feipoa, the wiki on x86 memory segmentation might confuse you some more 😁 https://en.wikipedia.org/wiki/X86_memory_segmentation

and yes, you'll need to shadow a 16k range if the rom is 10k.
Software like qemm was able to recuperate the 6k that was unused by the rom, though used by the shadow range for umbs. I think.

I think I've had enough of memory segmentation! Time to move on...

OK, on to the PROM on the 3c515-tx. The manual states that it can take up to 32K PROMs. I assume that is 32 Kilobytes and not kilobits. What I am puzzled by is that the boot ROM DIP-32. If the max is 32 KByte, why would they not use a DIP-28 socket? It got me wondering if the pinout is non-standard. The manual says the part number for the PROM is 3c515TRIROM, but there is very little reference to this online with the exception of some IC part quote places.

I have two of these 3c515-tx cards. I installed a Winbond W27E257-12 EEPRM, which is DIP-28, into the DIP-32 socket on the 3c515-tx with left-most 2 pairs of pins not being used on the socket. Here is a visual of a 3c515-tx: http://ironalbum.pvk13.org/3c515tx-comp_side.jpeg

I removed the Lo-Tech card, installed the 3c515-tx with the W27E257 installed, and booted the computer. I ran 3c515cfg which sees the network card, but the card's connection LED's don't light up. The config utility also cannot detect the network. Is my 3c515tx broken and did the installation of the W27E257 break it, or did the card come broken in the mail? Using my other 3c515-tx card (without the EEPROM installed), the LAN lights turn on and the cfg util can find my network.

EDIT: OK, the boot PROM in the defective 3c515-tx card works! I can set C8000h from within the DOS cfg utility. I then put a "burned" eeprom into my working 3c515-tx and it also works. It was kind of tricky to get the Base Address options to show up in the DOS cfg utility though. The 3Com DOS CFG utility has these options:

C0000h
C4000h
C8000h <<<---- I have selected this one
CC000h
D0000h
D4000h
D8000h
DC000h

[I also noticed how the 3com address convention agrees with the motherboard's convention and not the XTIDE convention]

So I guess that other 3c515-tx card came DOA. It is getting more difficult to get money back from eBay these days unless the hardware has been tested within 30 days. But since the config util detects the card and there just seems to be a send/receive issue over the port, any idea which component may be dead?

feipoa wrote on 2021-08-07, 13:41:

So the table on the Lo-tech website is also wrong? https://www.lo-tech.co.uk/wiki/Lo-tech_8-bit_ … Address_Mapping Or is there another explanation for this? The XTIDE configuration and EEPROM writing software for DOS also indicate these same truncated addresses.

So you are saying that when I select D000 as the address on the ROM card that I should enable ROM shadowing in the MB's BIOS for: D0000-D3FFF and D4000-D7FFF ? I have done so now and it appears to have solved my speed problem! The CF cards now benchmark where they should. I'm left baffled by the truncated address listings.

Hi, address of option ROMs has always, as far as I know, been referenced by segment address rather than linear address, hence C800h for example. Linear address space didn't appear in Intel's instruction set until the 80386 with its protected mode operation.

In terms of ROM shadowing, this is a very important aspect for machines supporting this in relation to achieving good throughput. I've added to Readme to the XTIDE Universal BIOS Binaries Download Centre which includes a link to a page with some information about this and configuring the BIOS as well.

feipoa wrote on 2021-08-08, 14:13:

So I guess that other 3c515-tx card came DOA. It is getting more difficult to get money back from eBay these days unless the hardware has been tested within 30 days. But since the config util detects the card and there just seems to be a send/receive issue over the port, any idea which component may be dead?

Wow, Guess I missed this.
There's only a few things, the connector, the rj45 transformer, and then the chip.
as an example you can check https://github.com/Manawyrm/ISA8019/blob/mast … evA/ISA8019.pdf

weedeewee wrote on 2022-07-16, 11:07:

Wow, Guess I missed this. There's only a few things, the connector, the rj45 transformer, and then the chip. as an example you […]
Show full quote

feipoa wrote on 2021-08-08, 14:13:

So I guess that other 3c515-tx card came DOA. It is getting more difficult to get money back from eBay these days unless the hardware has been tested within 30 days. But since the config util detects the card and there just seems to be a send/receive issue over the port, any idea which component may be dead?

Wow, Guess I missed this.
There's only a few things, the connector, the rj45 transformer, and then the chip.
as an example you can check https://github.com/Manawyrm/ISA8019/blob/mast … evA/ISA8019.pdf

I was eventually able to get the seller to send another one of those 3c515-tx cards. He didn't want the dead card, so it is sitting in my box of dead parts with little motivation to fix it.

Feipoa,

I can't seem to make this work. Did you put the DIP-28 in backwards as well as shifted right? Asked another way, what direction is the notch on the chip facing? Towards the slot brackets or towards notch on the socket?

EDIT: Tried it, definitely not backwards!! Cooked EEPROM pretty good. No matter what I do I can't get the 3c515CFG util to show me the addresses for the PROM. If I set 8k, and accept, it says that it's programing the card, but then if I close out of the config and reload the program it tells me it's disabled again.

I assume you are referring to using a 3c515-tx. There should be a notch on the EEPROM's socket, even if ever so slight. Align the EEPROM's notch with the notch on the socket. If you fried your EEPROM, get a new EEPROM chip.

To view the EEPROM address using 3c515cfg, you need to have an EEPROM installed and not have an address conflict with another card using C8000 (or whatever address you want to use). If the address isn't getting set, you may have a conflict. I recall needing to disable shadowing in the computer's BIOS for the address you want to set on the 3c515. Once you set the address in 3c515cfg, and confirm it is set by reloading the config, then you can turn on shadowing again.

In some cases, I needed to place the 3c515tx into another computer, ideally a socket 7, to set the EEPROM's address. Try this first.

feipoa wrote on 2024-12-16, 02:35:

I assume you are referring to using a 3c515-tx. There should be a notch on the EEPROM's socket, even if ever so slight. Align the EEPROM's notch with the notch on the socket. If you fried your EEPROM, get a new EEPROM chip.

To view the EEPROM address using 3c515cfg, you need to have an EEPROM installed and not have an address conflict with another card using C8000 (or whatever address you want to use). If the address isn't getting set, you may have a conflict. I recall needing to disable shadowing in the computer's BIOS for the address you want to set on the 3c515. Once you set the address in 3c515cfg, and confirm it is set by reloading the config, then you can turn on shadowing again.

In some cases, I needed to place the 3c515tx into another computer, ideally a socket 7, to set the EEPROM's address. Try this first.

Ah, so it's probably conflicting with the ATI videocard in the PC. It has a 40k BIOS Rom and rolls over into C8000. I'll try it in my BX440 and see if I can get it programmed.

Just for clairty:

o o x x x x x x x x x x x x x x
>
o o x x x x x x x x x x x x x x

The EEPROM should go like that, where > is the notch on the socket, o are open pins and x are occupied pins?

aspiringnobody wrote on 2024-12-16, 03:26:

Just for clairty: […]
Show full quote

Just for clairty:

o o x x x x x x x x x x x x x x
>
o o x x x x x x x x x x x x x x

The EEPROM should go like that, where > is the notch on the socket, o are open pins and x are occupied pins?

Correct.

feipoa wrote on 2024-12-16, 03:32:

aspiringnobody wrote on 2024-12-16, 03:26:

Just for clairty: […]
Show full quote

Just for clairty:

o o x x x x x x x x x x x x x x
>
o o x x x x x x x x x x x x x x

The EEPROM should go like that, where > is the notch on the socket, o are open pins and x are occupied pins?

Correct.

I was able to program the 3C515-TX in another computer... it allowed me to set it to CC00, and gave me a warning about no rom being present and/or a conflict with another rom in the system, but it saved the configuration to the flash on the card. However, when I put it in the 486PC -- XTIDE loaded -- but at C800 (overwriting the last little bit of the ATI Video BIOS).

I'm not really sure what to make of that development. It boots and runs windows fine, and I'm certainly happy to take the extra 8K of upper memory, but I feel like this isn't a good idea long term.

For reference, all that is contained in the last 8K of the ATI BIOS Rom is:
TI??1ATIAH1c....2ATI -- and after those 20 Bytes it is all FF (except for one stray B4 about halfway through. Maybe some kind of checksum bit or some such, but most likely bit rot?)

I guess when I have some time I'll try pulling out the card and changing it in the other PC to D000? Why would it be shifted 4K?

EDIT: setting it to CE00 has the same effect -- XTIDE to C800. Maybe it can only be aligned at 32k boundaries? Why give the option for other settings if they don't work? The card says (in the other PC) that it is set to CE00.

I might need to put together a PC with an ISA video card so that there aren't any conflicts in the C800+ area.

I guess it makes sense, but the PnP BIOS was changing the ROM address on its own. Why it was choosing to overwrite the ATI BIOS is unknown. Changing PnP OS to YES seems to have fixed it.

It has been a long time since I've looked at this XT-IDE business.

Sounds like you figured your issue out though, PnP BIOS was doing its own thing, and your VGA card's BIOS was exceeding a more conventional 32K boundary? What final XT-IDE BIOS address did you end up with?

I don't recall all the ROM shadow boundaries. It seems like some BIOSes are setup for 32K, others have that split in half. On my only system that uses XT-IDE, I don't care much about saving a few kilobytes in upper memory.

feipoa wrote on 2024-12-17, 01:33:

It has been a long time since I've looked at this XT-IDE business.

Sounds like you figured your issue out though, PnP BIOS was doing its own thing, and your VGA card's BIOS was exceeding a more conventional 32K boundary? What final XT-IDE BIOS address did you end up with?

I don't recall all the ROM shadow boundaries. It seems like some BIOSes are setup for 32K, others have that split in half. On my only system that uses XT-IDE, I don't care much about saving a few kilobytes in upper memory.

I ended up setting it to DE00; the System Rom starts a E000, so that seemed as good a place as any. I found out if you do 3C515CFG /NOPNP it lets you set the address more reliably. If I let the motherboard assign resources it still moves it to C800. I'm not really sure what to make of it. Maybe the people who programmed the BIOS PNP implementation knew ATI was crappy and intentionally overwrote those bytes?

For documentation purposes, what graphics card and motherboard are you using?

feipoa wrote on 2024-12-17, 05:05:

For documentation purposes, what graphics card and motherboard are you using?

GPU: ATI Rage II+ DVD
MOBO: Gateway MBDPCI003ABWW

That's a board I haven't run into before. Are there any documented or undocumented FSB settings for 40 or 50 MHz?

feipoa wrote on 2024-12-17, 09:33:

That's a board I haven't run into before. Are there any documented or undocumented FSB settings for 40 or 50 MHz?

Not that I know of, although I haven’t tried. My 5x86 isn’t stable at 160 and my POD 83 isn’t stable at 100. Lost the silicon lottery pretty good.

I can have a peak later, but AFAIK it only has two jumpers for the FSB so I’m not hopeful. There isn’t much on the retroweb for this board, not even the usual jumper guide.

Two jumpers with two positions per jumper, that's 4 possible combinations. A main disadvantage of these Intel chipset socket 3 boards is 33 Mhz max fsb.

feipoa wrote on 2024-12-17, 13:03:

Two jumpers with two positions per jumper, that's 4 possible combinations. A main disadvantage of these Intel chipset socket 3 boards is 33 Mhz max fsb.

It's actually 2 3-pin jumpers. So there are five permutations:

• none, none (33MHz)
• 1-2, 1-2 (default 25MHz)
• 1-2, 2-3 (no post)
• 2-3, 1-2 (default 33MHz)
• 2-3, 2-3 (33MHz, buggy)

I checked the no post position with the heatsink off on my POD 83 to reduce the multiplier, just in case it was 40MHz and not posting because the CPU couldn't handle it. Still didn't post. Possible that it is really 40MHz and it's not posting because the dividers are wrong, but I didn't get anything at all on the post code card, so I'm thinking it just wasn't generating a clock at all. I'd have to get the oscilloscope out to know more. I agree it would be nicer if it ran at 40MHz, but I don't own any CPUs that can handle it, despite owning several that should be able to handle it.

Thanks for checking. The Wiki article on Intel chipsets claims the Intel Aries chipset can handle up to 50 MHz, but there aren't any motherboards with a 40 or 50 MHz option.

These CPUs certainly can handle 40 MHz:

AMD Am5x86 at 3x40 and usually at 4x40
Cyrix Cx5x86 at 3x40

but it sounds like the motherboard doesn't output at 40/50 MHz. What clock generator is on that motherboard?