0
NicoD

Announcement : Odroid N2

Recommended Posts

Hi all. As regular I was looking for news about a new Odroid.
Finally the news arrived. The N2 will have the Amlogic S922X with 4xA73(@1.8Ghz) + 2xA53(@1.9Ghz), a Mali-G52 GPU and 4GB DDR4 RAM. Looks very promising. It should quite a bit faster than the RK3399's.
It does have a giant heatsink, so this thing will get very hot I suppose.
https://forum.odroid.com/viewtopic.php?f=176&t=33781&p=245930&hilit=n2#p245930

 

afbeelding.thumb.png.a457a9aa08b7eecfee06bc185b401db4.png
Please let me know what you all think of this. I find it very interesting.
Greetings, NicoD

Share this post


Link to post
Share on other sites

Seems like it combines many of the advantages of both XU4 (powerful processor, boot switch, fan connector, USB3) and C2 (standard 3.3v GPIO and UART interfaces, IR receiver, and hopefully low consumption). Apparently you can clock the four A73's up to 2Ghz, which means a lot of horsepower.

 

With regards to the RK3399 boards, seems like we lose the PCIe interface, but the price tag is also lower. Looks like a very interesting board.

Share this post


Link to post
Share on other sites
1 hour ago, JMCC said:

With regards to the RK3399 boards

Indeed. It's got some great advantages, but also some disadvantages. Only 1 USB3 root HUB. Powered with 7.5V-20V.
But the Horsepowers, RAM, power consumption, will be unmatched.

I already made a video about it. Here it is.

 

Share this post


Link to post
Share on other sites

Unfortunately Hardkernel recently degrades as a leader, and make more and more stupid mistakes. Showing comparative tests of a new model from their own unfortunate model N1 on the basis of DDR3 (which they failed to begin to produce), they humiliated themselves by showing that badly designed N1. Their N1 (RK3399) used the old DDR3 memory, and comparing it to the new model with DDR4 is stupid. You need to compare with models RK3399 DDR4, then it will be clear the difference between the chips S922 and RK3399. The lower location of the radiator, which will be in this form to warm the entire Board with the rest of the components above it. This is a questionable solution for a model that should use the full power of processors. Do not understand the rejection of PCi (NVME) and the use of non-standard eMMC modules that have reduced speed and reliability. This model makes a strange impression. 80 for one PCB Board, how much will the working set cost ?

Share this post


Link to post
Share on other sites
1 hour ago, balbes150 said:

they humiliated themselves by showing that badly designed N1.

 

It was not badly designed, still the N1 is a great board you can trust them when they say it was due to the DDR3 ram and if the new N2 reaches the same temperatures like the N1 does, which was nearly nothing then the heat sink should be easily cool the board and you should never see any throttling

Share this post


Link to post
Share on other sites
3 hours ago, balbes150 said:

they humiliated themselves by showing that badly designed N1. Their N1 (RK3399) used the old DDR3 memory, and comparing it to the new model with DDR4 is stupid. You need to compare with models RK3399 DDR4, then it will be clear the difference between the chips S922 and RK3399.

I don't fully agree. It would indeed have been better to see a comparison with a DDR4 RK3399. But what they tried to do is to show that this is a better choice for a next SBC to bring out than the unfortunate N1. So it shows why they went for something different.
 

 

3 hours ago, balbes150 said:

The lower location of the radiator, which will be in this form to warm the entire Board with the rest of the components above it.

The biggest advantage with this board is that everything is on the top of the board and easily accessible. I also think it's best to have it on the underside. As long as you can keep it cool enough. I agree it can heat up the board. I once did a one hour trial with the NanoPi M4 while throttling with the big heatsink. The board started smelling burned. I'm not going to try it again. That's why you need to raise the heatsink so air can flow underneath it.
 

 

3 hours ago, balbes150 said:

Do not understand the rejection of PCi (NVME)

Here I can follow. PCIe, M.2 or SATA would have been good.
But for my use I don't need it.
 

 

3 hours ago, balbes150 said:

and the use of non-standard eMMC modules that have reduced speed and reliability.

Non-standard eMMC modules? They've become the standard, didn't they? And they are among the fastest eMMC modules in SBC land. I've been using them for years and haven't had any problems.
I now use them on my Rock Pi4B without any issues.
 

 

2 hours ago, AreaScout said:

if the new N2 reaches the same temperatures like the N1 does, which was nearly nothing then the heat sink should be easily cool the board and you should never see any throttling

I suspect they are down-clocking their A73 cores to 1.8Ghz so it wouldn't throttle in the temperature tests. I expect it will be able to run a lot higher, but it might need some beefy cooler to keep it cool.  Just a guess. A73 should be able to do a lot more than that.
 

Share this post


Link to post
Share on other sites
33 minutes ago, NicoD said:

Non-standard eMMC modules?

I think the point is, if you could have NVMe or eMMC, go for NVMe.

 

Heatsink over is better than heatsink under due to basic physics.  Less dense hot air rises.  

 

Honestly I'm still waiting for a board that can replace my XU4 as my media server, but since Emby will handle hardware transcoding for the Samsung SoC, that becomes tougher.  

 

What is Amlogic's code number/name for this SoC?  

Share this post


Link to post
Share on other sites
17 minutes ago, TonyMac32 said:

I think the point is, if you could have NVMe or eMMC, go for NVMe.

Rather both if I could choose :)

 

25 minutes ago, TonyMac32 said:

What is Amlogic's code number/name for this SoC?  

Not much info to be found, I guess we'll have to wait for a datasheet. I wonder what the standard clocks are. Hardkernel likes to deviate from that.

I agree it isn't a perfect sbc. But it's again a step up in performance.
I hope the ARM community is still as strong as it once was for Odroid. It's been a while since they've needed to begin from scratch.
But if I need to put a bet on one company to give great software. Then I'll bet on Odroid every time.

We will have to see, but I'll surely buy it. It might be sufficient to replace my aging and power munching i5 2500K. (it can't play 4K H.265)
The H2 is totally forgotten for me.

Share this post


Link to post
Share on other sites
7 hours ago, balbes150 said:

Showing comparative tests of a new model from their own

It depends on the way you look at it. If you have a BMW and you really like that and the new BMW is launched you will compare your current model to the new one. You can sell the old on eBay and get the new Odroid - for example.

Share this post


Link to post
Share on other sites
15 hours ago, AreaScout said:

It was not badly designed, still the N1 is a great board you can trust them when they say it was due to the DDR3 ram and if the new N2 reaches the same temperatures like the N1 does, which was nearly nothing then the heat sink should be easily cool the board and you should never see any throttling

If the device was not it is released. Failure to release-a direct proof of error in the development. The indication as a reason, the lack of DDR3 memory, looks strange. Look at the market of products of TV boxes, a huge number of different models with DDR3 and DDR4, so there is no problem with getting any options. I do not want to discuss here why they did not release N1, it was not a technical reason.

 

12 hours ago, NicoD said:

I don't fully agree. It would indeed have been better to see a comparison with a DDR4 RK3399. But what they tried to do is to show that this is a better choice for a next SBC to bring out than the unfortunate N1. So it shows why they went for something different.

Users when choosing between RK3399 and S922x see this comparison. Most immediately identifies all produced RK3399 (not only N1) with this test and conclude that RK3399 - this junk, much inferior to N2. Few people will be able to estimate the real difference between the chips, if you correct the numbers of the results. Even on the forum says that the difference in DDR3 and DDR4 almost 35%. And now, take and increase the results RK3399 these 30-35% and will not exceed such difference in tests. Do you think the specialists who wrote the review do not know this ? They know it very well and I have no doubt that they have conducted many different tests, but decided to pretend that we compare our products. Let me remind you-not on sale N1 with RK3399 even with DDR3, so this comparison is specially designed and therefore it is a deliberate deception.

 

 

13 hours ago, NicoD said:

Here I can follow. PCIe, M.2 or SATA would have been good.
But for my use I don't need it.

This Board has a powerful processor that will be used to process large amounts of data. This data should be placed on large media (100 GB or More). This requires appropriate media. That eMMC module that Odroid offers is not such (it's a legacy from old systems when there was a big problem with eMMC). The eMMC module, which was invented in Odroid in the old days, is not standard and has an inflated cost. Having it on a new system instead of the standard options (SATA or NVME etc) is a bad decision. This is an attempt to force users to use their products at an inflated price with low characteristics, instead of using standard, and significantly better in parameters, modern devices.

 

13 hours ago, NicoD said:

They've become the standard, didn't they?

No, they didn't become standard. The fact that someone repeats their mistakes and uses them (by the way, to reduce their development and production costs at the expense of others, simply put-steal) does not make these modules a standard. :)

 

12 hours ago, NicoD said:

I hope the ARM community is still as strong as it once was for Odroid. It's been a while since they've needed to begin from scratch.
But if I need to put a bet on one company to give great software. Then I'll bet on Odroid every time.

As they do, everyone decides for himself, I no, do not impose. And this is only my private opinion. Unfortunately, recently they have made a number of serious (in my understanding) mistakes that have greatly changed my opinion about this company for the worse.

 

12 hours ago, NicoD said:

We will have to see, but I'll surely buy it. It might be sufficient to replace my aging and power munching i5 2500K. (it can't play 4K H.265)
The H2 is totally forgotten for me. 

At the moment, I am quite satisfied with RK3399 and the significantly changed position of Rockchip regarding Linux support.

For RK3399 already now, there is everything in working and ready to use form, images Armbian, LE, etc., and what will be with s922, is not yet known, and it is not clear how much time will pass that would be ready-made system.

 

10 hours ago, Tido said:

It depends on the way you look at it. If you have a BMW and you really like that and the new BMW is launched you will compare your current model to the new one. You can sell the old on eBay and get the new Odroid - for example. 

It would be so if N1 (RK3399) was produced and people had a choice what to take. Why didn't they make a comparison with the mass models of RK3399 that have been on the market for a long time ? But they add stunted and retarded RPi :)

Share this post


Link to post
Share on other sites
5 hours ago, balbes150 said:

If the device was not it is released. Failure to release-a direct proof of error in the development. The indication as a reason, the lack of DDR3 memory, looks strange. Look at the market of products of TV boxes, a huge number of different models with DDR3 and DDR4, so there is no problem with getting any options. I do not want to discuss here why they did not release N1, it was not a technical reason.

 

 

Well sometimes a very big competitor is buying all the RAM chips and you get out of parts, that's just how it could be and there are a ton of other possibilities that could happen that force you to stop production, again if HK says that the parts run suddenly out then you can be sure that HK was doing everything that it's not, but at the end they had to react.  And there was also some other reasons 

H2 was a different story but with some slightly comparable "scenes" with parts that suddenly not available,  but this board will be available soon (hopefully) 

Share this post


Link to post
Share on other sites

Regarding the lack of support for NVMe/SATA, I think it is as simple as the fact that the SoC doesn't seem to support them. Amlogic aims to release SoC's that are just enough for a standard TV box, they don't have things such as Chromebooks in mind when they design their SoC's (as, for example, Rockchip does for their higher-end SoC's). And that is the reason why we can buy this board much cheaper than any equivalent RK3399. Here you trade brute horsepower for extra features in the SoC.

 

[UPDATE]: Even though PCIe was never mentioned in the official announcements of S922X (see, for example here and here),  recently someone pointed out that they got a "leak" from Amlogic saying otherwise. According to that "leak", the SoC would have a single-lane PCIe interface, but it is pin muxed with USB3. In other words, you would need to disable USB3 in order to use PCIe, and that would still be single-lane.

 

Definitely, disabling USB3 in favor of a possible SATA/NVMe interface would be IMO a very bad choice for a general-purpose board. Maybe if they release in the future something like the HC1, based on S922X, it could make sense. However, given the fact that Amlogic has not officialy announced that PCIe support, and after the "fiasco" with the broken PCIe in Allwinner H6, I won't believe that these are real options for this SoC until I see it actually working.

 

You can find the whole discussion about the matter in this page from CNXSoft

Share this post


Link to post
Share on other sites

Some 7zip benchmarks. I always do this with all my sbc's, so I know how they compare to each other.
Left they compare it with the N1( @balbes150 not my fault :) )
I look to decompression, the NanoPi M4 does this in betweeen 6500 and 8500(armhf no oc, arm64 with oc)
The best I had until now for this task was the NanoPi T3+(8x1.4Ghz) reached 10100.
That seems to be the result with the "default" clock of the N2. But with OC of 6x 2Ghz it even goes over 11000.

 

All just numbers that I can't confirm if they've been generated right. But an indication of what we can expect. For this 7-zip benchmark the ram speed has no big influence.

afbeelding.png.e03d6ce917c9388ba79e891337f58054.pngafbeelding.png.3e8d7757b9828322eb729d93d3091f35.png


I would like to see what blender does. That's 100% cpu useage + ram speed. In Blender this will blow everything away(I think).

I just got the new Pine H64 and the OPi3 on the way. So these will have to fill the hole this announcement has created. I'm just crazy about fast SoC's for rendering.

Share this post


Link to post
Share on other sites
6 hours ago, rooted said:

OT but do you run a renderfarm using various SBC?

That works great with Blender for 3d. My NanoPc T3+ is best for that. For big renders I've used the T3+, Tinker and Odroid C2 together. Still took a long time for 30 seconds of 3d video in 1080p(2d). But I wouldn't want my pc to be occupied for that long. So it's great to be able to do.
But I don't use Blender that often anymore.
It's more for video rendering with Kdenlive I use them(and many other things). Also when traveling I use it to edit/render my travel images and put it online. All with solar energy.  sbc's are awesome for that. energy efficient but still very powerful.
The C2 did 10minutes video in 720p in 55minutes rendering.
The RK3399's do 10min in 1080p in 40minutes.
So every step up in performance is great for me. Now I want 1440p with an SBC. But the N2 I'll probably never take along on trips. Too big and bulky, and i's not 5V. For now the RockPi4B is my first choice with RK3399 and 4GB lpddr4. Certainly the ram speed makes a big difference.
The N2 will be for at home to offload render task from my pc, or to play games while my pc renders. And I need to be able to watch 4K h.256 since my pc can't handle it.(i5 2500K ->too old)

 

 

Share this post


Link to post
Share on other sites
On 2/15/2019 at 11:11 AM, Tido said:

pi day is coming soon  March 14, may be with Qualcomm Snapdragon 855

With such a chip, RPi will become big price. :)

 

On 2/15/2019 at 5:00 PM, AreaScout said:

Well sometimes a very big competitor is buying all the RAM chips and you get out of parts, that's just how it could be and there are a ton of other possibilities that could happen that force you to stop production, again if HK says that the parts run suddenly out then you can be sure that HK was doing everything that it's not, but at the end they had to react.  And there was also some other reasons 

Announcement N1 was a year ago, in the early autumn of his publicly refused. Do you believe that in 9 months it was impossible to solve the problem with memory ? All manufacturers of TV boxes and DyU produced at this time different models with DDR3 and DDR4 , so they were able to receive memory, and one HK could not solve this problem. But on the issue of junk on the basis of the intel they had and the time and money.

 

On 2/16/2019 at 12:56 AM, JMCC said:

Amlogic aims to release SoC's that are just enough for a standard TV box, they don't have things such as Chromebooks in mind when they design their SoC's (as, for example, Rockchip does for their higher-end SoC's).

Yes, they have always shown disdain for Source Linux. This is clearly seen in the example of S912-when they did not bother to release GPU support for Linux and limited to one android.

 

On 2/16/2019 at 5:49 PM, NicoD said:

For this 7-zip benchmark the ram speed has no big influence. 

Packaging and unpacking (with the right utility that is able to work in multithreaded mode with all cores at once) is mainly determined by the processor, so it is not surprising that 4(fast)+2(economical) cores are faster than 2(fast)+4(economical). By the way, you can look at the rk3368 - there are 8 cores at a frequency of 1500.

By the way, and looking at the result of this test, the question arises, if the S922 4 fast core (twice), why the result is better not twice, but only 20% ? AML can be deceiving, again with frequency and cores together ?

 

20 hours ago, NicoD said:

And I need to be able to watch 4K h.256 since my pc can't handle it.(i5 2500K ->too old)

Video playback 4K cope even s905 and rk3328. :)

Share this post


Link to post
Share on other sites
On 2/14/2019 at 6:49 PM, NicoD said:

The board started smelling burned. I'm not going to try it again. That's why you need to raise the heatsink so air can flow underneath it.

 

On 2/14/2019 at 7:29 PM, TonyMac32 said:

Heatsink over is better than heatsink under due to basic physics.

Well, the basic physics also allows you to flip the board by 180° like this:

https://reho.st/self/dbe8a2f2236732bc929bedb89fd627a5bf0d2bac.jpg

 

I also replaced the original thermal sheet, by a 1.5mm copper sheett + thermal grease.

40min to reach 80° at full load. Never smelled burned.

https://reho.st/self/028b5abe980510d83950ceb82cf36153e44f91d3.png

 

On 2/14/2019 at 3:10 PM, balbes150 said:

Do not understand the rejection of PCi (NVME) and the use of non-standard eMMC modules that have reduced speed and reliability.

 

The reason is simple, the Amlogic S922 doesn't support PCIe:

hardkernel_odroidn2_block.jpg

 

 

 

 

About the N2, the SoC seems to be very interesting, the lack of PCIe is indeed disturbing, but except for storage, there is no really need for it, USB 3 can do the job.

Personally I never liked the board design of the N- series. The boards are too big, they are huge, they are as big as an Intel NUC.

 

Do you think that FriendlyArm may release a board using the same SoC in the near future? Because I like the compact format of their M4 and Neo's.

Share this post


Link to post
Share on other sites

Sure, you can flip a car over to avoid tire wear too.... ;-)

Copper shim as also proposed by tkaiser and myself at various points, with tkaiser providing data.

I think friendlyelec probably won't be making any more Amlogic based boards after the K2's seemingly disappointing performance.

Sent from my Pixel using Tapatalk

Share this post


Link to post
Share on other sites
3 hours ago, Tantalum said:

Well, the basic physics also allows you to flip the board by 180° like this:

See photo of the Board and the casing in the subject of the announcement. There is clearly shown that the working position of the device is down by the radiator. The shape of the plastic case and the hole for connecting additional devices to the 40-pin connector confirm this. Also on the plastic case there are no hints of legs, that would indicate the lower location of the body on the support. If you do not have a plastic case, the device does not have any element that will allow you to install this device up the radiator without adding any racks\legs, that it would be stable. Yes, you can do everything yourself, and put it correctly (up the radiator), but it is not a regular (which the developers have provided) position for operation. I didn't find any mention from officials that I (as a regular user) should use the N2 up radiator. All these facts indicate that the developers originally planned this location and means that they made a gross mistake in the development. Perhaps tkaser is right when he wrote in the discussion that he had a feeling that the tests were done by an Intern (a stupid teenager). And I would add-and the development of the body and the cooling system was engaged in a schoolboy who does not know the basic laws of physics. Although it apparently was the normal - to place the radiator upside down manufacturers many Chinese TV boxes. But if the Chinese cheap TV boxes price and reputation appropriate, the HK has no right to make such mistakes (if you want to be considered a technological leader, which, in my opinion, it was previously).

 

To be fair, it must be said that almost all manufacturers make stupid mistakes in the development of the cooling system. Radiators upside down Nano PI M4, no elementary vents over the fan Firefly-RK3399, etc.

 

NanoPi-M4-Heatsink.jpg

 

 

Firefly-RK3399-Fully-Assembled.jpg

 

 

But there are holes under the SSD, which lies tightly on the plate

 

Firefly-RK3399-SATA-Drive.jpg

Share this post


Link to post
Share on other sites

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now
0