• Happy holidays, folks! Thanks to each and every one of you for being part of the Tom's Hardware community!

[SOLVED] OC noob, overkill rig, can’t POST with XMP

Jul 29, 2020
3
0
10
I’ve been wanting an upgrade for years and finally I could justify it. I built my dream machine:
Asus Z490-E Strix
i9-10900k
64gb G.Skill 3600 18Q (16x4)
ROG Thor 1200w PSU
ROG Ryujin AIO
EVGA GTX 1080 (to be upgrade this Fall)
Now I know a lot of it is way overkill, but the goal here was “I want it, I get it” as a kind of Treat-Yo’-Self for various reasons.

I’m having trouble with overclocking it, obviously because I have no idea what I’m doing. I’m trying to do my research as best as I can, reading guides and trying to understand how to properly leverage the Asus AI overclocking since I figure that’s my best shot at performance:safety ratio with the level I’m at.

Whenever I enable XMP, there is one profile, and it sets the DRAM frequency to 3600 MHz, and I’m thinking that’s the problem. I changed that alone and it seemed to run with a stable over clock, but then I subjected it to high and heavy load with a game that I was playing, and it began to crash every 10 to 15 minutes. I undid the overclock, and it ran like a breeze, with some minor frame rate drop due to the performance cut back.

So that’s where I’m at, an overkill computer in the hands of an absolute beginner, unsure what questions exactly to ask. What I know is that I am unable to set the target DRAM frequency to what I should be able to reach, even with the voltage set to 1.35v like it says on the box. I’m not 100% I’m actually overclocking the cpu either, but it did turbo up to 5.1 GHz or so on a few cores during initial benchmarks, although BIOS advertises target is 5300 MHz so ¯\(ツ)
Any advice, things to try, or articles to read?
 
Solution
4 DIMM’s is equally fast if not faster but the chances of these types of issues is increased as it puts more load on the memory controller.

If the memory is not on the motherboard QVL list that also increases the chance of issues.

From what I’ve read DDR4 is safe up to 1.45v and a few reviews even claimed it’s ok up to 1.50v but 1.45v seems the more commonly accepted safe limit. However do your own research as with any voltage increase the safe limit depends on the individuals perception of risk. Anyway try 1.40v, if it’s stable then try lowering 0.01v at a time to find the minimum stable voltage. If it’s not stable try increasing 0.01v at a time until you hit 1.45v or even 1.5v if you want to push it that far.
Your 16GBx4 RAM did it come from a single matched RAM kit or was more than 1 kit used? If it was from a single kit was the kit listed on the motherboard QVL list?

Running 4 DIMM’s is more likely to have problems running at higher speeds. You can try increasing the DRAM voltage as this might help.
 
It was a single kit, yea. I wasn’t aware that 4 sticks wasn’t as good as 2 for individual speeds. I would assume however, if they sell it in a pick of 4, and they market that pack of 4 as 3600 MHz, it should still be able to hit that? I also didn’t know I could up the voltage a bit more. I was worried about damaging it, so I’ll look into that. I know it also says on the box 1.35v, so I figure I probably shouldn’t need to overvolt it to hit that, right?
Thank you for your reply and insight!
 
4 DIMM’s is equally fast if not faster but the chances of these types of issues is increased as it puts more load on the memory controller.

If the memory is not on the motherboard QVL list that also increases the chance of issues.

From what I’ve read DDR4 is safe up to 1.45v and a few reviews even claimed it’s ok up to 1.50v but 1.45v seems the more commonly accepted safe limit. However do your own research as with any voltage increase the safe limit depends on the individuals perception of risk. Anyway try 1.40v, if it’s stable then try lowering 0.01v at a time to find the minimum stable voltage. If it’s not stable try increasing 0.01v at a time until you hit 1.45v or even 1.5v if you want to push it that far.
 
Solution
Cool, I got a second source on the voltage targets you proposed and tried 1.4, and then 1.45v and it started up at 1.45. I will try the lowering by .01v method, but how would I go about testing it? Is it a situation or ‘if it starts it works’? Or is it a stress/load kind of test? I know memtest86 is proposed a lot, and I’m running a Windows memory diagnostic right now (god save my poor free time as this thing proceeds to take an hour x_x). I’d rather not use memtest86, as I try to stick with OEM software as much as possible. What do you think?
 
The voltage is too high at 1.4v. You should be able to get a good 5GHz all core overclock at around 1.3v. Here is a video that might help as the BIOS for ASUS motherboards are all the same.

Start with a baseline and make sure temps are okay and in stress testing never in the 90's degrees up to 80's degrees is okay.

Just use Prime95 (Blend) Non AVX/Aida64 Extreme test and Cinebench R20 (which has AVX) and do some runs. In CinebenchR20 check scores against other 10900K's so that you know you are in the right ball park.

View: https://www.youtube.com/watch?v=_67m2pz7Emo


View: https://www.youtube.com/watch?v=vzKJNuctGQo


View: https://www.youtube.com/watch?v=Rqi_EO5f6KU
 
The voltage is too high at 1.4v. You should be able to get a good 5GHz all core overclock at around 1.3v. Here is a video that might help as the BIOS for ASUS motherboards are all the same.

Start with a baseline and make sure temps are okay and in stress testing never in the 90's degrees up to 80's degrees is okay.

Just use Prime95 (Blend) Non AVX/Aida64 Extreme test and Cinebench R20 (which has AVX) and do some runs. In CinebenchR20 check scores against other 10900K's so that you know you are in the right ball park.

View: https://www.youtube.com/watch?v=_67m2pz7Emo


View: https://www.youtube.com/watch?v=vzKJNuctGQo


View: https://www.youtube.com/watch?v=Rqi_EO5f6KU
We are talking about the DRAM voltage and not CPU.
 
Cool, I got a second source on the voltage targets you proposed and tried 1.4, and then 1.45v and it started up at 1.45. I will try the lowering by .01v method, but how would I go about testing it? Is it a situation or ‘if it starts it works’? Or is it a stress/load kind of test? I know memtest86 is proposed a lot, and I’m running a Windows memory diagnostic right now (god save my poor free time as this thing proceeds to take an hour x_x). I’d rather not use memtest86, as I try to stick with OEM software as much as possible. What do you think?
I keep stress testing to a minimum. From my experience stress testing is ok as an initial indication of stability but the real test is using the system. I’ve had systems pass all stress tests and then fail under real world usage. Memtest86 is a good initial test.