F5-8400J4052G24GX2-TZ5RK GSkill

[Hopper64]

Well I was able to pick up this kit but can’t get it stable on the Apex Encore with the 14900k. Best result was 1 minute 45 seconds of ramtest. Typically errors within 7 seconds. Using XMP1 with all timings on auto. 

VDD/VDDQ 1.5V.

MCC 1.5V.

SAV Manual Mode 1.37V.

CPU Input 2.0V.

Seemed to work much better with a bump in the CPU input Voltage. Reducing the SAV down to 1.15V doesn’t help. Tried several values in between of course. Lower value seemed to help some people with the same kit. Increasing the VDD/VDDQ induces as much instability as lower voltages. Tried reducing the MCC as well. Might be the IMC on this chip though. SP for the performance cores is 112 with the MC at 83 on bios 0601. Suggestions welcome. Continuing to tinker with free time. Thanks.

[iBruceypoo]

Simple XMP with MCE disabled is not stable?

I'd reduce variables and use your "known good" 13900KS for initial trials. Testing 3new components at this level can be complicated.

If you have your Apex Vanilla and 13900KS still setup - I'd mount the 8400 mem kit in that testbed, Braegnok is using his Apex white and getting good results with the 8400 kit - PM Sir Braegnok he's a fine gentleman! 😄

[Hopper64]

Yea I had considered that as well. Thanks.

[Silent_Scone]

Hi Hopper,

Maybe try the timings I used on the 8000MT kit.

1.61v VDDQ / 1.5v VDD

[Hopper64]

Thanks. The problem is that it fails even with loose XMP1 timings. My voltages are off target I guess or I have a poor IMC sample. Not sure. Appreciate your input. 🙂

Edit-Oh, just noticed your VDDQ is 1.61. I will try that when I get home from work. 😎

[Fragbert]

That is incredibly disappointing to hear - if you cannot get the RAM to work at its advertised XMP settings then full stop, it should be returned and exchanged. Per GSkill, their DDR5-8400 RAM's XMP settings are valid only on the Encore, there is no other MB on the list thus far! If GSkill stands by its XMP settings and its QVL validation process, and especially for the money paid, exchange that kit pronto. Any suggestions from support to try any kind of additional manual tweaking would be absolutely unacceptable, if we are to accept XMP and QVL as valid standards that the manufacturer supports.

[Silent_Scone]

Hi,

Can you explain how CPU variance fits into the sentiments above?

XMP is not automatic overclocking.The closer the memory timings are set to their lowest values or the higher the frequency exceeds the default specifications, the greater the likelihood that the user will need to manually fine-tune settings to ensure system stability.

Taken from GSKILL's FAQ:

[Fragbert]

Because there has to be accountability; else the whole XMP standards and QVL process collapses like a house of cards if manufacturers make excuses and pass the buck - especially if the customer is not overclocking their CPU, setting XMP with no other tweaks, and is using only parts in the manufacturers QVL.

Specifically with CPU variance, any kit that sets its XMP profile to 1.35v or under, there is simply no excuse if it doesn't work as advertised on a 12/13/14th gen system provided you are playing by their rules and QVL. For this particular instance the XMP profile is setting the voltage to 1.4v  which *may* be exceeding the recommended maximum of 1350 mv+5% (1417.5 mv) for the IMC, but if you are going to boldly proclaim and sell a RAM kit that sets itself to 1.4 or 1.45+ volts, that's on you to deal with the blowback (and probably factored into the pricing).

RAM manufacturers are absolutely selling XMP as automatic overclocking, don't kid yourself otherwise with their fine print. The OP could probably buy a GSkill DDR5-8000 kit and with diligence get it to work at 8400 at significant cost savings - but they didn't, they paid a price premium to get a kit advertised as working at DDR5-8400 right out of the box. At the very least, Gskill should either refund or better yet exchange and try again. The onus is on GSkill more than Intel or Asus, especially if we are dealing with only 1.4v on a K-class CPU.

[Silent_Scone]

Hello,

Nobody is trying to kid anyone, the information is clear in GSKILL FAQ. DRAM voltage has no bearing on the silicon variance of the CPU. Whilst we can throw our hands up and claim foul play these things are and always have been a known quantity when we run the system far out of spec (frustrating as that sometimes may be). The QVL is there to provide the best possible chance of sucess, but they can't make assurances. The UEFI allows for manual adjustment of key voltages for this exact purpose. Moreover, the signal variability between CPU samples can result in a lower or higher voltage on certain rails being needed. Meaning the auto rules can only get you so far. This is especially applicable when pushing memory beyond 8000MT as the margins are far tighter. That's why feedback is sometimes best in the form of sharing results, that way users can adjust some parameters up or down in a linear fashion. We all want to obtain the highest clocks possible, but we can't change the laws of physics when it comes to silicon variance (as nice as that would be!)

[Fragbert]

Nothing you stated has any bearing on the ability of the OP to rightfully demand an exchange - and the probability that they will get a better performing kit.

The QVL absolutely makes assurances - depending on the manufacturer, that is sometimes the only way to qualify for an exchange and/or refund.