| View previous topic :: View next topic |
| Author |
Message |
achtung049

Joined: 04 Feb 2014 Posts: 481 Location: Russia, Vologda
|
Posted: Sun Aug 09, 2015 7:38 am Post subject: Guide to Weitek chips, need little help |
|
|
Hi!
I think everyone here knows Weitek company. This company has not survived so long, but it made a lot of unusual and rare chips. I'm tired of every time to look for old datasheets to find out what I see - I decided to collect all the information in one topic. If you have any additions or you want to correct a mistake - please write here, and I will make changes to the list.
I hope this little guide to the marking and purpose will be useful to anyone except me
OK, let's go!
(I have omitted all markings everywhere except 4 digits indicating the purpose of the chip - this is enough I think. You can later specify more details)
Here is my research:
No information. Could not find any photos, datasheets and so on, only a few references. Purpose unknown. Existence of chip has not been confirmed.
Purpose known. Purpose chip is clearly indicated in different sources, but datasheets and photos not found.
Photo only. Existence of chip with this label confirmed, but I have no data, picture only.
FP is floating point, ALU is ALU
Wherever it is possible, the function of chips described according to the datasheets description - so often found less common terms, like a "integer processing unit" (instead CPU) and "Full floating point data path" (instead FPU)
IV Part I. Mysterious chips:
1004 Photo only. 1983
1010 No information.
2048 No information.
Part II. Early multipliers
1010 NMOS 16x16 Parallel Multiplier-Accumulator 1983
1516 NMOS 16x16 Parallel Array Multiplier 1984
IV 2010 CMOS 16x16 Parallel Multiplier-Accumulator 1986
2516 CMOS 16x16 Parallel Multiplier-Accumulator 1986
2517 CMOS 16x16 Parallel Multiplier-Accumulator 1983
2245 CMOS 16x16 Parallel Multiplier-Accumulator 1986
9017 Multiplier ??? 1984
Part III. x86 FPU
1067 Purpose known., possibly 80286 coprocessor
*1167 80386 replacement (one chip 1163+1164+1165 realisation) 1986
2167 Purpose known., possibly 80386 coprocessor
3167 32b Abacus 80386 FPU
4167 CMOS 32b Abacus 80486 FPU 1989
Part IV. Motorola 68k FPU (as far as it is known, used mainly with Moto)
1063 No information.
1064 32b FP Multiplier
1065 32b FP ALU
1066 32b 32-word register file (support chip for 1232/33, 2264/65, 1264/65 FP subsistem)
V 2264 32b FP Multiplier/Divider
2265 32b FP ALU
Part V. NS 32k FPU
3164 National Semiconductor 32k FPU (requires NS32580 FP controller for work)
Part VI. SPARC FPU (as far as it is known, used mainly with Sun)
1031 No information.
1032 32b FP Multiplier/Divider
1033 32b FP ALU
1066 No information.
1163 FP controller (80386-1064/65 interface) 1986
*1164 64b FP Multiplier/Divider (used in SUN-4)
*1165 64b FP ALU (used in SUN-4)
1231 No information.
1232 CMOS 32b FP Multiplier/Divider 1986
1233 NMOS 32b FP ALU 1986
1264 32b FP Multiplier/Divider 1986
1265 32b FP ALU 1986
**3170 32b FP Multiplier/Divider (used in SPARCstation 1)
**3171 32b FP ALU (used in SPARCstation 1)
**3172 Full FPU ??? (used in SPARCstation IPC / SLC)
Part VII. Different FPU for scientific calculations
3132 Full FP data path (FPU, used in Connection Machine-2)
3232 Graphics FP databus data adress ???
3332 Full FP data path
3364 64b full FP data path (used in QCD machines - ???)
Part VIII 7000 series FPU and CPU
7034 32b FP Multiplier ??? (used in CDC Cyber 910 with Motorola CPU)
7035 32b FP ALU ??? (used in CDC Cyber 910 with Motorola CPU)
7136 32b programm sequencer for 7137
7137 32b integer processing unit (CPU, MIPS based ???) 1986
Part IX. XL-8000 RISC processor series (Weitek own RISC architecture)
8000 integer processing unit (CPU) 1987
8032 32b single-precision FP processor 1987
V 8064 64b double-precision FP processor 1987
8136 program sequencing unit for XL-8000
8137 integer processing unit (CPU)
Part X. XL-8200 RISC processor series (PA-RISC based, used in printers???)
8220 Printer Controller - 32-bit RISC processor
8232 32b graphic FPU
8236 22b raster code sequencer (1/2 part of CPU for raster printing) 1988
8237 32b raster image processor (2/2 part of CPU for raster printing) 1988
Part XI. SPARC CPU series (last Weitek cpu's early 1990)
8601 SPARC CPU
8701 SPARC CPU
Power uP SPARC CPU (modified 8601 - may be 8701 in diff. pack. ???)
8720 SPARC Integrated Graphics Controller for accelerated 2D color displays
***Part XII. P9000 series graphic chips, 1991
P9000
P9001
VI P9100
P9130
Part XIII. VideoPower 5x86 series video chips
VideoPower 5086
VideoPower 5186
VideoPower 5286
Part XIV. PCI graphic controllers, Intel chipset, last Weitek chips
W464 64bit PCI graphic controller for 80486 1995
W564 64bit PCI graphic controller for PI 1995
Thank you that you have read!
Please indicate inconsistencies and add information if you have it. There is not enough information about this company, but their chips are expensive, rare and beautiful - I think would be useful to finally collect information about them in one place.
* 1st correction (thanks John and isa-d):
1164 64b FP Multiplier/Divider (used in SUN-4)
1165 64b FP ALU (used in SUN-4)
1163+1164+1165 is a complete replacement of the 386
** 2nd correction (thanks Birdman):
3170 Full FPU (used in SPARCstation 1)
3171 Full FPU (used in SPARCstation 1)
3172 Full FPU ??? (used in SPARCstation IPC / SLC)[/list]
*** 3rd correction (thanks Vlask):
P9001 ISA->PCI bridge (used with P9000 on PCI Diamond Viper)
P9130 video acceleration support chip (used with P9100).
IV 4th correction (thanks magictom and UMMR):
early Weitek chips is a second sources for TRW:
1004 possibly TDC1004 64-bit Digital Correlator
1010 confirm TDC1010 16 x 16 TTL Multiplier
2010 TMC2210 16 x 16 CMOS Multiplier
V 5th correction (thanks John again):
2264 32b FP Multiplier/Divider
2265 32b FP ALU
Is single precision and
2364 32b FP Multiplier/Divider
2365 32b FP ALU
Is double precision
8032 32b single-precision integer and FP full CPU
8064 64b double-precision integer and FP full CPU
VI 5th correction (thanks isa-d again):
P9100 Framebuffer used on Tadpole Sparcbook 3GX
Last edited by achtung049 on Tue Nov 03, 2015 7:32 am; edited 8 times in total |
|
| Back to top |
|
 |
CPUShack

Joined: 16 Jun 2003 Posts: 34259 Location: State of Jefferson, USA
|
|
| Back to top |
|
 |
isa-d

Joined: 16 Aug 2006 Posts: 2984 Location: Italy
|
|
| Back to top |
|
 |
isa-d

Joined: 16 Aug 2006 Posts: 2984 Location: Italy
|
Posted: Sun Aug 09, 2015 11:48 am Post subject: |
|
|
about this
Power uP SPARC CPU (modified 8601 - may be 8701 in diff. pack. ???)
its number is 2000 and without heatsink looks completely different than 8701
... but same pinout like 8601 and 8701 |
|
| Back to top |
|
 |
achtung049

Joined: 04 Feb 2014 Posts: 481 Location: Russia, Vologda
|
|
| Back to top |
|
 |
Birdman.

Joined: 13 Nov 2012 Posts: 833 Location: Finland
|
Posted: Sun Aug 09, 2015 3:16 pm Post subject: |
|
|
3170, 3171 and 3172 should all be complete SPARC FPUs. I've taken die shots of all three and they have some differences but many similar areas on their dies like most of datapath. I hope someone finds some information about their functional differences so that we can know why dies have those differences.
In addition to those three I've also released 3132 and 8601 die shots and I have 8136 and 8137 waiting for release. I hope those die shots are useful and give some new information. Unfortunately, some Weitek chips have been too rare and expensive for die shots so far like x86 coprocessors. |
|
| Back to top |
|
 |
achtung049

Joined: 04 Feb 2014 Posts: 481 Location: Russia, Vologda
|
Posted: Sun Aug 09, 2015 3:46 pm Post subject: |
|
|
| Thanks, updatet! |
|
| Back to top |
|
 |
Vlask
Joined: 19 Nov 2004 Posts: 41 Location: Czech Republic
|
Posted: Tue Aug 11, 2015 3:56 pm Post subject: |
|
|
Think that P9001 is only ISA->PCI bridge, not graphic chip. Never seen one standalone. Only in combination with P9000 on PCI Diamond Viper.
Same is P9130 - that should be video acceleration addon chip, not graphics chip. Used together with P9100. _________________ Not only mine graphics card collection at http://www.vgamuseum.info |
|
| Back to top |
|
 |
achtung049

Joined: 04 Feb 2014 Posts: 481 Location: Russia, Vologda
|
Posted: Wed Aug 12, 2015 2:12 am Post subject: |
|
|
Excellent, fixed! Thanks!  |
|
| Back to top |
|
 |
magictom

Joined: 14 May 2009 Posts: 2281 Location: Hawaii
|
Posted: Wed Aug 12, 2015 3:30 am Post subject: |
|
|
I have no hard proof of this, but always thought several of the early Weitek chips were clones of TRW signal processors, and followed their numbering convention. I.e.
Weitek TRW
1004 = TDC1004 64-bit Digital Correlator
1010 = TDC1010 16 x 16 TTL Multiplier
2010 = TMC2010 = TMC2210 16 x 16 CMOS Multiplier
It would be great to see some documentation to that effect, though .. |
|
| Back to top |
|
 |
UMMR

Joined: 23 Aug 2007 Posts: 381 Location: Udine, ITALY
|
Posted: Wed Aug 12, 2015 12:32 pm Post subject: |
|
|
| magictom wrote: | I have no hard proof of this, but always thought several of the early Weitek chips were clones of TRW signal processors, and followed their numbering convention. I.e.
Weitek TRW
1004 = TDC1004 64-bit Digital Correlator
1010 = TDC1010 16 x 16 TTL Multiplier
2010 = TMC2010 = TMC2210 16 x 16 CMOS Multiplier
It would be great to see some documentation to that effect, though .. |
Hi,
for the 1004 I have no infos but for the 1010 I can confirm it's a 16 by 16 multiplier derived directly from a TRW design.
Check this: http://archive.computerhistory.org/resources/access/text/2012/10/102746380-05-01-acc.pdf.
Paolo |
|
| Back to top |
|
 |
achtung049

Joined: 04 Feb 2014 Posts: 481 Location: Russia, Vologda
|
Posted: Wed Aug 12, 2015 2:34 pm Post subject: |
|
|
Wo-ow!
many thanks magictom and UMMR - it's really important piece of the puzzle! I was not able to find any information about these chips... |
|
| Back to top |
|
 |
CPUShack

Joined: 16 Jun 2003 Posts: 34259 Location: State of Jefferson, USA
|
|
| Back to top |
|
 |
CPUShack

Joined: 16 Jun 2003 Posts: 34259 Location: State of Jefferson, USA
|
|
| Back to top |
|
 |
CPUShack

Joined: 16 Jun 2003 Posts: 34259 Location: State of Jefferson, USA
|
Posted: Wed Aug 12, 2015 4:29 pm Post subject: |
|
|
also:
8000 integer processing unit (CPU) 1987
8032 32b single-precision FP processor 1987
8064 64b double-precision FP processor 1987
This is a Weitek architecture (RISC but not MIPS)
the XL8000 is integer only (software for FP)
The 32 and 64 are integer AND FP, not just FP _________________ New for 2025! The CPU Shack has a co-processor!
Visit The CPU Shack of microprocessor history and information. |
|
| Back to top |
|
 |
|