Thanks but I already ordered some form utsource.net. They're only 15ns but that's still good for 66MHz bus speed - and I'm not planning to overclock so that's fine by me. They're a little more expensive at $2.50 each but shipping was only $8 for basic registered post.

The part number is IS61C102415N and you'll need 9 of them, including the TAG RAM which is near the large SIS chip.

For that hot little number, you'd need 8 pieces of IS61C1024 (128K x 8) and 1 piece of W24512AK (64K x 8) for the TAG RAM, however it seems that another piece of IS61C1024 will also work as the TAG RAM. This will give you 1024 KB of L2 cache.

It can be -10, -12, or -15 (cache speed, lower the number, the faster. Units are nanoseconds). They are all the same price (for example, IS61C1024-10N).

DonutKing wrote:

They're only 15ns but that's still good for 66MHz bus speed.

You're running your front-side bus at 66 MHz? I thought I was the only one doing this extreme overclock. If you are going to attempt using cache wait states of 2-1-1-1, 10 ns cache is probably your best bet. Otherwise, 3-2-2-2 will work with 15 ns cache with a 66 MHz front-side bus. Are you sure you aren't using a 33 MHz front-side bus and a 2X CPU multiplier?

I'm running at 33MHz bus speed. I was referring to the rated speed of the chips 😉

Nice one, I've ordered me some too 😀

It is nice to see that (finally) more retro enthusiasts are experimenting with 1024 KB of cache. Once you get your pieces installed and properly jumpered, can you verify using CTCM7 that all your RAM is being cached? For example, if using 1024 KB of cache in write-back mode, that 128 MB of RAM is really being cached?

Booting into Windows 98 and running a few benchmarks, preferably Ziff-Davis CPUMark99 v1.0, could also be used as an indicator for all the RAM being cached. From the Ultimate 486 Benchmark Comparison (U4BC), Appendix 2, Colums A-V, you can match up which CPU you are using, and if all your RAM is cached, your CPUMark99 scores (row 57) should be ~50% greater than those found in Columns A-V. Note that you need to be using the fastest possible cache and memory timings for this to be a fairer comparison.

Unfortunately, I didn't have all the RAM cached when I did the U4BC, but test 57 is really the only test that was significantly affected by this. A typical example of a partially cached score was 3.82 for a Cyrix 5x86-133. This want up to 5.55 for a fully cached score. Note that colums A*, B*, and C* are fully-cached.

Thanks.

I've also discovered that the 402 BIOS for the 486SV2G works on my revision 1.2 board (supposedly its only for rev 2 and later)
Among other things this adds support for the Am5x86 (currently running a DX4-100 in mine, will try out an upgrade when I get some free time)

It also seems to speed up the boot, mine had about 10 seconds delay after memory test before it would go to the next screen, now it goes through straight away.

You can get the file here
http://www.driverstock.com/stor/files/6/pub/A … x4/sv2g0402.zip

My board had a UV-erasable EPROM so you'll need the proper eraser and programmer to upgrade.

DonutKing wrote:

My board had a UV-erasable EPROM so you'll need the proper eraser and programmer to upgrade.

Did you try 5V and 12V flashable BIOS chips (i.e. from SST, Winbond, or Intel)?

Whats the real world benefit of having all your ram cached? I recently got a VLB 486 board that can handle 1024kb of cache, the 256kb of cache installed is only 20ns and im running a POD83, I think the board can only take 64meg of ram (2 72 pin slots).

dirkmirk wrote:

Whats the real world benefit of having all your ram cached? I recently got a VLB 486 board that can handle 1024kb of cache.

I find that Windows feels about 2X as fast when my available RAM is all cacheable (on a 486). For this reason, I typically run the system with only 64 MB of RAM instead of 128 or 256 MB (w/512 KB of write-back cache). I've attached a Speedsys screenshot of my everyday-use computer. It indicates that cache access is an average of 50% faster than RAM access.

dirkmirk wrote:

I think the board can only take 64meg of ram (2 72 pin slots).

You might want to test 2 sticks of 64 MB FPM RAM (128 MB total) to see if it works. If it does, then I'd recommend sourcing 1024 KB of L2 cache. If the chipset only works with 64 MB, you can find 512 KB of double-banked cache pretty easily. The only way to ensure your board doesn't work with 128 MB of RAM is to test it; I never trust 486 motherboard manuals to identify all possible RAM configurations supported by the chipsets.

feipoa wrote:

DonutKing wrote:

My board had a UV-erasable EPROM so you'll need the proper eraser and programmer to upgrade.

Did you try 5V and 12V flashable BIOS chips (i.e. from SST, Winbond, or Intel)?

No because I don't have any 😀 I do have a programmer and a UV eraser though 😀

A make/model/part number might be helpful for others sourcing a UV eraser and programmer.

Just cheap Chinese jobs off ebay. UV Eraser has no model number but its about $15 delivered. Programmer is G540 USB

I find hot swapping 5V and 12V flashable chips to work well for my whole lot of 486 boards. I'm curious if any such boards require a UV-erasable piece for any reason?

I figure the UV-erasable chips might generally last longer than the flashable pieces, that is, if the flashable pieces use some kind of early floating gate technology.

I find hot swapping 5V and 12V flashable chips to work well for my whole lot of 486 boards. I'm curious if any such boards require a UV-erasable piece for any reason?

As far as I'm aware they shouldn't - at least most of the UV erasable EPROMs I've found on motherboards are of the 27 series (eg 27C512) and I've replaced these in the past with 28F and 29F EEPROM parts without problems.

Nice build mate!! like it..

feipoa wrote:

I find that Windows feels about 2X as fast when my available RAM is all cacheable (on a 486). For this reason, I typically run […]
Show full quote

dirkmirk wrote:

Whats the real world benefit of having all your ram cached? I recently got a VLB 486 board that can handle 1024kb of cache.

I find that Windows feels about 2X as fast when my available RAM is all cacheable (on a 486). For this reason, I typically run the system with only 64 MB of RAM instead of 128 or 256 MB (w/512 KB of write-back cache). I've attached a Speedsys screenshot of my everyday-use computer. It indicates that cache access is an average of 50% faster than RAM access.

dirkmirk wrote:

I think the board can only take 64meg of ram (2 72 pin slots).

You might want to test 2 sticks of 64 MB FPM RAM (128 MB total) to see if it works. If it does, then I'd recommend sourcing 1024 KB of L2 cache. If the chipset only works with 64 MB, you can find 512 KB of double-banked cache pretty easily. The only way to ensure your board doesn't work with 128 MB of RAM is to test it; I never trust 486 motherboard manuals to identify all possible RAM configurations supported by the chipsets.

Thanks for that info!

I took a chance on a cheap socket 3 mainboard from Italy, I could'nt understand what was written in the description but I got the sense that it was working.

http://www.ebay.com.au/itm/250970980604?ssPag … 49#ht_500wt_949

The board is a "Chicony 498B", the mainboard did'nt have any cpu settings silkscreened but I managed to find a manual for it and it supported the P24T, initially I was dissapointed as the bios did'nt showed the P24T as a "DX2-S 80mhz" or something like that and previous 2 attempts with a soyo sy25/p & EXP4044 showed the same damn thing and locked up constantly, however it appears to be working fine, (Speedsys 61.4, 3dbench 72, pcpbench 7.7, *Cirrus logic 5429 & 32mb ram". Speedsys reports the cpu as a P2T at 83mhz, The board has a UMC chipset and I think its a keeper, I just rechecked the manual and it does support 128meg, I might've assumed it supported 128meg with the 4 30 pin slots and 64meg with the 2 72 pin slots, sounds like 1meg of cache will make it run sweet!

Well, the price was certainly reasonable.

Did you try two sticks of 64 MB FPM SIMMs in the 72-pin slots?

OK the cache I ordered for this ASUS VL/I-486SV2GX4 has arrived so I'm ready to start playing around.

One thing that bothers me (it's superficial and it shouldn't) is that during the POST, the list of details on the screen include "IPEX blah blah". IPEX is the name of the company who put this PC together, and obviously they've updated the BIOS to include their name - does anyone know how to get rid of this?

I upgraded the BIOS on mine and it got rid of that 😀

If you mailed me your BIOS chip I could upgrade yours as well.