Remove rear case pci holder to dissipate better heat?

[Cuzzupino]

Hi,

I have an gtx 1080 MSI aero blower flashed for mac. I see that the rear hole to dissipate hot air of gpu are partially covered from a little part of case...

I wonder if I can trim away that part and what best tool to use to avoid damage... or if this heat port covered is a bypassable problem...

Some suggestions?

See photo below to better understand the “problem”

 

[yuzgen]

There ain't no problem. Don't cut your Mac Pro.

 

[AlexMaximus]

Hi,

I have an gtx 1080 MSI aero blower flashed for mac. I see that the rear hole to dissipate hot air of gpu are partially covered from a little part of case...

I wonder if I can trim away that part and what best tool to use to avoid damage... or if this heat port covered is a bypassable problem...

Some suggestions?

See photo below to better understand the “problem”

There is an excellent Hybrid Cooler out there from either EVGA/Nvidia. Alternative you could use the Corsair Hydrogen cooler as well. This should reduce temps for about 10 - 20 degrees.

I currently think about to refit a hybrid cooler on my 980Ti. In case you do not use Pixlas Mod, you should definitely use the EVGA Power Link to better balance power between those 6+8 or 8+8 GPU leads.

 

[Cuzzupino]

There is an excellent Hybrid Cooler out there from either EVGA/Nvidia. Alternative you could use the Corsair Hydrogen cooler as well. This should reduce temps for about 10 - 20 degrees.

I currently think about to refit a hybrid cooler on my 980Ti. In case you do not use Pixlas Mod, you should definitely use the EVGA Power Link to better balance power between those 6+8 or 8+8 GPU leads.

My vga is single 8 pin... I split power over 2 aux with proper cable. However I’m interested on better cooling options... but can I have problem with gpu flashed efi if I switch cooler?

 

[ActionableMango]

Is there a problem with it overheating?

Don't fix what ain't broke. Especially if the fix involves cutting up your case.

 

[AlexMaximus]

My vga is single 8 pin... I split power over 2 aux with proper cable. However I’m interested on better cooling options... but can I have problem with gpu flashed efi if I switch cooler?

I didn't do the conversion yet. However, I have watched dozens of YouTube Videos, regarding the swap. It looks to me, the PCB is always the same. You just need to check the compatibility on the EVGA webpage or on Corsair's.
I don't think a software change (EFI Flash) does have any effect on the cooling.
Only thing to note, the guy in the video stated in the comments, that he later on installed the air cooler on the radiator to blow out air in the front, not through the machine as described in the video. This mad a big difference.

 

[alphaod]

I don't think it's a problem. In a typical PC chassis you have air around the GPU that's warm from all the other components like the motherboard, CPU, power supply. So you have warm air to cool the GPU so you might want a more open vent.

In the Mac Pro it's compartmentalized the CPU with motherboard, and the optical drive with power supply have their own respective cool air to circulate. The PCI slots have cooler air from the front to cool the GPU so even if the vent is a little blocked it's still better than a typical PC chassis environment.

 

[davidmartindale]

Hi,

I have an gtx 1080 MSI aero blower flashed for mac. I see that the rear hole to dissipate hot air of gpu are partially covered from a little part of case...

I wonder if I can trim away that part and what best tool to use to avoid damage... or if this heat port covered is a bypassable problem...

Some suggestions?

See photo below to better understand the “problem”

Thats exactly how it would be on a standard PC that the card is designed for. It should not impact performance to any noticeable degree. So in short, no need to cut it.

 

[MriX]

There is an excellent Hybrid Cooler out there from either EVGA/Nvidia. Alternative you could use the Corsair Hydrogen cooler as well. This should reduce temps for about 10 - 20 degrees.

I currently think about to refit a hybrid cooler on my 980Ti. In case you do not use Pixlas Mod, you should definitely use the EVGA Power Link to better balance power between those 6+8 or 8+8 GPU leads.

Pretty neat this EVGA Power Link. Do you know if this also fits with a MSI R9 280X in a Mac Pro? Or is there another alternative for balancing between the 6 pin and the 8 pin?

 

[handheldgames]

Hi,

I have an gtx 1080 MSI aero blower flashed for mac. I see that the rear hole to dissipate hot air of gpu are partially covered from a little part of case...

I wonder if I can trim away that part and what best tool to use to avoid damage... or if this heat port covered is a bypassable problem...

Some suggestions?

See photo below to better understand the “problem”

A pair of tin snips will work without creating metal shavings. The edges bay be sharp, cover with electrical tape if necessary.

I’ve cut into the bottom of a 2009, through a hole made by removing the WiFi antenna, allowing for an external express card bus.

Mod away. I don’t think you will be effecting the structural integrity of the case.

 

[Cuzzupino]

I didn't do the conversion yet. However, I have watched dozens of YouTube Videos, regarding the swap. It looks to me, the PCB is always the same. You just need to check the compatibility on the EVGA webpage or on Corsair's.
I don't think a software change (EFI Flash) does have any effect on the cooling.
Only thing to note, the guy in the video stated in the comments, that he later on installed the air cooler on the radiator to blow out air in the front, not through the machine as described in the video. This mad a big difference.

I find the evga hybrid but seems that is not compatible with MSI aero GTX 1080, from corsair I can't find anything like evga hybrid, I see only cpu/gpu standalone watercooler with no cooling for the rest of vga, can you link me someone from corsair?
[doublepost=1533179512][/doublepost]the NZXT Kraken G12 GPU Adapter with corsair h75 seems a good option to me... only I don't understand where to connect fans cable... peraphs do I need somethings like this https://www.ebay.it/itm/PC-SATA-1-t...997184?hash=item1a3ea2a900:g:GeAAAOSwT2dZ5x4- ???

 

[h9826790]

Hi,

I have an gtx 1080 MSI aero blower flashed for mac. I see that the rear hole to dissipate hot air of gpu are partially covered from a little part of case...

I wonder if I can trim away that part and what best tool to use to avoid damage... or if this heat port covered is a bypassable problem...

Some suggestions?

See photo below to better understand the “problem”

In my own test, make the back end “more open” makes little to no difference in GPU cooling. (Tested with a PNY XLR8 1080Ti with the native air cooler).

What I did is not by cutting anything. But keep / remove the PCIe slot end metal covers. If that makes no difference, then I can’t see how cutting the cMP will help.

My 1080Ti generate way more heat than your 1080. If my 1080Ti doesn’t overheat in the cMP. Then no way a 1080 will overheat becasue of the cMP design.

If the bad cooling is coming from the poor cooler. Then mod the cMP won’t help. You should fix the cooler, not the cMP.

And as ActionableMango pointed out, if nothing goes wrong at this moment, don’t fix it.

Pretty neat this EVGA Power Link. Do you know if this also fits with a MSI R9 280X in a Mac Pro? Or is there another alternative for balancing between the 6 pin and the 8 pin?

You can do that simply by using 2 cables.

1) dual mini 6pin -> single 8pin

2) single 8pin FEMALE -> dual 6+2 pin

I used this combination to power my PNY XLR8 1080Ti (which also can’t use the EVGA PowerLink and have the following result.

As you can see, the power draw from the mini 6pins are very balanced.

Of course, you can always use the PowerLink as the bridge, and use some extension cables to connect the PowerLink output to the GPU’s input. That won’t looks tidy, but will work.

N.B. PowerLink provide 2 functions, apart from balance the power draw. It also has capacitors inside to smooth out the power spike. The “cables combination method” won’t provide the 2nd function. But you can go for some cables that has build in capacitor to provide the same function.

 

[MriX]

In my own test, make the back end “more open” makes little to no difference in GPU cooling. (Tested with a PNY XLR8 1080Ti with the native air cooler).

What I did is not by cutting anything. But keep / remove the PCIe slot end metal covers. If that makes no difference, then I can’t see how cutting the cMP will help.

My 1080Ti generate way more heat than your 1080. If my 1080Ti doesn’t overheat I’m the cMP. Then no way a 1080 will overheat becasue of the cMP design.

If the bad cooling is coming from the poor cooler. Then mod the cMP won’t help. You should fix the cooler, not the cMP.

And as ActionableMango pointed out, if nothing goes wrong at this moment, don’t fix it.



You can do that simply by using 2 cables.

1) dual mini 6pin -> single 8pin

2) single 8pin FEMALE -> dual 6+2 pin

I used this combination to power my PNY XLR8 1080Ti (which also can’t use the EVGA PowerLink and have the following result.
View attachment 774037
As you can see, the power draw from the mini 6pins are very balanced.

Of course, you can always use the PowerLink as the bridge, and use some extension cables to connect the PowerLink output to the GPU’s input. That won’t looks tidy, but will work.

N.B. PowerLink provide 2 functions, apart from balance the power draw. It also has capacitors inside to smooth out the power spike. The “cables combination method” won’t provide the 2nd function. But you can go for some cables that has build in capacitor to provide the same function.

Yes, that’s a super alternative and I don’t think the capacitors are quite necessary. I think the PowerLink won’t fit the Mac Pro, because the GPUs are pretty long with the power link and my conflicts with the PCI fan.
The tool you used for the graph is the paid tool Hardwaremonitor, isn’t it?

 

[AlexMaximus]

I find the evga hybrid but seems that is not compatible with MSI aero GTX 1080, from corsair I can't find anything like evga hybrid, I see only cpu/gpu standalone watercooler with no cooling for the rest of vga, can you link me someone from corsair?
[doublepost=1533179512][/doublepost]the NZXT Kraken G12 GPU Adapter with corsair h75 seems a good option to me... only I don't understand where to connect fans cable... peraphs do I need somethings like this https://www.ebay.it/itm/PC-SATA-1-t...997184?hash=item1a3ea2a900:g:GeAAAOSwT2dZ5x4- ???

Link:

https://www.corsair.com/us/en/Categ...A1-GPU-Liquid-Cooling-Bracket/p/CB-9060001-WW

 

[bsbeamer]

Do you notice a huge difference running with the side door on vs. side door off? Seems that would be the easiest first test without permanent damage. Have a client that will "vent" the side door when running renders overnight, but I do not notice any noticeable differences in render times.

 

[handheldgames]

Do you notice a huge difference running with the side door on vs. side door off? Seems that would be the easiest first test without permanent damage. Have a client that will "vent" the side door when running renders overnight, but I do not notice any noticeable differences in render times.

Running without the side panel defeats the air tunnel design of the CPU Bay Intake and Exhaust fans. Minimizing the fans ability to cool the CPU's, RAM and Chipset.

 

[TheIguana]

I used this combination to power my PNY XLR8 1080Ti (which also can’t use the EVGA PowerLink and have the following result.
View attachment 774037
As you can see, the power draw from the mini 6pins are very balanced.

Just out of interesting, what software are you running to get that data-set?

 

[h9826790]

Do you notice a huge difference running with the side door on vs. side door off? Seems that would be the easiest first test without permanent damage. Have a client that will "vent" the side door when running renders overnight, but I do not notice any noticeable differences in render times.

Running without the side panel defeats the air tunnel design of the CPU Bay Intake and Exhaust fans. Minimizing the fans ability to cool the CPU's, RAM and Chipset.

That has an assumption that the cMP has a very good air flow design. However, from my own test, it is not. (Of course, it's not bad either. But for GPU, it's definitely not super good. My same graphic card can run ~10C cooler in my Hackintosh. Same card, same setting, same loading, but 10C cooler.)

There is nothing special on the cMP's air flow / cooling design. The CPU won't run warmer with side panel door open.

For GPU, it very depends on the cooler type. For blower type, again, almost no difference with side panel ON or OFF.

For open fan type, my 1080Ti shows that can run up to 5C cooler with side panel open, but normal difference is less than 3C.

If you want the CPU / GPU run cooler, one of the very easy way to achieve that is just rotate the cMP for 90 degree (let the front panel face the ground, and the exhaust face the sky).

https://forums.macrumors.com/thread...ted-performance.1333421/page-43#post-19522505

This may be more effective than open the side panel or cut the back end.

Yes, that’s a super alternative and I don’t think the capacitors are quite necessary. I think the PowerLink won’t fit the Mac Pro, because the GPUs are pretty long with the power link and my conflicts with the PCI fan.
The tool you used for the graph is the paid tool Hardwaremonitor, isn’t it?

Just out of interesting, what software are you running to get that data-set?

https://www.bresink.com/osx/HardwareMonitor.html

 

[handheldgames]

That has an assumption that the cMP has a very good air flow design. However, from my own test, it is not. (Of course, it's not bad either. But for GPU, it's definitely not super good. My same graphic card can run ~10C cooler in my Hackintosh. Same card, same setting, same loading, but 10C cooler.)

There is nothing special on the cMP's air flow / cooling design. The CPU won't run warmer with side panel door open.

For GPU, it very depends on the cooler type. For blower type, again, almost no difference with side panel ON or OFF.

For open fan type, my 1080Ti shows that can run up to 5C cooler with side panel open, but normal difference is less than 3C.

If you want the CPU / GPU run cooler, one of the very easy way to achieve that is just rotate the cMP for 90 degree (let the front panel face the ground, and the exhaust face the sky).

https://forums.macrumors.com/thread...ted-performance.1333421/page-43#post-19522505

This may be more effective than open the side panel or cut the back end.

https://www.bresink.com/osx/HardwareMonitor.html

FWIW... Extensive RAM IO benefits from being cooled more than than Apple's default profile. The RAM will be warmer and slower with the panel removed.

RAM Bandwidth - modified cooling profile

RAM Bandwidth - Apple's default cooling profile

 

[h9826790]

FWIW... Extensive RAM IO benefits from being cooled more than than Apple's default profile. The RAM will be warmer and slower with the panel removed.

RAM Bandwidth - modified cooling profile
View attachment 774127
RAM Bandwidth - Apple's default cooling profile
View attachment 774128

Did you actually observe those numbers with panel ON / OFF?

Or your "modified cooling profile" mean higher fan speed (but both numbers are with side panel ON)?

 

[handheldgames]

Did you actually observe those numbers with panel ON / OFF?

Or your "modified cooling profile" mean higher fan speed (but both numbers are with side panel ON)?

Both of the benchmarks were with the side on. I'm still recovering from surgery and crawling under my desk to pull the side panel is on my list of things not to do. From what I remember, Intake/Exhaust runs around ~920RPM. The noise difference is inaudible. Testing on that Mac with the side off will have to wait.

 

[h9826790]

Both of the benchmarks were with the side on. I'm still recovering from surgery and crawling under my desk to pull the side panel is on my list of things not to do. From what I remember, Intake/Exhaust runs around ~920RPM. The noise difference is inaudible. Testing on that Mac with the side off will have to wait.

Yes please (of course after you fully recovered. I am sorry to hear that, hope you recover soon). From this test, you only shows us better cooling may provide better performance (I will explain why only "may"), but not showing us the cMP has better performance with the side panel ON. This is the whole idea of the debate.

I tried to reproduce your result, but unable to get similar outcome.

1) I can 100% confirm that DIMM will run cooler with side panel CLOSED.

Here is the test result.

Area 1: BlackMagic speed test (5GB) on a 10GB RAM drive with side panel closed.
Area 2: Keep the speed test running but remove the side panel.
Area 3: Keep the speed test running and put the side panel back on.

So, it's very clear that the DIMM will run few degree C cooler with side panel on. I didn't let the temperature go fully stabilised because that will take lots of time. But from the test result, I am quite confidence the difference is up to around 5C.

HOWEVER, there is no performance difference thought out the whole test.
The write speed alway around 2350MB/s, and write speed always around 2850MB/s.

And yes, that fan setting fit my own observation as well, the intake / exhaust fans can spin up to about 900 RPM without any detectable noise increase, and the booster fan can increase up to around 1100 RPM. Once beyond this line, I can hear the fan noise difference.

Last but not least, when you re-run the test. Please try to keep the Northbridge temperature fairly constant or low (in my test it only vary between 61-62C, so fairly constant), because the NB may affect the data copying performance as well.

So, the conclusion so far, keep the side panel ON may let the open fan type GPU run few degrees warmer, but let the DIMM run few degrees cooler. But as long as the GPU is not in thermal throttling yet, that few degrees shouldn't make any difference. On the other hand, that few degree difference on DIMM also show no performance difference at normal operating temperature.

I've check my DIMM's spec, once reach 85C temperature, there should be some performance penalty.

However, even with the side panel removed, 36C system ambient, that's still very far away from that 85C. So, should be similar to a good graphic card, there should be no real world "performance" difference regardless side panel ON / OFF.

 

[handheldgames]

Yes please (of course after you fully recovered. I am sorry to hear that, hope you recover soon). From this test, you only shows us better cooling may provide better performance (I will explain why only "may"), but not showing us the cMP has better performance with the side panel ON. This is the whole idea of the debate.

I tried to reproduce your result, but unable to get similar outcome.

1) I can 100% confirm that DIMM will run cooler with side panel CLOSED.

Here is the test result.
View attachment 774176

Area 1: BlackMagic speed test (5GB) on a 10GB RAM drive with side panel closed.
Area 2: Keep the speed test running but remove the side panel.
Area 3: Keep the speed test running and put the side panel back on.

So, it's very clear that the DIMM will run few degree C cooler with side panel on. I didn't let the temperature go fully stabilised because that will take lots of time. But from the test result, I am quite confidence the difference is up to around 5C.

HOWEVER, there is no performance difference thought out the whole test.
The write speed alway around 2350MB/s, and write speed always around 2850MB/s.
View attachment 774178

And yes, that fan setting fit my own observation as well, the intake / exhaust fans can spin up to about 900 RPM without any detectable noise increase, and the booster fan can increase up to around 1100 RPM. Once beyond this line, I can hear the fan noise difference.

Last but not least, when you re-run the test. Please try to keep the Northbridge temperature fairly constant or low (in my test it only vary between 61-62C, so fairly constant), because the NB may affect the data coping performance as well.

So, the conclusion so far, keep the side panel ON may let the open fan type GPU run few degrees warmer, but let the DIMM run few degrees cooler. But as long as the GPU is not in thermal throttling yet, that few degrees shouldn't make any difference. On the other hand, that few degree difference on DIMM also show no performance difference at normal operating temperature.

I've check my DIMM's spec, once reach 85C temperature, there should be some performance penalty.
View attachment 774183
However, even with the side panel removed, 36C system ambient, that's still very far away from that 85C. So, should be similar to a good graphic card, there should be no real world "performance" difference regardless side panel ON / OFF.
View attachment 774184

Thanks for running extended heat tests. I appreciate your technical details.

i should explain my ram and ramdisk in greater detail.

I'm running Samsung 2rx4 ddr3 16gb ecc server rdimms without heatsinks running at 1333mhz.

I usually run a raid 0 ramdisk created with a terminal script. While it sounds crazy, there are three hardware memory channels and it seems to work very well with 2 or 3 stripes, improving overall performance and write consistency. Raid 0 ramdisk 4k and large file performance when compiling code cant be matched by the best PCIe SSD's.

When I started bench-marking multiple raid 0 ramdisk slice and chunk configurations to find the fastest settings with the side off, I noticed the RAM quickly slow down after repeated tests. just like the thermally challenged sm951. Re-installing the side panel and setting up macs fan control to set the intake / exhaust based on dimm 3, cools the ram enough to thwart throttling.

Quickbench is one of the few disk benchmarks on mac os. It's suite of mixed file size and custom sequential tests helps to expose performance penalties or gains that may be difficult to quantify outside of a hunch or feeling.

 

[h9826790]

Thanks for running extended heat tests. I appreciate your technical details.

i should explain my ram and ramdisk in greater detail.

I'm running Samsung 2rx4 ddr3 16gb ecc server rdimms without heatsinks running at 1333mhz.

I usually run a raid 0 ramdisk created with a terminal script. While it sounds crazy, there are three hardware memory channels and it seems to work very well with 2 or 3 stripes, improving overall performance and write consistency. Raid 0 ramdisk 4k and large file performance when compiling code cant be matched by the best PCIe SSD's.

When I started bench-marking multiple raid 0 ramdisk slice and chunk configurations to find the fastest settings with the side off, I noticed the RAM quickly slow down after repeated tests. just like the thermally challenged sm951. Re-installing the side panel and setting up macs fan control to set the intake / exhaust based on dimm 3, cools the ram enough to thwart throttling.

Quickbench is one of the few disk benchmarks on mac os. It's suite of mixed file size and custom sequential tests helps to expose performance penalties or gains that may be difficult to quantify outside of a hunch or feeling.

May be we can look closer to the DIMM, I wonder if different DIMM make the difference in this case.

My DIMM is the Micron MT36KDZS2G72PZ-1G4E1HE. Also no heatsink and running at 1333Hz.

Of course, Benchmark software or the way to create the RAM drive may make the difference as well.

When I ran the test, there were some performance degradation (about 10% hit), but I quickly realise that's due to some other OS operation. Once finished, performance back to normal regardless the temperature.

I am more than happy to run more tests to find out why you have those result (I am still thinking should I buy QuickBench just for this test). But so far, in my own setup, I can't see how the side panel ON / OFF can affect the DIMM (and a graphic card that has decent cooler) performance under normal conditions.

 

[handheldgames]

May be we can look closer to the DIMM, I wonder if different DIMM make the difference in this case.

My DIMM is the Micron MT36KDZS2G72PZ-1G4E1HE. Also no heatsink and running at 1333Hz.
View attachment 774448
View attachment 774449

Of course, Benchmark software or the way to create the RAM may make the difference as well.

When I ran the test, there were some performance degradation (about 10% hit), but I quickly realise that's due to some other OS operation. Once finished, performance back to normal regardless the temperature.

I am more than happy to run more tests to find out why you have those result (I am still thinking should I buy QuickBench just for this test). But so far, in my own setup, I can't see how the side panel ON / OFF can affect the DIMM (and a graphic card that has decent cooler) performance under normal conditions.

We have similar RAM with the exception of the brand. I’m running Samsung whilst you’re running Micron. Both should perform similarly.

Generally speaking, keeping the side off has greatest effect on components with passive cooling, such a RAM and PCIe SSD’s.

It also lowers the noise generated from the rear of the cMP and the airflow through the chassis. While the video card may be pulling in warm air bellowing up from below, any loss in cooling will be minimized.

With the Mac Pro on its side, about 3.5 feet off the floor, on a standing desk with the side panel removed, I initially encountered the RAM speed throttling with the side panel removed. In this orientation, heat travels upward from the heat pipe, towards the RAM. I repeated the results with a 5,1 at my office, that's wedged under a lowered standing desk in the corner of a cube in a somewhat traditional, hide it under the desk approach.

The type of video card cooler will also impact the temps of the pci bay. Cards without a blower type exhaust will act as a space heater for any components vertically positioned above it.

Re-installing the side panel reduced throttling, allowing the RAM to quickly recover rather than slowing down.

I’ll admit, I’ve run my cMP without the side panel for years without worry or concern. Only after upgrading to RDIMMS and encountering a reproducible slow down running benchmarks did the the design of Apple’s wind tunnel design truly make sense.

I’ve attached a copy of my raid0 ramdrive scripts. One for a 8gb 2 stripe Ramdisk array, and one with a 9gb 3 stripe ramdisk array.

One more thing... On the 'bench' cmp I replaced the traditional PCIe slot covers with a black mesh version from a PC case, allowing for more airflow out of the PCIe bay.