Production Lazer3D LZ7 - Quiet Gaming Cube PC Case

K888D

SFF Guru
Original poster
Lazer3D
Feb 23, 2016
1,465
2,925
www.lazer3d.com
Apologies in advance for the long post!

Many of the latest Mini-ITX boards now include an M.2 slot on the underside of the board. These drives offer many advantages over conventional storage drives such as reduced component space inside the case and vastly increased data transfer speeds.

Those are the positives, now onto the negatives:

There seems to be allot of talk about how well these drives will perform in cramped and hot cases with reduced airflow, with concern for throttling and reliability. M.2 drives are known to run hot and their massive transfer rates causes drive temperatures to rapidly rise even further.

I have carried out a series of testing in the LZ7 to try and see whether throttling is a real issue, and also whether storage maximum performance can be maintained under heavy load stress conditions.

The test system being used:

SYSTEM SPEC
LZ7 Case
140mm Prolimatech Case Fan
Corsair SF450 PSU
Gigabyte B150N Pheonix-Wifi motherboard
Intel i5-6500
Noctua NH-L9i
8Gb Dual Channel 2133Mhz DDR4 RAM
Gigabyte GTX 960 4G OC

STORAGE DRIVE
Samsung PM951 256Gb M.2 PCIe SSD
Claimed Sequential Read/Write = 1000/280 (MB/s)

TEST SETUP
The PM951 SSD was benchmarked under various system conditions, including:
  • Idle system
  • Heavy load system (Simultaneous Prime 95 + Unigine Valley)
  • Various case fan speeds (0%, 50% and 100% speed)
  • Underside vents open
  • Underside vents closed

Each test state allowed the system and drive temperatures to settle before benchmarking was carried out.

Temperatures of the PM951 were recorded under each condition at idle drive state and also the maximum drive temp recorded during the CrystalDiskMark benchmarking test. The sequential read speed was recorded from each test to determine how much (if any) the drive throttled.

RESULTS


OBSERVATIONS

IDLE SYSTEM STATE
Straight away you can see that the PM951 exceeds its claimed sequential read speed by around 50% under normal conditions reaching a massive 1,576MB/s. Great stuff!

During the testing it was observed that throttling occurs in 2 steps, the first step is when the drive hits 70C, the PM951 throttles down to 315MB/s, around one third of its claimed 1000MB/s rating. The second step occurs at 75C where the drive takes a massive dip down to just 71MB/s. it may drop further above 80C but the drive never exceeded this temp.

Focusing on the M.2 Idle temps, you can see that without any active case cooling (0 rpm case fan) the M.2 Drive sits at around 60C when the system is completely idle. This is a worrying temperature sitting just 10C below its throttle temp without any effort.

Activating the system fan to 50% speed reduces the M.2 idle temp down by around 10C, taking the fan up to 100% brings it down a further 10C sitting at a more comfortable 38C.

Running the CrystalDiskmark test caused the PM951 to reach 72C and throttle to 315MB/s without any active cooling, it took less than a minute to pass 70C. With the case fan set to 100% the PM951 reached 59C maximum and maintained its full performance.

Now for the interesting part, covering the vents had the opposite effect to what I was expecting, the PM951 drive temperatures actually reduced in both idle and benchmarking conditions in the region of 5C - 10C. I will talk more about this later.

FULL SYSTEM STRESS STATE
Whilst the system is being put under heavy load and without any active case cooling the system temperature reaches a toasty 57C, the noctua NH-L9i is also nearly flat out trying to keep the CPU under control. Under these conditions the PM951 drive sits at a disturbing idle temperature of 67C, just 3C lower than its throttle point. As soon as the CrystalDisk benchmark is started the drive hits 70C and throttles, and within a minute the drive hits its 75C limit resulting in a read speed of just 71MB/s and maxing out at 80C by the end of the test.

A system case fan speed set to 50% (753 rpm) brings the drive temperature down by 10C during disk benchmarking, but it still experiences some throttling achieving a read speed of 607MB/s.

Taking the fan speed up to 100% (1072 rpm) reduces the idle drive temp by around 10C but doesn't bring the benchmark temperature down below 70C, however it does allow the drive to achieve almost full performance during the disk benchmarking at 1515MB/s.

Again, covering the underside vents up has a positive effect on M.2 drive temperatures, this is completely unexpected and will be discussed in the next section.

UNDER SIDE VENTS
The bottom panel of the LZ7 has a series of vents under the motherboard, the intention of these vents is to allow hot positive pressure air to escape out and apply a cooling effect to M.2 drives mounted on the underside of a motherboard.


However, from the testing carried out so far, it has been observed that covering up vents actually results in lower M.2 drive temperatures in almost all conditions. This is the complete opposite of what was expected.

I have put some serious thought into why this could be, and I think there are 2 possible reasons both of which may be contributing:

REASON 1
On the Gigabyte B150N board, the M.2 Drive is located along the edge near to the GPU which is the opposite side of the board to the case fan. The underside vents sit in between the fan and the drive, it is very possible that the fresh cool air from the case fan is escaping through the vents before it reaches the drive.

Covering the vents forces the cool air from the fan to instead travel all the way along the underside of the motherboard so it reaches the M.2 Drive, the air cools the drive and the heat is then carried toward the GPU side and exits the case.

A possible improvement to the design would be to remove the 'square' vent arrangement (middle of the picture above), and instead increase the width of the GPU underside vents (seen on the left of the picture above).

This theory explains the large difference in temperatures when the fan is running for open and closed vents.

REASON 2
The GPU vents directly out the bottom of the case, it is possible that some of this heat will make its way under the case and travel towards the middle vents, this hot air may be seeping upwards back into the case through the vents and contributing to the M.2 drive temperature. Covering the vents may be preventing this re-circulation from happening.

This theory explains why there is a small difference in temps when the case fan is switched off between open and closed vents.

CONCLUSION
Hopefully your still with me, sorry for the long post!

Without any active case cooling the M.2 PCIe PM951 SSD idles at around 60C with a system temp in the high 40's, this temperature rises to around 67C when the system is put under heavy load. With a throttle temp of 70C it doesn't take much storage activity before throttling occurs with read speeds dropping from 1576MB/s to 315MB/s in a low loaded system and as low as 71MB/s in a heavy loaded system.

Add in some active case cooling and the M.2 drive temperatures reduce to a more acceptable level and are able to sustain their full performance for much longer periods.

Under normal system usage with some level of active case cooling I don't expect that an M.2 PCIe drive will throttle unless you are transferring some really large files one after another.

However, a couple of hours of gaming with higher system temps the PM951 PCIe SSD does begin to struggle, even with good case cooling and ventilation the drive can hit its throttle temp very quickly during sequential reading.

Part of the reason for these high temperatures are that the drives are tucked away under the motherboard and shielded from most of the direct the case cooling. Directing airflow under the motherboard is intended to help with this.

Testing has shown that airflow under the motherboard is key to maintaining full drive performance and vent position play a part in directing the airflow, therefore a small design change will be made to the underside ventilation layout to improve M.2 Drive thermal performance.


*EDIT:
I forgot to mention that this testing was done in 25C ambient thanks to a brief encounter with something called a 'summer' in the UK.
 
Last edited:

iFreilicht

FlexATX Authority
Feb 28, 2015
3,243
2,357
freilite.com
What would be very interesting is how results looked without a fan but with the case in vertical orientation (GPU at the bottom like a normal tower). I would suspect that passive airflow could help the temperatures even if no fan was installed.
 

K888D

SFF Guru
Original poster
Lazer3D
Feb 23, 2016
1,465
2,925
www.lazer3d.com
What would be very interesting is how results looked without a fan but with the case in vertical orientation (GPU at the bottom like a normal tower). I would suspect that passive airflow could help the temperatures even if no fan was installed.

I will give it a go and let you know.
 

Phuncz

Lord of the Boards
Editorial Staff
Moderator
Gold Supporter
May 9, 2015
5,377
4,662
Very well done on the extensive testing !

But in my opinion, you are putting too much emphasis on the M.2 temperature throttling aspect. Just for another metric I'll do some tests:

My 950 Pro sits at 62°C idle right now, with 26°C ambient.

I fire off Malwarebytes Anti-Malware to scan my file system. When it's done checking the "file system" part and switches to heuristic analysis, it's temperature was still 62°C. Limited by my CPU most likely as I see it between 25 and 40% CPU load.

I do a cleanup of my C: drive with Windows' included utlity. I have no CCleaner or other cleaning software installed, I just depend on Windows Cleanup Utility and do it manually every few months. Scanning takes about 2 seconds, scanning for system files takes about 4 seconds and cleaning everything up (some few dozens of MB in most of the listed items, a lot of temporary files) takes about maybe 10 seconds. Drive temperature didn't budge.

I'll let Steam check my GTA 5 file integrity, 67GB of game data at about 610MB/s while my CPU is at 50% load (checksum calculation most likely). At about 80% in, it hits 70°C and half a minute later it throttles to 330MB/s read. I check my DOOM install at 55GB, the same behavior. Still, that 120GB of data is still read in a few minutes.

I copy a 5GB file to my desktop from my NAS. I'll copy that file from my desktop to another folder on my desktop. It takes about 5 seconds and my drive temp shoots up to 68°C for a short time. I don't have any other faster-than-SATA storage available, which isn't very common on SFF anyway.

I launch Diablo 3, play three different areas, ALT-F4'ed out, temperature is still same as idle.
I play GTA 5, grab the nearest sports car (had to chase one), run across the city at near max speed, dragged along some cops, stole some more cars, ALT-F4'ed out and the SSD was still at idle temps.

---------------------------------------------

While doing all this, I regularly saw my CPU go up, along with my GPU, a lot more than my SSD. Although both didn't throttle, I sure can feel the limits of both and the potential thermal throttle a lot more than I ever can with my SSD. My point is that it's easy to max out a GPU and sometimes a CPU while doing AAA games or doing some cleanup, but it's quite the challenge to get the SSD to feel limited during most of these periods.

It's the same thing with FurMark: it puts an unrealistic load on a component, many people take that huge graph jump as a real world problem and make wrong conclusions.

Maybe, in 5 to 7 years when PCIe storage is more common, maybe we'll see software actually using the capabilities where this might be a problem. But most laptops and desktop PCs sold still don't have a SATA SSD as a standard configuration.

tl;dr

Relax, it's not a problem.
 

K888D

SFF Guru
Original poster
Lazer3D
Feb 23, 2016
1,465
2,925
www.lazer3d.com
Very well done on the extensive testing !

But in my opinion, you are putting too much emphasis on the M.2 temperature throttling aspect. Just for another metric I'll do some tests:

My 950 Pro sits at 62°C idle right now, with 26°C ambient.

I fire off Malwarebytes Anti-Malware to scan my file system. When it's done checking the "file system" part and switches to heuristic analysis, it's temperature was still 62°C. Limited by my CPU most likely as I see it between 25 and 40% CPU load.

I do a cleanup of my C: drive with Windows' included utlity. I have no CCleaner or other cleaning software installed, I just depend on Windows Cleanup Utility and do it manually every few months. Scanning takes about 2 seconds, scanning for system files takes about 4 seconds and cleaning everything up (some few dozens of MB in most of the listed items, a lot of temporary files) takes about maybe 10 seconds. Drive temperature didn't budge.

I'll let Steam check my GTA 5 file integrity, 67GB of game data at about 610MB/s while my CPU is at 50% load (checksum calculation most likely). At about 80% in, it hits 70°C and half a minute later it throttles to 330MB/s read. I check my DOOM install at 55GB, the same behavior. Still, that 120GB of data is still read in a few minutes.

I copy a 5GB file to my desktop from my NAS. I'll copy that file from my desktop to another folder on my desktop. It takes about 5 seconds and my drive temp shoots up to 68°C for a short time. I don't have any other faster-than-SATA storage available, which isn't very common on SFF anyway.

I launch Diablo 3, play three different areas, ALT-F4'ed out, temperature is still same as idle.
I play GTA 5, grab the nearest sports car (had to chase one), run across the city at near max speed, dragged along some cops, stole some more cars, ALT-F4'ed out and the SSD was still at idle temps.

---------------------------------------------

While doing all this, I regularly saw my CPU go up, along with my GPU, a lot more than my SSD. Although both didn't throttle, I sure can feel the limits of both and the potential thermal throttle a lot more than I ever can with my SSD. My point is that it's easy to max out a GPU and sometimes a CPU while doing AAA games or doing some cleanup, but it's quite the challenge to get the SSD to feel limited during most of these periods.

It's the same thing with FurMark: it puts an unrealistic load on a component, many people take that huge graph jump as a real world problem and make wrong conclusions.

Maybe, in 5 to 7 years when PCIe storage is more common, maybe we'll see software actually using the capabilities where this might be a problem. But most laptops and desktop PCs sold still don't have a SATA SSD as a standard configuration.

tl;dr

Relax, it's not a problem.

I totally agree with what you are saying. I don't think throttling is an issue for these types of setups with active cooling. I am assuming that you are using a case with good airflow and case fans? Did I also see in your other post your M.2 SSD is mounted on the top of your motherboard and not the underside? If so then your usage may be close to best case for motherboard mounted storage.

I have seen in many other forums and threads people who are interested in buying a small case but are worried about thermal performance of their M.2 drive because of heat build up under the motherboard, I've also seen in some threads people are worried about 2.5" drives getting too hot as well.

I wanted to see if these concerns are actually real, and if so would the LZ7 be able to provide good enough cooling to counter the potential problem.

When the LZ7 is operated without its case fan then the M.2 drive definitely runs too close to its throttle temp, in some tests the drive hits 67°C before the disk benchmark is even started. With the case fan turned on however, M.2 drive performance is not an issue.

From the feedback received so far in the SURVEY for the LZ7, this is what people will most likely be putting inside the case:
  • 73% of people will be using a CPU greater than 70W
  • 100% of people will be using a GPU rated higher than 120W
  • 55% of people will be using a GPU rated higher than 150W
  • 73% of people will be using motherboard mounted storage
This means that most people's setups will be more tortuous than what I have been testing so far (65W CPU and 120W GPU). If I wanted to claim that this case provides good motherboard mounted storage cooling, then I had to check first. I think I can confidently say that it does.
 
  • Like
Reactions: Phuncz

MarcParis

Spatial Philosopher
Apr 1, 2016
3,469
2,504
Super review k888d!!

The most interesting point was about bottom vents and their unexpected effect on cooling..:) You also just demonstrate, once again, that passive cooling has its clear limit..:)
 

Phuncz

Lord of the Boards
Editorial Staff
Moderator
Gold Supporter
May 9, 2015
5,377
4,662
I'm hoping the M.2 PCIe SSDs will atleast get a strip of aluminium to better wick away the temperature of the bare controller chip, secondly to protect from possible EMI (don't know if it could be an issue), thirdly to protect the PCB from ESD when handling it and fourthly because I want that ugly barcode-laden sticker on the back where it belongs.

Plextor is probably going to be the first one doing this, but I hope others will follow suit. While this might be a problem for notebook users that potentially don't have the space for the added Z-depth, that's mostly OEM territory anyway but I don't think it would be very hard to offer a heatsink-less version for notebooks.

I am assuming that you are using a case with good airflow and case fans? Did I also see in your other post your M.2 SSD is mounted on the top of your motherboard and not the underside? If so then your usage may be close to best case for motherboard mounted storage.
It's on an Asus VII Impact board, which has a daughterboard for M.2 just above the GPU slot:



Theoretically it should get airflow from the CPU but Asus summized that 2,5GB/s of throughput would never need cooling and also placed the sound daughterboard about 5mm above it. You see the edge of the PCB just to the right on the picture above.

And then Asus decided it needed an efficient way to secure the new generation of 32Gbps SSDs:



So the controller, the chip on the left, is blocked by this aluminium wrap-around shield that actually holds the M.2 into place. So it's just as blocked, since the NAND chips themselves don't heat up as much: http://www.guru3d.com/articles-pages/samsung-950-pro-m-2-ssd-review,6.html
Before I got the drive I was contemplating on using a thermal pad between it and the shield, but since I don't notice the thermal throttling in my usage ever *, I just left it stock.
 
Last edited:

MarcParis

Spatial Philosopher
Apr 1, 2016
3,469
2,504
I agree with you I had the same space constraint on asus impact vi...however seemed to find a solution on asus impact viii...forget m.2 slot and go for U.2 plug...:(
At least there is one existing u.2 drive, but hard to find...however it's better than sata express drive...completely unexistant...:)
 

K888D

SFF Guru
Original poster
Lazer3D
Feb 23, 2016
1,465
2,925
www.lazer3d.com
Thank you to everybody who has filled in the LZ7 feedback SURVEY, I will keep it open for anybody who hasn't had chance to fill it in yet.

Here are the results for the LZ7 feedback survey, hopefully anyone else who is developing a case may find this data useful as well.

The first part is focused on how the LZ7 would be used and what type of system would be built into the case:

 
  • Like
Reactions: Phuncz