Thread recycling!

Hello there. As I'm sure you already know, the M919 is a very infamous board due to so many fake versions made and it's funky setup. That said, I commend you for still taking this piece of history seriously.

I have an early version of the board and that has a slew of issues of it's own, but I've never had an issue running the AMD 5x86 properly so I'm going to have to leave that question in the hands of someone more experienced with that catagory.

The PCI and ISA stability is fine as far as the board is concerned (in my experience), it ususally comes down to what cards you use. for example my TX2 IDE/SATA controller works like a charm at high PCI bus speeds, but I've seen others that choked at 40+MHz. If you don't trust the card, use the PCI or ISA dividers (I think the M919 has settings for 1:1 and 1:1/2 PCI and similar for ISA). 150 MHz is an issue that lies in the CPU more than the Motherboard in my experience as well. I've tested that board (with a good CPU) at 33, 40, 50,and of course the undocumented 60 and 66 MHz bus settings.

Could you please post some pictures and list the rest of the specifics? I.e. Ram type and quantity, CPU details, etc...

I'll see about getting some pictures tomorrow, as of now the current setup is:

CPU: AMD 5x86 133MHz
Memory: 4x 4MB 72pin IBM 70ns
Cache: None
Video: PCI S3 Trio 64V2 (Will be replaced with Mach64 since it's output is cleaner than this one)
SCSI: Adaptec AHA-1540 ISA (Planning on using this only for CDRom)

Oddly enough the SCSI hard drive attached now will not boot, but is fully accessible under DOS.
This is even after running sys.com and fdisk /mbr.

I would like to try for 50MHz as the increased memory bandwidth should help alleviate the lack of L2 cache, but I'm more worried about a horrible heat death for my only 133MHz 486 chip.
😀

I'm also going to guess that even at 25MHz no 486 could saturate the PCI bus so maybe the PCI divider would be fine in this case.

As requested:
😀

I had a v3.2 of the M919 board at one point but tossed it after I realised that v3.4 was less buggy.

The M919 generally does fine with a 40 and 50 MHz FSB because it implements and auto 2/3 multiplier onto the PCI frequency. You will never run your PCI bus at 40 or 50 MHz on an M919. I'd recommend running your chip at 50x3 for optimal PCI performance.

The AMD X5 works relatively well on the M919, however, I have two v3.4 boards, and one of them insists on setting the L1 cache of the AMD x5 into write-through mode without the ability to change it. The option is simply greyed out in the CMOS setup. I'm not sure why the other board allows for adjustment when the jumper settings on both boards are identical.

In which ways was your board buggy?
Just so I know what to look out for... 😀

The write-back/write-through option is grayed out in the bios though, but I assumed it was because I set it up as a regular AMD DX-4.
I'm not worried about the caches so much though, in fact, the main memory has more bandwidth than my other 486s L2 caches.

Well, image if you were using cache at 50 MHz? It would be even faster. Now imagine cache at 66 Mhz... w00t! That's the World's Fastest 486.

The WB/WT option is greyed out, but stuck in which mode? WT mode I assume.

It was more than 10 years ago that I tested and disgarded the v3.2 board. I no longer recall the exact symptom of bugginess. Brain memory is like that - you sometimes only remember it being buggy and getting fed up with it.

I believe it is stuck in WT mode, yes.
I think I read somewhere that you can do a motherboard mod to force the cache mode though.

So far though none of the memory I've found is willing to run at 50MHz with 0 waitstates without errors during memtest86.
It seems to be running fine at 40x3 though.

I only have one set of gold contact FPM RAM (64 MB/stick) which will work well at 50 Mhz and 0 WS. For all others, I needed to insert 1 WS on the read, or was it the write? I think it was the read. With 1 WS, I think 50 MHz is still your best bet overall due to the increase PCI performance and overall improved RAM throughput compared to 40 MHz w 0 WS.

feipoa wrote:

I only have one set of gold contact FPM RAM (64 MB/stick) which will work well at 50 Mhz and 0 WS. For all others, I needed to insert 1 WS on the read, or was it the write? I think it was the read. With 1 WS, I think 50 MHz is still your best bet overall due to the increase PCI performance and overall improved RAM throughput compared to 40 MHz w 0 WS.

Worked like a charm, thanks!
😀

1WS Read and it did 4 passes of memtest with no errors at 50MHz.

Can you reset?

0-1WS makes my m919 freeze instead if I reset, but otherwise it works fine (and quite fast).

Reset works but ctrl-alt-del will lock it up.