Memory clock not matching CPU-Z's RAM frequency

franz

franz

Active Member
Hi,

Why is there a discrepancy between the memory clock reported by HWiNFO and the DRAM frequency reported by CPU-Z? My RAM sticks are running at 3466 MT/s (2 x 1733 Mhz). The value reported by CPU-Z appear accurate while the one reported by HWiNFO is significantly off:

1585851847182.png

Even loading the machine does not make the memory clock reported by HWiNFO reach its expected value.

System: AMD Ryzen Threadripper 3970X on GIGABYTE TRX40 Aorus Xtreme, RAM is HYPERX Fury RGB 4x16 GB (HX434C16FB3AK4/64).

Franz
 
What does HWiNFO show for the Bus Frequency? If it's around 91 MHz, it could explain the Memory Clock: 1585 MHz (Memory Clock) divided by 17.33 (Memory Clock Ratio) equals ~91 MHz. But I'm not sure if HWiNFO calculates the memory clock like this.

Regards
Dalai
 
Bus clock is hovering around 99 MHz, going down as low as 87 MHz and as high as 105 MHz. But Memory Clock isn't fluctuating at all, which makes me think the two are unrelated.

Franz
 
Last edited:
This indeed looks like some issue with the BCLK measurement. Can you try to restart HWiNFO a few times and note what memory clock will it show?
 
Ah, that's interesting. I restarted HWiNFO three times, here are the Memory Clock values: 1729 MHz, 1737 MHz, 1720 Mhz.

I'm assuming you do some kind of calibration at startup? Could you elaborate a bit on the process? Out of curiosity really.
 
Memory clock depends on BCLK as was already stated, so first the actual BCLK needs to be measured. This is unfortunately sometimes not precise, especially under high system load. Moreover this measurement is not done periodically to avoid some potential issues caused by interaction with the operating system. So if such an imprecise BCLK is measured, it can cause such effects.
Unfortunately there's no other way how to do this. Later Intel CPUs can report the BCLK straight without the need to do such measurements. We asked AMD to implement a similar technique, but they didn't seem to be willing to do so yet.
 
Be sure to run latest UEFI/BIOS from Gigabyte for the board, and latest Chipset drivers straight from AMD driver site (not Gigabyte) for Win.
 
Ah, if only! I'm running a custom unreleased BIOS provided by AMD engineers, until Gigabyte gets its act together and fix their TRX40 Aorus Xtreme (and while we're at it, Master) BIOS...
 
@franz I'm having weird issues with a 3960x with an Aorus Xtreme--my bus clock always shows as below 99, usually between 98.5 and 98.9. This is despite my attempt to fix it to 100 Mhz (same issue existed with auto) in the Gigabyte bios. Should I assume that this is a measurement error, or is the Gigabyte bios simply providing an incorrect bus speed?
 
It's almost surely a measurement error. You can check the BCLK frequency in the BIOS, or with tools like AIDA64:

1589016956518.png
 
Reporting of actual BCLK can slightly vary on some systems especially under higher load. Only later Intel systems offer an accurate method to measure BCLK.
 
Martin said:
Reporting of actual BCLK can slightly vary on some systems especially under higher load. Only later Intel systems offer an accurate method to measure BCLK.
Click to expand...
Any idea what causes the variation? It's true this is my first AMD system, but I haven't seen any instances online of anyone seeing low BCLK's like mine, and I've been testing this at basically minimal load and with a range of applications that are attempting to sample BCLK (presumably in slightly different ways, since they each report different numbers--CPU-Z tends to be a bit higher than HWiNFO).
 
@Fatih While this is not the right topic --nor even the right forum-- to ask this type of question, the answer is yes as long as the HX434C16FB3AK4/64 memory kit is listed as compatible in the motherboard's Qualifier Vendor List (which you can find on the motherboard's page on Asus website). I didn't check.
 
You must log in or register to reply here.
Back
Top