Sidebar Sidebar
Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.

Patchy graphics - Is it the card?

SayCheese

SayCheese

macrumors 68000
Original poster
Jun 14, 2007
1,720
919
Oxfordshire, England
Hi all,

I have a problem with my Mac Pro custom build.
I think it's the graphics card although I'm not sure. Basically, I often get what I can only really describe as glitchy patches across the screen as I move the pointer across the screen.

I've only seen it in Photoshop and Lightroom. I'm using a graphics tablet but I've also been able to reproduce the effect with the mouse. I've previously run these programs without an issue, however, this has started in the last few days only. I've tried a restart to no effect.

I've attached a screen shot so you can see what I'm talking about.

My graphics card is a Nvidia GE Force GTX 980. The mac has more than enough RAM (96Gb), so it shouldn't be that.

If anyone has any suggestions, I'd appreciate it please.
 

Attachments

  • Screen Shot 2016-03-18 at 11.18.03.png
    Screen Shot 2016-03-18 at 11.18.03.png
    69.4 KB · Views: 150
Almost certainly hardware. If you have the D500 or D700 then it'll be fixed free-of-charge, even if you're out of warranty.
 
SayCheese said:
D500/700 as in the camera?
Click to expand...

Bugger I skimmed your thread, didn't realise you had the cMP. Sorry man, I should read more carefully in future.

But it does look like graphics anyway, do you have a spare GPU you can verify with?
 
keysofanxiety said:
Bugger I skimmed your thread, didn't realise you had the cMP. Sorry man, I should read more carefully in future.

But it does look like graphics anyway, do you have a spare GPU you can verify with?
Click to expand...

I was a little confused!
I do have a spare graphics card. Need to find time to swap it out though.
 
I get the same problem with my Macbook Pro. It doesn't last long but still a concern.
It's definitely a hardware issue.
 
See if you can replicate these artifacts in a more controlled environment, by using various stress tests individually based on specific APIs . Some good programs would be Cinebench for Open GL , LuxMark for Open CL and the Octane test octane_benchmark.ocs / RenderTarget PT for CUDA . It seems unlikely such a new card would fail so early unless you are really stressing it hard with renders .
 
I've seen these when I did 3D renders. You might want to check your GPU temp. I used SMC Fan Control.
 
I picked up a GTX 580 very cheaply a few years ago that was doing exactly the same thing. Temp readings seemed normal, but when I took it apart I found that the factory-applied thermal paste had been packed on like Oprah making pancakes and there were pockets of air trapped between the silicone and the heatsink. I cleaned it up and applied a layer of thermal paste a fraction of a millimetre thick, and it's worked perfectly ever since. I'm not sure whether removing the heatsink would void your warranty, as there are lots of after-market water-cooling systems on very expensive video cards out there, but if it's not under warranty it's worth checking out.
 
davidb367 said:
I've seen these when I did 3D renders. You might want to check your GPU temp. I used SMC Fan Control.
Click to expand...

SMC Fan Control is an older utility that stopped being supported after Mountain Lion OS X 10.8.x . The modern utility used these days is Macs Fan Control . All the same , does SMC Fan Control report the GPU temps of PC Editions installed in Macs, like the OP's Maxwell card ?
 
jamall said:
The very useful HWMonitor package with its accompanying kexts for specific functionality give a temp for my GTX 580 as well as many other readings.
Click to expand...

Much obliged . I'll install this on one of my test drives and check it out . It would be awesome to have the ability to grab the GPU Temp sensor data in an OS X session on non-Apple shipping graphics cards .
 
I had very much the same thing show up on 1 of the monitors in my 5 display setup. It was caused by an app, Tweetbot I think. After I upgraded it, the patchy graphics hasn't come back.

You can also get rid of the patchiness by dragging another app window across the screen, it's really like wiping it down (I used Firefox). But it comes back, so I had to find the root cause.

I also found that the nVidia web driver had reverted to the OSX driver. Switched that back too.
 
owbp said:
Isn't that a hackintosh tool that has to be installed with FakeSMC.kext?
I love it for Hacks, but never tried it on Mac...
Click to expand...

I tried, hard to setup correctly on a real Mac, and it seems you will lost some original SMC data (e.g. Fan speed) after installing the FakeSMC.

I asked if there is any way to get both data some time ago, from memory, Netkas reply only one or the other.

So, it's quite meaningless if we can check the GPU temperature but lost the original important data.

Jamall, do you have all the original reading with FakeSMC installed? May you share how you do it?
 
Machines said:
Aha , but not for a technician ;) .

We'll make a dedicated OS X install just to launch this utility . It's that valuable, if it really works .
Click to expand...

True, you can have different machines / boot partitions for different tests. Each of them may only build / setup for specific purpose.

However, that may gives you another problem. e.g. You now know the GPU temperature, but not know the PCIe fan speed (which actually may affect the GPU temperature). So, you should technically monitor both variables at the same time to confirm the system work as expected, but it seems very hard to achieve that in OSX.

If only need to know the temperature regardless other variables, then boot into Windows can already achieve that.

Anyway, do you know if there is any software can monitor the PCIe power draw in Windows? That's the variable missing in Windows.
 
Last edited:
h9826790 said:
However, that may gives you another problem. e.g. You now know the GPU temperature, but not know the PCIe fan speed (which actually may affect the GPU temperature). So, you should technically monitor both variables at the same time to confirm the system work as expected, but it seems very hard to achieve that in OSX.
Click to expand...

Hmmmm ... then how about we run HWMonitor with the kernel extensions for the GPU temps and Bresink Hardware Monitor for the case fan rotationals concurrently ? Or do you think the fan sensors will simply go offline due to a kext ? I would need to stay within OS X .
h9826790 said:
Anyway, do you know if there is any software can monitor the PCIe power draw in Windows? That's the variable missing in Windows.
Click to expand...

Not a software program , but something better :) A kill-a-watt device . And you can simply add and remove individual PCI cards (and also place them at idle and at load) and get a real time readout of the amperage, wattage and voltage in any given situation . My stations are brimming with Kill-A-Watt devices ...

31VRHeNmx0L.jpg
 
You could then also look into HWSensors (kozlek) and HWSenesors3 (slice) in that case, but HWMonitor didn't have fan controls with my Dell/Hack also. Was positive that it was due to its BIOS and MB at the time...

Edit: Didn't try HWSensors myself.
 
Machines said:
Hmmmm ... then how about we run HWMonitor with the kernel extensions for the GPU temps and Bresink Hardware Monitor for the case fan rotationals concurrently ? Or do you think the fan sensors will simply go offline due to a kext ? I would need to stay within OS X .

Not a software program , but something better :) A kill-a-watt device . And you can simply add and remove individual PCI cards (and also place them at idle and at load) and get a real time readout of the amperage, wattage and voltage in any given situation . My stations are brimming with Kill-A-Watt devices ...

View attachment 622129
Click to expand...

I guess there is a way to do it. However, so far, in my own experience, you can only get one or the other. e.g. Either the reading from fake SMC (GPU temp etc), or the value from the real SMC (fan RPM etc).

I think if you choose all the correct kext for Fake SMC, you can get all the readings at the same time, but I never success. And if install the wrong kext, it may cause Kernel panic during boot, which need some skill to recover (I assume this is a relatively easy job for you).

Anyway, the last time I try this at least a year ago, may be it's much easier to setup and everything works now.

Update:

I just read them again. It seems we can use FakeSMC to read the real SMC now.

http://www.insanelymac.com/forum/topic/309328-howto-acpi-access-for-monitoring/

My Mac is encoding some movies at this moment, so can't test it yet. But this is what I understand so far. When installing this software, we need to choose custom install in step 3, and then pick the correct option. e.g. For my 4,1 (W3680, HD7950).

Choose the Intel CPU, AMD card, ACPI Access, and Silent AppleSMC.
Screen Shot 2016-03-21 at 22.48.59.jpg
Screen Shot 2016-03-21 at 22.49.06.jpg

However, I have no idea do I need to install the kext for LPC chip, or which type to choose.
Screen Shot 2016-03-21 at 22.50.29.jpg

After that, if we get all the readings, then it's fine. Otherwise, we need to write something into ACPImonitor.kext/info.plist with correspondence of ACPI method and SMC keys to get what we want.
[doublepost=1458573744,1458570144][/doublepost]
Machines said:
Hmmmm ... then how about we run HWMonitor with the kernel extensions for the GPU temps and Bresink Hardware Monitor for the case fan rotationals concurrently ? Or do you think the fan sensors will simply go offline due to a kext ? I would need to stay within OS X .



Not a software program , but something better :) A kill-a-watt device . And you can simply add and remove individual PCI cards (and also place them at idle and at load) and get a real time readout of the amperage, wattage and voltage in any given situation . My stations are brimming with Kill-A-Watt devices ...

View attachment 622129
Click to expand...

It may be better to measure the overall power consumption, but can't measure each 6pin power source loading. The overall loading is not a concern for me. However, I would like to know how much power I am drawing from each of those line under stress (in Windows).
 
h9826790 said:
It may be better to measure the overall power consumption, but can't measure each 6pin power source loading. The overall loading is not a concern for me. However, I would like to know how much power I am drawing from each of those line under stress (in Windows).
Click to expand...

Maybe by using two external PSU , one for each power booster cable from the graphics card ? And a Kill-A-Watt device attached to each PSU ? That should do the trick . However, we are ,of course, not testing a normal System, but just an itemized draw from the video card's connectors .
 
Machines said:
Maybe by using two external PSU , one for each power booster cable from the graphics card ? And a Kill-A-Watt device attached to each PSU ? That should do the trick . However, we are ,of course, not testing a normal System, but just an itemized draw from the video card's connectors .
Click to expand...

Does ATI have an API to report power? The Nvidia SMI interface does.

Code: 
Mon Mar 21 13:51:09 2016
+------------------------------------------------------+
| NVIDIA-SMI 352.39     Driver Version: 352.39         |
|-------------------------------+----------------------+----------------------+
| GPU  Name        Persistence-M| Bus-Id        Disp.A | Volatile Uncorr. ECC |
| Fan  Temp  Perf  Pwr:Usage/Cap|         Memory-Usage | GPU-Util  Compute M. |
|===============================+======================+======================|
|   0  GeForce GTX TIT...  Off  | 0000:81:00.0     Off |                  N/A |
| 22%   35C    P0    69W / 250W |     23MiB / 12284MiB |      0%      Default |
+-------------------------------+----------------------+----------------------+
|   1  GeForce GTX TIT...  Off  | 0000:C1:00.0     Off |                  N/A |
| 22%   32C    P0    69W / 250W |     23MiB / 12284MiB |      0%      Default |
+-------------------------------+----------------------+----------------------+
|   2  GeForce GTX TIT...  Off  | 0000:C4:00.0     Off |                  N/A |
|  0%   36C    P0    54W / 250W |     23MiB / 12284MiB |      0%      Default |
+-------------------------------+----------------------+----------------------+
 
You must log in or register to reply here.
Register on MacRumors! This sidebar will go away, and you'll see fewer ads.
Top Bottom
Back