Thanks, I will check this version the next time I have access to this machine (hopefully in 2 weeks). Capacitors on the board look fine, I don't think it can report PSU voltages. I don't have NT installed on this system, might give it a try.
GL1zdA wrote on 2022-08-08, 05:57:I don't think it can report PSU voltages.
OK and I had in mind that they should be checked with a digital voltmeter.
Given you can't SIV for two weeks by then I will have released SIV 5.66 Beta-04 so please use SIV32L 5.66 Beta-04 or later..
Unfortunately, it still hangs the PC. Here are the logs.
GL1zdA wrote on 2022-08-12, 23:31:Unfortunately, it still hangs the PC.
Thank you for checking and posting the .LOG files. Have you checked the PSU voltages with a digitalvolt meter?
Things are very strange, the first run with the GPUs got stuck @ win_ini_sdms() trying to read the disk SMART information. When -NOGPUS was specified then win_ini_sdms() was OK and it got stuck trying to read the [ 0 - 07 - 0 ] which should be a Intel 82371EB PIIX4E Power Management Controller.
Given win_ini_sdms() caused an issue and then was OK I am confused and feel there must be some issue with the hardware.
Does doing SIV32L -DBGINI -DBGPCI -DBGGPU always result of the last line in SIV_DBGOUT.log being for win_ini_sdms()? My guess it it won't
Irinikus wrote on 2022-10-01, 17:54:It worked fine on My Pentium Pro, running windows XP Professional SP3:
I also have a DEC AlphaStation 255 @ 300MHz, fitted with a PowerStorm 4D60T that I could test it on in the future!
Thank you for checking SIV out, most things look good, but the DIMM sizes are not. How many DIMMs do you have and what size are they please? If you do Menu->File->Save Local and post the two files generated I should be able to fix the DIMM sizes. I will also try and get SIV to report the motherboard, which model is it?
I will be interested to see how SIV does on your AlphaStation 255 as I not tested SIV32A on that model, again please will you post the save files as I suspect I will need to make some small adjustments. Note you should get .TXT files from say SIV32X.zip in addition to the files from SIV32A.zip as the 5MB limit means I can't include them in SIV32A.zip.
The System has 6 x 32MB SIMMs installed.
The Motherboard is a: Chaintech 6ITM Socket 8
I'm still busy building this system and it doesn't currently have a USB header installed, it's also not currently linked to a network, so I'll get the .txt file to you when I can get it off the system!
I'm currently using a rewritable DVD-ROM to get information onto this machine, It's pretty crude, I know!!! 😀
Irinikus wrote on 2022-10-01, 18:53:The Motherboard is a: Chaintech 6ITM Socket 8
Thank you, with luck I will be able to find Chaintech 6ITM in the .DMI file and get SIV to report it. I suspect the BIOS version number might be possible, what is it? Some Socket 8 motherboards have the SMBIOS tables, but I guess 6ITM is a tad too old.
Looking at [Chipset MCH] I would have expected SIV to report 6 x 32MB and I am wondering why it does not. Once I get the save files I expect all will become clear.
The .TXT +.DMI files from my PentiumPro are only about 800KB, so if you have a floppy drive you could use that.
Irinikus wrote on 2022-10-01, 18:53:The System has 6 x 32MB SIMMs installed.
I have just upload the SIV 5.67 Beta-04s which may well fix the SIMM size reporting issue, does it?
To stand any chance of reporting the motherboard I will need the save files though.
Not sure if I shared these before: SIV public report from my Visual Workstation 540. It currently has 2 Pentium III 500 2M CPUs. I had 4 Pentium III 900 2M ES CPUs installed there before, have an old report from 5.60, it includes also private parts, but if you for some reason want it, I can share it via e-mail.
GL1zdA wrote on 2022-10-21, 23:11:I can share it via e-mail.
Thank you for the save files which are the first I have ever had from an SGI SG540. It's an unusual system as looking at [ PCI Bus ] there is no MCH or GPU!
Looking at [ SMB Bus ] I can see the Xeon PIRs, but not the DIMM SPD; Given these share the same addresses there must be an SMBus multiplexor. I also noticed an Analog Devices ADM1021 which will report the Xeon temperature, but there should be two, again the MUX may be hiding the other. I wonder what there was with 4 CPUs installed, please will you either PM or email be the other save file so I can check and then get SIV to report these temperatures?
When I looked at [ Windows ] I was surprised to see 32 CPUs are licensed rather than the usual 2, is this a special version of NT4 for SGI workstations? I also noticed it has SGI HAL, given this I guess W2K can't be used, have you tried to install W2K or later on it?
The Visual Workstation is as nonstandard as an x86 computer can be and still run Windows, that's why you might see odd things. The chipset is Silicon Graphics's custom design, it's actually a chipset from their O2 MIPS workstation adapted for Intel CPUs. Graphics is integrated in the same chip with the memory controller and uses part of RAM for its purposes. Video is also integrated into the chipset The feature they were always proud of was live video texturing: the graphics processor could directly use the memory, where video was digitized to, to get the texture. You see the theme: do as much with memory as you can without creating copies. I'm attaching a copy of SGI's paper describing the architecture of the Visual Workstation.
There are custom memory modules, basically DIMMs split in three parts, compatible only with the Visual Workstation 540 and its smaller brother, the VW 320 (Irinikus, who posted recently in this thread, has it). That's why there's probably no SPD or a nonstandard one.
As for Windows. The report was run on NT 4.0, that was installed from manufacturer provided recovery CDs. Since it was NT 4.0 Workstation, they had to customize it to allow more than 2 CPUs. This is actually trivial, the number of licensed CPUs is controlled just by one entry in SETUPREG.HIV. 32 CPUs is the maximum for NT, this workstation can use 4 at most. Windows 2000 has the necessary HAL (HALBORG.DLL) and runs on the Visual Workstation. The report with 4 CPUs was actually done on Windows 2000 Professional, I will e-mail it to you later. If you upgrade from the factory NT 4.0 installation, Windows 2000 will inherit the raised CPU limit. If you do a clean install, you have either to edit SETUPREG.HIV yourself before you do it, or do a bit more work to change it on an already installed system (it's protected like the other licensing registry entries, if you edit it, Windows will revert it - you basically have to do a backup, edit the backup and restore the edited backup).
GL1zdA wrote on 2022-10-22, 12:01:The chipset is Silicon Graphics's custom design
Thank you for all the information and the overview document. I wonder if it's possible to read the memory timings, do you know of a datasheet that lists what all the registers are and how to access them? Looking at Menu->System->ACPI + PnP->System HAL there are no Memory Addresses, only I/O Addresses so I guess these need to be used to access the SGI Memory Controller. I think I will wait and see what SIV reported on W2K as the PnP support should provide more insight. It may also be helpful to know what SIV reports for a VW 320 .
red-ray wrote on 2022-10-22, 17:28:GL1zdA wrote on 2022-10-22, 12:01:The chipset is Silicon Graphics's custom design
Thank you for all the information and the overview document. I wonder if it's possible to read the memory timings, do you know of a datasheet that lists what all the registers are and how to access them? Looking at Menu->System->ACPI + PnP->System HAL there are no Memory Addresses, only I/O Addresses, so I guess these need to be used to access the SGI Memory Controller. I think I will wait and see what SIV reported on W2K as the PnP support should provide more insight. It may also be helpful to know what SIV reports for a VW 320.
I'm not aware of any datasheets for this chipset. It was short-lived, they replaced it with generic PC workstations (the Visual Workstation 230, 330 and 550 are standard PCs), there was never much support from software for its features.
The best I can think of, what could resemble some documentation, is what was once supported in the Linux kernel. Here's the initial patch from 99 and here's the commit from 2014 removing it - things named "Cobalt" and "Lithium" look like something interesting. Cobalt is basically the memory controller and Lithium is the I/O controller.
I've e-mailed you the "full" reports from Windows 2000 with 4 CPUs.
GL1zdA wrote on 2022-10-22, 18:21:I've e-mailed you the "full" reports from Windows 2000 with 4 CPUs.
Thank you for the links and the W2K save files. When I looked at [ SMB Bus ] there were 4 x Xeon PIRs and 2 x ADM1021s which is what I expected there would be. I had a quick look at the Linux code and did not see anything relating to memory timings or an SMBus MUX 🙁
SIV did not report the GPU in W2K as there is no display interface. Given it can't report much anyway I don't feel it's worth working around this. I also noticed there are no monitors.
I have adjusted SIV to report the ADM1021s, would you like to try a test SIV now or wait 'till I release SIV 5.67 Beta-07 Wednesday or Thursday next week?
red-ray wrote on 2022-10-22, 21:17:Thank you for the links and the W2K save files. When I looked at [ SMB Bus ] there were 4 x Xeon PIRs and 2 x ADM1021s which is […]
GL1zdA wrote on 2022-10-22, 18:21:I've e-mailed you the "full" reports from Windows 2000 with 4 CPUs.
Thank you for the links and the W2K save files. When I looked at [ SMB Bus ] there were 4 x Xeon PIRs and 2 x ADM1021s which is what I expected there would be. I had a quick look at the Linux code and did not see anything relating to memory timings or an SMBus MUX 🙁
SIV did not report the GPU in W2K as there is no display interface. Given it can't report much anyway I don't feel it's worth working around this. I also noticed there are no monitors.
I have adjusted SIV to report the ADM1021s, would you like to try a test SIV now or wait 'till I release SIV 5.67 Beta-07 Wednesday or Thursday next week?
None of the tools I've tested (HWiNFO, AIDA64, CPU-Z) reported anything about the GPU, so that's not surprising. I also think it has a weird driver, as far as I can remember, Device Manager shows it with this "square gray icon" as a "non plug-and play device" and not as a graphics card, so maybe that's why there are no monitors.
I can wait for the next beta to test it.
Irinikus wrote on 2022-10-24, 11:05:Here's the SGI Visual Workstation 320
Thank you for checking out SIV then posting the save files. After looking at the save file I have made a couple of small adjustments to SIV which will be in V5.67 Beta-07. Having thought about the GPU not getting reported on W2K I plan to adjust Beta-07 so it should report it.
I noticed the speed reported by CPUZ 1.74 looks to be incorrect, I assume it should be around 1000 MHz rather than 1138.81 MHz, should it? I am also wondering if this has been fixed in CPUZ 2.03.0.x32, has it?
red-ray wrote on 2022-10-24, 20:25:I noticed the speed reported by CPUZ 1.74 looks to be incorrect, I assume it should be around 1000 MHz rather than 1138.81 MHz, should it? I am also wondering if this has been fixed in CPUZ 2.03.0.x32, has it?
Both 320 and 540 should work at 100 MHz FSB. There are no settings to change it, and I've never heard of anyone overclocking it, so It looks like a CPU-Z oddity.
On my machine, AIDA64 was the program that showed odd results. I've checked my station in 2017 with then current version 5.80, and it reported my 900 MHz Xeon running at 5x180 MHz. I think this was because the motherboard was jumpered to 5x, because Tanner Xeons required it, and I've run 500 MHz Xeons in this station before. I didn't change the settings for the 900 MHz CPU, since it would override any setting to 9x anyway, but it looks like it was somehow still "visible" to AIDA. AIDA64 reported the "factory" speed for this CPU as 1200 MHz (9x 133 MHz), but these parts operated at 100 MHz.
