zuldan wrote on 2026-03-18, 09:39:

tehsiggi wrote on 2026-03-18, 09:34:

Have you checked the traces going from the PCIe slot to the chipset? Kind of feels like the PCIe slot/lanes are having some issues.

Yep double checked them under the microscope.

It could be as simple as a failing SMD capacitor or resistor between the chipset and PCI-E but as you know, without a boardview or schematics it would be almost impossible to diagnose.

There are small series caps for the TX lines of the individual lanes. I marked them red on this picture:

Also, for a reference of NF4 SLI board implementation, you can take a look at the schematic of the MSI K8N Diamond (MS-7100):

zuldan wrote on 2026-03-18, 09:30:

Yep, not talking about the PCI-E version. I mean the PCI-E bandwidth.
The second PCI-E x16 slots runs at 2x because this is not the SLI version of the board. The BIOS and GPU-Z both confirm the card is running at x2 when plugged into the second PCI-E x16 slot. This is normal for this DFI board. I can change the slot to x16 but that requires a 0 ohm resistor installed on the chipset.
I've attached the report for you.

I guessed this, but felt I should confirm.

So it's OK in the 2nb slot, I wonder might the 1st slot be faulty given both AMD + NVidia CPUs have the issue? I would try setting the 2nd slot to x16 and see what happens.

Thank you for the save files, strangely SIV reported [402MHz Single DDR1 RAM], I wonder what changed.

Looking at [Machine] I suspect if you moved the DIMM from A3 Bank6/7 to A0 Bank0/1 you would have Dual Channel memory which should be faster.

I can't see what I need to check as it's only a partial save file, the last few lines are:

1[fat]  [windows]
2
3  SIV32X - running on \\WINDOWS-A330342 - Windows XP x32 Professional  V5.01  Build 2600  Service Pack 3
4
5  Unable to Report FAT Statistics
6
7  Status 15 The system cannot find the drive specified.

It's only 325KB when I expected 1 to 2 MB. Did the save complete? What did the [Save Local] panel look like?

If there was an issue with one section do Menu->File->Save->Save Selected, deselect that section and then press [Start].

tehsiggi wrote on 2026-03-18, 10:12:

There are small series caps for the TX lines of the individual lanes. I marked them red on this picture: […]
Show full quote

zuldan wrote on 2026-03-18, 09:39:

tehsiggi wrote on 2026-03-18, 09:34:

Have you checked the traces going from the PCIe slot to the chipset? Kind of feels like the PCIe slot/lanes are having some issues.

Yep double checked them under the microscope.

It could be as simple as a failing SMD capacitor or resistor between the chipset and PCI-E but as you know, without a boardview or schematics it would be almost impossible to diagnose.

There are small series caps for the TX lines of the individual lanes. I marked them red on this picture:

The attachment Screenshot 2026-03-18 at 11.09.46.png is no longer available

Also, for a reference of NF4 SLI board implementation, you can take a look at the schematic of the MSI K8N Diamond (MS-7100):

The attachment msi_ms-7100_rev_100_sch.pdf is no longer available

Thanks for that I’ll have a looksee!

red-ray wrote on 2026-03-18, 10:26:

So it's OK in the 2nb slot, I wonder might the 1st slot be faulty given both AMD + NVidia CPUs have the issue? I would try setting the 2nd slot to x16 and see what happens.

I changed the 2nd slot to x8 and the problem returned. Since both slots are having issues, I suspect the issue is closer to the chipset.

red-ray wrote on 2026-03-18, 10:26:

Thank you for the save files, strangely SIV reported [402MHz Single DDR1 RAM], I wonder what changed.

Hmm nothing has changed. Your program was reporting 0MHz but now it’s showing 400MHz. I’m pretty sure the BIOS was reporting dual channel. I’ll check next time I start the board up.

red-ray wrote on 2026-03-18, 10:26:

It's only 325KB when I expected 1 to 2 MB. Did the save complete? What did the [Save Local] panel look like?

If there was an issue with one section do Menu->File->Save->Save Selected, deselect that section and then press [Start].

I managed to run the Save Local once but didn’t realise it saved it to the directory the .exe was in. I then ran it again and the program kept crashing. Dinner was ready and I had had enough for the day 🤣. I’ll see what I can do tomorrow.

zuldan wrote on 2026-03-18, 10:57:

I changed the 2nd slot to x8 and the problem returned. Since both slots are having issues, I suspect the issue is closer to the chipset.

Hmm.. the 8 lanes of the second slot are very likely the "missing" 8 lanes from the first one, which makes me wonder.

If you have the second on 8 lanes, is the first one 8 lanes as well? Does the issue persist on the first slot too?

If you wiggle the card in the primary x16 slot does the system crash? If so, that's oxidised pins in the PCI-E slot.

red-ray wrote on 2026-03-18, 10:26:

It's only 325KB when I expected 1 to 2 MB. Did the save complete? What did the [Save Local] panel look like?

Fixed!

red-ray wrote on 2026-03-18, 10:26:

Looking at [Machine] I suspect if you moved the DIMM from A3 Bank6/7 to A0 Bank0/1 you would have Dual Channel memory which should be faster.

It's 100% in dual channel mode

tehsiggi wrote on 2026-03-18, 19:24:

zuldan wrote on 2026-03-18, 10:57:

I changed the 2nd slot to x8 and the problem returned. Since both slots are having issues, I suspect the issue is closer to the chipset.

Hmm.. the 8 lanes of the second slot are very likely the "missing" 8 lanes from the first one, which makes me wonder.

If you have the second on 8 lanes, is the first one 8 lanes as well? Does the issue persist on the first slot too?

Tried these various configurations with the card in each slot. I’ve never seen a problem like this ever! Been in IT for 20+ years.

MikeSG wrote on 2026-03-19, 07:05:

If you wiggle the card in the primary x16 slot does the system crash? If so, that's oxidised pins in the PCI-E slot.

The board has been washed and I've used proper electrical contact cleaner on both PCI-E slots (multiple times). I can't find any corrosion on the board. It's in very good condition.

zuldan wrote on 2026-03-20, 07:07:

Fixed!
It's 100% in dual channel mode

Thank you for the new save files, this time all is as it should be. I wonder what the issue was and guess we will never know for sure, my guess is a hardware issue. Likewise I suspect we will never know why SIV reported 0MHz as the memory speed, if it ever does this again please generate new save files.

I noticed you changed the DIMMs from 2 x 512MB to 1 x 1GB, was there an issue with the 512MB ones?

SIV reported Single as bit-9 (DualDimmEn) was not set, as I wrote this code over 20 years ago I can't remember why, I suspect Franck (CPUZ) told me this was how to detect Dual Channel and wonder, what does CPUZ report?

When I looked at the datasheet clearly SIV should be checking bit-16 (128-bit/64-bit) so I have updated the code and in test mode SIV V5.87 Beta-00 reported Dual. Beta-00 also tidies up the ITE IT8712 reporting.

red-ray wrote on 2026-03-20, 09:45:

Thank you for the new save files, this time all is as it should be. I wonder what the issue was and guess we will never know for sure, my guess is a hardware issue. Likewise I suspect we will never know why SIV reported 0MHz as the memory speed, if it ever does this again please generate new save files.

Will do!

red-ray wrote on 2026-03-20, 09:45:

I noticed you changed the DIMMs from 2 x 512MB to 1 x 1GB, was there an issue with the 512MB ones?

These motherboards are well known to be extremely picky about memory. The 1GB was crashing the system (and SIV).

red-ray wrote on 2026-03-20, 09:45:

SIV reported Single as bit-9 (DualDimmEn) was not set, as I wrote this code over 20 years ago I can't remember why, I suspect Franck (CPUZ) told me this was how to detect Dual Channel and wonder, what does CPUZ report?

CPU-Z is reporting dual channel

zuldan wrote on 2026-03-20, 07:18:

I’ve never seen a problem like this ever! Been in IT for 20+ years.

The closest I have seen is a 20% interrupt load on my Intel D5400XS (Skulltrail) system, it's been like this ever since I built it, I would like to know why/fix it, thus far I have not managed to. I still wonder, does this happen on all Skulltrail systems?

Perhaps you could try to see what happens under linux, specifically something like ubuntu. I don't expect the desktop with gnome to work well.. (lubuntu might be better, desktop is much more lightweight) - however I wonder what dmesg would print out on startup and if it lists any PCIe related errors.

No need to install, most distros can do live boot from USB.

zuldan wrote on 2026-03-20, 10:05:

Will do!

I have just released SIV V5.87 Beta-00 which should fix the channel reporting, does it please?, it should also tidy up the DFI LP NF4 ITE IT8712 reporting.

I was wondering what effect the very high interrupt load would have on SIV so did some checking, most of the time it has a minimal impact, but as I expected it causes issues for [Cache Latency].

tehsiggi wrote on 2026-03-20, 17:23:

Perhaps you could try to see what happens under linux, specifically something like ubuntu. I don't expect the desktop with gnome to work well.. (lubuntu might be better, desktop is much more lightweight) - however I wonder what dmesg would print out on startup and if it lists any PCIe related errors.

No need to install, most distros can do live boot from USB.

Tried Lubuntu, Slitaz and Puppy Linux, I couldn't get any of them to boot on this board. Some sat at "GRUB..." for 20 min or so.

Hmmm tempting....https://www.ebay.com/itm/134326054210

red-ray wrote on 2026-03-20, 19:50:

I have just released SIV V5.87 Beta-00 which should fix the channel reporting, does it please?, it should also tidy up the DFI […]
Show full quote

zuldan wrote on 2026-03-20, 10:05:

Will do!

I have just released SIV V5.87 Beta-00 which should fix the channel reporting, does it please?, it should also tidy up the DFI LP NF4 ITE IT8712 reporting.

I was wondering what effect the very high interrupt load would have on SIV so did some checking, most of the time it has a minimal impact, but as I expected it causes issues for [Cache Latency].

Here you go

zuldan wrote on 2026-03-21, 04:42:

tehsiggi wrote on 2026-03-20, 17:23:

Perhaps you could try to see what happens under linux, specifically something like ubuntu. I don't expect the desktop with gnome to work well.. (lubuntu might be better, desktop is much more lightweight) - however I wonder what dmesg would print out on startup and if it lists any PCIe related errors.

No need to install, most distros can do live boot from USB.

Tried Lubuntu, Slitaz and Puppy Linux, I couldn't get any of them to boot on this board. Some sat at "GRUB..." for 20 min or so.

Hmmm tempting....https://www.ebay.com/itm/134326054210

I'm out this weekend, but I'll try some on my 939 machine on monday.

tehsiggi wrote on 2026-03-21, 06:30:

I'm out this weekend, but I'll try some on my 939 machine on monday.

No rush at all. Hope you have a great time! I have zero Linux skills. It was my job to remove all Linux systems and replace with Windows 🤣

zuldan wrote on 2026-03-21, 04:44:

Here you go

Thank you, all the SIV changes worked as I intended. I noted that SIV no longer reported the CPU Power Levels, this is down to the _PSS ACPI methods no longer being there, I guess you changed some BIOS setting, did you? I looked in the manual and suspect CPU Spread Spectrum, I also noticed PCIe Spread Spectrum and wonder if changing this will affect the interrupt storm.

zuldan wrote on 2026-03-21, 06:40:

I have zero Linux skills. It was my job to remove all Linux systems and replace with Windows 🤣

Have you tried any other version of Windows? I run Windows 2008 R2 x64 Enterprise Server on my 2 x Dual Core AMD Opteron 270 (Italy) 2.00GHz [JH-E6] which is the Opteron variant of your AMD Athlon 64 X2 Dual Core 4400+ (Toledo) 2.20GHz [JH-E6] so I suspect Windows 7 x64 would work well provided you installed >= 4GB of memory and may be OK with 2GB.

red-ray wrote on 2026-03-21, 08:21:

Thank you, all the SIV changes worked as I intended. I noted that SIV no longer reported the CPU Power Levels, this is down to the _PSS ACPI methods no longer being there, I guess you changed some BIOS setting, did you?

I've been resetting the BIOS a lot and trying different settings.

red-ray wrote on 2026-03-21, 08:21:

I looked in the manual and suspect CPU Spread Spectrum, I also noticed PCIe Spread Spectrum and wonder if changing this will affect the interrupt storm.

I've tried CPU and PCI-E Spread Spectrum on and off for both but it makes no difference 🙁

red-ray wrote on 2026-03-21, 08:21:

Have you tried any other version of Windows? I run Windows 2008 R2 x64 Enterprise Server on my 2 x Dual Core AMD Opteron 270 (Italy) 2.00GHz [JH-E6] which is the Opteron variant of your AMD Athlon 64 X2 Dual Core 4400+ (Toledo) 2.20GHz [JH-E6] so I suspect Windows 7 x64 would work well provided you installed >= 4GB of memory and may be OK with 2GB.

I haven't tried other OS's but I suspect the same issue would occur because it's a hardware problem.

I think this problem has probably occurred for other people, but they haven't noticed it. I'm weird because I sit with Task Manager open all day 😉