If you have an old old pin cpu you don't care about, a hot air station, and the patience of a Saint: Theoretically, you could transfer the pins across to the chipset, put in your bodge wire, and have the space required to solder to the board with hot air.

The chipset would be on stilts, but it should work. It's only 100mhz

All you people telling me I can't do it has kinda made me want to try even harder 😀 I did get it off cleanly without any ripped pads so it's theoretically possible.

A reballing kit for this PCH is gonna cost more than the motherboard though, heh. Maybe I'll just buy a brand new balled PCH. Could even upgrade to a real X99 while I'm at it!

rasteri wrote on 2024-04-25, 10:24:

All you people telling me I can't do it has kinda made me want to try even harder 😀 I did get it off cleanly without any ripped pads so it's theoretically possible.

A reballing kit for this PCH is gonna cost more than the motherboard though, heh. Maybe I'll just buy a brand new balled PCH. Could even upgrade to a real X99 while I'm at it!

Reballing kit costing more than the mainboard? Just get yourself some balls ! leaded balls, though they'll have to be tiny 😉 They're cheap and plentiful.
Then you can start reballing it like some people do, very zen, very meditative, https://www.youtube.com/watch?v=L7k4ZkD2bpI
oh yeah don't forget to actually heat up the chip after placing all the balls, so the balls adhere to the pads and the reballing is actually done !

Is a real X99 chipset pin (ball?) compatible with the one that was originally on the mainboard? I'ld doubt it.

weedeewee wrote on 2024-04-25, 10:31:

Reballing kit costing more than the mainboard? Just get yourself some balls ! leaded balls, though they'll have to be tiny 😉 They're cheap and plentiful.

If I thought there was a decent chance of it working then I might try reballing it by hand 😀 The pins on that thunderbolt controller are really far apart compared to the PCH, though

Is a real X99 chipset pin (ball?) compatible with the one that was originally on the mainboard? I'ld doubt it.

They actually do have the same ballout. The exact same machinist motherboard has been found with loads of different PCHs (although maybe the BIOS is different)

rasteri wrote on 2024-04-25, 11:09:

If I thought there was a decent chance of it working then I might try reballing it by hand :) The pins on that thunderbolt contr […]
Show full quote

weedeewee wrote on 2024-04-25, 10:31:

Reballing kit costing more than the mainboard? Just get yourself some balls ! leaded balls, though they'll have to be tiny 😉 They're cheap and plentiful.

If I thought there was a decent chance of it working then I might try reballing it by hand 😀 The pins on that thunderbolt controller are really far apart compared to the PCH, though

Is a real X99 chipset pin (ball?) compatible with the one that was originally on the mainboard? I'ld doubt it.

They actually do have the same ballout. The exact same machinist motherboard has been found with loads of different PCHs (although maybe the BIOS is different)

Yeah, I wouldn't trust that for a moment. Several years ago you could still get an actual X99 C610 chipset, these days It's all desktop chipsets and the boardname is X99-whatever, just like you now personally experienced.

https://www.intel.com/content/dam/www/public/ … h-datasheet.pdf
https://www.intel.com/content/dam/www/public/ … h-datasheet.pdf

Chapter 6 on both, at first glance it looks like the C610 has way more ballpads.

anyway, If you feel the need to reball, just do it ! 😁

I bought SnPb 63/37 % balls and BGA stencils for a few $ on ebay/aliexpress. Of course buying a new/recycled balled PCH is an easier option...

weedeewee wrote on 2024-04-25, 11:23:

Chapter 6 on both, at first glance it looks like the C610 has way more ballpads.

Ah crap you're right, I was looking at the generic 9 series chipset docs rather than the X99.

RayeR wrote on 2024-04-25, 15:11:

I bought SnPb 63/37 % balls and BGA stencils for a few $ on ebay/aliexpress. Of course buying a new/recycled balled PCH is an easier option...

Yeah a new balled PCH is the easiest, and also surprisingly cheapest option.

Back in the real world though - I just remembered I have an H81 LGA1150 motherboard that I'm yet to try, and it turns out that in single thread performance the fastest LGA1150 processor is faster than the fastest LGA2011-v3 processor. So maybe a better option all round.

PCB is made, one day it'll get ordered with all the parts necessary too and any experiments to actually get this going...

Heh, nice 3-slot version. So now there are 3 HW versions 😀 Did you place some DC-DC iverter for negative rails?
EDIT: oh I see LM2575-12 it can work as inverter? I used pretty universal old MC34063, I think it's more than enough for a few tens of mA that usually soundcard or RS232 driver drains...

>Rasteri
not a surprrise that high clocked unlocked core i7 is faster than xeon at singlethread 😀

Jaa, I want those 3 slots that the physical space can give me, and it was a pain to get things routed. I could have used 4 layers but I had to try to see if 2 will work out and it did ~

I was looking at MC34063 but it is only slightly cheaper, needs a few more parts and I actually have not used it before so I went with a solution I have used before and I know will work perfectly. It is a bit of an overkill as far as current output capability goes but it'll take even the biggest analog stages on all the cards I can stuff in those slots 🤣.

Many switching converters can be turned negative if they have external "catch diode", in same way both MC34063 and LM257x and related can be. I have done it on some chips that didn't even mention it in the datasheet and it worked fine.

Tiido wrote on 2024-04-26, 10:31:

PCB is made, one day it'll get ordered with all the parts necessary too and any experiments to actually get this going...
LPC2ISA.png

One suggestion if you don't mind.
Design it to fit the standard atx case but with a spacer that fits under matx/mitx boards with lower standoffs.
Ie, a metal tab soldered on that sticks out, and the motherboard sits on top of.

I'll draw it later if you don't get what I mean

Forgive my amateur hour mspaint skillz

There are two holes to put plastic spacers into like you sometimes see used on AT cases+mobos, to support the board at the mobo end. You can screw those copper bits into those holes if you need further connectivity, they align with existing ATX screwholes too.

Ah, fair enough

>Tiido
I did it also on 2 layers but for project that fits on smaller pcb I consider 4 layers that are now cheap on JLCPCB so it result in less routing pain and better signal integrity due to planes. 😀 I thinking of making a version that will integrate also secondary superIO but far future when have more time...

I played with MC34063 at school and it seemed pretty universal. All possible topologies are described in datasheet and also some books. I digged those chips from various thrashed boards, they was used quite common so did't need buy them. LM2575 did't mention inverter topology in datasheet so I wonder a bit but if you have tested it works then ok...

Only benefit I would get from 4 layers at this point is a proper ground plane and a bit better power delivery, but the way signals are done, is already basically optimal, with minimal via use etc.
4 layers is starting to get pretty cheap indeed, perhaps one day I'll justify the jump in my head 🤣

Inverting is seen in some other LM257x datasheets, the entire "simple switcher" series is pretty much compatible on that regard. I have used them and this exact chip on another design doing this exact task with success so I know it will work fine ~

I've put my X99M Killer/3.1 and dISAppointment 0.2 into the case that previously housed my RUBY-9719VG2AR, as the test setup is now considered good enough to replace it.

For the past few days I've been working on preparing system drives for this system, installing and testing OSes I intend to run on it.

It seems I can continue using the existing cable (~40cm) for the time being, as it turned out to be just the right length for the case I'm installing it into, making the connection between the board and adapter adequately firm. Tested Duke Nukem 3D again after assembly, and the adapter and sound card (CT2950) did not fall off the bus like it used to.

On the other hand, I found some differences in behavior when booting to different DOS versions.
FreeDOS T2404: Unreliable. Duke Nukem 3D fails test on either IRQ or 16-bit DMA. DIAGNOSE fails on IRQ setup. Jazz Jackrabbit plays for only a brief while on startup then silence.
PC-DOS 7.1: DIAGNOSE passes all tests (IRQ, low and high DMA) in one shot. Duke Nukem 3D tests pass and plays fine in-game. Jazz Jackrabbit also plays fine in-game.

Before assembly I used a USB stick containing an old FreeDOS install (T2305), and the results were a bit better compared to my current T2404, though still not as reliable as with PC-DOS 7.1. There are indeed some changes to the kernel during the period, though I'm not sure which changes might have any influence on the reliability of the adapter and sound card... I'm using mostly same TSRs for both PC-DOS 7.1 and FreeDOS at the moment.

Existing issues such as DIGPAK (and games utilizing it), as well as SBDIAG (DMA and tada test), remain regardless of DOS kernel used. Additionally, Tyrian 2000's setup cannot enable sound at all.

PS: In case anyone's interested, I've uploaded a picture I took about where LDRQ1# is on X99M Killer/3.1. You need to remove the heatsink to be able to access it. The same board layout should also apply to the non-3.1 variant.

LSS10999 wrote on 2024-04-27, 15:51:

On the other hand, I found some differences in behavior when booting to different DOS versions.

I always seem to have unexplained problems with anything but normal DOS, especially on weird PCs.

Ha! Managed to find LDRQ1 on my H81 LGA1150 motherboard (dfi hd100) using the "set GPIO23 to output and flap it on and off" technique.

I hacked LPCEXP this time. Butchered source code attached.

BTW it isn't flapping at as high a frequency as I'd expect, maybe the watcom delay command doesn't allow as fine-grained control. But whatever, who cares 😀

Now off to wire up the rest of the LPC bus and test this thing!

LSS10999 wrote on 2024-04-14, 16:04:

I've found the answer. Actually port 21h is being used for IRQ stuffs and should not be touched.

So I'm having the same problem as you (interrupts not firing), except using your latest code that doesn't touch 21h doesn't seem to help.

DMA is working great though.

Any ideas? Maybe I can set 21h to the same value as one of my other boards...

(EDIT : to be honest I should probably check I've wired up SERIRQ properly)