I have a BUSCOMPUTER made in New York City @1992. It is a Hsing Tech Product bios# ami 30-0100-001437-060692-pc chip f. The bios has an advanced chipset setup with some terms I am not familiar with. "Cacheable Ram Address Range", It can go to 64mb . There is also "non-cache Block-1 size" that can go to 64,128,256,512. "non-cach block-1 base" up to 65024. The system has an amd386 40 with a mathco math processor .There is 128k of socketed cache ram on board. I checked the hardware99 list. Does any one know about these machines? People look at me strange when I tell them that I got a 386 on dsl using Arachne and it works fine! I would like to know because I dont have any 8mBx8 30 pin simms laying around to check it out.
- VCF South West - June 14 - 16, Davidson-Gundy Alumni Center at University of Texas at Dallas
- VCF West - Aug 2 - 3, Computer History Museum, Mountain View, CA
- VCF Midwest - Sept 7 - 8 2024, Schaumburg, IL
- VCF SoCal - Mid February 2025, Location TBD, Southern CA
- VCF East - April 2025, Infoage Museum, Wall NJ
-
Please review our updated Terms and Rules here
64meg 386?
- Thread starter Iconclass
- Start date
Vlad
Moderator
I have some goofy 486 around here somewhere that has an absurd amount of RAM slots I'm thinking can go pretty high like that. I think its pushing like 10+ 30-pin SIMM slots or something. Last I knew, it was almost fully loaded too, I'll have to finish it and see what happens.
NathanAllan
Veteran Member
I had a machine with a bunch of 30-pin simm slots, too, it was a monster. I couldn't take it in the move, unfortunately. iirc, it had something like 8 ISA slots, too. Wish I could help on specs, but can't, sorry. Sure was a neat machine.
Nathan
Nathan
vwestlife
Veteran Member
There were some 486 motherboards built with eight 30-pin and two 72-pin SIMM slots. 4 MB was the largest common capacity for 30-pin SIMMs, so that would be 32 MB plus whatever you could get from two 72-pin SIMMs.
When you have a large amount of RAM installed in an older PC, be sure to run CACHECHK.EXE to see if all of it is being cached. I have a late 486/5x86 board (PC Chips M919) and it doesn't cache anything above 64 MB. On the CACHECHK printout you can see that the memory access timing suddenly gets a lot slower above that point.
On the Apple side, I have a PowerMac 8100, that's a 110 MHz PowerPC 601 with 8 MB RAM built into the motherboard and eight 72-pin SIMM slots. Fully loaded, it can accomodate 264 MB of RAM! Not bad for a computer that was introduced in 1994.
The classic SE/30 also had a huge RAM capacity for its era (1989): eight 30-pin SIMM slots that can take the rare 16 MB 30-pin SIMMs, for up to 128 MB of RAM.
When you have a large amount of RAM installed in an older PC, be sure to run CACHECHK.EXE to see if all of it is being cached. I have a late 486/5x86 board (PC Chips M919) and it doesn't cache anything above 64 MB. On the CACHECHK printout you can see that the memory access timing suddenly gets a lot slower above that point.
On the Apple side, I have a PowerMac 8100, that's a 110 MHz PowerPC 601 with 8 MB RAM built into the motherboard and eight 72-pin SIMM slots. Fully loaded, it can accomodate 264 MB of RAM! Not bad for a computer that was introduced in 1994.
The classic SE/30 also had a huge RAM capacity for its era (1989): eight 30-pin SIMM slots that can take the rare 16 MB 30-pin SIMMs, for up to 128 MB of RAM.Druid6900
Veteran Member
Well, since his board is a PCCHIPS as well, it may well have the same problem.
Sharkonwheels
Veteran Member
There were some 486 motherboards built with eight 30-pin and two 72-pin SIMM slots. 4 MB was the largest common capacity for 30-pin SIMMs, so that would be 32 MB plus whatever you could get from two 72-pin SIMMs.
When you have a large amount of RAM installed in an older PC, be sure to run CACHECHK.EXE to see if all of it is being cached. I have a late 486/5x86 board (PC Chips M919) and it doesn't cache anything above 64 MB. On the CACHECHK printout you can see that the memory access timing suddenly gets a lot slower above that point.
On the Apple side, I have a PowerMac 8100, that's a 110 MHz PowerPC 601 with 8 MB RAM built into the motherboard and eight 72-pin SIMM slots. Fully loaded, it can accomodate 264 MB of RAM! Not bad for a computer that was introduced in 1994.
16mb 30-pin simms aren't THAT rare - You can usually find a set of 4 on fleaBay for around $35-40. I have a couple sets I keep just in case...
Apple did know how to make machines back then, didn't they?
I have a Quadra 840AV - I wish IT had 8 simm slots! Only 4 on the 'AV
For some reason, I've gotten into a vintage Mac mood lately - dragged the old hotrodded Q605 out of the garage (68MB RAM, 33MHz hack, CPU Fan, RC040, ethernet card)
because I hadn't seen the old friend in a while...also dug out my 6360 with the 6500/250 logic board...sweet small setup. Wouldn't mind finding a Sonnet L2 G3 upgrade for it...
Would like to find another Q840AV (or a case), a Q660AV, a Q650 (for A/UX), and a couple 'books here and there - 3400c, 5300ce for the logic board, Duo 2300c and 280c...
T
squirrel-steam
Experienced Member
A intel and amd 386 could address 4 gigs of ram, and manage 64TB of virtual memory, but most boards could only hold 16mb. Also, you must have at least 4, 8, or 12 (or any multiple of four,) sticks of 30 pin ram in a 30 pin system.
I ougth to have stated it has two banks of 30 pin simms for a total of eight. It currently has eight "4 meg simms" in it for a total of 32meg. The 8mBx8 chip is used on the ultra rare "8 meg" 30 pin simm. My statement was ,is there a good chance that the board can handle the 8 meg simms if the bios allows cacheing of up to 64 meg.I used the 8mBx8 description to test you guys knowledge. A knowledgable person would have figured out the board had room for eight 30 pinn simms. Does anyone know or has had expierence with such a system? I have seen on the net and researced the issue to know that 8meg 30 pinn simms at 60 ns do exist. Thankyou.
Last edited:
vwestlife
Veteran Member
On a processor with a 32-bit external bus, that is true (386DX, 486DLC, 486SX, 486DX, 5x86, etc.). But on a processor with a 16-bit external bus (386SX, 386SLC, 486SLC, etc.), you can use pairs of 30-pin SIMMs, instead of sets of four -- although on these processors, the maximum addressible RAM is only 16 MB (a limitation going back to the 286).
Speaking of which, I always found it interesting that the IBM 486SLC clearly had a 16-bit external bus due to its 386SX foundation, but most of the motherboards upon which it was placed had "32-bit" VESA Local Bus slots. The processor did the 32-bit I/O as two chunks of 16-bit data in a row, thereby negating any speed advantage that the VLB slots were designed to provide.
Anonymous Coward
Veteran Member
What's interesting is that I recently found an AMI 386SX motherboard with cache that requires the SIMMs to be installed in groups of 4 rather than in pairs. It uses some weird bank interleaving technique from what I've learned. Who knows, maybe there are even 386DX motherboards that allow you to install SIMMs in pairs by cutting the datapath back to 16 bits. I have Pentium boards that do this.
Getting back on topic....
Yes, I noticed that you clearly specified 8mbx8 density 30-pin SIMMs. But quite honestly I figured it was a typo. I do it myself from time to time. Not only have I never in my life seen an 8mb 30-pin SIMM, I have never seen a motherboard manual that listed support for it. I have seen 16meg 30-pin SIMM support from time to time, though it's not common in mainstream 486 boards made before 1992. I am not saying 8mb SIMMs do not exist, but you might be very hard pressed finding a board that supports them. I would say your best chances would be to find a board that supports 16mb 30-pin SIMMs hoping there is also support for the 8mb variety.
In general the size of the L2 cache is indicative of how much memory the board is capable of caching. The "rule of thumb" is 64kb for every 16megs. So all boards with 256kb cache should be able to cache 64mb. However I find with the last generation of 486 boards cache size is not always indicative of caching, and you can find boards with 512kb cache that only cache 32mb of RAM (The ALi PCI chipset comes to mind). Perhaps it was just some crazy marketing scheme. More cache=more better. What it really boils down to is what is listed as "supported" in the motherboard manual. If it's not in the manual, chances are it's not going to work without modification.
Getting back on topic....
Yes, I noticed that you clearly specified 8mbx8 density 30-pin SIMMs. But quite honestly I figured it was a typo. I do it myself from time to time. Not only have I never in my life seen an 8mb 30-pin SIMM, I have never seen a motherboard manual that listed support for it. I have seen 16meg 30-pin SIMM support from time to time, though it's not common in mainstream 486 boards made before 1992. I am not saying 8mb SIMMs do not exist, but you might be very hard pressed finding a board that supports them. I would say your best chances would be to find a board that supports 16mb 30-pin SIMMs hoping there is also support for the 8mb variety.
In general the size of the L2 cache is indicative of how much memory the board is capable of caching. The "rule of thumb" is 64kb for every 16megs. So all boards with 256kb cache should be able to cache 64mb. However I find with the last generation of 486 boards cache size is not always indicative of caching, and you can find boards with 512kb cache that only cache 32mb of RAM (The ALi PCI chipset comes to mind). Perhaps it was just some crazy marketing scheme. More cache=more better. What it really boils down to is what is listed as "supported" in the motherboard manual. If it's not in the manual, chances are it's not going to work without modification.
Last edited:
vwestlife
Veteran Member
One of the popular late 486/5x86 motherboards from M-Tech had 8 30-pin and 2 72-pin SIMM slots and could support up to 1 MB cache onboard. It came in both VLB and PCI versions. It was a good performer, but the huge cache was really just a gimmick. Benchbacks back in the day pretty much confirmed that 512K cache only had a small speed advantage over 256K, and 1 MB had virtually no advantage at all over 512K.
Amazingly M-Tech still has all the info on those boards up on their web site:
http://www.mtiusa.com/486.htm
The R07 has about as many slots and sockets you'll ever see on a motherboard!
Anonymous Coward
Veteran Member
If you only have 16megs of RAM in your system, 64kb L2 cache is all you can actually use. As far as I can tell, upgrading to 128kb, 256kb, 512kb or 1meg should have not have an impact on system performance. However, if you wish to upgrade to 32mb, only the first 16megs will be cached (therefore taking a big performance hit on memory access above 16mb) unless you add another 64kb of L2 cache to the system. Please correct me if I am wrong on this. I am pretty sure this was the case with all 486 motherboards, not just the m-techs.
The M-tech website is pretty ridiculous, and I really can't understand why they haven't changed it since 1997 or so. I think the company still exists, but they no longer produce motherboards.
The M-tech website is pretty ridiculous, and I really can't understand why they haven't changed it since 1997 or so. I think the company still exists, but they no longer produce motherboards.
Druid6900
Veteran Member
I have a couple of 386DX motherboards (the make and model escape me at the moment, but, when I'm in that workshop, I'll check on the board and edit this) that has 12 30-pin sockets and 36 DIP sockets for a total of 16MB.
You could run the RAM either way, using either 4 of the 30-pin sockets as bank 0 or the 36 DIP sockets as bank 0.
It is a 25/33MHz turbo board with 64K of cache.
You could run the RAM either way, using either 4 of the 30-pin sockets as bank 0 or the 36 DIP sockets as bank 0.
It is a 25/33MHz turbo board with 64K of cache.
The most 30 pin SIMM slots I have on 386/486 machines is 8, most have 4 30 pin and 2 72 pin, or just 72 pin on the late PCI 486 systems.
I think the way cache works must have changed over time, because you have p3's with 256K cache on the CPU that can cache a GB of RAM, so its a chipset function.
I think the way cache works must have changed over time, because you have p3's with 256K cache on the CPU that can cache a GB of RAM, so its a chipset function.
Anonymous Coward
Veteran Member
I think pipelined burst cache was likely the major cause of change in caching scheme. Having the L2 cache on die may change the situation as well.
UPDATE:
I've been doing reading about cache and tag RAMs. I'm still not quite certain exactly how all caches work, because it sounds like it might vary from board to board. Supposedly the actual DATA is stored in the cache RAM, and the lookup table is stored in the TAG RAM. It sounds like with at least some motherboards you only need to upgrade the TAG RAM in order to increase the cacheable area. Nevertheless, it seems that for 486 systems larger caches aren't beneficial unless you have a lot of main memory.
Today I ran a program called CTCM 1.7a on my AMI 386SX system which has 32kb L2 cache. It reported back the cacheable range was 64mb. This is surprising since SX systems only have a maximum of 16mb. Perhaps my board has some built in EMS memory capability I am not aware of...or maybe it's just a case of AMI over-engineering their products.
UPDATE:
I've been doing reading about cache and tag RAMs. I'm still not quite certain exactly how all caches work, because it sounds like it might vary from board to board. Supposedly the actual DATA is stored in the cache RAM, and the lookup table is stored in the TAG RAM. It sounds like with at least some motherboards you only need to upgrade the TAG RAM in order to increase the cacheable area. Nevertheless, it seems that for 486 systems larger caches aren't beneficial unless you have a lot of main memory.
Today I ran a program called CTCM 1.7a on my AMI 386SX system which has 32kb L2 cache. It reported back the cacheable range was 64mb. This is surprising since SX systems only have a maximum of 16mb. Perhaps my board has some built in EMS memory capability I am not aware of...or maybe it's just a case of AMI over-engineering their products.
Last edited: