feipoa wrote:

I was using an IBM 5x86C-100HF on my crappy HOT-433 rev 1-3 boards. Where did you find a rev 4 HOT-433? To confirm, the Rev. 4 has only single bank cache (4 cache sockets + 1 TAG), and this is what you have? If so, I'd say this is the board to be testing overclocked CPUs on. I've always had shoty luck with my rev 1-3 boards.

Could you let me know if the PS/2 mouse feature of the rev 4 actually works? it never worked on the rev 1-3 boards, as far as I could tell. Thanks.

That's the same CPU I've been using. I got the HOT-433 from a friendly Canadian. And yes, it's the Rev. 4 board with only 1 bank of cache (soldered onto the board, it's in my cache chips post). Also, to confirm, it says that it's a revision 4 HOT-433 on the corner of the board, which is visible in one of the pictures in said post.

As for the PS/2, I'm still hunting for the adapter to test it with.

Does it have 256 KB or 512 KB L2 cache? If it has only 256 KB cache, you could always desolder the cache and put 28 hole sockets in their place and upgrade it to 512 KB cache.

As for the PS/2 connector, you can use a DIN-6 female connector from digikey or anywhere. Just wire up the correct pins. From the HOT manual, the pin headers on the motherboard are:
Pin 1 = Data
Pin 2 = N/C
Pin 3 = GND
Pin 4 = Vcc (5V, I assume)
Pin 5 = Clock
Pin 6 = N/C

Then the pin's on the DIN-6 are oriented as per this wikipedia page,
http://en.wikipedia.org/wiki/PS/2_connector

Alternately, is there not an actual DIN-6 PS/2 mouse connector on the motherboard? It seems that some also come with the DIN-6 along with the headers.

It doesn't have the DIN-6 connector, I'll have to wire one together to a PS/2 bracket sometime when I'm feeling ambitious enough.
I had actually planned on using a serial - PS/2 adapter that I have laying around here somewhere to connect my mouse for now. I need to organize everything again though.

PS/2 to serial adapters only work with PS/2 mice that are specifically designed to send ps/2 as well as serial signals. Very few ps/2 mice support such signal switching. The ones that do will usually specify this on the bottom.

If your MB is lacking the actual PS/2 DIN, I wonder if the required PS/2 IC controller is even on the motherboard.

PS/2 to serial adapters only work with PS/2 mice that are specifically designed to send ps/2 as well as serial signals.

Good point there, I do have a mouse that does work with that though (I used it with my Soyo 4saw2 build from about a year ago). And I have a serial mouse that works great, if I need to use it.

If your MB is lacking the actual PS/2 DIN, I wonder if the required PS/2 IC controller is even on the motherboard.

Wish I knew which chip to look for (not that I'm going to spend a lot of energy messing with PS/2 mice right now when I'm stuck using an adapter so I don't need to hook up my AT keyboards).

Now for the bitter news: I pulled out my IBM 100HF CPU from the socket and discovered that while handling it, at some point one of the pins got bent completely around so it was pointed back at the CPU. I used some delicate tools and attempted to salvage it. At this point, I'm running some tests to see if it has been compromised at all.

Now for the slightly less bitter news: I accidently discovered the 60MHz and 66MHz bus settings. Sadly it won't quite finish POST with the 66MHz setting but I'm gathering some benchmarks at the 60x2 setting right now (as part of my stability tests).

EDIT: Picture proof.

Uploaded with ImageShack.us

Did you adjust your memory and cache timings for 60 MHz? What about your CPU voltage?

Did you adjust your memory and cache timings for 60 MHz?

Nope, I didn't bother to adjust anything yet, because it seemed to reflect little or nothing in previous experiences.

What about your CPU voltage?

I stuck to 4.0v.

It appears that I might have pushed things a bit too far now. After running it at 66x2 and 60x2 successfully (finally) and benchmarking it quite a bit, I rebooted and discovered that the hard drive became corrupted. I didn't lose much, just benchmark pics from Speed Sys. I'll probably wait until I can rest after this holiday rush before I start poking around to see what actually caused the corruption

P.S. The best benchmark that I could get at the 66x2 was just over 73, iirc. The cache and ram were much more efficient than before.

nemesis wrote:

It appears that I might have pushed things a bit too far now. After running it at 66x2 and 60x2 successfully (finally) and benchmarking it quite a bit, I rebooted and discovered that the hard drive became corrupted. I didn't lose much, just benchmark pics from Speed Sys. I'll probably wait until I can rest after this holiday rush before I start poking around to see what actually caused the corruption

Are you using the onboard PIO IDE? Perhaps try a controller on the PCI bus.

nemesis wrote:

P.S. The best benchmark that I could get at the 66x2 was just over 73, iirc. The cache and ram were much more efficient than before.

That is fantastic! You're up to my level now. I only got 76 because I turned branch prediction on. A score of 73.5 is normal for an optimised Cyrix 5x86-133. But what are your L1/L2/RAM times?

So it looks like we we may have another motherboard to add to the list of 2x66 Cyrix 5x86-133/3X supporters.

I can install Windows 98SE with no trouble as long as I don't OC beyond 120MHz as far as I can tell.

Are you using the onboard PIO IDE? Perhaps try a controller on the PCI bus.

Yes, I'm using onboard IDE controller, but due to your request I hunted high and low for a cheap decent PCI IDE controller and finally ordered a new one.
It might be fun to see if I can get the SATA to work on it properly too (card has IDE and SATA).[/i]

That is fantastic! You're up to my level now

WOW! Whats with you and "levels"!? I have to earn your respect just 'cause I'm not canadian born!? (Ok, I'm just kidding. I know what you ment. 🤣 )

But what are your L1/L2/RAM times?

The funny part is that I didn't see any benching for the L2, only for the L1 (was between 190 and 200MB/s iirc) and total memory throughput was ~55MB/s, iirc. I still havn't been able to test my 50ns sticks to see if they'll work on account that they still haven't arrived, so this is still my 16MB ram.

I can install Windows 98SE with no trouble as long as I don't OC beyond 120MHz as far as I can tell.

You may need to run the CPU at 3.85 V for it to be stable at 133 MHz. Be sure to use 3-2-3 for the L2 cache when using a 66 MHz FSB and 1ws/0ws (read/write) or slower for the DRAM wait states.

What PCI SATA/IDE controller are you using? As I noted in the World's Fastest 486 thread, item 10 (Other additions known to work well on newer UMC, PCI-based 486 systems), the Promise SATA/ATA host controller SATA150 TX2plus works quite well. It even has NT 4.0 drivers for SATA/ATA. SATA should work without any problem, though I'm not sure if a 2 TB drive will work or not. When I spoke with the company some time ago, they though 1 TB was the upper limit.

Your L1 and RAM speeds look appropriate, though I'm not sure what is up with the L2. Is your L2 cache 15 ns? Perhaps it is getting disabled? I always look at soldered cache with a questioning mind.

The PCI sata/ide controller that I'll be trying shortly is a dirt cheap rosewill card that I found on one of my favorite online computer shops. I won't go really crazy on drive size though,... maybe as much as 150GB if that's supported by the controller. I might try SCSI sometime again soon too, but I never got along with that standard 🙁 .

My L2 cache turned itself off at 66MHz, that's the only way it would boot, I'll mess with that later but as for now, I'm shopping for some at least 15ns 128kx8 chips at a very low price. The L2 cache reliability seems to be my biggest problem holding my speedsys score back anyway, so if I can fix that, then I'll be able to focus on the other timings.

EDIT: As a side note, I could have sworn that my hard drive performance was well over 8k at 66MHz, while it averages around 5k at 50MHz - 60MHz.
Any idea why that would impact it so much?

5000 KB/s to 8000 KB/s seems a little extreme based on the theoretical maximum bandwidth of a PCI bus.

If I recall, you are running your system with a 1/2 FSB-to-PCI divider in the BIOS, so

66 MHz * 1/2 * 32 bits * 1 byte / 8 bits = 132 Mbytes/sec
60 MHz ... = 120 Mbytes/sec
50 MHz ... = 100 Mbytes/sec

There is a 32% theoretical drop in going from your 66 MHz bus to a 50 MHz bus. 5K to 8K is a bit more than 32% though.

I recall your cache is soldered in. Are you going to desolder? There was a seller on eBay a few months ago who seemed to have an endless supply of ISSI 128Kx8 SRAM chips. I think they were $10-12 per chip, so you'd be looking at about $50 for a 4 pack with shipping.

Yes, I plan to desolder. I've been digging up my old low power solder kits so I can practice on my dead electronics.
It's a good thing I kept my broken boards too, that way I can use the sockets and maybe some of the chips if those are still good. I don't think I'll need to order a TAG chip since I have some that should be large enough (and fast enough) laying around here, but I can only find the 128k x 8 DIP chips on foreign markets, so that's going to take a while to ship in.

I double checked my BIOS and found that it in fact had reset to it's standard configuration so the PCI bus was actually at 1:1 during the benchmark @ 8k hdd. I'm surprised that my video card didn't burn 😖 .

A quick update on the PCI controller card: It requires Windows 2k to install, and because I don't have Win2K, I'll have to use it for one of my other builds or just save it until I get that version of Windows (which probably won't work on a 486 anyway).

I highly recommend the Promise SATA TX2 Plus card. Not only is it one of the few mainstream SATA/ATA cards to have NT4.0 drivers, it also has Windows 98 drivers.

2 SATA ports and 1 ATA port, which will work with ATA33 thru ATA133. If you later want to put it into a PCI-X system, this card also works with a 66 MHz bus. If my system wasn't all SCSI now, I'd probably switch to using this card. It would be fun to test this out a card like this an a 486 with a 66 MHz bus

I got mine a few years ago for $8 when these things were dirt cheap; there's currently one on e-Bay for quite a bit more at $30+. I don't beleive the SATA300 TX2 plus has NT4/W98 drivers.

You should be able to install Windows 2000 fine on a 486 as long as there aren't bus mastering issues. Some UMC-based motherboards I've tested have W2K bus mastering issues, while others don't.

It's a good thing I kept my broken boards too, that way I can use the sockets and maybe some of the chips if those are still good.

I recommend saving yourself the headache and find some new DIP sockets on Digikey for a few $. They still sell them new.

When desoldering the old cache, be very careful not to use much force when removing the old cache pieces as you can easiily rip off the PCB traces like this. Definately use some kind of solder sucker or vacuum pump. Be patient. If you are only using a manual hand pump as the sucker, plan on spending 1 hour per cache module.

I got my RAM today. Good news is that it works... bad news is that it only recognizes 8MB of the 64MB.

As far as digikey goes, I couldn't find any 32pin DIP sockets that would work. Maybe I was looking in the wrong places.

Did you try to use high density RAM? I find that the HOT-433 only likes low density RAM, while the Biostar does fine with high density or low density.

The DIP you are looking for,
2x16 = 32 pins (2 rows of 16)
2.54 mm pitch (space between each in-line pin)
7.62 mm row spacing

is from TE, and has part number 1-390261-0 or 2-390261-0. TE lists it as active on their website, the 2* part number.

Unfortunately, this has become a very unpopular item as you need to order 6,000 pieces. I checked digikey, mouser, newark, and future electronics. Nobody has stock anymore. This is the digikey link,
http://search.digikey.com/us/en/products/1-39 … 61-0-ND/1136406

Shoot an e-mail to Elcodis.com to see if they have them in smaller quantities. Them seem to be able to find very obscure items quickly.

I guess I figured that since I found my DIP socket for my RTC so easily, that the cache DIPs were just as easily available, but I was wrong.

If anybody is looking for the DIP socket for their RTC, it's part number 4824-6000-CP.
http://search.digikey.com/scripts/DkSearch/dk … Words=3M5467-ND

EDIT:
It just occured to me that you can line up two 2x8 DIPs. They are 18 cents each from digikey, or 7 cents each from Future Electronics. Note, the shipping fee is also cheaper at Future Electronics compared to Digikey.

From TE (1-390261-4):
http://search.digikey.com/us/en/products/1-39 … 0206-ND/1125624

or if you want the 3M ones (4816-3000-CP), 54 cents each.
http://search.digikey.com/us/en/products/4816 … 5463-ND/1133616

I hope this helps.

Some interesting events occured today. I have been having stability problems recently since I switched to the new RAM, so I switched back and decided to experiment with the cx5x86 registers again at 120MHz...
Well, it barely made it past POST before locking up. I was confused so I dissembled the computer again to see what I had done wrong this time. It turns out that I had accidently POSTed @ 60x3. Now the CPU won't overclock stabily at all and even dropped my 150MHz score down to 82.5.

I'm thinking the damage might be permanent, but I still plan on modding the motherboard when I get all the parts in hopes that the board itself is fine.

Did you try the CPU in a different motherboard?

The CPU is likely overheating at 4V if you have it inside a case with cover. I've recently discovered this fact of Case vs. noCase making a huge thermal difference with these Cyrix 5x86 CPUs. You might want to consider adding a case fan (photo).

I've angled my case fan (92mm) to point more towards the CPU/harddrive. Although the case doesn't have proper front vents, the fan seems to help. A good thermal test is to get the system stable with whatever test means you find suitable (i.e. playing Subspace for 1 hour in spectator mode, or running mp3's on repeat) without the case cover. Then, add the case cover and run the same test. If it fails this time, the problem is thermal.

If the high density RAM you tried is of the wrong voltage, that could permanently damage your motherboard's memory controller. I've done this once before.

If you are playing with registers again, I've noticed that some MB/CPU combinatios want RSTK_EN set off, and sometimes DTE set off. These features have little to no impact on performance.

Trying running the CPU at 133 MHz. Just remember to put the jumper on the 2X setting.

I now have two fully cased systems running at 2x66 MHz and am trying to get them thermally stable with the case on. The photo is of the uglier system.

Did you try the CPU in a different motherboard?

No, not yet. I don't have any suitable motherboards set up right now. 🙁

The CPU is likely overheating at 4V if you have it inside a case with cover.

The cover was always off when I'd overclock. Also, the CPU never felt very warm to the touch, and the nice looking blue heatsink fell off right after I got it, so I've been using a 12v fan on it the whole time.

If you are playing with registers again, I've noticed that some MB/CPU combinatios want RSTK_EN set off, and sometimes DTE set off. These features have little to no impact on performance.

Checked that already with no change.

Trying running the CPU at 133 MHz. Just remember to put the jumper on the 2X setting.

Will not POST anymore at all with any overclocking so far.

If the high density RAM you tried is of the wrong voltage, that could permanently damage your motherboard's memory controller. I've done this once before.

This very well may be the case... actually I'm almost certain it is. It turns out that the RAM is 3.3v and the Motherboard seems to only handle 5v sticks.

If that is in fact the case, I would need to change the controller. Is that possible? And if so, do you know which chip to look for?