repaxan wrote on 2020-05-23, 20:51:

tRFC reported in the BIOS (auto timings) is 59, which is what it should be, based on the SPD.

Thank you, looking at the logs the nForce fan speed now seems stable, but it was 8 times what it should be so I have adjusted the scaling in the attached SIV64X V5.49 Fans-03, looking at [Status] does it now look to be correct?

With all the other fans they are reported via the Nuvoton W83627DHG which directly reports the RPM so I don't need the -DBGMCH log for them and just needed to ensure the names are correct.

By searching the save file for 0x3B (59) I am pretty sure I found where tRFC is reported, Fans-03 should report Row Refresh Cycle Time (tRFC) 59 on the initial screen and if you change the setting in the BIOS the updated value should be reported.

I think I have now fixed all the issues so please may I have a new save file generated after SIV has been active for >= 10 minutes so I can check before I release Beta-03?

Yup you got the tRFC right, I changed it to 60 and it reflected in SIV.

Nforce fan is within expected ranges now (seems to vary between 4700 and 5300 for no reason even in the BIOS)

repaxan wrote on 2020-05-24, 20:17:

Yup you got the tRFC right, I changed it to 60 and it reflected in SIV.

Thank you for checking tRFC and confirming the nForce fan speed range now as expected, I will plan to release Beta-03 in the next couple of days. It's a same I can't get SIV to report the CPU current and will keep looking into how to do this.

I have never come across fan reporting like the nForce fan reporting before and it's strange it's used at all as the Nuvoton W83627DHG can report five fans. I am wondering, do you have any other interesting systems?

Depends what you mean by interesting, most of my stuff is 'not retro' but in particular I do have two AMD nForce boards which I'll take a closer look at the next time I have them on the bench.

Anyway, I'm dual booting XP and 7 on my Gigabyte EP45-UD3P now, and I noticed the memory clock reported by SIV in 7 jumps around when the system is idle. I saw a low of 645 but usually it shows something in the 700s range.

When I put the system under load (prime95) it shows 800. It also shows up correctly when idling in XP (probably missing OS or driver support for some power saving stuff).

These options are enabled in the BIOS: C1E, C2/C2E, C4/C4E, TM2, EIST.

repaxan wrote on 2020-05-24, 21:36:

Depends what you mean by interesting, most of my stuff is 'not retro' but in particular I do have two AMD nForce boards which I'll take a closer look at the next time I have them on the bench.

Anyway, I'm dual booting XP and 7 on my Gigabyte EP45-UD3P now, and I noticed the memory clock reported by SIV in 7 jumps around when the system is idle. I saw a low of 645 but usually it shows something in the 700s range.

To me interesting is multi-socket systems, systems with motherboards I have not added support for ([Sensor] panel says Generic on the Voltage/VCore header line) and if SIV "get's it wrong".

Is the FSB also changing, I suspect it will be, is the TSC Counter on Menu->Hardware->CPU Detail->CPU Clocks always 3.00GHz or does it drop when the memory speed drops?

SIV does show lower FSB when idle but it's something I never heard of happening.

repaxan wrote on 2020-05-25, 02:42:

SIV does show lower FSB when idle but it's something I never heard of happening.

Thank you for the screen shots and looking at then I can see that this all stems from the TSC speed averaged over two seconds dropping from about 3.60 GHz to such as 2.83 GHz when idle when it's supposed to be invariant. Before the motherboard died used to have an ASUS P5Q PRO Turbo which also has a P45 chipset, but with an Intel Core 2 Extreme QX6700 (Kentsfield) 2.66GHz [B3] CPU and with that I did not see this effect and also don't see it on my ASUS P5WDG2 WS Pro (82975X Chipset) and the same CPU. I wonder why it's happening on your system when both on my systems also had/have Windows 7 x64.

I spotted the CPUZ reported FSB is strange, when the CPU speed is 2400.08 MHz @ x6 then the FSB speed is reported as 399.98 Mhz rather than 2400.08 ÷ 6 = 400.01 MHz.

There's a SIV bug with listing the DIMM's supported frequency, this video is of CPU-Z but shows the same problem:

https://www.youtube.com/watch?v=SHLiZJZuDao

repaxan wrote on 2020-05-30, 04:21:

There's a SIV bug with listing the DIMM's supported frequency, this video is of CPU-Z but shows the same problem:

Thank you for the report and it all comes down to what is displayed. CPUZ displays the maximum frequencies that can be used for a given CAS Latency and by default SIV does the same, but in addition SIV can switch to reporting the rounded/JEDEC frequencies. Given this what do you now think?

I feel I should warn you that Typhoonburner is poorly engineered as last time I checked I found among other issues that it fails to use the Global\Access_SMBUS.HTP.Method lock to interlock SMBus access, given this if you choose to run it I can't/won't support SIV also being active at the same time. All of SIV + AIDA64 + CPUZ + HWiNFO + OHM + many others do use Global\Access_SMBUS.HTP.Method so it should be safe to run them at the same time. To check that a program uses the locks check Menu->Help->Lock Handle when it's active.

red-ray wrote on 2020-05-30, 06:17:

Thank you for the report and it all comes down to what is displayed. CPUZ displays the maximum frequencies that can be used for a given CAS Latency and by default SIV does the same, but in addition SIV can switch to reporting the rounded/JEDEC frequencies. Given this what do you now think?

I'm not understanding - how come changing the tCL can affect the frequency?

repaxan wrote on 2020-05-30, 21:35:

red-ray wrote on 2020-05-30, 06:17:

Thank you for the report and it all comes down to what is displayed. CPUZ displays the maximum frequencies that can be used for a given CAS Latency and by default SIV does the same, but in addition SIV can switch to reporting the rounded/JEDEC frequencies. Given this what do you now think?

I'm not understanding - how come changing the tCL can affect the frequency?

I don't understand what you are asking, in general the faster the RAM is running the higher tCL is. Have you read and understood https://en.wikipedia.org/wiki/Serial_presence_detect? If you wish to know the fine detail check the JEDEC specs available via https://www.jedec.org/committees/jc-45.

I've figured it out, your tools are taking the min tCL in nanoseconds, and the supported tCL bitmask, and essentially saying, if the memory was run as fast as possible for each supported tCL value, while not going below the min cycle time, what frequency would that result in.

Interesting choice to be sure, it's *technically correct* regardless of whether it's actually going to be a usable choice from the memory divider. Definitely a matter of taste kind of thing and very non obvious.

Also now I realize why this never used to happen with DDR2, the SPD for DDR3 is totally different...

repaxan wrote on 2020-05-31, 06:35:

Interesting choice to be sure, it's technically correct

I am happy you figured it out, thinking back to when I added SPD reporting to SIV I made sure that SIV reported the same as CPUZ and some time later added the ability to show the rounded clocks.

It comes does to personnel preference so if you prefer the rounded values the add -RSCS (see [About]) as a SIV start-up qualifier on Menu->Windows->SIV Qualifiers.

The 512 bytes of DDR4 SPD is again different to the 256 byte DDR3 with DDR5 being different again and is 1024 bytes in size.

Missing fan reporting on Intel D865GBF.

The integrated GPU isn't supported on Windows 7, so I did a 'Have Disk' install of the XPDM drivers. So there might also be some weirdness related to that.

repaxan wrote on 2020-06-01, 06:41:

Missing fan reporting on Intel D865GBF.

Thank you and the issue was SIV did not have support for the SMSC EMC6D102 sensor chip, does the attached SIV32X V5.49 D102-05 test SIV do any better please?

How do the Temperatures, Fan Speeds and Voltages compare to what HWiNFO reports?

I suspect the fan names are not ideal, if you tell me what they should be I will change them for the next beta.

I guess it would also be worth looking at Menu->Machine->GPU Info to ensure we both report the same GPU information.

HWiNFO is only reporting the high 8-bits of the 12-bit (averaged 10-bit AtoD) sensors so the SIV values are likely to be slightly higher. When I made SIV32L do the same it reported the same values with the SIV32X ones being slightly higher test mode.

It's reporting fine now (aside from caveat about the least significant bits that you mentioned).

The board has 'CPU fan', 'front fan', and 'rear fan', each of which SIV got right. There's pads for a fourth header, also labeled as 'rear fan', which I assume is what shows up as 'system fan'.

repaxan wrote on 2020-06-02, 02:18:

It's reporting fine now (aside from caveat about the least significant bits that you mentioned).

Thank you, I have just released SIV 5.49 Beta-05 and attached it to SIV support for 386/486/586 class + Alpha CPUs and 3dfx + S3 + SiS + Matrox + XGI + old ATI + NVidia GPUs - Testing Help

I spotted there is a new HWiNFO, but don't know if it's been fixed to use 12-bit sensors.

Hi! Thanks for your efforts with SIV.

I just ran it in a machine I'm currently working on, and it worked mostly fine

I did find a couple of mistakes, though;
-DDR is reported as Dual Channel, its only Single Channel;
-Temperatures are mislabeled: T2 AUX should be the CPU temp;
-I have no clue what the 3rd temperature, reported as T3 CPU, is;
-12 V rail is reported with approx. -7 V.

Thanks again for your efforts!

slivercr wrote on 2020-06-03, 20:30:

-12 V rail is reported with approx. -7 V.

Thank you for checking out SIV and posting the results. I am unclear where SIV is reporting the -12 V rail as I don't think it does. Can you post a screen shot of the BIOS monitoring screen?

Different motherboards use different mappings for some of the Nuvoton W83627THF inputs so I need to adjust SIV for this. I have now added MSI MS-7135 support which should fix the sensor reporting. Are the fan names correct?

Please try the attached SIV32X V5.49 Test-06 and post new save files. Ii should report Single Channel and I also added the reporting of the AMD RV570 [Radeon X1950 Pro (PCIe)] clocks.

Which Samsung monitor do you have? At the moment there is no entry for SAM065E in MONDEVS.txt.

red-ray wrote on 2020-06-03, 22:16:

slivercr wrote on 2020-06-03, 20:30:

-12 V rail is reported with approx. -7 V.

Thank you for checking out SIV and posting the results. I am unclear where SIV is reporting the -12 V rail as I don't think it does. Can you post a screen shot of the BIOS monitoring screen?
...

No no, sorry. I typed it in a confusing way. Its not reporting the -12 V, its reporting the 12 V with approximately 6 V or 7 V less than it should. The "-" was just to denote a new item in the list.

As soon as I get back home I'll take the BIOS picture and try this new test. Thanks!

EDIT: "AUX Fan" should be "NB Fan".

EDIT2: Instead of a new post I'll just edit this one.

The changes you mentioned were correctly implemented

• I can now see the frequencies for the x1950 Pro
• RAM now shows as Single channel,
• temperatures have the correct name, although I still don't know what that high temperature is (not reported in BIOS, see attached screenshot)

Weirdly enough, as I unplugged the Geforce2 MX the 12 V rail now reports a more credible voltage, in line with what the motherboard's own monitoring software reports.

Thanks for the changes! Will continue to report if needed, and I have a couple other boards to test this on which should be more interesting for you.

EDIT 3: The monitor! Forgot to tell you the info.
Its an old Samsung TV with VGA input, model UN32C4000PD.
(Keep it around because it handles 240p signals through YPbPr quite well).