Jonsbo N1 NAS Case + ASUS IOT H310i-IM-A R2.0 Industrial MOBO Custom NAS Assembly Review

[Peter Suh]

This time, it will be uploaded to the hardware section after upgrading from the 4th generation Haswell i3-4160T to the 9th generation Coffee Lake G5500,

which has been used as the main XPE for many years.

This is the H310i-IM-AR2.0 industrial motherboard of ASUS IOT, which starts at $197 7 months ago.

 

 

Like industrial boards, it has a lot of serial ports, but it was useless to me as I didn't know how to use it at the time. Haha.

As you can see, the m.2 wifi/M-Sata slot is visible on the upper left. Among them, M-Sata prepared another 128GB SSD from Samsung and utilized it as a read-write cache.

Two built-in Satas are visible in the lower right corner, three for M.2 NVMe slots, and four for M-Sata slots.

 

There are two serials, one HDMI, two DPs, two LANs, and four USB 3.0 ports on the back.

 

These are the parts to be replaced with the JONSBO N1 body equipped with the Haswell board, which is today's replacement target.

The cooler is a bit over-spec, but I have prepared a Noctua NH-L9x65.

Thermal paste uses ZF extreme, which is the highest virtual level, among the available THERMAGIC types.

I am planning to use one 256GB KIOXIA M.2 NVMe as a read-only cache for Volume 2, but I am worried about heat generation and rear interference,

so I have prepared a review M2F08 M.2 laptop heat sink.

At first, I was going to use Samsung M-SATA 256GB, but I changed it to 128GB because it was out of order.

I know it doesn't work in the synology, but I tried to install it under M-SATA because one of the m.2 Bluetooth/WIFI modules, DW1830, was left.You can't see it because it's covered.)

 

 

This is the Gigabyte Z87N-WIFI Intel Dual-LAN ITX motherboard for nearly two years.

This board was also a model that was very difficult to find in the used market, but I'm attached to it because I even paid extra money to buy it.

It's hard to see because it's covered by a cooler, but it has a lot of high-end capacitors, so I've been using it stably.

 

The KIOXIA M.2 NVMe is installed on the exclusive heat sink for the M2F08 M.2 laptop.

There was only one shake on the front end of the memory module, so if it was installed as it is, there was no gap in the rear part, so there was a possibility of board interference again.

Therefore, the heat-resistant rubber ring shown on the screen was inserted in the middle to adjust the spacing.

 

This is the back of the review M2F08M.2 laptop heat sink with KIOXIA M.2 NVMe.

 

Apply THERMAGIC ZF Extreme Thermal Paste.

I've used it a few times in the past, but it's a thermal paste that I can't really get used to. It spreads so poorly that everything sticks to the spatula.

It's a little suspicious that the liquid ingredient has dried up in the thermal paste.

 

 

It is the process of assembling a Noctua cooler.

It's my first time using Noctua, but I understand why it's Noctua.

I couldn't upload the cooler packaging on the photo, but the completeness of each package and the subform is no joke.

There is also a separate AMD kit and sufficient thermal paste, and a cable for controlling cooler speed is also attached, but it is omitted from the picture.

 

 

It looks like it is equipped with Noktua's cooler backplace.

It's said that it'll be a big trouble if you put it on the front and back, but in fact, it doesn't seem to make a difference^^

 

 

It looks like a mounting bar with a cooler attached to it.

You can change the shape by turning it 90 degrees, considering the direction of the wind.

In my case, the Jonsbo N1 basic cooler blew in from the front to the back, and I decided the direction as follows.

 

 

H310i-IM-A seems to be installed fine on Jonsbo N1.

The case is cramped, so it's a little easier if you push it gently with a back panel inserted in the back of the board.

Install the Dell Perc H200 card and cover the five hard drives installed in the Jonsbo N1. 

 

 

This is what I saw from the top of the Jonsbo N1.

 

 

This is what I saw from the right side of the Jonsbo N1.

 

 

Finally, I applied it with SSD/M.2 NVMe cache.

I thought the SSD and M.2 NVMe would be bundled with RAID1 in read-write mode, but I couldn't create a cache like that.

DS3622xs+ does not show NVMe officially supported or not in the outline.

 

 

 

Edited October 31, 2022 by Peter Suh

[IG-88]

nice compact build

 

the cpu fan has not much space to suck in air, might run on higher rpm because of that, also kind of contradicting the airflow as it pushes warm air in both directions (front an back), maybe inverting the direction and drill some holes in the side make the cpu fan push air out of the housing?

if the front fan before the disks is pushing air in, then it might interesting to turn the psu to have the psu fan face inside the housing and push air out from the housing instead of taking in fresh air and just blow it out in the back without having any "use" for venting the case

 

in my tests the lsi sas controller seemed to have less performace as the write back cache of the disks can't be enabled as its possible with ahci controller (but i did not test this thoroughly)

as you already have two sata ports onboard it might be possible to swap the lsi for a jmb585 (you only would need 3 ports), might also lower the power consumption (and heat) as the lsi does take some power compared the a jmb585 (or asm1166) and the two lanes pcie 3.0 would be good enough

 

the m.2 nvme will not make much difference as its just read cache and in most cases there is not much gain in that and even with two m.2 nvme and r/w cache - without 10G nic (or at least 2.5G nic) there will be no usable up in performance as the 1G nic will limit the transfers

 

there are m.2 to pcie 4x adapters with 20-30cm cable that can be used to attach a standard pcie card, might be o option the replace the m.2 nvme to a 10g nic (its also possible to use that for a storage controller but if there a re any stability problems with that construct a nic is less risky

there might me more usable performance by replacing the nvme ssd with 10G nic and having that directly connected to a 10G nic on the "main computer" (and let other things like media streaming run through 1G or wifi)

and if you want you can still squeeze one or two 2.5" sata ssd's and there will still be some gains

 

14 hours ago, Peter Suh said:

DS3622xs+ does not show NVMe officially supported or not in the outline.

i'm sure its supported as syno sells pcie cards for that as optional components

https://www.synology.com/en-global/products/DS3622xs+ (also mentioned in the spec sheet)

"...

Accelerate storage performance

Enable SSD caching to speed up HDD arrays with an M.2 NVMe SSD adapter card and SNV3000 series SSDs, or install SAT5200 SATA SSDs in drive bays to increase performance..."

https://www.synology.com/en-global/products/M2D20

[dimakv2014]

20 часов назад, Peter Suh сказал:

This time, it will be uploaded to the hardware section after upgrading from the 4th generation Haswell i3-4160T to the 9th generation Coffee Lake G5500,

which has been used as the main XPE for many years.

This is the H310i-IM-AR2.0 industrial motherboard of ASUS IOT, which starts at $197 7 months ago.

 

 

Like industrial boards, it has a lot of serial ports, but it was useless to me as I didn't know how to use it at the time. Haha.

As you can see, the m.2 wifi/M-Sata slot is visible on the upper left. Among them, M-Sata prepared another 128GB SSD from Samsung and utilized it as a read-write cache.

Two built-in Satas are visible in the lower right corner, three for M.2 NVMe slots, and four for M-Sata slots.

 

There are two serials, one HDMI, two DPs, two LANs, and four USB 3.0 ports on the back.

 

These are the parts to be replaced with the JONSBO N1 body equipped with the Haswell board, which is today's replacement target.

The cooler is a bit over-spec, but I have prepared a Noctua NH-L9x65.

Thermal paste uses ZF extreme, which is the highest virtual level, among the available THERMAGIC types.

I am planning to use one 256GB KIOXIA M.2 NVMe as a read-only cache for Volume 2, but I am worried about heat generation and rear interference,

so I have prepared a review M2F08 M.2 laptop heat sink.

At first, I was going to use Samsung M-SATA 256GB, but I changed it to 128GB because it was out of order.

I know it doesn't work in the synology, but I tried to install it under M-SATA because one of the m.2 Bluetooth/WIFI modules, DW1830, was left.You can't see it because it's covered.)

 

 

This is the Gigabyte Z87N-WIFI Intel Dual-LAN ITX motherboard for nearly two years.

This board was also a model that was very difficult to find in the used market, but I'm attached to it because I even paid extra money to buy it.

It's hard to see because it's covered by a cooler, but it has a lot of high-end capacitors, so I've been using it stably.

 

The KIOXIA M.2 NVMe is installed on the exclusive heat sink for the M2F08 M.2 laptop.

There was only one shake on the front end of the memory module, so if it was installed as it is, there was no gap in the rear part, so there was a possibility of board interference again.

Therefore, the heat-resistant rubber ring shown on the screen was inserted in the middle to adjust the spacing.

 

This is the back of the review M2F08M.2 laptop heat sink with KIOXIA M.2 NVMe.

 

Apply THERMAGIC ZF Extreme Thermal Paste.

I've used it a few times in the past, but it's a thermal paste that I can't really get used to. It spreads so poorly that everything sticks to the spatula.

It's a little suspicious that the liquid ingredient has dried up in the thermal paste.

 

 

It is the process of assembling a Noctua cooler.

It's my first time using Noctua, but I understand why it's Noctua.

I couldn't upload the cooler packaging on the photo, but the completeness of each package and the subform is no joke.

There is also a separate AMD kit and sufficient thermal paste, and a cable for controlling cooler speed is also attached, but it is omitted from the picture.

 

 

It looks like it is equipped with Noktua's cooler backplace.

It's said that it'll be a big trouble if you put it on the front and back, but in fact, it doesn't seem to make a difference^^

 

 

It looks like a mounting bar with a cooler attached to it.

You can change the shape by turning it 90 degrees, considering the direction of the wind.

In my case, the Jonsbo N1 basic cooler blew in from the front to the back, and I decided the direction as follows.

 

 

H310i-IM-A seems to be installed fine on Jonsbo N1.

The case is cramped, so it's a little easier if you push it gently with a back panel inserted in the back of the board.

Install the Dell Perc H200 card and cover the five hard drives installed in the Jonsbo N1. 

 

 

This is what I saw from the top of the Jonsbo N1.

 

 

This is what I saw from the right side of the Jonsbo N1.

 

 

Finally, I applied it with SSD/M.2 NVMe cache.

I thought the SSD and M.2 NVMe would be bundled with RAID1 in read-write mode, but I couldn't create a cache like that.

DS3622xs+ does not show NVMe officially supported or not in the outline.

 

 

 

Nice setup but kind of expensive with no energy saving, ASRock J4125B-ITX motherboard would be cheaper and it already has J4125 cpu preinstalled, it has only 10watt TDP (25% less perfomance difference compared to G5500 with 54watts) but it can run fanless, no m.2 slot however, but still nice setup :)

[Peter Suh]

6 hours ago, IG-88 said:

nice compact build

 

the cpu fan has not much space to suck in air, might run on higher rpm because of that, also kind of contradicting the airflow as it pushes warm air in both directions (front an back), maybe inverting the direction and drill some holes in the side make the cpu fan push air out of the housing?

if the front fan before the disks is pushing air in, then it might interesting to turn the psu to have the psu fan face inside the housing and push air out from the housing instead of taking in fresh air and just blow it out in the back without having any "use" for venting the case

 

in my tests the lsi sas controller seemed to have less performace as the write back cache of the disks can't be enabled as its possible with ahci controller (but i did not test this thoroughly)

as you already have two sata ports onboard it might be possible to swap the lsi for a jmb585 (you only would need 3 ports), might also lower the power consumption (and heat) as the lsi does take some power compared the a jmb585 (or asm1166) and the two lanes pcie 3.0 would be good enough

 

the m.2 nvme will not make much difference as its just read cache and in most cases there is not much gain in that and even with two m.2 nvme and r/w cache - without 10G nic (or at least 2.5G nic) there will be no usable up in performance as the 1G nic will limit the transfers

 

there are m.2 to pcie 4x adapters with 20-30cm cable that can be used to attach a standard pcie card, might be o option the replace the m.2 nvme to a 10g nic (its also possible to use that for a storage controller but if there a re any stability problems with that construct a nic is less risky

there might me more usable performance by replacing the nvme ssd with 10G nic and having that directly connected to a 10G nic on the "main computer" (and let other things like media streaming run through 1G or wifi)

and if you want you can still squeeze one or two 2.5" sata ssd's and there will still be some gains

 

i'm sure its supported as syno sells pcie cards for that as optional components

https://www.synology.com/en-global/products/DS3622xs+ (also mentioned in the spec sheet)

"...

Accelerate storage performance

Enable SSD caching to speed up HDD arrays with an M.2 NVMe SSD adapter card and SNV3000 series SSDs, or install SAT5200 SATA SSDs in drive bays to increase performance..."

https://www.synology.com/en-global/products/M2D20

 

 

@IG-88 Thanks for the advice on making a better NAS.

 

I use XPE for home use as a PLEX DIRECT streaming server and file server, and it is rare that large amounts of data are moved between NAS and PC.

Seven months later, Jonsbo N1 specs have changed.
As you advised, there is not much space in the N1 case, so the Noctua cooler doesn't seem to be of much use, so I changed to the Intel 1151 type bundled cooler.

 

 

Noctua NH-L9x65 is currently installed on my main Hackintosh PC, below 11th gen motherboard.
https://github.com/PeterSuh-Q3/Gigabyte-Z590i-Aorus-Ultra-OC8.4-Monterey

 

Currently, the front fan is in charge of cooling at high speed, and the temperature of the 5 HDDs is operating satisfactorily coldly.
There are wide side perforated holes on both sides of the chassis, and a lot of warm air flows out when you touch the perforated holes.
Perhaps the structure of the front fan and PSU fan is concentrated toward the central CPU and exits through the CPU through the hole on both sides.

 

The HBA, Dell Perc H200, also has a problem with overheating, and there is a problem that it cannot work with Apollo Lake, Gemini Lake, and V1000 platforms.
So I replaced it with the AHCI SATA controller of the Marvel chipset with 6 SATA ports purchased from Ali Express, which I had been using since the DSM 6.
In REDPILL, only 1~4 out of 6 ports operate. The other two ports do not work.

 

 

Currently not using M.2 NVMe.
TCRP FRIEND caused a problem and the M.2 slot is empty with the detached state.
As you said, the M.2 slot is equipped with an AHCI SATA controller to further expand the two SATA ports.

 

If this happens, there will be room for the PCIE slot, and it would be good to use a network controller for this part.
However, as mentioned above, since it is a home NAS, 10G NIC seems to be over-spec.
Peripherals including my switch and PCs are still staying on 1G Nic.
The Intel 11th Gen motherboards mentioned above have an Intel Chipset NIC with 2.5 Gbps.
For network communication with this PC, I have only a plan to install it as a NIC of 2.5G.

 

When this happens, I plan to make the disk cache two 2.5" sata ssd.
It's still for home use, so buying a genuine Synology M.2 NVMe controller for cash is burdensome.^^

[Peter Suh]

55 minutes ago, dimakv2014 said:

Nice setup but kind of expensive with no energy saving, ASRock J4125B-ITX motherboard would be cheaper and it already has J4125 cpu preinstalled, it has only 10watt TDP (25% less perfomance difference compared to G5500 with 54watts) but it can run fanless, no m.2 slot however, but still nice setup :)

 

I also want to consider energy saving, but energy saving doesn't mean much since I'm already using 5 HDDs, 1 SDD, and 1 M-Sata.
Even if the CPU is changed to a lower power consumption and the cooler disappears, the overall power consumption is not significantly different.
Electricity is expensive in Korea. So I have to think about possible energy savings, but at the moment I have no choice.

[dimakv2014]

10 минут назад, Peter Suh сказал:

 

I also want to consider energy saving, but energy saving doesn't mean much since I'm already using 5 HDDs, 1 SDD, and 1 M-Sata.
Even if the CPU is changed to a lower power consumption and the cooler disappears, the overall power consumption is not significantly different.
Electricity is expensive in Korea. So I have to think about possible energy savings, but at the moment I have no choice.

Well it still minus 44watts and minus cpu cooler which drains about 2-3 watts) plus motherboard with cpu around 120$ so its economy for sure, but since you have no choice its a different story, the main thing that it works for you

 

[Peter Suh]

44 minutes ago, dimakv2014 said:

Well it still minus 44watts and minus cpu cooler which drains about 2-3 watts) plus motherboard with cpu around 120$ so its economy for sure, but since you have no choice its a different story, the main thing that it works for you

 

 

I have a power consumption meter.
Have you ever measured the difference between the theoretical power consumption of the CPU and the actual power consumption?
From what I have experienced in the past, there was a difference between the theoretical and actual figures.

 

I haven't been able to measure the power consumption of my N1 NAS for a long time.
And, I have a separate J1900 MOBO of 10W.
This board also doesn't have a CPU cooler.
Let's configure this J1900 MOBO under the same conditions as N1 and measure the difference in power consumption between the two NAS.

 

It will be a fun experiment.^^

[Peter Suh]

The power consumption measured by Jonsbo N1Case and ASUS IOT H310Mi-IM-A R2.0 + MOBO + G5500 introduced in the review above and
The separately prepared ECS BAT-I V1.2 J1900 discloses the measured power consumption under the same disk conditions.

 

 

Because the captured photo is a Korean-style power meter, Hangul is included.
Each item will be explained in advance.
Power consumption is in the upper left, voltage is in the upper right, current is in the lower left, and usage time is in the lower right.

 

As you can see in the picture, my main XPE Jonsbo N1, G5500-54W (1 cache SSD, 5 3.5 inch HDD) consumes 45.2 W in IDLE.

 

 

 

 

ECS MOBO's J1900-10W (1 cache SSD, 5 3.5-inch HDDs) used in the test consumes 42.8 W in IDLE.

 

 

The difference in power consumption between the two XPEs is 2.4 W, which is not much difference.

 

For a long time, I've seen the XPE community buy low-power CPUs at high prices.

 

Even among users, it is said that there is no need to insist on a low-power CPU.

 

If possible, low power is advantageous, but expensive,
It might be worth thinking about whether a lower-power CPU that pays a higher price actually saves a lot of power.

Edited November 1, 2022 by Peter Suh

[superma87]

TDP is not power consumption, so @dimakv2014considerations are useless.

 

Plus, newer CPUs have a way better power management than old ones or low power ones(those does not have the need to cut power even more)

 

Please refer to this video to see what really matters about power consumption https://www.youtube.com/watch?v=MucGkPUMjNo

 

[dimakv2014]

В 01.11.2022 в 14:14, Peter Suh сказал:

The power consumption measured by Jonsbo N1Case and ASUS IOT H310Mi-IM-A R2.0 + MOBO + G5500 introduced in the review above and
The separately prepared ECS BAT-I V1.2 J1900 discloses the measured power consumption under the same disk conditions.

 

 

Because the captured photo is a Korean-style power meter, Hangul is included.
Each item will be explained in advance.
Power consumption is in the upper left, voltage is in the upper right, current is in the lower left, and usage time is in the lower right.

 

As you can see in the picture, my main XPE Jonsbo N1, G5500-54W (1 cache SSD, 5 3.5 inch HDD) consumes 45.2 W in IDLE.

 

 

 

 

ECS MOBO's J1900-10W (1 cache SSD, 5 3.5-inch HDDs) used in the test consumes 42.8 W in IDLE.

 

 

The difference in power consumption between the two XPEs is 2.4 W, which is not much difference.

 

For a long time, I've seen the XPE community buy low-power CPUs at high prices.

 

Even among users, it is said that there is no need to insist on a low-power CPU.

 

If possible, low power is advantageous, but expensive,
It might be worth thinking about whether a lower-power CPU that pays a higher price actually saves a lot of power.

Good point, but 3.5 hdd consumes around 5-8watt, so basically j1900 on idle is not much less that G5500 but what it shows on load? On maximum load J1900 should show 10watt for cpu only which will still be 44watt less on average and thats already a huge difference, unless you want to use server on idle all the time. But I agree from idle to medium it wouldn't be very big difference, it all depends how much load on it. I personally like low power cpus for being fanless, unfortunately I live among them and extra noise all the time is kind of annoying.

Btw I already built system for DVA3219 with your JOT loader, currently have dva3221 on it without gpu, tomorrow finally will pick up msi gtx1650 gddr5 low profile to test it sorry took me long enough but in a War state its a little hard to get stuff :(...

Also been struggling with old e5-2660v2, dva1622 works on all loaders baremetal or vm through esxi7 but dva3221 doesnt want to, thought maybe its possible to emulate cpu flags/instructions to somehow to make it work, but dont know which will work best for this Proxmox, Unraid or QEMU.

Edited November 3, 2022 by dimakv2014

[Peter Suh]

2 hours ago, dimakv2014 said:

Good point, but 3.5 hdd consumes around 5-8watt, so basically j1900 on idle is not much less that G5500 but what it shows on load? On maximum load J1900 should show 10watt for cpu only which will still be 44watt less on average and thats already a huge difference, unless you want to use server on idle all the time. But I agree from idle to medium it wouldn't be very big difference, it all depends how much load on it. I personally like low power cpus for being fanless, unfortunately I live among them and extra noise all the time is kind of annoying. Btw I already built system for DVA3219 with your JOT loader, currently have dva3221 on it without gpu, tomorrow finally will pick up msi gtx1650 gddr5 low profile to test it :).

Also been struggling with old e5-2660v2, dva1622 works on all loaders baremetal or vm through esxi7 but dva3221 doesnt want to, thought maybe its possible to emulate cpu flags/instructions to somehow to make it work, but dont know which will work best for this Proxmox, Unraid or QEMU.

 

 

Since the J1900 for testing has already been dismantled, it seems difficult at the moment to test the power in the peak state.

Probably not as much as 44 watts at peak, but expect the G5500 / J1900 to see more power differential between the two CPUs.

 

If you use a Surveillance platform such as dva3221/dva3219/dva1622, it is highly likely to operate 24 hours a day, 365 days a year.

In this case, in the case of a home server, the fan noise seems to be more painful.

If you prefer low noise, the fanless products of the J series may be advantageous.

 

When looking for the CPUs of the Surveillance-type genuine Synology models,

both DVA3221 / DVA3219 use Intel Atom C3538, and the GPU is GTX 1650 / GTX 1050 Ti, respectively.

Genuine Synology seems to be similarly fanless. Rather, the GPU coolers seem to make noise while the GPU is actively processing.
Aren't GPU coolers for low-profiles running faster? I've never used a low profile GPU so I don't know.

 

Most of the Xeon CPUs use a lot of power, so the cooler must be good with high performance and the noise is likely to be severe.

Since I'm not in an environment where I'm going to use Surveillance, my biggest concern these days is direct streaming of 4K video from a plex.
Since there is no Android-type set-top such as Shield or Mibox, I watch it through LG Smart TV. I mainly use XPLAY, a WEBOS-based software.
In DSM 6, 4K video was directly streamed, but there were restrictions on high-definition video, but most of it was playable.
After changing to DSM 7, there was a problem that not even a single video could be played.
So, I am preparing for a comparative test by installing DSM 6 separately.

Edited November 3, 2022 by Peter Suh

[dimakv2014]

1 час назад, Peter Suh сказал:

 

 

Since the J1900 for testing has already been dismantled, it seems difficult at the moment to test the power in the peak state.

Probably not as much as 44 watts at peak, but expect the G5500 / J1900 to see more power differential between the two CPUs.

 

If you use a survival platform such as dva3221/dva3219/dva1622, it is highly likely to operate 24 hours a day, 365 days a year.

In this case, in the case of a home server, the fan noise seems to be more painful.

If you prefer low noise, the fanless products of the J series may be advantageous.

 

When looking for the CPUs of the Surveillance-type genuine Synology models,

both DVA3221 / DVA3219 use Intel Atom C3538, and the GPU is GTX 1650 / GTX 1050 Ti, respectively.

Genuine Synology seems to be similarly fanless. Rather, the GPU coolers seem to make noise while the GPU is actively processing.
Aren't GPU coolers for low-profiles running faster? I've never used a low profile GPU so I don't know.

 

Most of the Xeon CPUs use a lot of power, so the cooler must be good with high performance and the noise is likely to be severe.

Since I'm not in an environment where I'm going to use Surveillance, my biggest concern these days is direct streaming of 4K video from a plex.
Since there is no Android-type set-top such as Shield or Mibox, I watch it through LG Smart TV. I mainly use XPLAY, a WEBOS-based software.
In DSM 6, 4K video was directly streamed, but there were restrictions on high-definition video, but most of it was playable.
After changing to DSM 7, there was a problem that not even a single video could be played.
So, we are preparing for a comparative test by installing DSM 6 separately.

Why I bought low profile because of a slim itx case with tfx power supply and it can only fit low profile gpus and has no pci-e power connectors so power from pcie lane only, it was two choices gigabyte or msi but since my other normal size MSI gtx1660 gpu works fine with dva and msi was cheaper choice was to buy it. Will compare how loud it is compared to full size, but full size is loud for sure when SS starts recording on 6 cameras with Face Rec ))) too bad its impossible to control RPM of gpu coolers through bios or dsm... Temps are good so I would better get quieter solution with little bit of temp raise, gpu is never loaded at full, mostly about 47%

 

As for xeons, i wouldn't say noctua coolers are that loud, but its not silent for sure )) 95watts at full load... But I have two cpus just laying around and wanted to make use of them. Baremetal is definitely no good for DVA3221, esxi7 also doesnt want to run it, tried you loader as well no go, hope I can fake instructions using Proxmox, QEMU or Unraid, some has to work I hope )) otherwise will have to stick to bunch of dva1622 VMs for camera licenses in CMS (40 threads with 128gb ram) J4105 or J4125 are completely silent with ssds and no gpu) so comparing xeons pointless )

 

Plex can load cpu and gpu, whatever has to do with transcoding/rendering it will load everything for sure, but thats kind of strange that dsm7 doesn't want to play anything maybe it has to do with the version of plex, I never used plex, I usually go to online cinema website and watch everything there ) maybe I don't understand the point of Plex and Jellyfin in my specific case, but everyone talks about it and its tempting )

 

 

Edited November 3, 2022 by dimakv2014

[dimakv2014]

22 часа назад, Peter Suh сказал:

 

 

Since the J1900 for testing has already been dismantled, it seems difficult at the moment to test the power in the peak state.

Probably not as much as 44 watts at peak, but expect the G5500 / J1900 to see more power differential between the two CPUs.

 

If you use a Surveillance platform such as dva3221/dva3219/dva1622, it is highly likely to operate 24 hours a day, 365 days a year.

In this case, in the case of a home server, the fan noise seems to be more painful.

If you prefer low noise, the fanless products of the J series may be advantageous.

 

When looking for the CPUs of the Surveillance-type genuine Synology models,

both DVA3221 / DVA3219 use Intel Atom C3538, and the GPU is GTX 1650 / GTX 1050 Ti, respectively.

Genuine Synology seems to be similarly fanless. Rather, the GPU coolers seem to make noise while the GPU is actively processing.
Aren't GPU coolers for low-profiles running faster? I've never used a low profile GPU so I don't know.

 

Most of the Xeon CPUs use a lot of power, so the cooler must be good with high performance and the noise is likely to be severe.

Since I'm not in an environment where I'm going to use Surveillance, my biggest concern these days is direct streaming of 4K video from a plex.
Since there is no Android-type set-top such as Shield or Mibox, I watch it through LG Smart TV. I mainly use XPLAY, a WEBOS-based software.
In DSM 6, 4K video was directly streamed, but there were restrictions on high-definition video, but most of it was playable.
After changing to DSM 7, there was a problem that not even a single video could be played.
So, I am preparing for a comparative test by installing DSM 6 separately.

Got gpu, had to reinstall SS first because it gave error first, after that works good.

After 6 hours of testing was surprised by low profile quietness, low temps (highest was 60c in a closed case enviroment) under very heavy load.

Stress test:

6 pseudo cameras with .mp4 files 4K 60fps Face Recognition with a lot of people walking on street, thats like having 12 8MP cameras 30fps each on Face Rec.

VRAM randomly loaded up till 92-99%, gpu however never exceeded 53%.

So my verdict is GTX1650 gddr5 low profile is much quieter than full size GTX1660 despite 2 small coolers which is quiet surprising, didn't hear any noise difference between idle and heavy load on low profile.

Tomorrow will test out DVA3219 finaly)

Edited November 3, 2022 by dimakv2014

[dimakv2014]

В 03.11.2022 в 02:35, Peter Suh сказал:

 

 

Since the J1900 for testing has already been dismantled, it seems difficult at the moment to test the power in the peak state.

Probably not as much as 44 watts at peak, but expect the G5500 / J1900 to see more power differential between the two CPUs.

 

If you use a Surveillance platform such as dva3221/dva3219/dva1622, it is highly likely to operate 24 hours a day, 365 days a year.

In this case, in the case of a home server, the fan noise seems to be more painful.

If you prefer low noise, the fanless products of the J series may be advantageous.

 

When looking for the CPUs of the Surveillance-type genuine Synology models,

both DVA3221 / DVA3219 use Intel Atom C3538, and the GPU is GTX 1650 / GTX 1050 Ti, respectively.

Genuine Synology seems to be similarly fanless. Rather, the GPU coolers seem to make noise while the GPU is actively processing.
Aren't GPU coolers for low-profiles running faster? I've never used a low profile GPU so I don't know.

 

Most of the Xeon CPUs use a lot of power, so the cooler must be good with high performance and the noise is likely to be severe.

Since I'm not in an environment where I'm going to use Surveillance, my biggest concern these days is direct streaming of 4K video from a plex.
Since there is no Android-type set-top such as Shield or Mibox, I watch it through LG Smart TV. I mainly use XPLAY, a WEBOS-based software.
In DSM 6, 4K video was directly streamed, but there were restrictions on high-definition video, but most of it was playable.
After changing to DSM 7, there was a problem that not even a single video could be played.
So, I am preparing for a comparative test by installing DSM 6 separately.

Nice Job Peter!

SUCCESSFUL:

DVA3219 with Friend DSM 7.1.1-42962

Used commands:

curl --location "https://github.com/PeterSuh-Q3/tinycore-redpill/raw/main/my.sh.gz" -O;tar -zxvf my.sh.gz;
./my.sh
./my.sh DVA3219F
sudo reboot

 

AI is NOT working.

Tried to create Face Rec task shows "network is busy"  uninstall/install SS didnt help.

 

Question, how can I edit SN/MACs (it just doesnt allow me to and skips to sata settings) should I press N when it asks if serial is ok?

 

 

 

[Peter Suh]

5 hours ago, dimakv2014 said:

Question, how can I edit SN/MACs (it just doesnt allow me to and skips to sata settings) should I press N when it asks if serial is ok?

 

 

 

If you have a separate SN/MAC
Please edit user_config.json before using my.sh to build the loader.

 

And if you use the noconfig option like below
user_config.json is read and used only
Changes to VID/PID/SN/MAC/SataportMap/DiskIdxMap etc. are not processed.

 

If you do ./my.sh , the above is explained in Help.

 

./my.sh DS3219F noconfig

Edited November 5, 2022 by Peter Suh

[dimakv2014]

32 минуты назад, Peter Suh сказал:

 

 

If you have a separate SN/MAC
Please edit user_config.json before using my.sh to build the loader.

 

And if you use the noconfig option like below
user_config.json is read and used only
Changes to VID/PID/SN/MAC/SataportMap/DiskIdxMap etc. are not processed.

 

If you do ./my.sh , the above is explained in Help.

 

./my.sh DS3219F noconfig

Thanks. Worked it out differently: right before sata config changed user config, AME doesn't help, tried all Mods currently testing Jun mod no luck with AI on gtx1650 still, its locked to Pascal gpus. More detailed info in Development DVA3219 topic 

https://xpenology.com/forum/topic/63946-dva3219-loader-development-thread/page/2/#comment-363096

Edited November 5, 2022 by dimakv2014