Merge pull request #48 from lanefu/moar-phytium

Phytium D2000 benchmarks before and after RAM upgrade

Results.md

@@ -91,8 +91,8 @@ So do **not** rely on collected numbers unless you carefully read through all th
 | [Orange Pi PC 2](http://ix.io/3MQJ) | 1370 MHz | 5.10 | Focal arm64 | 3500 | 114830 | 637410 | 1070 | 3680 | - |
 | [Orange Pi Plus 2](http://ix.io/1iX4) | 1300 MHz | 4.14 | Stretch armhf | 2890 | 21480 | 25250 | 830 | 3240 | - |
 | [Orange Pi Prime](http://ix.io/2kTH) | 1370 MHz | 5.4 | Buster | 3590 | 89210 | 637980 | 1180 | 3540 | - |
-| [Phytium D2000](http://ix.io/3GrW) | 2300 MHz | 5.13 | Impish arm64 | 16680 | 448610 | 828450 | 2920 | 7310 | - |
-| [Phytium D2000](http://ix.io/3Sl9) | 2300 MHz | 5.4 | Kylin arm64 | 16970 | 437820 | 828520 | 4090 | 16290 | - |
+| [Phytium D2000 1xSO-DIMM](http://ix.io/445T) | 2300 MHz | 5.19 | Jammy arm64 | 16390 | 427440 | 827090 | 2820 | 6490 | - |
+| [Phytium D2000 2xSO-DIMM](http://ix.io/446h) | 2300 MHz | 5.19 | Jammy arm64 | 16670 | 427490 | 828130 | 3480 | 16110 | - |
 | [PineH64](http://ix.io/1jEr) | 1800 MHz | 4.18 | Stretch arm64 | 4650 | 123400 | 836900 | 1380 | 5530 | 5.62 |
 | [PineH64](http://ix.io/26Ph) | 1800 MHz | 5.4 | Buster arm64 | 4710 | 116900 | 839870 | 1420 | 5560 | 7.10 |
 | [Quartz64](http://ix.io/3rUb) | 1800 MHz | 5.13 | Buster arm64 | 4840 | 165250 | 845490 | 2980 | 7650| - |
@@ -193,7 +193,7 @@ So do **not** rely on collected numbers unless you carefully read through all th
 * Gigabyte H270-T70 numbers are for one [blade module equipped with two Cavium ThunderX CN8890 (48 cores each)](https://twitter.com/linux_chenxing/status/1449603003057532930). With different DRAM config/settings results vary (see [here](http://ix.io/3Owk) and [there](http://ix.io/3Owv)).
 * Hugsun X99 is an [overclocked RK3399 TV box](https://github.com/ThomasKaiser/sbc-bench/issues/20), just to show the effect of overclocking the A53 cores to 1.8 GHz and the A72 to 2.1 GHz on an RK3399.
 * Jetson Nano was [properly powered with 5V/5A via barrel plug](https://forum.armbian.com/topic/9921-nvidia-jetson-nano/?do=findComment&comment=78467) (when powering through Micro USB the board enters a lower consumption/performance profile)
-* Phytium D2000 consists of 8 [custom 64-bit ARMv8-compatible FTC663 cores](https://en.wikipedia.org/wiki/FeiTeng_(processor))
+* Phytium D2000 consists of 8 [custom 64-bit ARMv8-compatible FTC663 cores](https://en.wikipedia.org/wiki/FeiTeng_(processor)). The 2 numbers above only differ by RAM config: one or two 16GB DDR4 SO-DIMMs. While memory bandwidth differs significantly latency does not and as such the `7-zip` benchmark is almost unaffected while the `openssl` test does not depend on memory performance by design. Though other tasks that are sensitive to memory bandwidth might benefit a lot from a dual channel memory config.
 * NanoPi NEO4 numbers: 1st result is from my NEO4 N°1 running with a [NanoPi M4 image](https://github.com/armbian/build/blob/1c00822819f7fdfeac57bff8f991be526ca1add7/config/sources/rk3399.conf#L91). This NEO uses the vendor supplied thermal pad between SoC and heatsink. 2nd number from my 2nd NEO4 this time using NEO4 settings (`rk3399-nanopi4-rev04.dtb` loaded) with a copper shim between heatsink and SoC which as usual improves 'thermal performance' a lot. Since memory bandwidth and especially latency is too low another test needed with my NEO4 N°2, this time again with M4 settings (`rk3399-nanopi4-rev01.dtb` loaded) and an additional fan. Memory performance restored, slightly better performance due to colder SoC. 4th result made with 4.19.0-rc4. Please be aware that RK3399 memory performance numbers differ alot between 4.4 and mainline kernel for yet unknown reasons!
 * ODROID-N2 number should be taken with a grain of salt since made with a [pretty early software stack](https://forum.armbian.com/topic/9619-announcement-odroid-n2/?do=findComment&comment=72764). Most probably scores will slightly improve over time. 'Overclocked' executions with both CPU clusters set to 2.0 GHz showed reliability issues most probably due to DVFS undervoltage (cpuminer quit almost immediately [here](http://ix.io/1BrG) while it ran only 50 seconds [there](http://ix.io/1Bsz) -- this tool since being a load generator checking for data corruption can also be used for reliability testing but I would prefer our [StabilityTester](https://github.com/ThomasKaiser/StabilityTester) instead)
 * Rock Pi S is based on RK3308 Quad Cortex-A35 but the above numbers are not typical for A35 since the SoC design is severly limited: only a 16-bit RAM bus and 589MHz(*2) DDR clock in Rockchip’s DDR loader