VOGONS


Reply 520 of 577, by repaxan

User metadata
Rank Newbie
Rank
Newbie

ECS NForce4-A754, with some wrong voltage readouts and broken fan reporting, should be one NB fan at ~4500 RPM, one CPU fan at ~3500 RPM. Actually the CPU fan shows up as 0 RPM in the BIOS so I am not sure what to think about this.

Reply 521 of 577, by red-ray

User metadata
Rank Oldbie
Rank
Oldbie
repaxan wrote on 2020-06-12, 03:38:

ECS NForce4-A754, with some wrong voltage readouts and broken fan reporting, should be one NB fan at ~4500 RPM, one CPU fan at ~3500 RPM. Actually the CPU fan shows up as 0 RPM in the BIOS so I am not sure what to think about this.

Thank you for the report and posting the save file, when voltages need to be adjusted a screen shot of the BIOS monitoring screen often helps.

I have adjusted things in the attached SIV64X V5.49 Oops-08 test SIV, how does it look please?

To me it looks like the CPU Fan has a dodgy tachometer sensor as in the save file it's sample count is 63 when the NB Fan is 294. I also spotted VBAT +2.83 and suspect the CR2032 may need replacing.

Were you aware the Intel SSDSA2BW080G3H may have issues, check the End-to-End Error Detection Count.

Last edited by red-ray on 2020-06-13, 05:02. Edited 1 time in total.

Reply 522 of 577, by repaxan

User metadata
Rank Newbie
Rank
Newbie
red-ray wrote on 2020-06-12, 07:26:
Thank you for the report and posting the save file, when voltages need to be adjusted a screen shot of the BIOS monitoring scree […]
Show full quote
repaxan wrote on 2020-06-12, 03:38:

ECS NForce4-A754, with some wrong voltage readouts and broken fan reporting, should be one NB fan at ~4500 RPM, one CPU fan at ~3500 RPM. Actually the CPU fan shows up as 0 RPM in the BIOS so I am not sure what to think about this.

Thank you for the report and posting the save file, when voltages need to be adjusted a screen shot of the BIOS monitoring screen often helps.

I have adjusted things in the attached SIV64X V5.49 Oops-08 test SIV, how does it look please?

To me it looks like the CPU Fan has a dodgy tachometer sensor as in the save file it's sample count is 63 when the NB Fan is 294. I also spotted VBAT +2.83 and suspect the CR2032 may need replacing.

Were you aware the Intel SSDSA2BW080G3H may have issues, check the End-to-End Error Detection Count.

Voltages look right now.

The fans are definitely fine and the issue is on the board. Both the CPU and 'CASFAN1' don't report properly, only the NB fan header seems to work. If you don't mind speculating, what can cause this issue?

CR2032 is actually ~2.7v on the meter.

I'm out of good SSDs so this one serves as my 'bench disk'. You'll probably be seeing it again :)

Reply 523 of 577, by red-ray

User metadata
Rank Oldbie
Rank
Oldbie
repaxan wrote on 2020-06-13, 02:54:

I'm out of good SSDs so this one serves as my 'bench disk'. You'll probably be seeing it again 😀

Thank you for checking out the test SIV and posting the results. Given 1.2 Volts is reported by the BIOS I have made SIV better report it in the attached SIV64X V5.49 Tidy-08 test SIV.

I don't really know what issue with the CPU Fan might be, were it my system I would be inclined to attach a oscilloscope to the two fan tachometer signals and check both looked good. The CPU Fan seems to report a reading about 14% of the time, but the range is 892 to 6,026 RPM which seems wrong. I further spotted the range of T3 CPU is -5.0°C to 66.0°C so maybe the ITE IT8712 is faulty.

I have recently been improvising the Radeon X series GPU reporting so am wondering if you have any of those GPUs.

Last edited by red-ray on 2020-06-13, 07:06. Edited 1 time in total.

Reply 524 of 577, by repaxan

User metadata
Rank Newbie
Rank
Newbie

1.2v shows now.

This time I'm only using one dual rank DIMM and I noticed it incorrectly reports as dual channel.
If it matters, there's 1 channel with 3 slots, and the manual has a fancy chart but the gist of it is that with up to 3 ranks it supports DDR400, otherwise it drops down to DDR333 (actually 320).

I have a X850 and X1950 which I can try out sometime this weekend, maybe.

Reply 525 of 577, by red-ray

User metadata
Rank Oldbie
Rank
Oldbie
repaxan wrote on 2020-06-13, 06:10:

This time I'm only using one dual rank DIMM and I noticed it incorrectly reports as dual channel.
If it matters, there's 1 channel with 3 slots, and the manual has a fancy chart but the gist of it is that with up to 3 ranks it supports DDR400, otherwise it drops down to DDR333 (actually 320).

I have a X850 and X1950 which I can try out sometime this weekend, maybe.

What do CPUZ + HWiNFO report for the channels? If I know this then I can ask around to try and figure out why.

Tidy-08 has been tested with X1050 + X1050 + X1650 + X1950 on Windows 7 x64 using the inbuilt drivers which support D3D meaning the GPU Utilisation was reported. Do you need an x32 SIV for the tests?

Reply 526 of 577, by repaxan

User metadata
Rank Newbie
Rank
Newbie
red-ray wrote on 2020-06-13, 06:25:

What do CPUZ + HWiNFO report for the channels? If I know this then I can ask around to try and figure out why.

They both report single channel.

red-ray wrote on 2020-06-13, 06:25:

Tidy-08 has been tested with X1050 + X1050 + X1650 + X1950 on Windows 7 x64 using the inbuilt drivers which support D3D meaning the GPU Utilisation was reported. Do you need an x32 SIV for the tests?

Nope, I'll be using my X58 board, so Win7 64-bit.

Reply 527 of 577, by red-ray

User metadata
Rank Oldbie
Rank
Oldbie
repaxan wrote on 2020-06-13, 06:36:

Nope, I'll be using my X58 board, so Win7 64-bit.

Good, I guess it has multiple PCIe x16 slots so please try with both installed at the same time. I guess you could make things even trickier for SIV by also having an NVidia GPU installed.

Reply 528 of 577, by repaxan

User metadata
Rank Newbie
Rank
Newbie

In the meantime here's my HTPC, a Gigabyte 78LMT-USB3 reporting 5v as 4.1v (not reported in BIOS, but measures 5.1v).

The PSU in this thing is a little sketchy so this was actually pretty concerning.

Last edited by repaxan on 2020-06-13, 19:38. Edited 1 time in total.

Reply 529 of 577, by red-ray

User metadata
Rank Oldbie
Rank
Oldbie
repaxan wrote on 2020-06-13, 06:36:

They both report single channel.

I think it's fixed in SIV64X 5.49 Beta-08, is it please?

Last edited by red-ray on 2020-06-13, 13:13. Edited 3 times in total.

Reply 530 of 577, by red-ray

User metadata
Rank Oldbie
Rank
Oldbie
repaxan wrote on 2020-06-13, 07:14:

In the meantime here's my HTPC, a Gigabyte 78LMT-USB3 reporting 5v as 4.1v (not reported in BIOS, but measures 5.1v).

The PSU in this thing is a little sketchy so this was actually pretty concerning.

Given the BIOS fails to report +5 then I suspect that +5 is not available via the SIO so SIV64X 5.49 Beta-08 does not report it at all and should also tidy up the other SIO things. Does all now look OK please?

I was pleased to see SIV correctly dealt with only one of the four DIMMs having MTS support.

file.php?id=85787

Attachments

Reply 531 of 577, by repaxan

User metadata
Rank Newbie
Rank
Newbie

Both issues are fixed now.

For the 754 board, earlier you said 'I further spotted the range of T3 CPU is -5.0°C to 66.0°C so maybe the ITE IT8712 is faulty.'
But I noticed that in the BIOS it reports a stable CPU temperature.

Reply 532 of 577, by red-ray

User metadata
Rank Oldbie
Rank
Oldbie
repaxan wrote on 2020-06-13, 19:46:

Both issues are fixed now.

But I noticed that in the BIOS it reports a stable CPU temperature.

Thank you for checking and good, but the BIOS does not report Min + Max values 🙁 Given both the BIOS and SIV see issues with the CPU Fan speed something is wonky with the hardware and I feel this can't be sensibly worked around in software.

Reply 533 of 577, by repaxan

User metadata
Rank Newbie
Rank
Newbie

Here's just the X1950 XTX for now, I'm having trouble getting the system to POST consistently and it seems to be caused by the card.

SIV has trouble reading the GPU clocks, also it correctly figures out triple channel memory but doesn't realize there's 3 dimms.

I also just realized that GPU-Z screenshot shows 80c at idle, kind of concerning...

Last edited by repaxan on 2020-06-18, 07:50. Edited 2 times in total.

Reply 534 of 577, by red-ray

User metadata
Rank Oldbie
Rank
Oldbie
repaxan wrote on 2020-06-16, 07:16:

SIV has trouble reading the GPU clocks, also it correctly figures out triple channel memory but doesn't realize there's 3 dimms.

Thank you for the report and I will try and adjust SIV for the X1950.

The reason SIV is only reporting two DIMMs is that the SMBIOS (see [Machine]) is only reporting two DIMMs, further Windows is only seeing 8GB rather than 12GB. I suspect there is something wonky, what do you think?

I was pleased to see SIV reported the ASP0800 in the System Power Usage section.

Update: With ATI GPUs there are two ways they can be reported which are directly or via ATIADL and your system SIV was using ATIADL, had it not been then SIV would probably have done better. It seems that ATIADL fails to report the clocks for the X1950 so I have adjusted the SIV code so when ATIADL fails to report the clocks then SIV will report them directly. I have been pondering this for some time, but had not got around to adjusting the code.

Please try again with the attached SIV64X V5.50 DACs-00 test SIV which also should remove the secondary device and tidy a few odds and ends.

When I uninstalled all the ATI software C:\Windows\system32\atiadlxx.dll was not expunged and I had to delete it manually.

I just looked at the GPUZ reported clocks and am wondering why the current clocks are different on the two panels 🙁

file.php?id=86024file.php?id=86025

Last edited by red-ray on 2020-06-17, 18:40. Edited 1 time in total.

Reply 535 of 577, by repaxan

User metadata
Rank Newbie
Rank
Newbie

Okay so I just reseated the memory over and over until all 12 GB were detected in both the BIOS and Windows. Memory slots might be going bad :(
I also noticed the GPU fan connector was coming loose, thankfully before I ran any benchmarks.

I tried running 3dmark06 with SIV open but I got a black screen and a 'display driver stopped responding but recovered' balloon. I tried again and this time there was no graceful recovery and I had to do a hard reset. Not sure if this is supposed to be supported.

I used DDU to clear drivers of the previous card, an HD 3000-something. The timestamp on the atiadlxx.dll file is from 2010 so it's correct for the latest available drivers.

The 506/594 in the GPU-Z sensor panel is right (idle state).

Last edited by repaxan on 2020-06-18, 07:50. Edited 2 times in total.

Reply 536 of 577, by red-ray

User metadata
Rank Oldbie
Rank
Oldbie
repaxan wrote on 2020-06-17, 04:23:

Okay so I just reseated the memory over and over until all 12 GB were detected in both the BIOS and Windows. Memory slots might be going bad 🙁

Thank you for the new report and I am pleased to hear you sorted out the memory.

SIV said GDDR3 rather than GDDR4 so I plan to adjust this, but before I do can you provide a link to exactly what the GPU is? I suspect it's one of the ones listed by https://www.techpowerup.com/gpu-specs/?genera … sort=generation, but it may not be listed at all.

Do the SIV reported GPU clocks also change as the GPU state changes? In the save file they did not.

I fixed SIV V5.50 Beta-00 to say Xeon W3690 rather than XeonW3690 and I am wondering what to do about the reported power levels. Looking at Menu->System->ACPI + PnP->ACPI Eval the CPU _PSS method reports 4474.0 Watts for 3468 MHz which is clearly impossible. This information is setup by the BIOS and I suspect all stems from MSR 000001AC (TURBO_POWER_LIMIT) being set to 0xFFFFFFFF. On my Core i7 Extreme 980X (Gulftown) the MSRs are as below and the CPU _PSS methods report the same powers.

file.php?id=86102

Initially on my system I did not install any ATI software at all with Windows 7 using it's inbuilt Radeon drivers. Maybe if you did the same things would be more stable. You may need to use dism + pnputil to remove them, I typically do dism /online /get-drivers /format:table | find "ATI" | sort /+17 then pnputil -f -d oem<nn>.inf

Attachments

Reply 537 of 577, by repaxan

User metadata
Rank Newbie
Rank
Newbie

It's a normal X1950 XTX, not the crossfire master version, but I totally forgot that on this card I have a modded VBIOS with a 'baked in' overclock. Oops.
Just so it didn't cause any trouble I restored the original VBIOS, but it didn't fix the SIV/3dmark06 lockups. Neither did playing with the drivers.
Once or twice the benchmark did make it a few seconds before it froze, but IIRC that was with the 2010 drivers.
On the other hand, with the Windows built in drivers, GPU-Z thinks the GPU clock doesn't change from idle.

I have the CPU overclocked to 31x core and 25x uncore, so you shouldn't really be seeing the 3468 MHz.

One weird thing about this system is that I have to keep the BIOS option for VT-D (which is on a separate page just by itself) disabled. Otherwise Windows 8.1 or 10 won't boot which might be related to Meltdown/Spectre mitigations. Do you know anything about this?

Reply 538 of 577, by red-ray

User metadata
Rank Oldbie
Rank
Oldbie
repaxan wrote on 2020-06-18, 07:49:

It's a normal X1950 XTX, not the crossfire master version, but I totally forgot that on this card I have a modded VBIOS with a 'baked in' overclock. Oops.
Just so it didn't cause any trouble I restored the original VBIOS, but it didn't fix the SIV/3dmark06 lockups. Neither did playing with the drivers.

One weird thing about this system is that I have to keep the BIOS option for VT-D (which is on a separate page just by itself) disabled. Otherwise Windows 8.1 or 10 won't boot which might be related to Meltdown/Spectre mitigations. Do you know anything about this?

Thank you for the clarification, with the stock VBIOS it's clear the card is a Radeon X1950 XTX so I have adjusted the attached SIV64X V5.00 R580+-01 test SIV in light of this, please may I have new save files? I think some of the Radeon names in PCIDEVS.txt may be slightly skewed, so I am slowly trying to fix them.

AFAIK SIV uses much the same code to read Radeon R500 GPU clocks as HWiNFO64, does HWiNFO64 also trigger the lockup?

I expect the MSR 000001AC (TURBO_POWER_LIMIT) being set to 0xFFFFFFFF is related to the BIOS power limit settings, what are they? Mine are as below. I wonder, is it possible to set at least half sensible limits, still get the most out of the CPU and for SIV to report plausible Power Limits?

I run Windows 7 x64 on my ASUS P6X58D Premium with VT-D enabled, a few years ago I used to also run W10 insider builds on it, but W10 stopped working at some stage so I just used a different system and never figured out why. I have never tried 8.1 on it but suspect it would run. I could easily try booting the install DVD and run SIV from the pre-install recovery command window, what happens with that on your system?

Attachments

Last edited by red-ray on 2020-06-20, 09:08. Edited 1 time in total.

Reply 539 of 577, by repaxan

User metadata
Rank Newbie
Rank
Newbie

Aside from the screenshot I posted there are no other options for adjusting the power limit, turbo durations, etc.

I've reinstalled the latest 2010 ATI drivers so that the GPU isn't stuck in idle clocks. HWInfo shows clocks properly, SIV being open during the benchmark still causes lockups.

update: I tried several other graphics loads (OCCT, Unigine, Furmark, even just the built in WinSAT) which still locked up the system if SIV was open in the background. I noticed that only the Unigine demos showed the expected 648 MHz in GPU-Z, the others stayed at 506.