps, so it cannot have been "smooth, stutter-free performance".

You don't need VSync on for smooth frame locked performance. I have a 144Hz gaming monitor and have never seen screen tear resulting from VSync although I have never gone above 144fps.

 

[MikaMac]

I usually find leaving v-sync off and allowing higher-than-monitor-refresh fps produces more pleasing results to my eye than leaving vsync on and getting frequent dips into the 50's as a result. Depends on the game though. All I can vouch for is that 60+ looked pretty good and had little noticeable tearing to my eyes in BF4. I could always turn on vsync and see if that plays havoc with crossfire though.

(edit: just tried with vsync on, very nice results, pinned at 60fps with occasional flutters below that)

 

[Irishman]

I maxed everything out on a fully loaded Mac Pro and games run perfect on it.

Good to know that perfection does have a price tag - $10,000!

 

[iBug2]

Good to know that perfection does have a price tag - $10,000!

The games would run even better on the 4 core model with D700, which costs 4000$.

 

[Sinx2oic]

The games would run even better on the 4 core model with D700, which costs 4000$.

I read the 8 core would be fastest as it's got the fasted turbo over two cores which most games utilise more than four cores.

 

[MikaMac]

The games would run even better on the 4 core model with D700, which costs 4000$.

I wouldn't expect to see much of a difference regardless of CPU. 3D games aren't typically effected by the cpu in any appreciable way unless the cpu is so bad as to be a bottleneck.

 

[Beta Particle]

You don't need VSync on for smooth frame locked performance. I have a 144Hz gaming monitor and have never seen screen tear resulting from VSync although I have never gone above 144fps.

The only displays which do not, are Nvidia's new G-Sync displays. (which require Nvidia GPUs)
Anything else will either suffer from stuttering or tearing if the framerate is not locked to the refresh rate.

You cannot present 80fps at 144Hz evenly, thus it will introduce judder.
If that display supports 72Hz, you could sync to that for a smooth gameplay experience though - assuming it does not drop below 72fps.
V-Sync would also allow things to run at 72fps when the display is operating at 144Hz, but repeating frames like this will cause the image to judder.

The games would run even better on the 4 core model with D700, which costs 4000$.

And better still on a Haswell system with a 780Ti (or a pair of them) for a lot less than $4000...

I usually find leaving v-sync off and allowing higher-than-monitor-refresh fps produces more pleasing results to my eye than leaving vsync on and getting frequent dips into the 50's as a result.

V-Sync won't let the game drop to 50fps - it will drop from 60 to 30. (assuming a 60Hz display)
Triple buffering would allow for that to happen though.

But if you're frequently dropping below 60, your system is not fast enough for the game/settings.
You need faster hardware, or to reduce the graphics quality.

Depends on the game though. All I can vouch for is that 60+ looked pretty good and had little noticeable tearing to my eyes in BF4.

If you're running at a variable framerate, you can't make the claim that the gameplay is smooth and stutter-free. Anything which does not match your refresh rate stutters.

Nvidia's G-Sync fixes this by using a continuously variable refresh rate, that is controlled by the framerate.

 

[evilpaddy]

I read the 8 core would be fastest as it's got the fasted turbo over two cores which most games utilise more than four cores.

Waiting on my 8c D700. Happy to feedback when it arrives...

 

[Anim]

The only displays which do not, are Nvidia's new G-Sync displays. (which require Nvidia GPUs)
Anything else will either suffer from stuttering or tearing if the framerate is not locked to the refresh rate.

You might get some slow down in FPS (e.g. half your refresh rate) with VSync on and a too demanding graphic quality. What is more common is frame tearing when your FPS exceed your refresh rate.

So I guess I must be in the sweet spot on my system with 144Hz I would drop to 72fps if I have VSync on and if I can't quite get to 144fps. Which I never have turned on but 72fps would not cause a visual loss in smoothness either. Maybe it drops to 120fps because the monitor also supports 120Hz?

On the flip side, If I exceed 144fps that would introduce tearing but at 144fps you would hardly see it happen, it's just too fast a frame rate and the exact experience I have, I don't get frame halving or tearing on my system. And stuttering would be totally unacceptable for any game I play, I simply wouldn't play it.

Anim

 

[Derpage]

Entirely depends on the game, and drivers at the time. One cannot make a blanket statement like that.

I've run multi and single GPU setups, and my last Xfire experience was extremely pleasant. Further tests are needed, and at the moment MP's are in very short supply with people that can sit down and just analyse this for us.

Single and Dual also mean nothing, it's the end result that does. Over all performance.

It has little to do with drivers or the game and more so the monitor being used with the gpu.