Looks like a healthy drop in temp without cover. I hadn't seen that in my tests. Maybe like you said, fan under the hs allows too much leakage around the fins and hot air recirculation. Cover off eliminates that.
Completed The MI-6 Microtower Case: Performance in 6.7L
- Thread starter firewolfy
- Start date
Yeah that's the most possible explanation.
I guess I'll try to tidy up the cables, put 80mm fan, close the case & try again later. If still too hot.. then I'll try to reverse NH-L12's fan orientation.
Speaking of which, anyone have info / review about push VS pull on these C-type coolers? My google-fu didn't return much.
I know on the NH-L12S Noctua reversed the fan direction so it nows blows away from the motherboard, might be a good idea to try it. Mine blows toward the motherboard but I have a bottom 92mm fan and can really feel heat coming out the top of the case, without the bottom fan may work better to blow the heat out of the case.
Really? They reversed the fan direction on NH-L12S?? Hmm I wonder why. If Noctua engineers did it, they sure had some good reasons for it.
Yeah I'm itching to reverse my fan direction.. This means the bottom case fan act as an input. But let me just add case fan first and run a test again. Will keep you guys updated.
In my current build, I have the fan on top of the cooler as pull, and it was by far the best config along with the bottom fans as intake and top fans as exhaust: https://smallformfactor.net/forum/threads/i-think-i-found-a-little-gem-realan-e-i7.6175/
A few things may have contributed to this result in my case (pun etc):
1) the side panel perforations do not cover the hole fan (only half of it), which really does not help when the fan is used as intake.
2) The bottom fans are very nicely aligned and adjacent to the heatsink. Without such fans, I wouldn't 100% vouch for this setup.
In sum, test for your case and see. My intuition is that pull would not be significantly worse, and it might even be better because it seems like in your particular case, the problem is that the heat is getting trapped. Using the fan as pull would be the shortest path for the heat to be exhausted out of the system.
EDIT: A few additional thoughts:
1) Pull may work better than you think in your case due to the orientation of the fan on NH-L12: in intake config, the blades are facing the fins, restricting the airflow for the fan. In effect, for your case, intake is pull, and exhaust is push. Fans do push better than they pull. So, while I am disadvantaging the fan in my setup, you'd be doing yours a favor (when you use it as exhaust).
2) I'd try and fit a slim 120mm fan there instead of the 92mm fan. Would make a difference (if it fits).
3) Definitely use an intake fan at the bottom if you can.
Last edited:
I have seen this done, the slim Noctua 120mm fan does fit below the heatsink but ram clearance is reduced. @ignsvn is using really low profile ram and it would probably fit.
I wonder if the AORUS Gaming Box version of the Gigabyte GTX 1080 Mini GPU (aka the Hulk) would fit into the MI-6...?!?
That is a 130mm fan on there (as opposed to the stock 90mm fan seen on the GTX 1070 Mini to the left) and there is no shroud (unless you count the Gaming Box enclosure as the shroud)...
This, with an AIO-enabled top section, and possibly a fan mount over the motherboard for pumping plenty of fresh air into the chassis & over the RAM & PCH, maybe a X299 system in a MI-6...?!?
That is a 130mm fan on there (as opposed to the stock 90mm fan seen on the GTX 1070 Mini to the left) and there is no shroud (unless you count the Gaming Box enclosure as the shroud)...
This, with an AIO-enabled top section, and possibly a fan mount over the motherboard for pumping plenty of fresh air into the chassis & over the RAM & PCH, maybe a X299 system in a MI-6...?!?
A lot of tuning later...
Here are a few temps I'm getting on my I7-8700 after I've tuned it to allow for 4.3GHZ speed on all cores while encoding. Right now that includes XMP with 3200mhz DDR4, Noctua NH-L12 with only the 92mm fan, bottom Noctua 92mm intake fan with LNA, CPU set for 95 Watts long term power, and a -.12 on the VCORE. Ambient temp was 72F. The system has nothing around it to impede airflow.
With this setup, AIDA64 did not crash or throttle on stability tests.
Converting a 1.5GB 48 FPS Go Pro file to HEVC at maximum quality in Adobe with CUDA turned on. This is a 2.5 hour long rendering test that uses all 6 cores and 12 threads. The CPU max turbo bounces between 4.0 and 4.3 GHZ, with 4.2 being most common. The highest temp observed was 99C for a few seconds on two of the cores. All others were either 95 or 96C. The average temp was about 88C.
Obviously this is borderline throttling. So my next step is to upgrade the stock 92mm fan of the NH-L12 from the 1600RPM model to the 2000RPM model.
For reference: Leaving the CPU at stock settings, with XMP DDR4-3200 enabled, and multi-core enhancement off, the same rendering turbos between 3.4 and 3.6 GHZ, at an average temp of 75C on the cores.
My next test will be stock settings as listed in the above paragraph, but with the max long term TDP at 75 watts instead of the stock 65.
After that, I will rerun the tests with the top off to see if there is a notable difference.
Keep in mind for anyone looking at this case, for gaming and day to day things, it runs fine at 4.3 GHZ stock all day long. I'm really trying to push the edge of ITX tech by doing an Encoding and video editing machine in such a small case on compact air cooling alone, while keeping it silent at idle, and very very quiet under load without delidding. It's a fun challenge.
That said, I might delid the thing soon.
Here are a few temps I'm getting on my I7-8700 after I've tuned it to allow for 4.3GHZ speed on all cores while encoding. Right now that includes XMP with 3200mhz DDR4, Noctua NH-L12 with only the 92mm fan, bottom Noctua 92mm intake fan with LNA, CPU set for 95 Watts long term power, and a -.12 on the VCORE. Ambient temp was 72F. The system has nothing around it to impede airflow.
With this setup, AIDA64 did not crash or throttle on stability tests.
Converting a 1.5GB 48 FPS Go Pro file to HEVC at maximum quality in Adobe with CUDA turned on. This is a 2.5 hour long rendering test that uses all 6 cores and 12 threads. The CPU max turbo bounces between 4.0 and 4.3 GHZ, with 4.2 being most common. The highest temp observed was 99C for a few seconds on two of the cores. All others were either 95 or 96C. The average temp was about 88C.
Obviously this is borderline throttling. So my next step is to upgrade the stock 92mm fan of the NH-L12 from the 1600RPM model to the 2000RPM model.
For reference: Leaving the CPU at stock settings, with XMP DDR4-3200 enabled, and multi-core enhancement off, the same rendering turbos between 3.4 and 3.6 GHZ, at an average temp of 75C on the cores.
My next test will be stock settings as listed in the above paragraph, but with the max long term TDP at 75 watts instead of the stock 65.
After that, I will rerun the tests with the top off to see if there is a notable difference.
Keep in mind for anyone looking at this case, for gaming and day to day things, it runs fine at 4.3 GHZ stock all day long. I'm really trying to push the edge of ITX tech by doing an Encoding and video editing machine in such a small case on compact air cooling alone, while keeping it silent at idle, and very very quiet under load without delidding. It's a fun challenge.
That said, I might delid the thing soon.
Last edited:
Update:
I’m skipping the 75 watt update in favor of the 85 watt. At 85 watts TDP, the system will turbo to 3.8ghz, with occasional bursts to 4ghz. This had CPU temps at about 85C, with occasional jumps to 88-89C.
Encoding HEVC in Handbrake yielded about the same.
Switching to NVIDIA Cuda HEVC, the CPU stayed at 4.3GHZ at about 65C, with the GPU (1070) at 2.1GHZ.
I can now go back undervolt it by a -.05 and drop the temps a bit more. That will be the next test.
I’m skipping the 75 watt update in favor of the 85 watt. At 85 watts TDP, the system will turbo to 3.8ghz, with occasional bursts to 4ghz. This had CPU temps at about 85C, with occasional jumps to 88-89C.
Encoding HEVC in Handbrake yielded about the same.
Switching to NVIDIA Cuda HEVC, the CPU stayed at 4.3GHZ at about 65C, with the GPU (1070) at 2.1GHZ.
I can now go back undervolt it by a -.05 and drop the temps a bit more. That will be the next test.
Update 2:
Sadly the 85 Watt TDP did not survive the BF1 gaming test. It quickly got into the very high 90s, and began to throttle. Dropping back down to the stock 65 watt longterm, I still saw it rise to the 90s on occasion. Max Turbo was about 4.2 GHZ on all 6 cores. This was tested at Ultra detail, at 1920x1200, and 2560x1440, on GeForce 1070, playing 64 player servers. So it seems that about 65-75 Watt longterm is all I'm going to get without a delid for Battlefield 1.
It seems that Frostbyte engine games can really push the CPU.
So if you're encoding, doing actual work, or browsing the web, 85-90 watts is perfectly fine with the same setup. For gaming, you have to back down to basically stock long term wattage.
However, I have one last trick. I can undervolt the CPU. I already know that this CPU is very stable when undervolted (reverse Silicon Lottery?), so I'm going to drop it by a full -0.1 and see what happens.
Let's all hope that Intel's next chip handles the heat better, and that some all copper coolers show up on the market.
Sadly the 85 Watt TDP did not survive the BF1 gaming test. It quickly got into the very high 90s, and began to throttle. Dropping back down to the stock 65 watt longterm, I still saw it rise to the 90s on occasion. Max Turbo was about 4.2 GHZ on all 6 cores. This was tested at Ultra detail, at 1920x1200, and 2560x1440, on GeForce 1070, playing 64 player servers. So it seems that about 65-75 Watt longterm is all I'm going to get without a delid for Battlefield 1.
It seems that Frostbyte engine games can really push the CPU.
So if you're encoding, doing actual work, or browsing the web, 85-90 watts is perfectly fine with the same setup. For gaming, you have to back down to basically stock long term wattage.
However, I have one last trick. I can undervolt the CPU. I already know that this CPU is very stable when undervolted (reverse Silicon Lottery?), so I'm going to drop it by a full -0.1 and see what happens.
Let's all hope that Intel's next chip handles the heat better, and that some all copper coolers show up on the market.
Last edited:
Update 3:
Sorry to do this in parts, but this is the culmination of 2 days of tuning / benchmarking, and I stopped counting at 30 for the benchmarks. That was yesterday.
So I began to undervolt using the negative offset.
- .050 Negative offset didn't yield anything substantial enough to warrant talking about.
- .075 Did offer some benefits, but again, it wasn't allowing me to claw to a solid 4.3GHZ while gaming and not overheating.
- .100 Made a big difference. It's amazing how much a little voltage change could make. It allowed me to return to 80 to 85 Watt longterm wattage. This allows the Adobe Premier HEVC encoding to run at an all core Turbo of 3.8 to 4.0 GHZ (3.9 most of the time). Max temps where in the mid 80s, with one core bumping 90C for a few seconds. BF1 ran at 4.3GHZ with no deviation, and stayed in the 70C to low 80C range.
Final stats:
80 Watt Max Long Term Wattage.
XMP DD4-3200
VCORE Voltage Offset -0.100
Stock 92mm Noctua CPU fan at 1600 RPM (Intake) when over 65C
- Tuned for silence in normal tasks. 40% Fan speed below 60C.
92x15 mm Noctua Case Fan at 1600 RPM (Intake) when over 65C (Low Noise Adapter) (Loudest Fan in the case)
- Tuned for silence in normal tasks. 40% Fan speed blow 65C
MSI ITX 1070 set for Fan 0 RPM below 50C.
-Stock Fan speed
- Overclocked by +200mhz (average between 2.0 and 2.1GHZ boost clock)
With this setup, you can expect:
85C+- When doing Adobe Encoding.
75 to 85C When doing heavy gaming.
39-41C Mechanical Hard drive temps using the bracket.
55C+- M.2 SSD for Asus Stryx boards with heatsink in place.
Recommendations:
1. Manage your expectations. This case is well designed and built, but not magical.
2. Swapping to a 2000RPM 92mm fan on the NH-L12.
3. Running an Intake fan at the bottom.
4. Getting custom length cables to keep airflow clean.
5. Be careful of high ambient temperatures. I tested at 72F to 76F
6. Understand that while the CPU may be sold as 65 Watts, Intel has designed them to Turbo at a higher wattage. I've seen up to 125 Watts stock for short bursts. This may be the cause of any initial bursts of throttling before the CPU settles in at it's long term wattage.
7. Be ready to tune the BIOS.
CONCLUSION:
Intel really made it a challenge to air cool these Coffee Lake CPUS in small form factor rigs. Anything smaller than NCASE M1 is going to require some tuning time. If you want a drop in and forget it, you might want to consider the T-Series of CPUs.
The MI-6, when paired with a Noctua NH-L12, can absolutely be a nearly silent with the latest Coffee Lake CPUs. However, to get it there, you're going to have to put the work in to fine tune it. I got lucky with a CPU that can undervolt really, really well. Your mileage may very.
The one caveat to this is the 8700K. That CPU will likely need more aggressive cooling as stock it's 95 Watts.
Sorry to do this in parts, but this is the culmination of 2 days of tuning / benchmarking, and I stopped counting at 30 for the benchmarks. That was yesterday.
So I began to undervolt using the negative offset.
- .050 Negative offset didn't yield anything substantial enough to warrant talking about.
- .075 Did offer some benefits, but again, it wasn't allowing me to claw to a solid 4.3GHZ while gaming and not overheating.
- .100 Made a big difference. It's amazing how much a little voltage change could make. It allowed me to return to 80 to 85 Watt longterm wattage. This allows the Adobe Premier HEVC encoding to run at an all core Turbo of 3.8 to 4.0 GHZ (3.9 most of the time). Max temps where in the mid 80s, with one core bumping 90C for a few seconds. BF1 ran at 4.3GHZ with no deviation, and stayed in the 70C to low 80C range.
Final stats:
80 Watt Max Long Term Wattage.
XMP DD4-3200
VCORE Voltage Offset -0.100
Stock 92mm Noctua CPU fan at 1600 RPM (Intake) when over 65C
- Tuned for silence in normal tasks. 40% Fan speed below 60C.
92x15 mm Noctua Case Fan at 1600 RPM (Intake) when over 65C (Low Noise Adapter) (Loudest Fan in the case)
- Tuned for silence in normal tasks. 40% Fan speed blow 65C
MSI ITX 1070 set for Fan 0 RPM below 50C.
-Stock Fan speed
- Overclocked by +200mhz (average between 2.0 and 2.1GHZ boost clock)
With this setup, you can expect:
85C+- When doing Adobe Encoding.
75 to 85C When doing heavy gaming.
39-41C Mechanical Hard drive temps using the bracket.
55C+- M.2 SSD for Asus Stryx boards with heatsink in place.
Recommendations:
1. Manage your expectations. This case is well designed and built, but not magical.
2. Swapping to a 2000RPM 92mm fan on the NH-L12.
3. Running an Intake fan at the bottom.
4. Getting custom length cables to keep airflow clean.
5. Be careful of high ambient temperatures. I tested at 72F to 76F
6. Understand that while the CPU may be sold as 65 Watts, Intel has designed them to Turbo at a higher wattage. I've seen up to 125 Watts stock for short bursts. This may be the cause of any initial bursts of throttling before the CPU settles in at it's long term wattage.
7. Be ready to tune the BIOS.
CONCLUSION:
Intel really made it a challenge to air cool these Coffee Lake CPUS in small form factor rigs. Anything smaller than NCASE M1 is going to require some tuning time. If you want a drop in and forget it, you might want to consider the T-Series of CPUs.
The MI-6, when paired with a Noctua NH-L12, can absolutely be a nearly silent with the latest Coffee Lake CPUs. However, to get it there, you're going to have to put the work in to fine tune it. I got lucky with a CPU that can undervolt really, really well. Your mileage may very.
The one caveat to this is the 8700K. That CPU will likely need more aggressive cooling as stock it's 95 Watts.
Only problem with the Pallas is it is 68mm tall, which is the absolute max interior height. This will put the fan directly next to the side panel and I would expect turbulence.
Thanks for the great write up and the hours spent.
I keep thinking about the fan being only 92mm, slim, and doing some recirculation because of the distance from the cover. Cant fit a 120mm fan under there can you?
The 92mm fan on the heatsink is actually full size. There is some recirculation from two sides of the fan that don't have a rubber grommet at the attachment point. However, I can definitely feel cold air being pulled through the side panel. Honestly, I think that it's a simple matter that a 92x25 at 1600 RPM is not going to pull enough air.
The bottom fan does a good job moving air out of the top of the case, which also benefits from natural convection. If it didn't, the mechanical HD I have in the quick bracket would overheat due to the backflow of the GPU while gaming. That's not happening, and it stays cool. The 92mm fan at the bottom is, in my opinion, a must if you use a GPU.
I did another round of heat testing. I've been encoding Adobe HEVC for about 12 hours now, and measured the temp difference with the top on and top off. The result?
.
.
.
There was no difference on any temp sensor that wasn't margin of error (+- 1C). Now the case itself stayed cooler, but that's not an issue. The case getting toast itself is a by product of it functioning as a heatsink for the case interior. If the components inside show no temp difference, than the case is doing it's job perfectly.
Later today, I'll do the same test with BF1 and PUBG.
Last edited:
UPDATED: I did some BF1 tests with the top on and off.
Geforce 1070 with I7-8700 tuned as above. Game was at 1920 x 1200 with VSYNC OFF and frame limiter set to 200. Ultra Detail levels. Play testing was done on 64 Player Conquest servers.
Removing the top of the case yielded a few results.
GPU Temp: I saw a 2-3C reduction in GPU temperature. However, the GPU MHZ did not deviate nor did the fan speed.
CPU Temp: I saw a 3-5 C reduction in CPU temps. This did not change the Turbo of the CPU, which stayed at 4.3 GHZ the entire time.
HDD Temp: I saw a 2C reduction in the temp of the mechanical HD in the quick mount slot. It's a Seagate 5400 RPM model.
Overall: while the temp differences with the top off are measurable, repeatable, and tangible, they did not correlate with a change in performance on heavy gaming load. As such, I'll be keep the top of the case on.
Tonight or tomorrow, I will rig up an external exhaust fan to test to see if there is a tangible difference. I will mount it to top of the case from the outside, but keep it sealed so as to only draw air from the inside.
Geforce 1070 with I7-8700 tuned as above. Game was at 1920 x 1200 with VSYNC OFF and frame limiter set to 200. Ultra Detail levels. Play testing was done on 64 Player Conquest servers.
Removing the top of the case yielded a few results.
GPU Temp: I saw a 2-3C reduction in GPU temperature. However, the GPU MHZ did not deviate nor did the fan speed.
CPU Temp: I saw a 3-5 C reduction in CPU temps. This did not change the Turbo of the CPU, which stayed at 4.3 GHZ the entire time.
HDD Temp: I saw a 2C reduction in the temp of the mechanical HD in the quick mount slot. It's a Seagate 5400 RPM model.
Overall: while the temp differences with the top off are measurable, repeatable, and tangible, they did not correlate with a change in performance on heavy gaming load. As such, I'll be keep the top of the case on.
Tonight or tomorrow, I will rig up an external exhaust fan to test to see if there is a tangible difference. I will mount it to top of the case from the outside, but keep it sealed so as to only draw air from the inside.
Below is an update regarding the MI-6 work going on. I sent it out as a short newsletter to website subscribers:
Hi, I wanted to touch base with you and let you know where CCD is going with the MI-6 mini-ITX case, revisions and improvements planned and the next production batch.
So with the first batch all completed in March I learned a few lessons. The biggest lesson was with the finish. The original plan for the outside surface finish was a challenge: the black anodized brushed finish proved difficult to get good quality appearance, and a customer option was added during the batch to select black powder coat finish instead. The batch was completed but several delays occurred in order to maintain a quality finish.
I am working on a plan for a new exterior finish: Black powder coat with slight texture on the sides and front of the case, and anodized natural brushed aluminum on the top. I am still working out details, but this is a very nice look. I will have photos up on the webpageand at SFF and [H] in the next few days.
Other Improvements Planned
I am working on a minor case modification to support the cover and prevent it from flexing inward when gripped near the top. This gives a more solid feel to the case.
Another mod is to the rear PCI opening for graphics cards. It came up in a review by Tekeverything that video cables with thick connectors could interfere with the case opening. The opening has been enlarged the 1mm needed to ensure clearance.
Some owners of the first batch mentioned difficulty in mating the GPU card into the PCIE riser connector. It looks like a tolerance issue, and I've made a dimensional adjustment to prevent this issue from coming back up.
Still More Planned Changes to Improve the MI-6
The MI-6 can hold a 92mm slim fan in the bottom of the case, right in front of the PSU. This location is great for supplying cool air t to internal components such as HDDs. The fan mounting is flush against the bottom of the case ventilation slots and that generates some noise at medium and higher fan speeds. The fix? A fan spacer thick enough to raise the fan off the vents, but thin enough to not impact the PSU or front I/O board. The spacer is designed to match the vent slot pattern to the round fan for smooth flow.
THE NEXT BATCH
Timing is up in the air right now, due to the revisions and improvements going on, but I can say that the next batch size is planned for 100-200 cases. A lot of the manufacturing and assembly processes have been worked out and it makes sense ramp up a bit and be able to get costs down and reduce the case price. I've kind of let the cat out of the bag on the pricing, so I'll tell you that I'm getting new quotes from my suppliers and then can flesh out how much we can cut the price tag. -- To a less stratospheric price, if I say so myself.
Thanks for your time, and let me know if you have any question s or comments.
Fred
CCD
Hi, I wanted to touch base with you and let you know where CCD is going with the MI-6 mini-ITX case, revisions and improvements planned and the next production batch.
So with the first batch all completed in March I learned a few lessons. The biggest lesson was with the finish. The original plan for the outside surface finish was a challenge: the black anodized brushed finish proved difficult to get good quality appearance, and a customer option was added during the batch to select black powder coat finish instead. The batch was completed but several delays occurred in order to maintain a quality finish.
I am working on a plan for a new exterior finish: Black powder coat with slight texture on the sides and front of the case, and anodized natural brushed aluminum on the top. I am still working out details, but this is a very nice look. I will have photos up on the webpageand at SFF and [H] in the next few days.
Other Improvements Planned
I am working on a minor case modification to support the cover and prevent it from flexing inward when gripped near the top. This gives a more solid feel to the case.
Another mod is to the rear PCI opening for graphics cards. It came up in a review by Tekeverything that video cables with thick connectors could interfere with the case opening. The opening has been enlarged the 1mm needed to ensure clearance.
Some owners of the first batch mentioned difficulty in mating the GPU card into the PCIE riser connector. It looks like a tolerance issue, and I've made a dimensional adjustment to prevent this issue from coming back up.
Still More Planned Changes to Improve the MI-6
The MI-6 can hold a 92mm slim fan in the bottom of the case, right in front of the PSU. This location is great for supplying cool air t to internal components such as HDDs. The fan mounting is flush against the bottom of the case ventilation slots and that generates some noise at medium and higher fan speeds. The fix? A fan spacer thick enough to raise the fan off the vents, but thin enough to not impact the PSU or front I/O board. The spacer is designed to match the vent slot pattern to the round fan for smooth flow.
THE NEXT BATCH
Timing is up in the air right now, due to the revisions and improvements going on, but I can say that the next batch size is planned for 100-200 cases. A lot of the manufacturing and assembly processes have been worked out and it makes sense ramp up a bit and be able to get costs down and reduce the case price. I've kind of let the cat out of the bag on the pricing, so I'll tell you that I'm getting new quotes from my suppliers and then can flesh out how much we can cut the price tag. -- To a less stratospheric price, if I say so myself.
Thanks for your time, and let me know if you have any question s or comments.
Fred
CCD
Another suggestion I can think of is to create more exhaust holes behind, just next to the motherboard IO cutout. Not sure if it will weaken the frame or increase the price too much tho.
Man, except the powder coating, I'm so jealous of these updates..
Nice update @firewolfy, I am curious to see what prices on v2.0 will be.
I did something similar with my 92mm ID-Cooling fan. The blades extended out slightly from the frame so they were hitting on the bottom of the case when installed and making a horrible racket. I have a silicon gasket from a 92mm Noiseblocker fan and installed that between the case and the fan. Got rid of the racket and I am guessing also helped with turbulence.
Good luck with future versions of the case.
I did something similar with my 92mm ID-Cooling fan. The blades extended out slightly from the frame so they were hitting on the bottom of the case when installed and making a horrible racket. I have a silicon gasket from a 92mm Noiseblocker fan and installed that between the case and the fan. Got rid of the racket and I am guessing also helped with turbulence.
Good luck with future versions of the case.
