Is it mobo cache? The stuff you can put in 386/486 mobos? I wonder if the -12J means 12ns?

Yes, no, and yes. Unfortunately those chips won't work for you (at least not without an incredible amount of trouble), as those are SOJ chips, and what you'd need for most 386/486 boards are DIPs.

Shame, too, as that's a really good price.

I decided to buy some DIPs! 😀 There's a French seller who has lots of them. They're only "slow" 15ns, but hey that's probably OK.

http://cgi.ebay.co.uk/ws/eBayISAPI.dll?ViewIt … em=180569448299

Also, I am nearly finished testing these contaq 386 mobos. A couple of things have occured. Firstly, one of the boards is bad. There's a stain on one of the Contaq chipset chips, like a sticker was once there. When I run memtest, it crashes with a strange unexpected interupt. I removed all components (CPU, RAM, etc) from this board, and put them all in to another identical Contaq mobo, with the same jumper settings and CMOS settings, and I get no problems.

The other thing I have noticed is that I can't seem to get the same "maxed out" speeds any longer. I don't know why. It could be something to do with the 100 MHz oscillator. If I remove it, and put in the 80 MHz osci, then everything is OK - even with all settings (jumpers and BIOS) set to max. I can use the 100 MHz osci, but not at the mobo's top bus speed any more.

I never intend to use this stuff at max speed. I just want to test this stuff to its very limit, then if it works, I know I've got some good hardware. Have you seen that film called K-19 the Widowmaker. The one where Harrison Ford is the Russian nuclear sub captain. He orders the boat to go down to it's max depth, just to see if it will work. Although my 386 isn't nuclear powered, it's roughly the same principle. 😉

RG 100 have you bought any 15ns from that french seller or not yet? I had found sth better than this, and I'm going to check that right now.

Markk wrote:

RG 100 have you bought any 15ns from that french seller or not yet? I had found sth better than this, and I'm going to check that right now.

Yes I just bought some. You say you've found something better. Something faster?

Markk wrote:

RG 100 have you bought any 15ns from that french seller or not yet? I had found sth better than this, and I'm going to check that right now.

edit : A couple of months ago, I'd been searching for cache chips, and I found a guy from Germany that used to sell sets of 50 chips for 10euros. And I saw that he had at least 2 sets for sale, and after one of those were sold, he immediately relisted another one. But unfortunately he hasn't got anything for sale right now. 😢

I have managed to get this 386 mobo stable again, at maximum speed. That is, using a 100 MHz osci, with the bus speed jumpers set to maximum. This is the "/4 option", which I think means 100/4 = 25 MHz bus speed. I did two things to get this to work:

Firstly, you see Donut King's original post. If you look at the 6th photo, you can see the cache chips. I have 15ns cache chips and also the default factory installed 15ns tag chip. However, there's another chip there, labelled Q8888, which is a 20ns chip. I managed to find a 15ns chip, and replaced the default factory 20ns chip. That stops the BIOS POST displaying "bad cache", whenever you attempt to run the mobo at the very fastest rate possible.

The other crucial thing I did was abandon all compact flash devices. I have come to the conclusion that they are problematic, when used with ISA based EIDE controllers. Now, I have a Western Digital 80 GB Caviar drive in place, connected with a 40-pin IDE cable. It works every time without issue, when used with the EIDE controller, at maximum bus speed.

I am using the same kind of CPU that Donut King is using, that is one of those CPUs on a "pin grid package". It seems to work just fine, and produces identical Speedsys and other benchmark scores, compared to another CPU I have used, which was the ceramic based AMD 386DX-40.

Okay, I have been trying to figure out what the hell you guys are doing, so I went back and read all 14 pages of the thread. I didn't realise that there are 3 different people showing off their 386 systems in here.

I really think RG100 has been playing with fire, yet somehow has avoided having anything explode 😳

In my personal experience, running ISA bus higher than 8-10MHz is asking for trouble. Though, I have to admit that I normally use SCSI controllers, which I find don't work reliably above 8MHz. Many people in the past have ran 16MHz and 20MHz, but normally discover quirks down the road. 25 MHz is just insane. Granted, there may be configurations where you can pull it off. Video adapters *normally* don't have issues if you are running chipsets that do double duty for ISA and VLB. Anything made after 1992 should work fine. The same would go for IDE controllers. I would stick with later hardware. I think you're at some point going to run into major problems with soundcard, network card and SCSI controllers. I would be very careful dropping any old card into a 25MHz ISA slot...you could seriously fry something.

50MHz overclocks were pretty common for 386DX-40 CPUs in the early 90s. However, I would argue that running at 40MHz would give you better overall performance. I have discussed this in detail in my experiments with my 50MHz bus 486s. It's really a matter of SRAM cache. A 40MHz system will perform best with all of the memory wait states cranked down to their lowest settings. This is normally possible when you use 15ns SRAM coupled with 15ns tag RAMs. As you have already discovered, running at 50MHz requires additional wait states. Unfortunately, it is normally not possible obtain SRAMs faster than 15ns. 12ns actually exists, but is really uncommon. I have personally owned quite a few boards equipped with 12ns tag RAMs, and that definetly makes a difference, but my memory throughput was still not as high as a fine tuned 40MHz system. Now, having both 12ns SRAM and tag RAMs may make a difference, but I think what would be ideal is 10ns chips. Unfortunately, these are not available as DIPs. I did come up with a way of converting surface mount chips to DIP at one point, but I figured the costs really outweighed the benefits.

Though I think replacing that Q8888 20ns chip was a great idea. You should try to locate a 12ns part if possible. Also, I am not 100% certain, but you might also benefit from using faster DRAM. I believe that 50 or 60ns DRAM was normally recommended for 50MHz EISA 486 systems in the early 90s. I currently use 60ns 16meg SIMMs. I'd ideally like to have 50ns parts, but they are normally only available in low density SIMMs.

I have a suggestion for a future upgrade path. Rather than running everything out of spec, why not try playing around with some of the hybrid 386-486 chips? TI486SXL2-50 chips may be of interest to you, because by default they run with a 1X multiplier. I suspect they would work well with a 50MHz system board. They also have integrated 8kb L1 cache, which is tough to get going, but will bring you up to true 486 level speeds if you get it going. I also experienced large gains in video performance when making the upgrade.

I've got a couple of these boards and one of those 486DLC chips made by Cyrix I think but was kinda waiting for my Fastmath fpu to arrive.
AC, you happen to know if these 486DLC's will work with a Cyrix Fastmath fpu? Can you just mix 'em together?
I've been trying to obtain such a clock doubler for 386 but haven't been successful so far (partly because it's not really high on my priority list).

AC, you also happen to know about the 386's compatibility with 16meg simms? The manual states it works with 1mb and 4mb simms, but makes no mention of the 16mb simms.

I have no experience with 386's except for what I've read on the net. Only have real experience with 486 and up, so the 386 is somewhat unchartered territoty for me 😜

AC, you happen to know if these 486DLC's will work with a Cyrix Fastmath fpu? Can you just mix 'em together?

Supposedly some earlier 387's had issues along side 486DLC chips, but for the most part they're cross compatible. It works the other way, too... I have a DLC coprocessor running alongside my 386.

I've got two identical IIT 4C87DLC-40 fpus. They work fine with 386-40 cpus. And I've also seen fpus by the same company, labeled 3C87-40. But I don't know if they would work also with 486DLCs.

Cheers you 2! 😉

I would suggest to read the Cyrix 386 guide.
http://blackroses.textfiles.com/programming/cyrix.cp
and this for performance
http://www.cpu-info.com/coprocessor/index2.ph … tasheet&tabid=4

If I remember correctly it is much faster if you use a DLC CPU/ DLC FPU combination.

It is well documented that the early generation Cyrix Fasmath FPUs do not work properly with Cyrix 486DLC chips, but the 486DLC should work fine with pretty much any other FPU out there. Unless it appears that your Fasmath is quite old (old style logos) it will likely work fine.

Probably your best bet at finding a clock doubled 386 chip would be either TI 486SXL-40 or the TI 486SXL2-50. Despite their different labels they're exactly the same chip, and clock doubling can be enabled with a CPU config program. Though it can be a little tricky to get the clock doubling to work. Other clock doubled chips are the Cyrix DRx and the IBM DLC2/DLC3.

Not very many 386 boards will take the 16meg SIMMs. Not very many 486 boards take them either. If your board manual doesn't specifically say it supports them, it likely won't work.

elianda wrote:

I would suggest to read the Cyrix 386 guide. http://blackroses.textfiles.com/programming/cyrix.cp and this for performance http: […]
Show full quote

I would suggest to read the Cyrix 386 guide.
http://blackroses.textfiles.com/programming/cyrix.cp
and this for performance
http://www.cpu-info.com/coprocessor/index2.ph … tasheet&tabid=4

If I remember correctly it is much faster if you use a DLC CPU/ DLC FPU combination.

That was a good read! Thanks 😉
Now I know to check the revision of the cpu's I intend to be using together 😀

"TI 486SXL-40 or the TI 486SXL2-50"
I thought these were 386SX?
But I agree with you that the clock doublers are the ones to have, but since they aren't that easy to find, I'll just work with the 'standard' DLC for now 😉

Regrettably I won't be doing much hardware testing for the time being, I got too much going on in RL atm, including a friend of mine who needs my set of hands to help him move house.

But the stuff that's being posted the last couple months is soo much good reading! Hereby an official THANK YOU from yours truly to all the people who've spend hours doing all the testing and posting about it on the forum! 😉

Thanks!! 😁

Anonymous Coward wrote:

I really think RG100 has been playing with fire, yet somehow has avoided having anything explode 😳

Hehe, just another day at chez RG100's. 😉 Actually, I've noticed that the Cirrus Logic video card's signal now "wobbles" a bit. If you look at the monitor, you can see the pixels "shimmering" from side to side. However, the video cable is near a PSU, so it could just be that.

What exactly can go wrong, if I continue with my mad 25 MHz bus experiments? Can I damage or degrade the video card? I'm not bothered about any aspect of the mobo, because I purposefully bought a handful of them, for insane experiments. The sound card seems OK, although I only tested a CT1600 for about 5 minutes. I only have the EIDE IO controller, made in 1996! Nothing SCSI based, and no network cards either.

Anonymous Coward wrote:

50MHz overclocks were pretty common for 386DX-40 CPUs in the early 90s.

Did the oc'ers use faster oscis, or... Here's something I don't understand: if I alter the "bus speed" jumper settings, why doesn't that affect the clock speed of the CPU? (I guess there's something fundamentally different about all of this, in comparison to increasing the FSB jumper setting on a more modern modo, eg Pentium 1 and upwards.)

Anonymous Coward wrote:

However, I would argue that running at 40MHz would give you better overall performance. I have discussed this in detail in my experiments with my 50MHz bus 486s.

Please can you post a link to this info, because I would be interested in reading it. Thanks.

retro games 100 wrote:

Here's something I don't understand: if I alter the "bus speed" jumper settings, why doesn't that affect the clock speed of the CPU? (I guess there's something fundamentally different about all of this, in comparison to increasing the FSB jumper setting on a more modern modo, eg Pentium 1 and upwards.)

We are talking about cpus older that those who implemented clock multiplication. The bus speed I suppose refers to at bus, or isa bus, which is the connection of the expansion cards with the system. And it was designed to run at 8MHz I think. Overclocking this may make the cards faster, if they can tolerate such a speed, but has no effect on the cpu, which runs at it's "native" speed.

Anonymous Coward wrote:

However, I would argue that running at 40MHz would give you better overall performance. I have discussed this in detail in my experiments with my 50MHz bus 486s.

I think the 486s are a completely different thing. First of all, the fact that there was a 50MHz cpu, and there were boards designed to work with that, is a lot better than having to overclock cpus to work on a faster speed than they were designed to.
A week ago, I tried to run a 486 board in various speeds, just to compare it to the 386. I used the same CL isa card and a 486/33 running at 25MHz, but it was a total failure. While in both systems I had set the ISA bus a lot higher than 8MHz, on the 486 the card's performance was less than the half compared to the 386 board. I really don't know why. But when I replaced it with a VL bus CL card, it was two times faster than the fastest isa score I had achieved on the 386.... And then I tried some different cpus, and I was really impressed by the performance of the 50MHz 486. The video cards had the highest score using that. I thought that because of the 50MHz bus, it would be near the dx2/66 cpu, but 3dbench had a different opinion, and showed a significant speed improvement with the dx2/66.

Intel 386 was built on 1.5um wafer, while am386 uses a 0.8um wafer. It really wasn't difficult to push the am386 CPU to 50MHz. ISA expansion cards shouldn't be an issue either, because normally you can set the divider to keep the bus in spec.

The real question is whether or not the motherboard can handle it. Hell, even many 486 boards couldn't do it. However, there are probably a number of chipsets out there designed for use in both 386 and 486 systems designed for 50MHz. Provided you have fast enough SRAM and DRAM and do a good job with your system timings, it should be no more of a challenge than a 486DX-50.

That being said, there were reasons why the 50MHz 486 systems never took off. They were often a bit flaky. For the high end EISA workstations this wasn't a problem, because they always used the best components. The real problem was with consumer grade systems that used VLB and cheaped out with 70ns DRAM and 20ns SRAM. VLB at 50MHz was a bitch, even with wait states.
-----
The TI486SXL CPU despite the name, is a 386DX replacement. I think it was named as to compete with the intel 486SX. There actually is a 386SX version as well. It is called TI486SXLC. In any case, these are fast and interesting chips, but the configuration can be a real bitch. They are quite particular about FPUs, and I know that a number of people have had issues getting the clock doubling going while an FPU is installed. My personal recommendation is to use the control software for the Cyrix DRx2 with these chips. All but one of the CPU registers are the same (the one to enable clock doubling).

Yes, that's exactly what I mean. The cpu may handle the extra speed, but the "regular" 386 board I don't know. But that's an interesting idea to use a board designed for 486 that could use a 386 also.
And sth else, a little irrelevant, but I reminded of that now. I had set a 486 board to be used with a 50MHz cpu. Then I thought to try the other 486s I have to see if they would start. The 33 and the dx2/66 that both have 33MHz bus, wouldn't start at all. The only one that posted and worked ok was the AMD486dx/40!

Looking at our 3dbench database, user aitotat has a 386 board with VLB using an Opti chipset and 256KB Cache. It has the highest score.

Amigaz and Keropi have scores for that Cyrix chip (SX-20 board, 40 MHz CPU clock) and it performs like a 386DX-33. IMO that Cyrix chip would be the one to use for overclocking, as you could overclock mostly the CPU and not run the board and other components out of spec.

Apart from that just get a 486 as it's more flexible and can scale a lot higher. A 386 should be left at stock just like DonutKing's machine! Poor vintage hardware 🤣

Maybe a 20MHz to 40MHz Cyrix SRx2 clock doubler upgrade will give you 386DX-33 power, but the system overall will be much much slower. Memory throughput on most 386SX system is the pits.