VOGONS


Reply 40 of 60, by Stiletto

User metadata
Rank l33t++
Rank
l33t++
feipoa wrote:

Unfortunately, the pdf is too large to attach in its entirety. The full PDF is 19 MB and there is a 5 MB upload limit. I have attached just Appendix D from the SXL reference guide.

You can link, you know. Bitsavers and Archive.org aren't going anywhere any time soon 😉

http://www.bitsavers.org/components/ti/TI486/ … rence_Guide.pdf
https://archive.org/details/bitsavers_tiTI486 … eGuide_19799789

"I see a little silhouette-o of a man, Scaramouche, Scaramouche, will you
do the Fandango!" - Queen

Stiletto

Reply 41 of 60, by debs3759

User metadata
Rank Oldbie
Rank
Oldbie
Stiletto wrote:
You can link, you know. Bitsavers and Archive.org aren't going anywhere any time soon ;) […]
Show full quote
feipoa wrote:

Unfortunately, the pdf is too large to attach in its entirety. The full PDF is 19 MB and there is a 5 MB upload limit. I have attached just Appendix D from the SXL reference guide.

You can link, you know. Bitsavers and Archive.org aren't going anywhere any time soon 😉

http://www.bitsavers.org/components/ti/TI486/ … rence_Guide.pdf
https://archive.org/details/bitsavers_tiTI486 … eGuide_19799789

Thanks

See my graphics card database at www.gpuzoo.com
Constantly being worked on. Feel free to message me with any corrections or details of cards you would like me to research and add.

Reply 43 of 60, by PiotrUU

User metadata
Rank Newbie
Rank
Newbie

Hello!
I was looking for information about motherboards with 486sxl support (pga168) - I found information about motherboards: DataExpert EXP4349 (manual), Micronics LX30WB (manual), Vision Technologies VLR486 (stason.org). Are there other motherboards for this processor? Do you work on any Motherboard?

Reply 45 of 60, by PiotrUU

User metadata
Rank Newbie
Rank
Newbie

I am looking for information that other motherboard models will allow to run the 486sxl processor.
I have a processor (sxl2-50) - I am looking for any compatible motherboard.

Reply 46 of 60, by feipoa

User metadata
Rank l33t++
Rank
l33t++

The EXP4349 will run the PGA-168 version of the SXL2. I tested this configuration before my motherboard died. The BIOS on that board has issues though, it will only let you enter the BIOS if the battery has been drained or there is some other form of CMOS mismatch. Even though it says, "Press DEL to enter BIOS", it won't enter BIOS. To get a CMOS mismatch, I had to remove the BIOS and put in another chip.

Any reason you don't want to use the PGA-132 SXL2? There are more motherboards which will work with this chip.

Plan your life wisely, you'll be dead before you know it.

Reply 48 of 60, by dionb

User metadata
Rank l33t++
Rank
l33t++

Bump.

Recently received a TI 486SXL2-66. I figured my DataExpert OPTI-4965SX-2VL, which explicitly supported Cx486DLC would stand a good chance of supporting it. I was wrong. The system boots up with Cx486DLC jumper settings, but hangs in POST on the message: "A20 Gate Stuck in Disabled State (A20=0) - System Halted". Got to love MR-BIOS error messages, but looks like this is one more board we can strike off. I'm running MR-BIOS for OPTi 496SX v160.

Reply 49 of 60, by feipoa

User metadata
Rank l33t++
Rank
l33t++

Your board has a 3.6 V Vcc option? Did you mess around with the cyrix DLC register settings? I think there was an A20 option in there.

Plan your life wisely, you'll be dead before you know it.

Reply 50 of 60, by dionb

User metadata
Rank l33t++
Rank
l33t++
feipoa wrote on 2020-02-07, 23:36:

Your board has a 3.6 V Vcc option? Did you mess around with the cyrix DLC register settings? I think there was an A20 option in there.

No 3.6Vcc option, just used an interposer (3.45V is close enough) - also tried on naked 5V with same result. As for the register settings - can't set anything if it hangs before completing POST.

Reply 52 of 60, by kool kitty89

User metadata
Rank Member
Rank
Member

I've got a TI 486SXL-40 that was purchased as-is several months ago and so far haven't gotten to POST. It's in good cosmetic condition and I assumed it wasn't dead, but it's failed to POST in the several boards I've tried it in. (starting with my OPTi 495SX based 3486 board then a Soyo SiS 471 and possibly an SiS 496 based board)

All attempted at 5V (except the 496 board which would have been at 3.46V at the time) and tried at lower speeds with no POST, so I assumed it was either dead or very picky about what board to use.

Much more recently I picked up a Ti SXL2-66 which so far has also failed to POST in 3 SiS based Socket 3 boards and a Symphony chipset 3486 board set to 486SX and then DX mode. (and I've stopped swapping CPUs into the OPTi board for now as it's in a fairly cramed AT case and I've ended up bending pins while extracting CPUs: I'll use it for testing again when I swap another board into there)

The SXL2-66 was also purchased as-is, but supposedly was pulled from a board in working condition many years ago and left in storage, so seemed likely to be fine. (print and heatsink were in good shape, though pins needed a little adjusting, I think some happened in transit due to lots of padding, but not much rigidity)

However, I don't believe I attempted to run either of these CPUs with a 3486 board set to 386 (or 486DLC) mode, and I've also found that many of the CPU-model/version specific jumpers don't impact things much while others are required so long as all those are the same socket type, while the socket-select ones are much more important. (386 settings work for DLC CPUs, 486SX and DX settings often work for SX or DX CPUs, though DX CPUs seem to be the least picky)

In any case, I don't think I ever tried using 386 (or 486DLC) jumper select settings for one of these Ti PGA-168 socket CPUs, but maybe that could impact things. (I suppose if it's an issue with differences in bus protocol, that could make the difference)

The boards I have might be too old and too new to explicitly list the PGA-168 SLX or Cyrix 486S CPUs. I have 3 386/486 boards now and manuals with 2 of them (the OPTi 495SX and UMC based US 3486), and no mention of these models. I think only the OPTi 3486 manual explicitly mentions the Cyrix or TI 486DLC, though with that board I haven't found any functional or benchmark performance difference between the 386 and 486DLC settings when a DLC is installed. (OTOH, Cyrix 486DX2, IBM DX2 BLs, and ST's DX4 CPUs seem to have no problems running at 5V as 486DX CPUs in these 1991 or 1992 boards)

I'll have to try again with the TI PGA168 CPUs with 386 or 486DLC mode set and see if that does anything. I'd made the assumption those jumpers were socket-specific selections, but given the boards tend to not work at all when both CPU sockets are populated, it makes some sense that both sockets are still active on the I/O side of things. (I remember getting the OPTI board to post in 486DX mode with both sockets populated at one point, but may be wrong and did not attempt to boot an OS at the time: it was also quite clear that both sockets remained powered and should be cooled accordingly: aside maybe from a 386DX and some low-clocked 486s ... I'm certainly not keen on using a DLC at 33 or 40 MHz without heatsink on it)

I considered getting a known-good SXL PGA168 CPU for testing purposes, but it seemed like a not-great investment at around $30 shipped on the low end there. (a few as-is auctions have gone for less, but that wouldn't solve my problem)

I was actually hoping the SXL2-66 would behave more like a Cyrix DX2-66 and be pretty hassle free, compatible with early (or at least Early non-Intel 486) and late gen 486 boards and apparently tolerant of a wide range of voltages while also being pretty overclockable. (the Cyrix/IBM/ST DX2 and Intel DX seem to be the most flexible models I've tried there: that goes for Intel DX-33 and DX-50 CPUs working at 3.3 or 3.45V without issues)

Instead, it seems like they may behave more like DLC/DRX type CPUs just in the PGA-168 socket, and rather being plug-in compatible with typical Intel and AMD 486 compatible boards, might just be for 386/486 chipsets/boards that omit the 386DX and 387DX sockets for a cheaper, simpler layout but the same 386-bus mode compatibility requirements. IBM's SLC/DLC and BL 386-derived CPUs probably relied on chipset and board layouts like those too, but they were all mated pairs, so matching CPU to board isn't the same sort of variable.

And I don't have a bunch of working 386 socket boards to try the 486DLCs in, but so far they've been pretty hassle free too, albeit given the lack of 'Heatsink and Fan Required' on the package label, I wonder if these were typically used without a heatsink or fan (beyond case mounted fan) as they get extremely hot when left bare and without excessive airflow. (they do seem to work when burning hot, but it can't be good for them, or especially some other components in the case ... especially electrolytic capacitors or even the plastic of the socket itself)

I'm going to assume the (much less common) TI 486SXL CPUs in PGA-132 form factor are mostly plug-in compatible in 386 or 486/486 boards without some of the quirks of these PGA-168 CPUs.

Hmm, but come to think of it, if these SXL CPUs are fully electrically compatible with the 386 bus, they could potentially be adapted back to the PGA-132 socket via an interposer/adapder board.

Maybe Socket-2 boards would be in the right age range for supporting these sorts of CPUs as well. I do have a Biostar Bioteq board that looks like Socket 2, though it still needs a BIOS replacement. (got it to post with a mis-matched BIOS and is in good cosmetic condition, but currently not useable otherwise)

Reply 53 of 60, by Deunan

User metadata
Rank Oldbie
Rank
Oldbie
kool kitty89 wrote on 2020-06-06, 10:59:

I'm going to assume the (much less common) TI 486SXL CPUs in PGA-132 form factor are mostly plug-in compatible in 386 or 486/486 boards without some of the quirks of these PGA-168 CPUs.
Hmm, but come to think of it, if these SXL CPUs are fully electrically compatible with the 386 bus, they could potentially be adapted back to the PGA-132 socket via an interposer/adapder board.

I have a couple of PGA-132 SXLs, these do indeed work in a 386 socket without any issues. Basically these are like the Cyrix DLC except sport 8kB of cache and clock doubler. Even the SXL has a clock doubler, not just SXL2 - but you're still limited by the chip top frequency unless it's overclocked.

AFAIR the PGA-168 variant was just Ti attempt to make something for the (more and more common) 486 mobos but without having to invest in making a "true" 486 core that supports bursts and smarter cache invalidation. So it's the same as the -132 variant but in different package and so yes, an interposer of sorts would probably work. That also means a mobo needs to have support for the SXL or else it won't work - it might be possible to hack that by cutting and rewiring some signals but frankly, why bother? It would be slower than a 486SX at the same clock and getting a Cyrix '87 to work with that would still be way inferior to a 486DX chip.

Reply 54 of 60, by Anonymous Coward

User metadata
Rank l33t
Rank
l33t

Was the SXL really TI's attempt to avoid making a true 486? I think the SXL came out only slightly before the TI486DX2/4.

"Will the highways on the internets become more few?" -Gee Dubya
V'Ger XT|Upgraded AT|Ultimate 386|Super VL/EISA 486|SMP VL/EISA Pentium

Reply 55 of 60, by kool kitty89

User metadata
Rank Member
Rank
Member

It doesn't seem like avoiding making a full 486 as much as having the existing 486SXL core available (and having expanded the Cyrix 486DLC from 1kB to 8kB L1 cache) and simply producing it in the PGA-168 form factor.

I'd assume the logic was to be compatible with boards very similar to existing 386/486 boards, but with the PGA-132 386 and 387 sockets omitted. Same chipsets, same feature sets, but using a single, common CPU socket with the added function of enabling 486DLC electrical compatibility over PGA-168.

The PGA-168 SXL seems a lot more common than the PGA-132 ones, so compatible boards must also have been common for the time.

Presumably TI started making TI branded 486DXs as soon as Cyrix had their DX core ready.

The TI 486SXLC appears to have been a straight up QFP-100 386SX pin compatible design with the same 8k cache internally. The 486SXLC2-50 seems to have been the fastest of those produced. (and the only one that's been available on ebay in any numbers for the last year or so: seems like some warehouse in China got access to a bunch of new old stock of those parts: I bough a few of them and they don't show signs of faking or remarking, though I'd have to actually mount one on a 386SX board to know if they work)

Edit: the ones on ebay currently are the SXLC2-G50-PQ model, which someone on CPUworld mentioned is a 3.3V part, though x86guide doesn't mention that:
http://www.cpu-world.com/CPUs/80386/Texas%20I … LC2-G50-PQ.html
http://www.x86-guide.net/en/cpu/Texas-Instrum … cpu-no2263.html

That does seem odd given TI seems to have otherwise stuck with Cyrix's use of the letter V to denote low voltage parts. (like the SXLC2-V40)

Though this 1998 datasheet implies it's 3.3V https://www.datasheetarchive.com/pdf/download … term=TI486SXLC2

And apparently the SXLC2 removed the direct-mapped cache mode in favor of only using the 2-way set-associative scheme. (maybe all the 8kB TI cache SXL/SXLC CPUs did that)

Mouser and some other suppliers listing the part direct the datasheet info to a page on TI's website that no longer exists.

It seems like there's at least a few hundred of the things still in stock with internet order component suppliers:
https://octopart.com/ti486sxlc2-g50-pq-texas+ … ruments-7637217

So I guess it's pretty reasonable to assume those ones in China came from liquidation or clearance of old stock. (and given that 1998 datasheet, they were probably being used in embedded systems applications long after they were irrelevant in the desktop or notebook PC markets, except maybe in niche/developing markets)

And either my TI 486SXL is dead, or this US3486 just doesn't support it as switching to 386 mode didn't do any good. (it'll run a 486DLC at 50 MHz though, which is nice)

Reply 56 of 60, by Deunan

User metadata
Rank Oldbie
Rank
Oldbie
Anonymous Coward wrote on 2020-06-08, 00:03:

Was the SXL really TI's attempt to avoid making a true 486? I think the SXL came out only slightly before the TI486DX2/4.

The timing might have been affected by some other factors. Or, perhaps, it was a big enough batch of unsold SXL chips that Ti wanted to get rid of ASAP knowing the new DX line of CPU is going to make them obsolete.

The reason I think that is that SXL requires some external logic for cache invalidation to even work in 486 socket at all, and not suffer terrible performance loss by doing a full flush on each memory refresh cycle. If Ti were serious about this chip they'd implement this themselves. But obviously it wasn't economical to redesign already existing core and create new mask sets, etc. Also apparently they didn't add anything external during the packaging process - no, they just tried to convince mobo makers to do it for them. But who would bother if a 486SX could be installed without all that trouble? And by that time wasn't all that expensive as even Intel was trying to undercut AMD's 386DX40 on price. One could say that was their first attempt to bankrupt AMD.

But anyway, some CPU upgrades came with external logic that fixes SXL, this is a photo I once found on ebay. Doing it this way however limits your choices of CPU cooling solutions.

So in the end the PGA-168 SXL is pretty much incompatible with most 486 mobos. Then why even bother making it? I see only two logical reasons, they didn't have a proper 486 core ready or they wanted to get rid of inventory glut. But for the latter it would make much more sense to add the extra logic (even externally) to make sure it works with existing mobos. But instead they've argued for extra configuration on new mobos.

Attachments

  • Ti486SXL2-50.jpg
    Filename
    Ti486SXL2-50.jpg
    File size
    227.82 KiB
    Views
    1798 views
    File license
    Fair use/fair dealing exception

Reply 57 of 60, by gonzo

User metadata
Rank Member
Rank
Member

I have tested the Cx486S without any success, too, with:

- one HOT-433 Rev. 1 -> there is POST (1-st POST-page and BIOS-pages possible), but the 2-nd POST-page (about the system-summary) is never complete, so the system hangs

- one HOT-433 Rev. 4 -> POST-/BIOS-pages and DOS-boot are ok, but the L1 cache is always disabled (as shown per CHKCPU; and as shown by a CPU-score of about 7-10 in System Speed Test 4.78, so the CPU de-grades to a better 386...). I tried to use/insert JP24 (usually not for Cx-CPUs on this board) - so one (single) time I got the L1 working in WT-mode (but never again later...).

- one ECS UM8810PAIO Rev.1.1 and one Rev.2.1, also one GA-486VC Rev.1A -> similar problems like the HOT-boards above. If even a DOS-boot is possible, Speedsys hangs on its main-screen showing the system-data etc. (only some upper few rows of this screen are visible). Using "speedsys /sp" for the ECS-board does not help...

I also have an interposer with Cx487-40 inside -> until now, there is no use possible together with the Cx486S-40 under no conditions on no one of those boards (if luky, only the first or both POST-pages are visible).

For me it's difficult to understand, why the Shuttle- and the ECS-boards offers jumper-settings especially for this CPU, but it never works with them?
I tried to use some different (older/newer) BIOSes of those boards, and also two different CPUs (one with the green cooler, and one without it) - again withhout success. Do this CPU maybe need an additional interposer itself, like the Ti-SXL2?

In fact, the HOT-433 Rev.4 was for me the only one board able to boot in DOS with the Cx486S, with L1 automatically disabled, and without the Cx487-interposer.
Here can be found an article of using the same combination (I assume it is Rev. 1.x of this board, not 4.0), but it is not clear how the jumper-settings are, and if the L1 is operating in WT or WB-mode:
https://www.cpushack.com/2021/02/25/the-486-c … locking-part-2/

By the way - I have a Ti486SXL2-66 here, but it does not boot with the above boards under any conditions (as this can be expected due to the experience in this thread).

Reply 58 of 60, by feipoa

User metadata
Rank l33t++
Rank
l33t++
gonzo wrote on 2023-01-29, 16:25:

I have tested the Cx486S without any success, too, with:

Try an m912 motherboard w/ cx486s. I have a thread showing the success of this pair somewhere on vogons.

Plan your life wisely, you'll be dead before you know it.

Reply 59 of 60, by gonzo

User metadata
Rank Member
Rank
Member

Yes, feipoa, I already know your M912-information about the Cx486s.
Sadly, I do not have this board here... 🙁
The reason I posted my Cx486S-experience above is, not at least, to prevent somebody else to waste time for testing with the other boards shown above 😉