Efinix FPGA Field Guide
There are two publicly available series of Efinix parts, Trion, and Titanium (sometimes branded Trion Titanium or QuantumX). Trion is made on SMIC’s 40nm planer process and is referred to as OPH and in some places more specifically as antman. Titanium is made on TSMC’s 16nm FF process and is referred to in the toolchain as OPX.
Marketing name to die layout is described in pt/db/devicemap.csv1 with additional color commentary in arch/devicefamilies.xdb2
Trion Parts
The Trion part naming scheme has marketing brain so I have invented my own name for each die3 4 5 6 7.
| Name | Vendor Name | Die Name | Subfamily | Marketing Names | Packages | Interface IP | Notes |
|---|---|---|---|---|---|---|---|
| Baby Jimothy | Q1 | oph_23x82_b0_d0oph_23x82_b1_d1 oph_23x82_b1_d1_io2 | V1 | N/A | N/A | N/A | Shuttle Test Chip and/or for toolchain CI |
| Cozumel | Q5 | oph_59x162_b3_d1 | ?? | N/A | N/A | N/A | Shuttle Test Chip? |
| Jimothy | Q8 | oph_77x162_b3_d1 | V1 | T4, T8 | BGA81, BGA49 | JTAG, PLL, OSC, GPIO | Jimothy is substantially different from the other chips from an interface IP Perspective |
| Rebecca | Q20 | oph_159x242_b17_d3 | V2 | T8, T13, T20 | LQFP144, LQFP100, BGA256, BGA169, WLCSP80 | JTAG, CONTROL, CLOCK_MUX, PLL_FB, GPIO_DDIO, LVDS, MIPI | |
| Rocket | Q35 | oph_129x482_b12_d5 | V2 | T20, T35 | BGA400, BGA324, BGA256 | JTAG, CONTROL, CLOCK_MUX_26x8, PLL_FB, GPIO_DDIO, LVDS, MIPI, DDR16, H264 | |
| Nook | Q120 | oph_337x642_b33_d10 | V2 | T55, T70, T85, T120 | BGA784, BGA576, BGA484, BGA324 | JTAG, CONTROL, CLOCK_MUX_26x8, PLL_FB, GPIO_DDIO, LVDS, MIPI, DDR32, H264 | Unclear if T70 exists |
Titanium Parts
Devices listed in the order they were taped out in
| Name | Vendor Name | Die Name | Subfamily | Marketing Names | Packages | Interface IP | Notes |
|---|---|---|---|---|---|---|---|
| AJ | QX60P | opx_218x322_b16_d10 | V1 (_es), V2 | Ti35, Ti60, Tz50 | BGA256, BGA225, BGA100, WLCSP64 | JTAG_DRCK, CONTROL, SEU, HPOSC, PLL_COMPLEX, DYNAMIC_CLOCK_MUX, HVIO, HSIO | Broken EFuse |
| Livia | QX180 | opx_333x642_b40_d20 | V2 | Ti90L, Ti120L, Ti180L, Ti180 | BGA529, BGA484, BGA361 | JTAG_DRCK, CONTROL, SEU, HPOSC, PLL_COMPLEX_TI180, DYNAMIC_CLOCK_MUX6TO1MUX, REGIONAL_CLOCK_MUX4TO1MUX HVIO, HSIO, MIPI_ADV, DDR32_ADV | DDR Partially or Fully Broken, Discontinued |
| Dr Melfi | ?? | opx_333x642_revB_b40_d20 | V2 | Ti90J, Ti120J, Ti180J, Tz110, Tz170 | BGA529, BGA484, BGA400, BGA361 | JTAG_DRCK, CONTROL, SEU, HPOSC, PLL_COMPLEX_TI180, DYNAMIC_CLOCK_MUX6TO1MUX, REGIONAL_CLOCK_MUX4TO1MUX HVIO, HSIO, MIPI_ADV, DDR32_ADV | Seems to just change some clock routing to make DDR work properly |
| Salvatore | QX375 | opx_466x964_b56_d28 | V2 | Ti165, Ti240, Ti375, Tz200, Tz325 | BGA1156, BGA900, BGA529C, BGA484 | JTAG_DRCK, CONTROL, SEU, HPOSC, PLL_V5, DYNAMIC_CLOCK_MUX8TO1MUX, REGIONAL_CLOCK_MUX_2TO1MUX, REGIONAL_CLOCK_MUX_4TO1_TI375, PMA_CLKMUX, HVIO, HSIO, HSIO_MAX, QUAD, QUAD_PCIE, SERDES_WRAP, SOC, MIPI_ADV, DDR32_ADV | |
| Jimmy | QX135 | opx_334x484_b40_d20 | V2 | Ti85, Ti135, Tz75, Tz100 | BGA676, BGA576, BGA484, BGA441 | JTAG_DRCK, CONTROL, SEU, HPOSC, PLL_V5, DYNAMIC_CLOCK_MUX8TO1MUX, REGIONAL_CLOCK_MUX_2TO1MUX, REGIONAL_CLOCK_MUX_4TO1_TI375, PMA_CLKMUX_2TO1MUX, HVIO, HSIO, HSIO_MAX, QUAD, QUAD_PCIE, SERDES_WRAP, SOC, TSD, MIPI_ADV, DDR32_ADV | Unclear what TSD is, some kind of DAC? |
| Christopher | QX125 | opx_268x484_b24_d12 | V3 | Ti95, Ti125 | BGA324, BGA225 | JTAG_DRCK, CONTROL, SEU_V2, CORE_REGSCAN, HPOSC, FPLL_V2, DYNAMIC_CLOCK_MUX_8TO1MUX_V2, REGIONAL_CLOCK_MUX_2TO1MUX, BYTECLK_X5, BYTECLK_X6, BYTECLK_PLL_FB_CTRL, PHYCLK_GATE, HVIO, HSIO, HSIO_MAX2 | Unclear what CORE_REGSCAN is, seems to be some kind of scanchain readout for the PCR Regs? |
| QX70 | opx_220x324_b16_d10 | V3 | ?? | ?? | Likely to be Ti40/Ti70 | ||
| QX1000 | opx_934x1284_b90_d30 | V2 | ?? | ?? | Likely Ti440/Ti660/Ti800 | ||
| Big Tony | QX2000 | opx_1652x1442_b200_d100 | ?? | ?? | ?? | Likely Ti1600/Ti2000 | |
TitaniumUltra
The two parts listed in this family are opu_268x484_b24_d12 (Tu125) and opu_934x1284_b90_d30 (Tu1000), many of the files are identical to the corresponding QX125 or QX1000. It is not clear if the Ultra branding was dropped, or if these are in here for toolchain test of a die shrink. The scant available timing info does seem to imply its on TSMC16FF as well.
Tachyon
This has timing info implying its on TSMC16FF. It does not appear to have been taped out and looks very Titanium shaped
SSN
SSN are Titanium shaped parts on Samsung 10LPP. The two parts are ssn_125x162_b4_d4 (SSNS10LPPQ10B) and ssn_281x322_b8_d8 (SSNS10LPPQ50B). These are eFPGAs and contain only JTAG, CONTROL, and a USERIO blocks, with USERIO presumably being intended for.
OPH28
This is on TSMC28HPC+, it appears to be a die shrunk Trion that was dropped in favor of Titanium after a test tapeout of oph28_117_162_2p75_dpram (OPH28EX10DP)
-
All paths are relative to the toolchain root directory given by
$EFINITY_HOMEunless otherwise noted ↩ -
See TODO for how instructions on converting XDB files to plain text ↩
-
I have only included devices here which appear to have been actually taped out at some point, Cozumel may or may not have been taped out, this is unclear but it is not listed as a Squish device like
Q16/Q25/Q50/Q100/Q150/Q250/Q350/Q500/Q750/Q1000which appear to be purely for toolchain test ↩ -
I have not included packages that appear to only exist for internal test and bringup purposes ↩
-
I have not included Q40 even though it appears to have likely been taped out and be canonically Trion as it was never released, we have only the boundary description for it, and the tooling considers it to be a separate family for some reason (it was taped out likely on a sanechips shuttle and may have been a very early prototype that was substantially different) ↩
-
Yes Rocket/Nook really have a hard H264 encoder I am not on crack ↩
-
Hard IP names taken from
pt/db/die/block_models/names with different parts of the same IP merged (eg mipitx/rx), sorting out the exact revisions and quantities of each will take a minute ↩