PC Geforce GTX 960 4GB running in Mac Pro 3,1 (2008) - benchmarks for comparison

[lewis_cooper]

I'd like to see if I have bought the best card for my budget (≤ £200) for my 2008 Mac Pro 3,1.

I've installed and am successfully running the 4GB version of the Gigabyte Nvidia Geforce GTX 960 which is an unflashed PC card so no boot screen.

I will admit that I bought the card on the assumption that 4GB memory must be better than 2GB. I am aware that this model of mac is limited to PCIe 2.0 and that it may potentially throttle the performance of any card I install. The GTX 960 is definitely an improvement over my previously installed 5770, though I have to say the improvement was somewhat underwhelming.

What I would like to do is benchmark my card and then compare the same values to other cards installed in the same model Mac Pro. So far I have not found any results from the setup I have.

I want to know if a card with a Mac EFI (whether genuine or flashed) such as the the GTX 680 cards on eBay would give me better results than an unflashed PC card despite only have 2-3GB memory.

If you can point me in the right direction(s) I'd be very grateful.

 

[ActionableMango]

I doubt you'll find any benchmark comparisons on a 3,1 with a wide range of older and newer video cards. It's too obscure of a setup.

Flashing will not noticeably improve performance.

One more thing. I might be remembering this incorrectly, but I thought there was a "slot 1 bug" in the 3,1 that slows down performance of certain video cards. I cannot remember the details, but you might want to try slot 2 anyway.

 

[Fl0r!an]

I could throw in a R9 280 or a GTX 770 in my cMP 3,1. Just tell me what you're interested in, I can run some benchmarks.

I already made some benchmarks in this machine a while back. Have a look for my posts from last year.

 

[devon807]

I could throw in a R9 280 or a GTX 770 in my cMP 3,1. Just tell me what you're interested in, I can run some benchmarks.

I already made some benchmarks in this machine a while back. Have a look for my posts from last year.

[doublepost=1464951733][/doublepost]I have a MSI Gaming 4G GTX 960

 

[gpzjock]

Try Heaven 4.0 or Valley 1.0 benchmarks, they may give you a performance metric of sorts.

 

[pastrychef]

Benchmark results from a hackintosh would only help to confuse him more.

 

[SoyCapitanSoyCapitan]

Yeah never put Hackintosh Unigene benchmarks against cMP. The CPU assists a lot in those tests.

GPU bound tests, cMP versus like similar cMP. That's the best way to isolate GPU performance in Macs.

 

[gpzjock]

The Hack benches are for example purposes only. If anyone with a cMP would do the same, the OP should see some suitable comparisons. My MP 3,1 doesn't want to run Heaven 4.0 or Valley 1.0 for some reason, possibly Snow Leopard.

 

[lowendlinux]

To be fair to gpz though I don't think a 3,1 is going to throttle a 960 it's not that fast of a card.

 

[h9826790]

To be fair to gpz though I don't think a 3,1 is going to throttle a 960 it's not that fast of a card.

I think it's very clear that's depends on software. At least CineBench is a well known software that CPU limiting. So, the 960 can be throttled in 3,1, just how bad the situation is.

 

[gpzjock]

I was rather hoping Fl0r!an or Devon would step up with some Heaven or Valley runs, guess not yet.

 

[pastrychef]

Here are some examples of how a 5,1 with X5680 CPUs bottleneck a GTX 980:

 

[gpzjock]

I have a MP 3,1 bench but it is Heaven 3.0 not current; you would need the earlier version to compare:

Pastry, do you have a Windows build on the same Mac Pro, to show the difference between OpenGL in Mac OS X and DirectX drivers in Windows?

 

[lowendlinux]

There's my passive 960 2GB paired with an i3 it should be close to a 3,1

 

[pastrychef]

Here are results from Windows 10 OpenGL. Same machine I used for above benchmarks.

 

[666sheep]

View attachment 634572 There's my passive 960 2GB paired with an i3 it should be close to a 3,1

Won't be close. 3,1 has about a half of single threaded performance of i3 4170 and this is what affects overall score in Heaven/Valley.
I'm getting slightly better results with 5677 and GTX 780 than pastrychef with his 5680s and GTX 980. Same OS and same presets in above benchmarks.

If something other than 3,1 is comparable, it would be 2,1 with 5365s or 4,1 with E5520s.

 

[ActionableMango]

I use 3440x1440 which is about 2.4 times as many pixels as 1080p. I saw big increases in Unigine benchmarks when upgrading GPUs. First when moving from an HD5770 > GTX680, and again moving from GTX680 > GTX980. (I always do A/B hardware benches in both Windows and OS X to make sure I'm actually seeing expected performance increases.)

So I suspect that the cMP might be CPU bound at lower resolutions, but not higher ones. I don't really understand how that could be, but it seems to be the case. I can say that in my case, 3440x1440 with a W3690, I am still seeing big increases with GPU upgrades up to the GTX980. I know the CPU is old though, so I don't expect that this will continue much longer if at all.

Unfortunately I do not keep older hardware around so I cannot replicate the results again to post here. I can only run a test with what I currently have.

 

[Synchro3]

So I suspect that the cMP might be CPU bound at lower resolutions, but not higher ones. I don't really understand how that could be, but it seems to be the case. I can say that in my case, 3440x1440 with a W3690, I am still seeing big increases with GPU upgrades up to the GTX980. I know the CPU is old though, so I don't expect that this will continue much longer if at all.

Unfortunately I do not keep older hardware around so I cannot replicate the results again to post here. I can only run a test with what I currently have.

Read this: https://forums.macrumors.com/threads/nvidia-gtx-1080.1971528/page-5#post-22931463

 

[Asgorath]

So I suspect that the cMP might be CPU bound at lower resolutions, but not higher ones. I don't really understand how that could be, but it seems to be the case. I can say that in my case, 3440x1440 with a W3690, I am still seeing big increases with GPU upgrades up to the GTX980. I know the CPU is old though, so I don't expect that this will continue much longer if at all.

At higher resolutions, the GPU has enough work to do so it never goes idle while the CPU sits around waiting for the GPU every frame. Or, in other words, the CPU is fast enough to give the GPU enough work each frame so that the GPU is the bottleneck.

At lower resolutions, the GPU finishes rendering a frame before the CPU has finished sending the commands for the next frame, so the GPU goes idle. Or, in other words, the CPU isn't fast enough and it becomes the bottleneck.

There's always a cross-over point where the GPU starts to matter, which will depend on how powerful your CPU and GPU are, as well as how many pixels you are rendering. Unfortunately, it's very easy to end up being CPU limited on Mac and thus your GPU won't improve benchmark scores.

 

[gpzjock]

"Here are results from Windows 10 OpenGL. Same machine I used for above benchmarks."


Thanks Pastry, that shows a real difference a driver makes. About 20 average fps gained from Windows OpenGL drivers alone. Can you do DirectX ones too please?

 

[Fl0r!an]

Thanks Pastry, that shows a real difference a driver makes. About 20 average fps gained from Windows OpenGL drivers alone. Can you do DirectX ones too please?

Different machine with different GPU, but I uploaded Heaven benchmarks (El Capitan OpenGL vs. Win 10 DirectX) in this thread yesterday: https://forums.macrumors.com/threads/the-ultra-settings-thread.1974637/#post-22983396

Both benchmarks were done in Ultra@1440p. Difference is a measurable 12% (=3FPS in my case), nothing I'd care about.

With Extreme preset it's even less than that: 60 FPS vs. 63.8 FPS (~ 6%)

I'll dig out some 3,1 benchmarks later!

 

[gpzjock]

Thanks Florian!
This made interesting reading: http://barefeats.com/winvosx.html

 

[lewis_cooper]

Just thought I'd revisit this and add my benchmarks since I actually went ahead and bought the unflashed GTX 960 4GB. Slight plot twist though I've also changed up the mac pro to a 6-core 3.33ghz 2010 model (5,1 flashed from 4,1).

What I'm trying to decide now is, is it worth upgrading the GPU to a flashed GTX 980 4GB.

The apps I'm hoping to see gains in are Adobe Premiere, Lightroom and Photoshop. What do you think?