Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
Status
The first post of this thread is a WikiPost and can be edited by anyone with the appropiate permissions. Your edits will be public.
Status
Not open for further replies.
Isn't it easier then to select the Macos drivers in the Nvidia Driver Manager before updating and reselect with screensharing the NVidia Web driver after the system update and rebooting?
I tried that, but my Mac didn't recognize the graphics card with the stock drivers. It only recognized it as "Nvidia Chip Model" with the stock drivers.
 
I tried that, but my Mac didn't recognize the graphics card with the stock drivers. It only recognized it as "Nvidia Chip Model" with the stock drivers.
Isn't that the whole point of the exercise? You then install the latest web driver after updating OS X and enable it again
 
Isn't that the whole point of the exercise? You then install the latest web driver after updating OS X and enable it again
Well, I’ve noticed that a lot of folks will try each new web driver which comes along, hoping that the new one will be better than the last. I was hoping that I could just use the stock driver and avoid that endless search for the best driver.
 
Well, I’ve noticed that a lot of folks will try each new web driver which comes along, hoping that the new one will be better than the last. I was hoping that I could just use the stock driver and avoid that endless search for the best driver.
When and if the stock drivers support the maxwell cards there will be no way not to notice :)
 
I recently got an eGPU. I noticed that Minecraft seems to only use my CPU even though I have the eGPU plugged in. Does anyone here know of a program which will allow me to force Minecraft to use the dedicated eGPU instead of just using the processor? Would GFXCardStatus suffice?
[doublepost=1452709027][/doublepost]
When and if the stock drivers support the maxwell cards there will be no way not to notice :)
I read somewhere that the GT 740 is just a rebranded Kepler card.
 
740 = 640

In fact I found a way to mod a 740 to use a 640 rom.

eGPU typically only works when displayed is plugged in to it for games or OpenGL apps in OSX

For GPGPU the display doesn't need to be connected.

In Win there is frequently a way to use the eGPU to accelerate internal screen.
 
to back further i guess what MVC is saying i Have DDR3 based GT 740 in my Mac pro 1,1 that will auto int even in Lion on stock drivers (OS X does not know what to call it but full acceleration is there heh) the GT 740 is a nifty card tho plays GTA 5 well on my Mac pro and is apparently about as fast as an ATI 5770 (though I do have one of those coming in the post since I need boot screens on my 1,1) if anyone is wondering its a EVGA single slot super clocked GT 740 looks a lot like the GT 640 MVC has on his site. the card will work with stock nvidia drivers fine but the name is Nvidia chip model other wise it works fine (but i installed web drivers since they perform better and no harm in doing so :) )


upload_2016-1-14_9-12-22.png
 
  • Like
Reactions: bogg
I can say that my one (GK107 DDR3) runs fine in OS X on both stock and web drivers (this card was originally got for a hackintosh i was running then to upgrade from a Geforce 210 :) )
 
740 = 640

In fact I found a way to mod a 740 to use a 640 rom.

eGPU typically only works when displayed is plugged in to it for games or OpenGL apps in OSX

For GPGPU the display doesn't need to be connected.

In Win there is frequently a way to use the eGPU to accelerate internal screen.
I always have my monitor connected to my eGPU. I have my MacBook connected to the monitor via DVI and I have the eGPU connected to the same monitor via VGA so I can use the same monitor and get slightly better video quality most of the time (I only use the eGPU for games).
[doublepost=1452790715][/doublepost]
I can say that my one (GK107 DDR3) runs fine in OS X on both stock and web drivers (this card was originally got for a hackintosh i was running then to upgrade from a Geforce 210 :) )
That’s interesting. Which brand is it?
 
Luxmark usually can accurately reflect the GPU's performance in OpenCL, not OpenGL, in this area, AMD card always stronger than Nvidia in OSX (that's why using the AMD card for something like FCPX is a better choice).
Is this still true? Nvidia has made a large improvement with the GTX 9xx series and OpenCL. If you take a look at LuxMarks top list, you see a lot of 980 (TI) and 970 in the Top 100 test results. My 970 is about 5-10 times faster than my two year older 660 TI (on OS X).
 
Is this still true? Nvidia has made a large improvement with the GTX 9xx series and OpenCL. If you take a look at LuxMarks top list, you see a lot of 980 (TI) and 970 in the Top 100 test results. My 970 is about 5-10 times faster than my two year older 660 TI (on OS X).

I assume it's still true, I didn't have time to go through the Top 100 Luxmark 3.0 results.

However, on the first page (top 20 for multiple CPU + GPU), only 2 entries are from OSX. Both of them are by AMD GPUs. Even though it's very clear that they are Hackintosh, but they are running OSX.

And If I further limit the result to single GPU, the result looks like this.
Screen Shot 2016-01-15 at 02.11.19.jpg


I didn't check which OS for each entry, but I think it's quite safe to assume AMD still has advantage on OpenCL.
 
Last edited:
Here's another OpenCL Benchmark application, from heise:
http://www.heise.de/mac-and-i/downloads/65/1/1/8/0/8/7/8/Mac___i_OpenCL_Benchmark.app.zip

What are your results when you run the tests: File -> "Run OCL Bench..."?
Are there nMP users here to check it?

Here are results from an i7 3770k Hackintosh with GTX 970:

GeForce GTX 970, VW 1
4096 x 4096 pixel in 74.69 ms
224.63 megapixel per second
943.45 GFLOPS

GeForce GTX 970, VW 4
4096 x 4096 pixel in 100.37 ms
167.16 megapixel per second
702.06 GFLOPS

GeForce GTX 970, VW 8
4096 x 4096 pixel in 138.89 ms
120.79 megapixel per second
507.32 GFLOPS

Intel(R) Core(TM) i7-3770K CPU @ 3.50GHz, VW 1
4096 x 4096 pixel in 3478.91 ms
4.82 megapixel per second
20.25 GFLOPS

Intel(R) Core(TM) i7-3770K CPU @ 3.50GHz, VW 4
4096 x 4096 pixel in 905.35 ms
18.53 megapixel per second
77.83 GFLOPS

Intel(R) Core(TM) i7-3770K CPU @ 3.50GHz, VW 8
4096 x 4096 pixel in 668.39 ms
25.10 megapixel per second
105.42 GFLOPS

All OpenCL devices together, VW (
1,
8
)
4096 x 4096 pixel in 83.50 ms
200.92 megapixel per second
843.85 GFLOPS
 
Sure your 970 can do better job in this OpenCL test then my 7950. But I realise that this test may be very PCIe bandwidth sensitive. I have 2x 7950, the 1st one installed in the normal x16 slot (slot 1), and the 2nd one installed in the x4 slot (slot 3, due to cooling reason). In other test (Luxmark, Unigine, etc), both card's performance are within 5%. But in this test, the difference is huge (>30%).

And their performance is quite job dependent. e.g. the 970 is doing better job in the VW 1, but fall behind in VW 8.


AMD Radeon HD 7950 Compute Engine, VW 1
4096 x 4096 pixel in 87.55 ms
191.62 megapixel per second
804.80 GFLOPS

AMD Radeon HD 7950 Compute Engine, VW 4
4096 x 4096 pixel in 112.70 ms
148.87 megapixel per second
625.23 GFLOPS

AMD Radeon HD 7950 Compute Engine, VW 8
4096 x 4096 pixel in 118.72 ms
141.31 megapixel per second
593.52 GFLOPS

AMD Radeon HD 7950 Compute Engine, VW 1
4096 x 4096 pixel in 117.78 ms
142.45 megapixel per second
598.28 GFLOPS

AMD Radeon HD 7950 Compute Engine, VW 4
4096 x 4096 pixel in 142.92 ms
117.39 megapixel per second
493.04 GFLOPS

AMD Radeon HD 7950 Compute Engine, VW 8
4096 x 4096 pixel in 148.98 ms
112.61 megapixel per second
472.96 GFLOPS

Intel(R) Xeon(R) CPU W3690 @ 3.47GHz, VW 1
4096 x 4096 pixel in 2638.43 ms
6.36 megapixel per second
26.71 GFLOPS

Intel(R) Xeon(R) CPU W3690 @ 3.47GHz, VW 4
4096 x 4096 pixel in 728.49 ms
23.03 megapixel per second
96.73 GFLOPS

Intel(R) Xeon(R) CPU W3690 @ 3.47GHz, VW 8
4096 x 4096 pixel in 704.27 ms
23.82 megapixel per second
100.05 GFLOPS

All OpenCL devices together, VW (
1,
4,
8
)
4096 x 4096 pixel in 57.00 ms
294.34 megapixel per second
1236.22 GFLOPS

done.
 
Last edited:
Mac & i OpenCL Benchmark 1.0 (1)
2016-01-16 03:19:19 +0000
MacPro5,1
(
{
deviceAdressBits = 32;
deviceID = "<00270201 00000000>";
deviceName = "GeForce GTX 980";
deviceType = 4;
deviceVersion = "OpenCL 1.2 ";
globalMemSize = 4294967296;
localMemSize = 49152;
maxClockFrequency = 1215;
maxComputeUnits = 16;
maxWorkGroupSize = 1024;
preferedVectorWidthFloat = 1;
},
{
deviceAdressBits = 64;
deviceID = "<ffffffff 00000000>";
deviceName = "Intel(R) Xeon(R) CPU X5680 @ 3.33GHz";
deviceType = 2;
deviceVersion = "OpenCL 1.2 ";
globalMemSize = 68719476736;
localMemSize = 32768;
maxClockFrequency = 3330;
maxComputeUnits = 24;
maxWorkGroupSize = 1024;
preferedVectorWidthFloat = 4;
}
)

GeForce GTX 980, VW 1
4096 x 4096 pixel in 64.73 ms
259.20 megapixel per second
1088.66 GFLOPS

GeForce GTX 980, VW 4
4096 x 4096 pixel in 90.15 ms
186.11 megapixel per second
781.66 GFLOPS

GeForce GTX 980, VW 8
4096 x 4096 pixel in 122.71 ms
136.72 megapixel per second
574.23 GFLOPS

Intel(R) Xeon(R) CPU X5680 @ 3.33GHz, VW 1
4096 x 4096 pixel in 1388.03 ms
12.09 megapixel per second
50.77 GFLOPS

Intel(R) Xeon(R) CPU X5680 @ 3.33GHz, VW 4
4096 x 4096 pixel in 383.34 ms
43.77 megapixel per second
183.81 GFLOPS

Intel(R) Xeon(R) CPU X5680 @ 3.33GHz, VW 8
4096 x 4096 pixel in 363.24 ms
46.19 megapixel per second
193.99 GFLOPS

All OpenCL devices together, VW (
1,
8
)
4096 x 4096 pixel in 63.42 ms
264.54 megapixel per second
1111.05 GFLOPS

done.
 
anyone know if there is there any love for El Capitan and expresscard egpu using PE4L v2.1 and setup 1.3?
I don’t know, but ExpressCard isn’t the best option if your Mac has Thunderbolt due to bandwidth, and since it costs the same to go the ExpressCard->Thunderbolt route since the only ExpressCard to Thunderbolt adapter costs $150.
 
Mac & i OpenCL Benchmark 1.0 (1) Radeon 5870
MacPro5,1


ATI Radeon HD 5870, VW 1
4096 x 4096 pixel in 208.61 ms
80.42 megapixel per second
337.78 GFLOPS

ATI Radeon HD 5870, VW 4
4096 x 4096 pixel in 176.99 ms
94.79 megapixel per second
398.13 GFLOPS

ATI Radeon HD 5870, VW 8
4096 x 4096 pixel in 166.21 ms
100.94 megapixel per second
423.94 GFLOPS

Intel(R) Xeon(R) CPU W3680 @ 3.33GHz, VW 1
4096 x 4096 pixel in 2804.89 ms
5.98 megapixel per second
25.12 GFLOPS

Intel(R) Xeon(R) CPU W3680 @ 3.33GHz, VW 4
4096 x 4096 pixel in 773.30 ms
21.70 megapixel per second
91.12 GFLOPS

Intel(R) Xeon(R) CPU W3680 @ 3.33GHz, VW 8
4096 x 4096 pixel in 740.37 ms
22.66 megapixel per second
95.17 GFLOPS

All OpenCL devices together, VW (4,8)
4096 x 4096 pixel in 164.18 ms
102.19 megapixel per second
429.20 GFLOPS

In this test a GTX 980 performs significantly better.
 
I don’t know, but ExpressCard isn’t the best option if your Mac has Thunderbolt due to bandwidth, and since it costs the same to go the ExpressCard->Thunderbolt route since the only ExpressCard to Thunderbolt adapter costs $150.

I have only expresscard... no thunderbolt. The setup should be very similar but I'm stuck on whether or not to modify the kexts or not. It has been said that this vidock OR pe4L setup works with old macbook pros that have expresscard slot without much modification (SIP or equivalent must be off). There are lots of tutorials for the expresscard to thunderbolt OSX setup. But I'm working with pe4L 2.1b and gt240 (osx native drivers) on a 4530s with 10.11.
 
Status
Not open for further replies.
Register on MacRumors! This sidebar will go away, and you'll see fewer ads.