DXZeff's Computing Forum
DXZeff's Computing Forum
Search | Statistics | User Listing Discord | Forums | Language
You are logged in as a guest. ( logon | register )

UMC U5SD ... Not actually a DX class processor?
View previous thread :: View next thread
   Computer Speak -> Hardware (DOS & Windows)Message format
 
waybacktech
Posted 2017-02-28 6:03 AM (#253)
Subject: UMC U5SD ... Not actually a DX class processor?



IDT WinChip

Posts: 237
10010025
Location: USA

Has Wikipedia lied again? Not to mention a few other websites which list information about the UMC U5SD processor as having an FPU? Well unless none of my motherboard correctly support a UMC U5SD and are leaving the FPU disabled, this certainly isn't adding up. Testing several motherboards, with chipsets including UMC, Intel, SIS and Symphony, this processor does not apparently actually contain an FPU. I found a 486 Benchmarking video on Youtube, from a guy in Russia, which this processor came from, and I checked the numbers on his U5SD chip and it is not the same one I have, and I looked at the information displayed on his screen, and the software he ran also listed his as not having an FPU. Thoughts?

Another interesting thing I discovered is that the Cyrix cache utility for their SLC/DLC/Drx2 chips does indeed detect this chip as being a Cyrix processor and utility says "setup complete!" where as other processors, including Cyrix 586, says Cyrix cpu not found. Interesting as some software and Bios does detect the U5SD chip as a Cyrix DLC/SLC or a TI processor, while others list it as a UMC U5S processor, or simply UMC. Fluke or could there be some kind of hybrid Cyrix/Intel design contained in these UMC processors? No way to know for certain either way.

Certainly had no idea that the U5SD would actually lack an FPU, but that seems to be the case, unless there is some kind of utility that has to be ran ( like the Cyrix cache enable utility ) for the FPU on the U5SD chip to function.





(umc_U5SD.jpg)



(UMC_U5SD_2.jpg)



(UMC_U5SD_3.jpg)



(UMC_U5SD_4.jpg)



(UMC_U5SD_5.jpg)



Attachments
----------------
Attachments umc_U5SD.jpg (76KB - 738 downloads)
Attachments UMC_U5SD_2.jpg (134KB - 718 downloads)
Attachments UMC_U5SD_3.jpg (181KB - 670 downloads)
Attachments UMC_U5SD_4.jpg (116KB - 681 downloads)
Attachments UMC_U5SD_5.jpg (106KB - 688 downloads)
Top of the page Bottom of the page
DXZeff
Posted 2017-02-28 6:42 AM (#254 - in reply to #253)
Subject: Re: UMC U5SD ... Not actually a DX class processor?



TM Crusoe

Posts: 618
500100
Location: Hull, UK
...slides away from the Wikipedia page quietly... That is indeed a mystery, because if there is no difference, then the question is, what is with the model number change?

The only thoughts I have at this point are, how does it actually hold up against a Cyrix 486S and is there any noticable gain/loss against the U5S/U5SX? If it is somehow different to the latter, I wonder if it would work with the 487S where the U5S does not, or else if UMC planned a similar upgrade but never actually made it. Unlikely but one has to wonder. In reality I have a suspicion the Cyrix identification might just be down to how certain boards and software detect processors as I've seen a few chips identify as Cx/Ti486 processors before when the code has no idea what it is talking to. One example being PCCheck which has some weird detection routine which seems to place priority more on Family/Stepping than CPUID. Said program reports the Pentium IV as a Cyrix 486SLC. Other programs simply give up as soon as CPUID fails, and many only contain AuthenticAMD or GenuineIntel but no instance of UMCUMCUMC (chortles childishly at the middle of the CPUID), and because Cyrix chips did not report a CPUID the program then falls back to the same code and assumes a Cyrix.

Of course, it is also entirely possible they copied.

If this does turn out to lack an FPU I guess I'll have to modify the Wikipedia page and assume the other sources were incorrect. This also puts a further damper on finding a DX or even learning about them as those who own them generally dislike talking to others and even then, will likely refuse to power things on.


Still, it's an extremely rare chip, so there is that.

Edited by DXZeff 2017-02-28 6:44 AM
Top of the page Bottom of the page
waybacktech
Posted 2017-02-28 4:35 PM (#256 - in reply to #254)
Subject: Re: UMC U5SD ... Not actually a DX class processor?



IDT WinChip

Posts: 237
10010025
Location: USA
I was just teasing you a little bit there about Wikipedia lying as I know who wrote up that Wiki article

Chiplist.com list the following for this cpu -
UMC U5SD CPU

Intel i80486DX CPU pin compatible UMC U5S CPU.

25 MHz: August 1994.
33 MHz: 2.25 W, August 1994.
40 MHz: August 1994.

Manufacturing: 0.6 micron CMOS.

ID: DH = 0x04 (family ID), DL = 0x1X (model ID, revision).

CPUID: family=0x4, model=0x1.

Pin compatible with a 486DX? What could that mean? Were there 486DX only motherboards that existed? Maybe these are defective U5D's? Speculation at the moment is all I really have.

Somewhat off topic but still relevant perhaps to UMC U5SD, but I found also a motherboard manual online stating the voltage for this chip was actually 3.45V and not the usual 5V. More on topic, manuals I have seen online with jumper settings specifically called out for a U5SD have no different settings from that of the U5S counterpart. Usually there will be some sort of a jumper change between an SX and DX processor, even it is just 1 jumper, however this is not the case with my M919 as either SX or DX class processors work with jumper settings set for their respective manufacturer, so who the hell knows on that. So many 486 boards made, so many jumper settings!

As far as the bios ID as a cyrix or IT processor, I am not worried about that, I am sure it is just simply not understanding what a UMC processor is. But I find it interesting that Cyrix'x own utility thinks it is a Cyrix chip as well, which is the first time I have noticed this with a UMC chip, albeit first time I have actually ran any of my UMC chips on this hard drive since installing the cache utility for the 486DLC chips, and no the utility is not causing the FPU to not be recognized, I already check that, before someone asks

M919 - Bios date sometime in 1996, newest 486 BIOS I have on any board - bios detects cpu as UMC U5S. No FPU detected. Jumper settings for Cyrix processor indeed work with this UMC chip. I mention this because usually they will only work with Cyrix processors, AMD/Intel usually won't post or exhibit strange behavior if board is left configured for Cyrix. U5SD on the other had, seems happy. I was trying to see if cpu jumper settings would make a difference in enabling the FPU, so that is how I discovered this.

Micronics 486 board with Intel Saturn 420 Chipset ( Board from HighTreason ) Sees processor as a 486SX running at 100Mhz ( lol ) while actually operating at 33Mhz. No FPU detected.

Shuttle Hot 433 V4 UMC 8881 chipset ( basically the same as M919 but without VLB ) sees processor as a TI chip, no FPU enabled. ( Bios update needed obviously since it sees the chip as a TI )

SIS 471 and 2 491 chipset boards ( one being Chicony CH-491 ) - Don't recall bios processor ID, but none detected FPU.

Last board so far is the Symphony based board shown in these photos. Same results. No FPU. Quake will not run. NSSI runs it's FPU benchmark in emulation, same as an Intel SX chip.


Benchmark wise, I have only chalked down the topbench score of 226 @ 33Mhz, which is in line with Intel/AMD/Cryix DX50 scores, and feels like a DX50 chip, including the rapid memory test. Next benchmark I was starting to run was config to measure Dhrystone/KWhetstone performance, and that's when I stopped. No Kwhetstone's being displayed.

Sure seems to me this chip does not have an FPU. While I have not tested every board that I own with this chip yet, none of the board I consider to be my best boards will recognize or enable the FPU if it does exist, which leads me to think it does not have such functionality. I can't believe they would design a DX class processor and it only work on a very limited number of boards. But I suppose that is a possibility.

I think Paul you might be on to something with the 487 / external FPU upgrade idea. While it might be a fluke that Cyrix keeps coming up in relation to this processor, they might have intended this chip to be used with the 487 Cyrix add-on board that you have which fits between the cpu and socket. That might explain the "Cyrix" nature of this chip. U5SD - U5S upgradable to a D (DX class) with Cyrix or some other FPU add on board, or Intel 487. i80486DX Pin Compatible UMC U5S description I posted at the beginning of this replay leads me to such conclusions at this moment.

Either way, I am a bit disappointed actually... was all set to pit this against the usual DX suspects and see it blow them away in Quake with a 486DX face off.


Edited by waybacktech 2017-02-28 4:50 PM
Top of the page Bottom of the page
waybacktech
Posted 2017-02-28 5:15 PM (#257 - in reply to #253)
Subject: Re: UMC U5SD ... Not actually a DX class processor?



IDT WinChip

Posts: 237
10010025
Location: USA
Some other tidbits from chiplist about UMC CPU's in general. This particular list was compiled in 1995. http://ftp.lanet.lv/ftp/mirror/x2ftp/msdos/programming/faq/pc_chip....

2.36 UMC 486 CPU

The UMC 486 CPU does violate some of Intel's patents and will therefore not be
sold in the USA.

CPUID: "UMC UMC UMC".


2.36.1 UMC U5S CPU

Intel i80486SX CPU instruction/pin compatible, no FPU (Floating Point Unit).

8 kbyte cache.
4 deep write buffer.

25 MHz: August 1994.
33 MHz: 2.25 W, August 1994.
40 MHz: August 1994.

Manufacturing: 0.6 micron CMOS.

ID: step level A: DH = 0x04 (family ID), DL = 0x23 (model ID, revision).

CPUID: family = 0x4, model = 0x2.


2.36.2 UMC U5SD CPU

Intel i80486DX CPU pin compatible UMC U5S CPU.

25 MHz: August 1994.
33 MHz: 2.25 W, August 1994.
40 MHz: August 1994.

Manufacturing: 0.6 micron CMOS.

ID: DH = 0x04 (family ID), DL = 0x1X (model ID, revision).

CPUID: family = 0x4, model = 0x1.


Edited by waybacktech 2017-02-28 5:17 PM
Top of the page Bottom of the page
waybacktech
Posted 2017-02-28 5:28 PM (#258 - in reply to #257)
Subject: Re: UMC U5SD ... Not actually a DX class processor?



IDT WinChip

Posts: 237
10010025
Location: USA
Multiprocessor support!
Lots of information and some answers. Looks like an email of a datasheet for the UMC U5S series.

http://datasheets.chipdb.org/UMC/U5S.txt

— SU.HARDW.PC.MOTHERBOARD (2:5020/299) —————————————— SU.HARDW.PC.MOTHERBOARD —
From : Sergey Kutikin 2:5020/324.16 Sun 19 Mar 95 03:55
Subj : UMS U5S è êýø
————————————————————————————————————————————————————————————————————————————————

ß ïîäóìàë,ìîæåò áûòü ýòî áóäåò èíòåpåñíî:
-------------------------------------------------------------------------
(C) UMC 1993,1995 U5S Introduction
-----------------
Compatible with Industry-Standard 486 SX System Management Mode

* 32-bit A/D buses and registers * GREEN CPU: fully static
* Dynamic 8/16/32-bit data-bus sizing design with SMM power
* Virtual,demand-paged memory managment management.
* On-chip 8 Kbyte,integrated,four-way * Accessible with interrupts.
associative instruction/data cache. * Automatic save/restore of
interrupted states.
Faster Execution * Transparent to operating
system.
* Executes frequently used instructions
30% to 48% faster than 486SX processor Binary Compatible with x86
at same clock rate. Software
* Burst data bus transfers: 80 MBytes/sec * Executes 486SX instructions.
at 25 MHz,106 MBytes/sec at 33 MHz, * Supports Windows,WindowsNT,
128 MBytes/sec at 40 MHz. MS-DOS,OS/2 and UNIX.

Standard 168-Pin PGA and 208-Pin LQFP Debug and Test Support
Package * Built-in self test.
* Hardware debug and test
0.6m(ïîäpàçóìåâàåòñÿ - "ìþ"CMOS Process registers.

5V or 3.3V Operating Voltage Multiprocessor Support
* Cache coherency.
* Support for level-2 cache.
* Multiprocessor instructions.
------------------------------------------------------------------------------
With my best regards,Sergey
--- GEcho 1.02+
* Origin: -= Breeze Ltd. =- (2:5020/324.16)

— SU.HARDW.PC.CPU (2:5020/299) —————————————————————————————— SU.HARDW.PC.CPU —
From : Vladimir Zhevnyak 2:5030/85.21 Sun 02 Jul 95 23:53
Subj : Re:U5S
————————————————————————————————————————————————————————————————————————————————

Äyìàþ, íàäî-áû ëþäÿì ïîìî÷ü, pàçüÿñíèòü ÷òî-ëè!
À òî ëåæèò ïîíèìàåøü ëè y ìåíÿ íà ñòîëèêå êíèæêà "U5S GREEN CPU"
1994-1995 DATA BOOK United Microelectronics Corporation.

Chapter 1: INTRODUCTION:
------------------------
The U5S GREEN CPU micriprocessors are a family of high-performance,
32-bit, static implementations of 486 SX architecture with energy-efficient
system management functions added. All processors ih the family support
the 486 SX instruction set at execution speed that, for most instructions,
are 30% to 48% faster than the 486 SX processor at same clock rate.

The ten processors ih the U5S family offer a mix of 25,33 or 40 MHz
operating speed for either the SX or DX sockets. The LV processors in the
family also support 3.3V operating voltage. All processors support a
power-saving System Management Mode. All processors are binary-compatible
with 486 SX operating system and application software, including software
that runs under Windows(R), WindowsNT(R), MS-DOS(R), OS/2(TM) and UNIX(TM).

The 8K byte on-chip cache and memory managenent logic implement fast,
coherent, demand-paged virtual memory management. The cache is software -
transparent so to be compatible with older x86 software.

All processors use a 1x external clock. The processors that operate at 5V
have a power consumption of 2.6W at 33MHz. Those that operate at 3.3V have
a power consumption of 1W at 33MHz. The static design supports clock freq-
uencies that can range dynamically down to zero to reduce power consumption.

The System Management Mode (SMM) in all processors is an added layer of
operating mode, above the standart Real, Protected, and Virtual-8086 modes.
SMM operates in an isolated memory space. It is transparent to all
operating-system and application software and support firmware-configured
function like intelligent, energy-efficiant power management functions.
The static design future, which allows clock frequencies to vary down
to zero, plus the optional SMM, support the design of energy-efficiant,
battery powered portable systems with extended battery life. The SMM is
implemented through a non-maskable System Management Interrupt (SMI)
which has higher priority than the NMI interrupt.

The improvements in the U5S instruction latencies relative to 486 XS
latencies, the compatibility of instruction execution and pinouts, and
the SMM for power management, make the U5S family of processors ideal for
supporting the broad base of existing x86 application and system software
on a wide range of system configurations, from desktop to portable.

Ôy, àæ âñïîòåë!
Çäåñü âñå íàïèñàíî è ñêàçàíî. Äyìàþ òåìy ïpî U5S ïîpà íàêpûâàòü !?

--- Îãîëåííûé Äåäyëÿ çèìîé 41 ãîäà.
* Origin: Cobra & Brosers Station (2:5030/85.21)

— SU.HARDW.PC.CPU (2:5020/299) —————————————————————————————— SU.HARDW.PC.CPU —
From : Aleksandr Konosevich 2:5004/9 Sat 06 Jan 96 18:27
Subj : CPU FAQ





* On-chip 8Kbyte, integrated, four-way associative instruction/data cache.
* 0.6m CMOS process
* 5V or 3.3V Operating Voltage
* Support Windows(r), WindowsNT(r), MS-DOS(r), OS/2(r) and UNIX(r)
* Built-in self test

U5 S D LV - SUPER33
^ ^ ^ ^ ^
| | | | | Performance VS. Speed :
| | | | \ Super25: Superior Performance at 25 MHz
| | | | Super33: Superior Performance at 33 MHz
| | | | Super40: Superior Performance at 40 MHz
| | | |
| | | \ Voltage Supply :
| | | : 5V Operating Voltage
| | | LV : 3.3V Operating Voltage
| | |
| | \ Pin-Out Configuration and Power Consumption :
| | : SX PGA Socket with SMM Power Management Pins Built-in
| | D : DX PGA Socket with SMM Power Management Pins Built-in
| | F : LQFP Pin-out with SMM Power Management Pins Built-in
| |
| \ Series Code :
| S : 486SX Function Compatible
|
\ Family code :
U5 : UMC Microprocessor, Green CPU

Part-specific pin table :

Pin # Part U5S Part U5SLV Part U5SD

G15 SMIRDY# SMIRDY# SMIRDY#
A13 SMI# SMI# SMI#
A15 NMI NMI N.C.
B15 N.C. N.C. NMI
R17 SMIADS# SMIADS# SMIADS#

------




Edited by waybacktech 2017-02-28 5:32 PM
Top of the page Bottom of the page
DXZeff
Posted 2017-03-01 1:08 PM (#259 - in reply to #258)
Subject: Re: UMC U5SD ... Not actually a DX class processor?



TM Crusoe

Posts: 618
500100
Location: Hull, UK
Very interesting. I'm not sure why they'd do that, presumably compatibility with certain boards?

I'm not sure what I can add other than I certainly have some time with the MediaWiki interface to look forward to, shame that this information never came up in any search I did (damn Google and their custom results crap) as it would probably have saved a lot of hassle.
Top of the page Bottom of the page
waybacktech
Posted 2017-03-01 3:48 PM (#261 - in reply to #259)
Subject: Re: UMC U5SD ... Not actually a DX class processor?



IDT WinChip

Posts: 237
10010025
Location: USA
I'd bet money these are defective U5D's. That's the only thing that would make sense to me as to why they would be "DX Pin Compatible", unless it will work with that 487S chip you have. Doesn't say in these docs anything about FPU add-on, but it looks like this was only a partial write up from a book or spec document or something. I'd like to see more of the "book" on this processor.. Still though, it is rare version if only by it's markings. Now I see why this chip was $15 shipped from Russia. Good price for any UMC 486 chip though. The CPU ID chart gives claim to the "Super" variants being faster than, I am assuming, the normal Green chips. Certainly worth a test.

Probably a good thing you didn't have someone on those message boards you posted to con you out of a lot of money to get one of these. I guess the U5D is just the unicorn of 486 processors, made of unobtainium, and only the specially chosen Acolytes possess the precious

Edited by waybacktech 2017-03-01 3:50 PM
Top of the page Bottom of the page
DXZeff
Posted 2017-03-02 10:43 AM (#262 - in reply to #261)
Subject: Re: UMC U5SD ... Not actually a DX class processor?



TM Crusoe

Posts: 618
500100
Location: Hull, UK
That was also a thought which entered my mind. I think the Super was just marketing, because the U5SX I have performs identically to the U5S at the same clock. It is strange how all the non-Super ones appear to be from 1995 onwards, where the Super marking is generally from 1993-1994.

I am doubtful as to whether the 487S would work, but it is certainly a possibility even if unintentional. I can try to dig that out for you if you want? I doubt if I'm ever going to use it and I could probably source another if I really needed to. I do not know if the motherboard requires specific support for the device or not as I've seen some jumper lists mention it, but never stopped to pay attention to whether the settings were the same as other chips on the list. It does occur to me that I have no idea how well the FPU performs, it would be amusing if it did work but turned out to be worse than the internal emulation.

Luckily it doesn't seem as though rip-offs are common on CPU-World. I had someone rob me of money elsewhere a few years ago, but had paid for it as a "Purchase" and not a "Gift" so I was able to back-pay it. The jerk had the nerve to then PM me and claim that he still had the money, so it didn't matter that I had back-paid it or some BS. A few minor incidents elsewhere, but that was for small things like disks that were cheap enough I generally didn't care do do much more than make sure nobody else tried buying from them... Not about to bludgeon somebody over £1.99.

I question the validity of some that claim to own them. Given they only show grainy pictures at best, refuse to answer questions, post better images or, oftentimes, even communicate at all. Always seems fishy to me. Then again, there are a lot of truly anti-social people on the web so you never know.
Top of the page Bottom of the page
brassicGamer
Posted 2017-03-24 1:00 PM (#386 - in reply to #262)
Subject: Re: UMC U5SD ... Not actually a DX class processor?


Slow 8088

Posts: 4
0
DXZeff - 2017-03-02 10:43 AM

That was also a thought which entered my mind. I think the Super was just marketing, because the U5SX I have performs identically to the U5S at the same clock. It is strange how all the non-Super ones appear to be from 1995 onwards, where the Super marking is generally from 1993-1994.


I could verify this as I have a U5SX 486-33 and a U5S-SUPER33. May basic guess is that they're identical but when the lawsuit happened they rebadged them and used 'Super' instead of '486'.
Top of the page Bottom of the page
CyrixInstead
Posted 2017-05-10 11:46 AM (#572 - in reply to #253)
Subject: Re: UMC U5SD ... Not actually a DX class processor?



Intel P5

Posts: 85
252525
Location: UK
I just installed a UMC U5S Super 40 CPU in my Zida Tomato 4DPS motherboard. In fact that is how I found my way here! A google for Jumper settings (no listing for the UMC in the manual) brought me here top link. as a matter of interest, it works great with Intel SX jumper settings and the BIOS correctly identifies it.
Great little chips these

Top of the page Bottom of the page
Jump to forum :
Search this forum
Printer friendly version
E-mail a link to this thread

(Delete all cookies set by this site)
Running MegaBBS ASP Forum Software
© 2002-2024 PD9 Software