Real World Comparison of Odroid N1 vs XU4

Post Reply
hominoid
Posts: 315
Joined: Tue Feb 28, 2017 3:55 am
languages_spoken: english
ODROIDs: C2, XU4, MC1, N1, N2
Location: Lake Superior Basin, USA
Has thanked: 10 times
Been thanked: 22 times
Contact:

Real World Comparison of Odroid N1 vs XU4

Unread post by hominoid » Mon Feb 19, 2018 6:26 am

Here is a very early and quick real world comparison of the Odroid N1 vs XU4. Of course it needs to be looked at in context for many reasons. The Odroid N1 is an engineering sample and not a released product, is running Debian on interim kernel 4.4 and there has not been adequate time to fully tune the N1's OS or crypto algorithms, to name just a few. It does suggest some very interesting results!

The test is a head to head comprised of a single Odroid N1 and OEM XU4 pool mining (stratum server) Verium(VRM) at sustainable frequency settings. At an ambient temperature of 71f (21.66c), this means 1.7Ghz for the OEM XU4 running in the 70c's and 1.99Ghz for the N1, which I never saw exceed 51c. I think it's a refrigerator in disguise...
Odroid N1

Code: Select all

odroid@odroid-n1:~$ uname -a
Linux odroid-n1 4.4.112 #2 SMP Thu Feb 8 21:25:35 -02 2018 aarch64 GNU/Linux

odroid@odroid-n1:~$ odroid-cpu-control -l
CPU0: governor ondemand current 408.00MHz       min 408.00MHz [408.00MHz]       max 1.51GHz [1.51GHz]
CPU1: governor ondemand current 408.00MHz       min 408.00MHz [408.00MHz]       max 1.51GHz [1.51GHz]
CPU2: governor ondemand current 408.00MHz       min 408.00MHz [408.00MHz]       max 1.51GHz [1.51GHz]
CPU3: governor ondemand current 408.00MHz       min 408.00MHz [408.00MHz]       max 1.51GHz [1.51GHz]
CPU4: governor ondemand current 408.00MHz       min 408.00MHz [408.00MHz]       max 1.99GHz [1.99GHz]
CPU5: governor ondemand current 408.00MHz       min 408.00MHz [408.00MHz]       max 1.99GHz [1.99GHz]

odroid@odroid-n1:~$ ~/cpuminer-fireworm -o stratum+tcp://stratum.poolsloth.com:3333 -u xxxx -p xxxx --randomize --no-redirect -t 9

 Verium Miner forked from cpuminer 1.4 {fireworm} by fireworm@github **
              credits to tpruvot et al. & effectsToCause et al. **

[2018-02-18 18:31:05] Starting Stratum on stratum+tcp://stratum.poolsloth.com:3333
[2018-02-18 18:31:05] HugePages unavailable (22)

[2018-02-18 18:31:05] 9 miner threads started, using scrypt algorithm.
[2018-02-18 18:31:09] Stratum difficulty set to 0.025
[2018-02-18 18:31:09] stratum.poolsloth.com:3333 scrypt² block 181936
[2018-02-18 18:32:39] Total: 538.110 H/m
[2018-02-18 18:32:59] accepted: 1/1 (100.00%), 0.00837 kH/s yes!
[2018-02-18 18:33:10] Total: 479.410 H/m
[2018-02-18 18:33:43] Total: 530.087 H/m
[2018-02-18 18:35:03] Total: 512.673 H/m
[2018-02-18 18:35:10] accepted: 2/2 (100.00%), 0.00822 kH/s yes!
[2018-02-18 18:36:04] Stratum difficulty set to 0.0171756
[2018-02-18 18:36:39] Total: 534.344 H/m
[2018-02-18 18:37:08] accepted: 3/3 (100.00%), 0.00829 kH/s yes!
[2018-02-18 18:37:31] accepted: 4/4 (100.00%), 0.00932 kH/s yes!
[2018-02-18 18:38:09] Total: 558.247 H/m
[2018-02-18 18:39:40] Total: 536.414 H/m
[2018-02-18 18:41:00] accepted: 5/5 (100.00%), 0.00915 kH/s yes!
[2018-02-18 18:41:02] Total: 537.398 H/m
[2018-02-18 18:41:21] accepted: 6/6 (100.00%), 0.00825 kH/s yes!
[2018-02-18 18:42:40] Total: 555.318 H/m
[2018-02-18 18:44:06] Total: 533.703 H/m
[2018-02-18 18:44:48] accepted: 7/7 (100.00%), 0.00930 kH/s yes!
^C[2018-02-18 18:45:08] SIGINT received, exiting
Odroid XU4

Code: Select all

root@c3n0:~# uname -a
Linux c3n0 4.14.5-92 #1 SMP PREEMPT Mon Dec 11 15:48:15 UTC 2017 armv7l armv7l armv7l GNU/Linux

root@c3n0:~# odroid-cpu-control -l
CPU0: governor performance      current 1.40GHz min 200.00MHz [200.00MHz]       max 1.40GHz [1.40GHz]
CPU1: governor performance      current 1.40GHz min 200.00MHz [200.00MHz]       max 1.40GHz [1.40GHz]
CPU2: governor performance      current 1.40GHz min 200.00MHz [200.00MHz]       max 1.40GHz [1.40GHz]
CPU3: governor performance      current 1.40GHz min 200.00MHz [200.00MHz]       max 1.40GHz [1.40GHz]
CPU4: governor performance      current 1.70GHz min 200.00MHz [200.00MHz]       max 1.70GHz [2.00GHz]
CPU5: governor performance      current 1.70GHz min 200.00MHz [200.00MHz]       max 1.70GHz [2.00GHz]
CPU6: governor performance      current 1.70GHz min 200.00MHz [200.00MHz]       max 1.70GHz [2.00GHz]
CPU7: governor performance      current 1.70GHz min 200.00MHz [200.00MHz]       max 1.70GHz [2.00GHz]

root@c3n0:~# ~/cpuminer-fireworm -o stratum+tcp://stratum.poolsloth.com:3333 -u xxxx -p xxxx --randomize --no-redirect -t 4 -1 2 --cpu-affinity-stride 1 --cpu-affinity-default-index 4 --cpu-affinity-oneway-index 0

 Verium Miner forked from cpuminer 1.4 {fireworm} by fireworm@github **
              credits to tpruvot et al. & effectsToCause et al. **

[2018-02-18 18:31:05] Starting Stratum on stratum+tcp://stratum.poolsloth.com:3333
[2018-02-18 18:31:05] Binding thread 0 to cpu index 0
[2018-02-18 18:31:05] Binding thread 1 to cpu index 0
[2018-02-18 18:31:05] HugePages unavailable (22)

[2018-02-18 18:31:05] Binding thread 2 to cpu index 0
[2018-02-18 18:31:05] Binding thread 3 to cpu index 0
[2018-02-18 18:31:05] 6 miner threads started, using 'scrypt²' algorithm.
[2018-02-18 18:31:05] Binding thread 4 to cpu index 0
[2018-02-18 18:31:05] Binding thread 5 to cpu index 0
[2018-02-18 18:31:09] Stratum difficulty set to 0.025
[2018-02-18 18:31:09] stratum.poolsloth.com:3333 scrypt² block 181936
[2018-02-18 18:31:43] Total: 388.377 H/m
[2018-02-18 18:32:14] Total: 387.199 H/m
[2018-02-18 18:32:45] Total: 387.127 H/m
[2018-02-18 18:33:16] Total: 384.155 H/m
[2018-02-18 18:33:47] Total: 385.000 H/m
[2018-02-18 18:34:18] Total: 385.126 H/m
[2018-02-18 18:34:49] Total: 384.142 H/m
[2018-02-18 18:35:20] Total: 383.299 H/m
[2018-02-18 18:35:51] Total: 383.115 H/m
[2018-02-18 18:36:22] Total: 384.423 H/m
[2018-02-18 18:36:54] Total: 385.171 H/m
[2018-02-18 18:37:25] Total: 385.309 H/m
[2018-02-18 18:37:35] accepted: 1/1 (100.00%), 0.00640 kH/s yes!
[2018-02-18 18:37:39] accepted: 2/2 (100.00%), 0.00639 kH/s yes!
[2018-02-18 18:37:44] accepted: 3/3 (100.00%), 0.00639 kH/s yes!
[2018-02-18 18:37:56] Total: 383.180 H/m
[2018-02-18 18:38:27] Total: 382.897 H/m
[2018-02-18 18:38:58] Total: 382.540 H/m
[2018-02-18 18:39:29] Total: 383.798 H/m
[2018-02-18 18:40:00] Total: 383.192 H/m
[2018-02-18 18:40:31] Total: 383.481 H/m
[2018-02-18 18:41:02] Total: 383.795 H/m
[2018-02-18 18:41:33] Total: 384.514 H/m
[2018-02-18 18:42:04] Total: 383.588 H/m
[2018-02-18 18:42:35] Total: 383.282 H/m
[2018-02-18 18:43:07] Total: 382.776 H/m
[2018-02-18 18:43:38] Total: 383.951 H/m
[2018-02-18 18:44:09] Total: 384.540 H/m
[2018-02-18 18:44:13] accepted: 4/4 (100.00%), 0.00642 kH/s yes!
[2018-02-18 18:44:17] Stratum difficulty set to 0.0169173
[2018-02-18 18:44:29] accepted: 5/5 (100.00%), 0.00642 kH/s yes!
[2018-02-18 18:44:40] Total: 385.162 H/m
^C[2018-02-18 18:45:04] SIGINT received, exiting
The average hash rate for the N1 was 531.57 H/m and 384.35H/m for the XU4. That is a 38.3% increase in hash rate for the Odroid N1 in real world operations. I only spent a relatively short amount of time quickly going through a bunch of thread and core combinations on the Odroid N1, so there is bound to be some room for improvement. Even though the Odroid N1 was running 9, 3-way threads for the test, I was successful in running 24 1-way threads. I didn't try any more because the performance was decreasing. It just demonstrates the flexibility and advantage of having 4GB of memory. As more time allows, a 6-way thread test is in order. And for the record, even with the XU4 running at 2Ghz, the hash rate was lower, and unsustainable.

Code: Select all

root@c3n0:~# odroid-cpu-control -s -M 2.0G
CPU0: max 1.40GHz [1.40GHz] -> 1.40GHz [1.40GHz]
CPU1: max 1.40GHz [1.40GHz] -> 1.40GHz [1.40GHz]
CPU2: max 1.40GHz [1.40GHz] -> 1.40GHz [1.40GHz]
CPU3: max 1.40GHz [1.40GHz] -> 1.40GHz [1.40GHz]
CPU4: max 1.70GHz [2.00GHz] -> 2.00GHz [2.00GHz]
CPU5: max 1.70GHz [2.00GHz] -> 2.00GHz [2.00GHz]
CPU6: max 1.70GHz [2.00GHz] -> 2.00GHz [2.00GHz]
CPU7: max 1.70GHz [2.00GHz] -> 2.00GHz [2.00GHz]
root@c3n0:~# ~/cpuminer-fireworm -o stratum+tcp://stratum.poolsloth.com:3333 -u xxxx -p xxxx --randomize --no-redirect -t 4 -1 2

 Verium Miner forked from cpuminer 1.4 {fireworm} by fireworm@github **
              credits to tpruvot et al. & effectsToCause et al. **

[2018-02-18 20:37:32] Starting Stratum on stratum+tcp://stratum.poolsloth.com:3333
[2018-02-18 20:37:32] HugePages unavailable (22)

[2018-02-18 20:37:32] 6 miner threads started, using 'scrypt²' algorithm.
[2018-02-18 20:37:36] Stratum difficulty set to 0.025
[2018-02-18 20:37:36] stratum.poolsloth.com:3333 scrypt² block 181963
[2018-02-18 20:37:42] accepted: 1/1 (100.00%), 0.00429 kH/s yes!
[2018-02-18 20:38:31] accepted: 2/2 (100.00%), 0.00642 kH/s yes!
[2018-02-18 20:38:48] Total: 356.060 H/m
[2018-02-18 20:39:58] Total: 357.322 H/m
[2018-02-18 20:41:01] Total: 353.908 H/m
[2018-02-18 20:41:02] accepted: 3/3 (100.00%), 0.00590 kH/s yes!
[2018-02-18 20:41:32] accepted: 4/4 (100.00%), 0.00611 kH/s yes!
[2018-02-18 20:42:12] Total: 347.295 H/m
^C[2018-02-18 20:42:18] SIGINT received, exiting
Another good point of reference for comparison is KaptainBlaZzed's hardware hash rate comparison sheet for VRM. On the second sheet, accessed by the tab at the bottom, there is a comparison of other SBC's. Again, some context is in order. My XU4 posted hash rate is for solo mining (get-work not a stratum server) and is a sustainable hash rate at 1.7Ghz for an OEM XU4. If someone has improved their cooling, has better OS or crpto-algorithm tuning, they could have better hash rates. At the far right of the spread sheet is the Hashes/Watt results that also shed some light on the efficiency of the SBC's. One other important number missing is the Hashes/Dollar (capital cost). It is another area that the Odroid's in general are at or near the top.

User avatar
odroid
Site Admin
Posts: 32498
Joined: Fri Feb 22, 2013 11:14 pm
languages_spoken: English
ODROIDs: ODROID
Has thanked: 180 times
Been thanked: 347 times
Contact:

Re: Real World Comparison of Odroid N1 vs XU4

Unread post by odroid » Mon Feb 19, 2018 9:31 am

Thank you for sharing the VRM hash rate test result. It is very useful to understand how badly the PoP memory could affect the CPU cooling.

Anyway, I think the XU4 and MC1 seem to be more cost-effective than N1 for mining rig. ;)

hominoid
Posts: 315
Joined: Tue Feb 28, 2017 3:55 am
languages_spoken: english
ODROIDs: C2, XU4, MC1, N1, N2
Location: Lake Superior Basin, USA
Has thanked: 10 times
Been thanked: 22 times
Contact:

Re: Real World Comparison of Odroid N1 vs XU4

Unread post by hominoid » Mon Feb 19, 2018 10:08 am

odroid wrote:Thank you for sharing the VRM hash rate test result. It is very useful to understand how badly the PoP memory could affect the CPU cooling.

Anyway, I think the XU4 and MC1 seem to be more cost-effective than N1 for mining rig. ;)
Yes, PoP memory design does affect cooling and allows for a smaller foot print. My goal was to give some idea of the real world performance of the N1 and naturally many people are familiar with the XU4's performance. The MC1/HC1/HC2 with better heatsinks and cooling make an excellent choice for clusters and will continue to be cost effective. As far as the cost effectiveness of the N1 to any particular use, it is to early for me to say since it's not a released product and has no MSRP. The N1's use will be different, but it will ultimately be determined by it's customers.

Edited for completeness.

elatllat
Posts: 1572
Joined: Tue Sep 01, 2015 8:54 am
languages_spoken: english
ODROIDs: XU4, N1, N2
Has thanked: 24 times
Been thanked: 64 times
Contact:

Re: Real World Comparison of Odroid N1 vs XU4

Unread post by elatllat » Mon Feb 19, 2018 12:25 pm

Is Verium dependant on RAM speed? (if there were an N1 with LPDDR4 would that tip the efficacy scale).

hominoid
Posts: 315
Joined: Tue Feb 28, 2017 3:55 am
languages_spoken: english
ODROIDs: C2, XU4, MC1, N1, N2
Location: Lake Superior Basin, USA
Has thanked: 10 times
Been thanked: 22 times
Contact:

Re: Real World Comparison of Odroid N1 vs XU4

Unread post by hominoid » Tue Feb 20, 2018 11:51 am

elatllat wrote:Is Verium dependent on RAM speed? (if there were an N1 with LPDDR4 would that tip the efficacy scale).
Yes some what, but as you know higher frequency ram is not necessarily faster, depending on many things like address strobe timings etc of the modules in question. Point in case, I have two identical Dual AMD Opteron 6276 servers running the same OS image. The one with 1333Mhz memory is considerably faster then the one with 1600Mhz memory simply due to memory timing(longer CAS). Higher frequency ram being slower than lower frequency ram is more common place than what most people think. Other factors can also play a larger role such as memory channel utilization, cooling and system tuning(hugh pages etc) that effects the operational performance.
As far the whole efficiency thing, I have not taken any position on the N1 because it is to early to tell, we don't have complete information. Once we have everything we can run the numbers and make a proper analysis. I only brought it to light because it is an area that the whole Odroid family of products do very well at compared to all but the newest and most expensive Intel servers. But, the Intel server capital efficiency is much much lower because the cost per hash goes through the roof at thousands of dollars per processor.

brodie59
Posts: 1
Joined: Fri Sep 20, 2019 4:21 pm
languages_spoken: english
Has thanked: 0
Been thanked: 0
Contact:

Re: Real World Comparison of Odroid N1 vs XU4

Unread post by brodie59 » Sat Sep 21, 2019 3:31 pm

Hardkernel had in plan an RK3399 based product, the Odroid-N1, however, due to changes in the hardware landscape, they moved to a new project FirstBankCard, the N2, with “faster CPU/GPU cores and native DDR4 support” (see here). This new SBC is planned to be released in early 2019.
Last edited by brodie59 on Mon Sep 23, 2019 1:20 pm, edited 1 time in total.

elatllat
Posts: 1572
Joined: Tue Sep 01, 2015 8:54 am
languages_spoken: english
ODROIDs: XU4, N1, N2
Has thanked: 24 times
Been thanked: 64 times
Contact:

Re: Real World Comparison of Odroid N1 vs XU4

Unread post by elatllat » Sat Sep 21, 2019 11:41 pm


Post Reply

Return to “General Chat”

Who is online

Users browsing this forum: No registered users and 0 guests