HWiNFO v8.01-5415 Beta released

[Martin]

HWiNFO v8.01-5415 Beta available.

Changes:

• Added workaround for Windows Defender messing up Effective Clock and C0 Residency counters.
• Added improvements for Auto Start.
• Enhanced support and monitoring of Moore Threads GPUs.
• Updated PresentMon to v2.0.0.

 

[VictorVG]

I don’t like something, but with the errors of the RAM controller it is clear - you need to complete the experiment with the RAM and change LRDIMM in one of the channels so that the pairs are the same iMB, and not different as it is now.

The phenomenon gets out, provided that we launch the score at the same time using CPU and GPU, for example, video score in Avidemux v2.8.2.24076 (nightie 03/16/2024) X64 use hardware codec "Nvidia H.264" with the filter "A.I. Enchance (programmatic)" task "Two -fold increase in FSRCNN with noise reduction "and such things are strange because there is no operation of trottling sensors and you can assume programmatic errors of reading the state of CPU control circuits.

The phenomenon is observed with v7.73-5390.

 

[Martin]

I don’t like something, but with the errors of the RAM controller it is clear - you need to complete the experiment with the RAM and change LRDIMM in one of the channels so that the pairs are the same iMB, and not different as it is now.

The phenomenon gets out, provided that we launch the score at the same time using CPU and GPU, for example, video score in Avidemux v2.8.2.24076 (nightie 03/16/2024) X64 use hardware codec "Nvidia H.264" with the filter "A.I. Enchance (programmatic)" task "Two -fold increase in FSRCNN with noise reduction "and such things are strange because there is no operation of trottling sensors and you can assume programmatic errors of reading the state of CPU control circuits.

The phenomenon is observed with v7.73-5390.

Maybe running some other tool along with HWiNFO that might be resetting the throttling indicators?

 

[VictorVG]

No, I use only HWiNFO64 for hardware monitoring - other tool don't needed.
Most of all, under suspicion, the possible errors of the RAM controller because In one of the banks there are LRDIMM memory modules with different iMB (Isolated Memory Buffer):

Memory: 131072 MBYTES @ 933 MHZ, 13-13-13-32
DIMM0 - 32768 MB PC14900 DDR3 SDRAM - Samsung M386B4G70DM0-CMA4
DIMM1 - 32768 MB PC14900 DDR3 SDRAM - Samsung M386B4G70DM0-CMA3
DIMM2 - 32768 MB PC14900 DDR3 SDRAM - Samsung M386B4G70DM0-CMA4
DIMM3 - 32768 MB PC14900 DDR3 SDRAM - Samsung M386B4G70DM0-CMA4

The manufacturer and brand of the iMB chip is the only difference between the Samsung M386B4G70DM0-CMA1/-CMA2/-CMA3/-CMA4 LRDIMM model's. I had previously refused -CMA4, ordered a couple of the same to replace, and a couple -CMA3 came. So I put one on a sample. It will have to put a pair of -cma3 in one bank, and in the other -CMA4 because The Xeon E5-2697v2 has two independent two-channel RAM controller operating in four channel mode and one controller now works with errors due to various iMB microcircuits, and I have no spare set of four identical LRDIMM.