01:32 | Artox | Ke: power draw? I can report that my 8040 with 3 spinning disks draws 30W according to the multimeter on the wallsocket |
01:33 | Artox | but apart from harddisk traffic, the cpu is idle in that state |
07:00 | Ke | Artox: which cpu frequency? |
07:02 | Ke | Artox: also, which power supply? |
07:14 | jnettlet | the 10Gbps phys can also draw quite a bit of power. |
12:41 | Artox | Ke: 1300MHz, and a 12V laptop switching PSU |
12:42 | Artox | and a custom 12V-5V dcdc converter for the harddisks, similar to what the picoPSU devices use |
12:43 | Artox | precisely: https://vps.jm0.eu/box/IMG_20170520_134745.jpg |
12:44 | Artox | over load I have so far seen 35W |
12:44 | Artox | s/over/under |
12:45 | Artox | but mind what j nettlet said, there can be more components drawing power (only one eth port is in use here) |
13:02 | topi` | jnettlet: do you remember which kind of DIMMs should I buy for the MachhiatoBIN board? I want to get 16 GB |
13:03 | topi` | I see the provided Kingston DIMM is a 2400mhz part. Is it single or dual chan? |
13:03 | topi` | DDR3 or DDR4? |
13:06 | topi` | cDd |
13:06 | topi` | I guess the 8040 ddr controller supports 16GB sticks just fine |
13:08 | Ke | can you get dual chan on a single slot? |
13:09 | topi` | "DDR4 DIMM slot with optional ECC and single/dual chip select support |
13:09 | topi` | what is a single/dual chip select? |
13:20 | Ke | notably, not everything mentioned is actually supported (yet) |
13:51 | topi` | yeah, ECC comes to mind |
14:09 | jnettlet | topi`, Marvell hasn't finished ECC support for the memory controller |
14:10 | topi` | you mean software or hardware? |
14:13 | jnettlet | firmware...software |
14:14 | topi` | :) |
14:16 | topi` | jnettlet: do you have a macchiatobin around at your desk ? |
14:17 | topi` | probably not the same version as mine - I have rev 1.2 board |
14:27 | jnettlet | topi` I have an original dev model yes. But I am in Stockholm right now :) |
14:27 | topi` | you are getting closer :) |
14:28 | jnettlet | mine is acting as a part of my high speed testing network. 10Gbps to my build server, and then a 2.5Gbps uplink from my ClearFog Pro router. |
16:09 | Artox | Ke: update: under full load I have now seen 42W |
16:10 | Ke | I guess it's still not clocked to 2GHz |
16:10 | Ke | the cpu cores that is |
16:10 | Artox | nop, 1300MHz |
16:11 | Artox | I don't plan to clock it any higher for now |
16:11 | Ke | wondering a bit, how is this not something that can't be done dynamically |
16:11 | Artox | there are DIP switches to set up the clock |
16:12 | Ke | I have always known that dynamic frequency scaling seems to be somehow different from non-dynamic, but it has never been clear, what it the part that can't be done dynamically |
16:12 | Artox | and I think I have a SoC that is meant to be driven at 1300MHz |
16:12 | Ke | Artox: yes, I know very well |
16:12 | Artox | I really don't know |
16:12 | Ke | how do you say, meant to? |
16:14 | Ke | interestingly dynamic frequency scaling vs. static, many manufacturers actually package 2 different sets of a53 cores clocked at different frequencies |
16:15 | Artox | I thought it was supposed to run at that speed, but I never asked specifically. |
16:16 | Ke | it may be the default, many places explicitly mention 2GHz and there are config options for 2GHz |
16:17 | Ke | may be that you have to throttle the cpu speed with the default cooler at full load, which is not wanted as it may cause unpredictable performance |
16:17 | Ke | many arm devices are not actually able to run at full speed for a prolonged perioid of time |
16:18 | Ke | I think it's common for x86 cpus as well nowadays |
16:19 | Artox | yes |
16:21 | Ke | you could just try the higher frequencies and check the temperatures |
16:21 | Artox | yes, but it isn't of much importance to me |
18:06 | jnettlet | Ke, the soc dynamic frequency scaling. The dip switches set the socs max performance |
18:07 | Ke | you are missing predicate verb in the first sentence |
18:08 | Ke | SoC does? |
18:08 | Ke | or can do |
20:22 | jnettlet | Ke, sorry yes. I am on mobile right now |
20:32 | Ke | should there be a difference with armada 8040 max at 2GHz clocked at 1.3GHz compared to armada 8040 max at 1.3GHz clocked at 1.3GHz |
20:32 | Ke | ie. is there any actual reason to do the clock speed with jumpers |
20:37 | jnettlet | no there is no difference. However depending on the board some chips will be binned to different speeds and the jumpers allow us to have a single board. |
20:42 | jnettlet | the second batch of chips support 2GHz but Marvell is still sorting out the memory timings, so the boards are shipping at 1.3Ghz and 800Mhz DDR speeds |
21:59 | rabeeh | Artox : 12v to 5v? |
22:00 | rabeeh | not really needed; only 12v is taken from the ATX power supply |
22:00 | rabeeh | everything else is generated on the mobo itself |
22:08 | Artox | rabeeh: yes |
22:08 | Artox | for the peripherals |
22:08 | Artox | harddisks |
22:08 | Artox | fans |