search

ThomasKaiser / sbc-bench

Simple benchmark for single board computers
BSD 3-Clause "New" or "Revised" License
675 stars 78 forks source link

Phytium D2000 Benchmark #26

Closed lanefu closed 3 years ago

lanefu commented 3 years ago

| | MHz | 5.13 | Impish arm64 | 16720 | 448270 | 828340 | 2962980 | 8158030 | - | http://ix.io/3FTK |

Ran by PtitSeb yesterday

ThomasKaiser commented 3 years ago

A bit sad that this ran with Impish and he didn't manage to build mhz. The crypto scores look a bit weird (but since Impish the only other references collected with OpenSSL 1.1.1l and ARMv8 Crypto Extensions are the A53 @ 2.0 GHz and A73 @ 2.4 GHz on ODROID N2+: http://ix.io/3DtN

        type             16 bytes     64 bytes    256 bytes   1024 bytes   8192 bytes  16384 bytes
FTC663: aes-256-cbc     386155.67k   644563.61k   774073.94k   808907.43k   826922.33k   828205.74k
   A53: aes-256-cbc     162742.46k   427874.47k   717046.87k   877222.91k   938270.72k   942221.99k
   A73: aes-256-cbc     378114.54k   847673.83k  1187437.06k  1313454.42k  1363293.53k  1366933.50k
lanefu commented 3 years ago

A bit sad that this ran with Impish and he didn't manage to build mhz. The crypto scores look a bit weird (but since Impish the only other references collected with OpenSSL 1.1.1l and ARMv8 Crypto Extensions are the A53 @ 2.0 GHz and A73 @ 2.4 GHz on ODROID N2+: http://ix.io/3DtN

        type             16 bytes     64 bytes    256 bytes   1024 bytes   8192 bytes  16384 bytes
FTC663: aes-256-cbc     386155.67k   644563.61k   774073.94k   808907.43k   826922.33k   828205.74k
   A53: aes-256-cbc     162742.46k   427874.47k   717046.87k   877222.91k   938270.72k   942221.99k
   A73: aes-256-cbc     378114.54k   847673.83k  1187437.06k  1313454.42k  1363293.53k  1366933.50k

Yeah some of the #'s seemed weird He wasn't too familiar with the tool. Either. I pre-ordered one. I'll provide more tests if things work out.

Also FYI if you didn't catch my comment on cnx. It can support up to 32 gigs of RAM so pretty interesting device.

ThomasKaiser commented 3 years ago

Would be interesting if a dual-channel RAM config results in memory performance getting somewhat decent.

Good luck with this thing. For me it has two huge disadvantages each of them being a noisy fan.

ThomasKaiser commented 3 years ago

@lanefu are you able to obtain grep ^CPU /proc/cpuinfo from the device? I am about to add ARM core details to sbc-bench v0.8.5.

lanefu commented 3 years ago

K I asked them to run it... will share after they respond.

just an FYI:

I had some back and forth with some honeycomb people about the SBCbench benchmarks. (I'm relaying info here, no horse in this race... i asked them to open an issue, but wasn't interest in a deeper conversation)

  1. Concern that 7zip single core test needed different argument. was running same number of threads as CPU count, but just on 1 core. Their point was the -mnt 1 argument to 7zip should be -mnt=1
  2. Concern that honeycomb test was at 1.9ghz. They provided a fresh run, but.... OS info was...fedora? http://ix.io/3GcC
lanefu commented 3 years ago

this is from FT2000/4 which should be the 4 core predecessor to the D2000.. they're gonna give me D2000 info later today 

CPU implementer : 0x70
CPU architecture: 8
CPU variant     : 0x1
CPU part        : 0x663
CPU revision    : 3
CPU implementer : 0x70
CPU architecture: 8
CPU variant     : 0x1
CPU part        : 0x663
CPU revision    : 3
CPU implementer : 0x70
CPU architecture: 8
CPU variant     : 0x1
CPU part        : 0x663
CPU revision    : 3
CPU implementer : 0x70
CPU architecture: 8
CPU variant     : 0x1
CPU part        : 0x663
CPU revision    : 3
ThomasKaiser commented 3 years ago

The FT2000 core is already in my list (started with arm-chip-id.db from @tparys and then adding other stuff I found):

41:Arm
41/810:ARM810
41/920:ARM920
41/922:ARM922
41/926:ARM926
41/940:ARM940
41/946:ARM946
41/966:ARM966
41/a20:ARM1020
41/a22:ARM1022
41/a26:ARM1026
41/b02:ARM11 MPCore
41/b36:ARM1136
41/b56:ARM1156
41/b76:ARM1176
41/c05:Cortex-A5
41/c07:Cortex-A7
41/c08:Cortex-A8
41/c09:Cortex-A9
41/c0d:Cortex-A17
41/c0f:Cortex-A15
41/c0e:Cortex-A17
41/c14:Cortex-R4
41/c15:Cortex-R5
41/c17:Cortex-R7
41/c18:Cortex-R8
41/c20:Cortex-M0
41/c21:Cortex-M1
41/c23:Cortex-M3
41/c24:Cortex-M4
41/c27:Cortex-M7
41/c60:Cortex-M0+
41/d01:Cortex-A32
41/d03:Cortex-A53
41/d04:Cortex-A35
41/d05:Cortex-A55
41/d07:Cortex-A57
41/d08:Cortex-A72
41/d09:Cortex-A73
41/d0a:Cortex-A75
41/d0b:Cortex-A76
41/d0c:Neoverse-N1
41/d0d:Cortex-A77
41/d0e:Cortex-A76AE
41/d13:Cortex-R52
41/d20:Cortex-M23
41/d21:Cortex-M33
41/d22:Cortex-M55
41/d41:Cortex-A78
41/d42:Cortex-A78AE
41/d4a:Neoverse-E1
41/d4b:Cortex-A78C
42:Broadcom
42/00f:Broadcom Brahma B15
42/100:Broadcom Brahma B53
42/516:Broadcom ThunderX2
43:Cavium
43/0a0:Cavium ThunderX
43/0a1:Cavium ThunderX 88XX
43/0a2:Cavium ThunderX 81XX
43/0a3:Cavium ThunderX 83XX
43/0af:Cavium ThunderX2 99xx
44:DEC
44/a10:DEC SA110
44/a11:DEC SA1100
46:Fujitsu
46/001:A64FX
48:HiSilicon
48/d01:Kunpeng-920
49:Infineon
4d:Motorola/Freescale
4e:NVidia
4e/000:NVidia Denver
4e/003:NVidia Denver 2
4e/004:NVidia Carmel
50:APM
50/000:APM X-Gene
51:Qualcomm
51/00f:Qualcomm Scorpion
51/02d:Qualcomm Scorpion
51/04d:Qualcomm Krait
51/06f:Qualcomm Krait
51/201:Qualcomm Kryo
51/205:Qualcomm Kryo
51/211:Qualcomm Kryo
51/800:Qualcomm Falkor V1/Kryo
51/801:Qualcomm Kryo V2
51/c00:Qualcomm Falkor
51/c01:Qualcomm Saphira
53:Samsung
53/001:Samsung Exynos-m1
56:Marvell
56/131:Marvell Feroceon 88FR131
56/581:Marvell PJ4/PJ4b
56/584:Marvell PJ4B-MP
61:Apple
61/022:Apple Icestorm
61/023:Apple Firestorm
66:Faraday
66/526:Faraday FA526
66/626:Faraday FA626
69:Intel
69/200:Intel i80200
69/210:Intel PXA250A
69/212:Intel PXA210A
69/242:Intel i80321-400
69/243:Intel i80321-600
69/290:Intel PXA250B/PXA26x
69/292:Intel PXA210B
69/2c2:Intel i80321-400-B0
69/2c3:Intel i80321-600-B0
69/2d0:Intel PXA250C/PXA255/PXA26x
69/2d2:Intel PXA210C
69/411:Intel PXA27x
69/41c:Intel IPX425-533
69/41d:Intel IPX425-400
69/41f:Intel IPX425-266
69/682:Intel PXA32x
69/683:Intel PXA930/PXA935
69/688:Intel PXA30x
69/689:Intel PXA31x
69/b11:Intel SA1110
69/c12:Intel IPX1200
70:Phytium
70/662:Phytium FT2000PLUS
70/663:Phytium S2500
70/???:Phytium D2000
c0:Ampere

Now looking into the 7-zip issue...

ThomasKaiser commented 3 years ago

Concern that 7zip single core test needed different argument. was running same number of threads as CPU count, but just on 1 core. Their point was the -mnt 1 argument to 7zip should be -mnt=1

They are absolutely right. I tried to fix that over 3 years ago with v0.5.2 but failed since I used mmt 1 instead of mmt1 or mmt=1. The more cores the CPU is equipped with the more single-threaded results were affected. And not just lower scores but execution time magnitudes longer.

Now in the process of measuring 'wrong' vs 'right' to decide which result list entries need to be thrown away (though I know of noone except myself who looks at those single core scores regularly).

tparys commented 3 years ago

The FT2000 core is already in my list (started with arm-chip-id.db from @tparys and then adding other stuff I found):

As much as I'd like to take credit for that database, it's really part of the util-linux package. If you have new additions, they should probably go there.

lanefu commented 3 years ago

Okay ptitSeb said D2000 output was identical to F2000, just 8 cores instead of 4

On Sat, Nov 27, 2021, 9:29 PM tparys @.***> wrote:

The FT2000 core is already in my list (started with arm-chip-id.db from @tparys https://github.com/tparys and then adding other stuff I found):

As much as I'd like to take credit for that database https://github.com/util-linux/util-linux/blob/master/sys-utils/lscpu-arm.c, it's really part of the util-linux package. If you have new additions, they should probably go there.

— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/ThomasKaiser/sbc-bench/issues/26#issuecomment-980824824, or unsubscribe https://github.com/notifications/unsubscribe-auth/ABEVNQCK3YA4542L2JAIJPLUOGHZPANCNFSM5IXK3JAA . Triage notifications on the go with GitHub Mobile for iOS https://apps.apple.com/app/apple-store/id1477376905?ct=notification-email&mt=8&pt=524675 or Android https://play.google.com/store/apps/details?id=com.github.android&referrer=utm_campaign%3Dnotification-email%26utm_medium%3Demail%26utm_source%3Dgithub.

ThomasKaiser commented 3 years ago

ptitSeb said D2000 output was identical to F2000, just 8 cores instead

Which is reasonable and also an indication that util-linux people confuse SoCs with cores :)

ThomasKaiser commented 3 years ago

@tparys I found relying on regs/identification/midr_el1 not that reliable (missing on some systems where I tested in the meantime). I came up with reading /proc/cpuinfo into an array like this:

ARMTypes=($(awk -F"0x" '/^CPU implementer|^CPU part/ {print $2}' /proc/cpuinfo))

And then like that

ThomasKaiser commented 3 years ago

@lanefu With cda7a50885bab18f1bea6e13733a6012bf84b3ab I removed entries from systems with more than 8 cores from results list. Also replaced few with 6/8 cores.

In case you're still in touch with the Solidrun guys it would be great if they can submit a new result made with v0.8.6 (same for the Phytium of course). And in case you've still access to the AWS instance with 32 N1 cores or this X-Gene thing, then fresh numbers are highly appreciated.

lanefu commented 3 years ago

New D2000 benchmarks. Solidrun said he'd try to get to it this week

In case this device is not already represented in official sbc-bench results list then please consider submitting it at https://github.com/ThomasKaiser/sbc-bench/issues with this line: | | MHz | 5.13 | Impish arm64 | 16680 | 448610 | 828450 | 2920 | 7310 | - | http://ix.io/3GrW |

On Sun, Nov 28, 2021, 2:32 PM ThomasKaiser @.***> wrote:

@lanefu https://github.com/lanefu With cda7a50 https://github.com/ThomasKaiser/sbc-bench/commit/cda7a50885bab18f1bea6e13733a6012bf84b3ab I removed entries from systems with more than 8 cores from results list. Also replaced few with 6/8 cores.

In case you're still in touch with the Solidrun guys it would be great if they can submit a new result made with v0.8.6 (same for the Phytium of course). And in case you've still access to the AWS instance with 32 N1 cores or this X-Gene thing, then fresh numbers are highly appreciated.

— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/ThomasKaiser/sbc-bench/issues/26#issuecomment-981138864, or unsubscribe https://github.com/notifications/unsubscribe-auth/ABEVNQEFDXWB62DBUJFJ6K3UOJ7VZANCNFSM5IXK3JAA . Triage notifications on the go with GitHub Mobile for iOS https://apps.apple.com/app/apple-store/id1477376905?ct=notification-email&mt=8&pt=524675 or Android https://play.google.com/store/apps/details?id=com.github.android&referrer=utm_campaign%3Dnotification-email%26utm_medium%3Demail%26utm_source%3Dgithub.