X99 board without PCI slot, PCI-E only, almost no HW device detection after PCI bus device addition - dead end? PCI is working, but..

[LoneCrusader]

15 hours ago, ruthan said:

  Ok i tested it, result almost fine, but not without issues. Probably because of setup /p i  parameters (i wonder could it be combined with other switches?, i didnt use others for this installation) , more devices were detected, even PCI bus and IDE controlelrs, videocard, but only again with generic names, any specific Intel XXX (but devices has 1-06-2015 driver date, so LoneCrusades inf are used)..  I copied infs to windows installation inf folder, removed PCI bus and added it back, but its still the same.

SETUP /p i is virtually mandatory for any new system and also for those going back a few years now. Windows 98 supported ACPI to a point when it was introduced but modern systems are using newer implementations of ACPI that are completely incompatible with 9x. There is no way to fix this short of writing a new ACPI.SYS driver from scratch. The 2K version uses NTOSKRNL.EXE functions for memory allocation and such that cannot be backported to 98SE in any reasonable manner, if at all. ACPI is pretty useless anyway; APM will handle things just fine in most cases. At most you might get the "It is now safe to turn off your computer" screen and have to switch power off manually when you shutdown, but this is just like the old AT days when 9x was young.
The /p i switch should be usable with other switches with the correct syntax, but I've not tried this.

If the INF's have the 1-06-2015 date then it didn't pick up the newer ones I linked for you during install.

15 hours ago, ruthan said:

     Once again i had to change IRQ for COM1 device from IRQ 4 to 3 to, to make USB controller start.

   USB 3.5, i have problem to locate USBD.SYS, during installation its copied somewere but dialog is so quick that i dont know where.. I can extract installation package with 7zip as in case NUSB3.6, was i cant copy file to right place.. I scanned whole disc, include archives for this file, but except original file(alter nUSB3.5 installation) -06/08/2000) and file in USP3 - 3.56  (01/04/2001) i didnt find any.

    There 5 USB PCI unknow devices, none recognized automatically.. 
    But they are detected manually by Device manager driver search, now im testing them.
    Update: Controllers are detected fine, but there some problem with hubs and devices, devices are detected but after reboot there is freeze, related to PCIto PCI bridge enumerating..

DO NOT mix USP3 files with NUSB! They do not share the same approach and use different selections of files.
If you wish to use USP3 then you should add this later after other things have been sorted out.
USBD.SYS should be extracted to \WINDOWS\OPTIONS\CABS or something like that by NUSB. If not you can manually extract it from the NUSB package.
NUSB3.6 uses the Windows ME SYSDM.CPL which is known to have some minor cosmetic bugs under 98SE. Other bugs may be lurking undiscovered this is why I recommended 3.5 for now.

Not sure about having to change the IRQ or about the freeze. I haven't seen this but it may depend on your BIOS/hardware.

15 hours ago, ruthan said:

   I wonder if install MB with /p i to ingore AHCI, its that reason for not working automatic shutdown, could it be fixed, or its Nvidia 7950GT problem (i still have only VGA driver for it).

ACPI, not AHCI. ACPI = power management. AHCI = disk controller mode.

Yes, this is the reason for no automatic power shutoff at shutdown as I mentioned above. You will just have to deal with this on newer systems where ACPI is incompatible with 9x. Leaving ACPI enabled brings on far worse problems. I have an old Intel G41 chipset board where 98SE will not even boot if it is installed without the /p i switch. (This gives you an idea how long ACPI has been incompatible.)

15 hours ago, ruthan said:

   Exclamation mark - there is a message that BIOS dont assign IRQ to device.

I have that too. Windows 9x doesn't know what DMI2 is. Probably no fix for it, but it remains to be seen whether or not it causes any actual problem besides an error in the Device Manager Don't worry about this for now.

13 hours ago, ruthan said:

USB Composite Device errors are NTKERN.VXD Code 2 - loader could not load device driver.

Update: Even all unknow devices could be manually detected as Intel devices with 2017 driver.. But i wonder why they arent  detect during installation, i just copied new infs, should i also deleted the old ones?

 

4 hours ago, ruthan said:

When i installed nUSB 3.6 there not exclamation marks for USB devices, but at least USB audio is not working and i got 2 minutes waiting just for desktop icons, something is not right with USB..

USB Composite Devices are somewhat unexplored territory. Not sure how to solve this. Simply removing them and reinstalling them might help.

What are these devices so we can see if anyone has successfully used them before with 9x?

Overwriting or renaming the older INFs to some other file extension such as .TXT should work.
If the new INFs were not picked up and copied during installation you may have to manually put them in \WINDOWS\INF.

 

4 hours ago, ruthan said:

  About infs, except USB3 controllers, there is only device which were not recognized:

As rloew said that is the HDA controller. No working 9x drivers (and it keeps defeating all our efforts to backport one from 2K, so it doesn't look promising.)

The USB3 controllers also have no working drivers of course. I added a generic do-nothing entry to my USB.INF to move them under the USB section and prevent them from showing up as Unknown Devices.

[ragnargd]

4 hours ago, LoneCrusader said:

SETUP /p i is virtually mandatory for any new system and also for those going back a few years now. [...]
The /p i switch should be usable with other switches with the correct syntax, but I've not tried this.

I never used "setup /p i" on any of my builds - perhaps, i assume, i was either rather lucky, or had the tendency to avaid such platforms anyway...

But looking back, i never kept systems with a bit more modern intel chipsets, as not only was ACPI a problem, but practically everything, from USB2 to SATA. I always had systems, when Intel, that were full to the brim with add-ins cards, as so much things weren't working from the chipsets themselves. The only system i hated more than Intel was NVidias NF3, but not because the lack of backward-compatibility (which it had in spades), but the lack of forward compatibility (weren't able to run W7 64bit on it by any means, and even Vista was painful).

Funny enough, on the AM3+ and AM4 platforms, ACPI was and is working really well. Even more: On AM4, everything was fu**ed up, BUT ACPI was the ONLY thing detected and working, you could cleanly shut off the system, hibernate, suspend-to-disk, you name it. AMD is always worth some surprises... Perhaps, if the memory-problem of AM4 can be sorted out by some dark magic later (not that i expect that to happen), AMD may have stayed so close to standards, that many things could/would work, which are way out on Intel...

[LoneCrusader]

15 hours ago, ragnargd said:

I never used "setup /p i" on any of my builds - perhaps, i assume, i was either rather lucky, or had the tendency to avaid such platforms anyway...

But looking back, i never kept systems with a bit more modern intel chipsets, as not only was ACPI a problem, but practically everything, from USB2 to SATA. I always had systems, when Intel, that were full to the brim with add-ins cards, as so much things weren't working from the chipsets themselves. The only system i hated more than Intel was NVidias NF3, but not because the lack of backward-compatibility (which it had in spades), but the lack of forward compatibility (weren't able to run W7 64bit on it by any means, and even Vista was painful).

Funny enough, on the AM3+ and AM4 platforms, ACPI was and is working really well. Even more: On AM4, everything was fu**ed up, BUT ACPI was the ONLY thing detected and working, you could cleanly shut off the system, hibernate, suspend-to-disk, you name it. AMD is always worth some surprises... Perhaps, if the memory-problem of AM4 can be sorted out by some dark magic later (not that i expect that to happen), AMD may have stayed so close to standards, that many things could/would work, which are way out on Intel...

I used to prefer AMD back in the days of Super Socket 7 and K6-II (I even have an ultra-rare 570MHz one of these), but once I stepped up from them to Pentium 4 I never looked back. I've not had a lot of experience with AMD systems since then, but a friend of mine had a prefab Compaq machine that was AMD and it was junk (very slow, but to be fair I don't remember what the CPU specs were). Also it seemed that when multi-core CPU's became mainstream AMD relied too much on multiple cores instead of raw clock speed (which to me is more important, especially for 9x) and at least for a while did not keep up with Intel. I've tried only three AMD systems since then myself and they each have at least one weird, annoying problem or another under 9x (but they're all nForce chipset based as well, so that may be part of the issue). So I'm not much of an AMD fan these days. To each his own I suppose. YMMV.

[rloew]

I have had the opposite experience. As a general rule, AMDs were more stable, (I only have one with a nForce chipset), had more features, and less expensive.
Clock speeds have caught up, but differences in architecture make comparisons difficult. For example, AMD CPUs are faster for Multiply/Add operations while slower for Loops.

I bought my Intel Z87 a few years ago to compare speeds, after hearing people say that Intel was faster, and to identify possible issues with my Products with Intel CPUs.
The M5A97 AMD was faster than the Intel doing my Multi-Core Matrix Algebra program after adjusting for the lower Clock Speed.
I saw the same difference when I bought my AMD970 SLI Krait. In this case I had to adjust for the higher Clock Speed.

AMD CPUs have a number of MSRs and APIC Functions that come in very handy. I had to make multiple changes to my Multi-Core and MEMORY64 Packages to support Intel CPUs.

The Z87 was the first Computer I found that needed SPLIT8MB to Boot.

Edited August 28, 2017 by rloew

[dencorso]

SPLIT8MB splits the lower 8 MiB above the 1st MiB to avoid them from being taken by what exactly? I know you've documented it, but please do remind me which issue it was created to address.

[ruthan]

Quote

DO NOT mix USP3 files with NUSB! They do not share the same approach and use different selections of files.

   I didnt i at least in this installation (back i did it for lots of others, because i never read dont do it, USP3 has own USB stack but is only optimal, you dont have to install it),
i only pointed that such filename is also in the pack.

Quote

If the new INFs were not picked up and copied during installation you may have to manually put them in \WINDOWS\INF.

 I copied them, but search for driver or remove it  was needed to redirection, i expected with this will be done during installation, but its only small problem, important is that infs are working great..

Quote

Yes, this is the reason for no automatic power shutoff at shutdown as I mentioned above. You will just have to deal with this on newer systems where ACPI is incompatible with 9x. Leaving ACPI enabled brings on far worse problems. I have an old Intel G41 chipset board where 98SE will not even boot if it is installed without the /p i switch. (This gives you an idea how long ACPI has been incompatible.)

    Whole OS power management is complicated, but could be just coded pure dirty ASM shutdown for fix to shutdown computer? 
 

Quote

USB Composite Devices are somewhat unexplored territory. Not sure how to solve this. Simply removing them and reinstalling them might help.

  I never got, what is it, is expected that is something like USB virtual device logic or something lie that, because is between USB controllers in device manager stuff. 
  Im using KVM USB switch and USB audio cards are if im not work detected like USB composite device (HID + Audio) i dont know why.

 

Summary for X99 build with this board-  if there no other how to fix USB2 suggestion, is that it can work, but USB2 addon card is needed is probably the same as 2 Z97 MBs were i have very similar issues there.  New Intel USB 3 /2 controller problematic even in never OSes and nUSB would probably need some adjustments to handle it properly.

 

Edited August 28, 2017 by ruthan

[rloew]

SPLIT8MB punches holes in the first 16MiB of RAM so that large blocks cannot be allocated from it. This was designed to keep Gigabit NIC Drivers out of it.
The later /M Option blocks this area during Initialization keeping the Registry from being placed in this area. This is usually enough to leave room for the NIC.
Some recent systems seem to need both to leave enough DMA RAM space for other software.
VFAT allocates 1 small block of DMA space as a reserve buffer. If this fails, VFAT aborts because there is no space left, so VFAT acts as the Canary.

[ragnargd]

19 hours ago, LoneCrusader said:

I used to prefer AMD back in the days of Super Socket 7 and K6-II, but once I stepped up from them to Pentium 4 I never looked back. ... So I'm not much of an AMD fan these days. To each his own I suppose. YMMV.

For the same reason as yours my "primary" systems (i.e. that of my kids) have always been intel - more bang and less watt per core. Ryzen didn't really change that, even though i recognize the great advances AMD made, AND the fact, that many modern games are more and more optimized for multi-core. Had this happened like 5 years before, the FX series had a better reputation as well, i guess... but that just wasn't the case.

My "preference" is only for the legacy systems, where compatibility rules over performance.

My w98se-flagship, the FX 8350 on 890FX, replaces the heating in my room (and adds to my illumination, with all that fancy rgb-strips inside... :), but my main gaming rig will remain an i7 for the time being.