Jump to content

Memory overclocking and stress test questions

A week orso ago I made THIS post asking about what ram kit would be better for my pc, tldr: On recommendation by @RONOTHAN## I kept my current ram kit.

With that said, I have a few questions regarding overclocking, voltages, die density? and stress tests as this is getting interesting to me.

INFO:
The ram kit I use: DDR4 Corsair Vengance LPX 32GB(2x16) 3200MHz CL16 (16-18-18-36) (CMK32GX4M2E3200C16)
I've ran this Thaiphoon Burner program to find out what ''die'' I had, results said: A-die
I'm currently testing 3600mhz, I have set the timings manually to 16-18-18-36(stock/xmp), (when I left these on auto my motherboard would set lower timings by it self).
Bios updated to the latest F22 (motherbard listed below)

Questions:
1. How good is ''A-die''? and how does it compare to others like ''B-die'' etc?
2. What are considered safe VCCIO & VCCSA voltages? (I currently have both on 1.3V)
3. Beside lottery being a factor, what do you think i can/should be able to achieve with this kit?
4. MHz or lower timings? Obviously you want both but should i focus on getting lower timings or higher mhz? (mainly use this pc for gaming/editing)
5. I'm running TestMem5 with the Extreme 1 @anta777 profile for testing, 1. is it good? 2. how long do i run it for? 3. how many cycles should i run? 4. i should aim for 0 errors right?  5. should i run something different then the anta777 profile intead?
6. memtest86, should i use this for stability testing? if yes are 4 passes enough? (i downloaded memtest86, put it on a USB and it only does 4 passes).
7. if i pass either TestMem5 or Memtest86 does this mean I achieved stability i can rely on on a daily basis?
8. do i run both these to be sure my ram OC is stable? or should i run different tests?

I know it's allot of question but I would appriciate it very much if you guys could answer these.

I'm finding out fast how time consuming ram overclocking is, assuming changing one setting at the time is the way to go.

Any tips are very much welcome!
Thx in advance

PC Specs:
CPU I7-12700KF P-cores: 5.1ghz/E-cores:3.8ghz 1.29V LLC:Turbo
MOBO Gigabyte Z690 Gaming X DDR4
COOLING Noctua NH-D15
RAM Corsair Vengeance 32GB 2x16 3200mhz CL16
GPU Asus RTX 3080 TUF Gaming OC (undervolted at 0.925v at 1980MHz)(+300 Memory clock)
PSU EVGA 850 G2
CASE Fractal design Torrent white TG
SSD Samsung 850 EVO 250GB
Windows 10 version:21H2

Recent build: Fractal Design - Torrent reviewMeshify C / The 1080TI Strix Noctua modDefine S X58 Xeon build  / Specs: i7-14700KF 5.8Ghz - ASUS TUF RTX 4080 super - G.Skill Ripjaws 32GB 4000mhz CL18 -  Gigabyte Z690 Gaming X d4 - Torrent Fractal Design white - EVGA 850W Supernova G2 80+ Gold - Noctua D15

Link to comment
Share on other sites

Link to post
Share on other sites

5 minutes ago, Wildgg said:

1. How good is ''A-die''? and how does it compare to others like ''B-die'' etc?

Depends on who make it and what its density is. 8Gb Hynix A die (also refered to as AFR) is notoriously terrible, but say their 16Gb variant of it (IIRC called AJR) from what I've seen isn't quite as bad, though it's still not great. 

 

8 minutes ago, Wildgg said:

2. What are considered safe VCCIO & VCCSA voltages? (I currently have both on 1.3V)

Below about 1.45V on both. Just saying, 12th gen doesn't have a VCCIO, they have a VDDQ TX voltage, which does behave a bit differently. There is debate on what is safe for the VDDQ TX voltage, some say 1.55V, others say 1.45V. I'm personally more in the 1.45V camp, especially since on the two boards I've used there's been no reason to raise it above 1.4V in general as raising it above that doesn't help stability (both MSI boards on DDR4 and DDR5 with my particular 13700K), but ASUS boards IIRC like that voltage really high so just keep that in mind you might need to run 1.5V or higher for full stability. Just remember that they both like to sweet spot, so odds are your fully stable settings will be using something below that aforementioned 1.45V mark. 

 

12 minutes ago, Wildgg said:

3. Beside lottery being a factor, what do you think i can/should be able to achieve with this kit?

Again, depends on who made it. 8Gb AFR you'll be lucky to get 3600 CL18 stable, Samsung A die the same thing, and 16Gb AJR should be able to do something like 4000 CL18 IIRC. 

 

15 minutes ago, Wildgg said:

4. MHz or lower timings? Obviously you want both but should i focus on getting lower timings or higher mhz? (mainly use this pc for gaming/editing)

MHz* for one main reason: every variant of A die I'm aware of does not really do good memory timings, so it's not like you're gonna get lower timings anyway. You do want to try and get it clocked as high as you can, but you also want to make sure that you are still in Gear 1, so odds are you're gonna clock your memory up until you hit 4200MT/s (if you can get there in the first place), then start tightening the timings as far as you can. 

 

19 minutes ago, Wildgg said:

5. I'm running TestMem5 with the Extreme 1 @anta777 profile for testing, 1. is it good? 2. how long do i run it for? 3. how many cycles should i run? 4. i should aim for 0 errors right?  5. should i run something different then the anta777 profile intead?

  1. Yes
  2. As long as it takes, it'll stop on its own
  3. IIRC it should run 4 loops, but it tends to change time to time for whatever reason in my experience. 
  4. Yes
  5. Running multiple stress tests isn't a bad idea, though if your just gonna run one TestMem5 is not a bad pick. Personally I do TestMem5, HCI Memtest to 700%, and Y Cruncher VST for 2 hours, but as 5 different people and you'll get 5 different answers on what stress tests to run. 

 

22 minutes ago, Wildgg said:

6. memtest86, should i use this for stability testing? if yes are 4 passes enough? (i downloaded memtest86, put it on a USB and it only does 4 passes).

It's not a bad idea, I don't personally use it since it's a lot less convenient than the other aforementioned stress tests, but it's not a bad stress tests, and 4 passes should be fine. 

 

24 minutes ago, Wildgg said:

7. if i pass either TestMem5 or Memtest86 does this mean I achieved stability i can rely on on a daily basis?
8. do i run both these to be sure my ram OC is stable? or should i run different tests?

Yeah, I'd be comfortable dailying that OC if they both pass on everything except Intel DDR5, as that tends to run into memory controller instability at high frequency (anything above 6600MT/s), where you really do want to run VST to make sure the memory controller doesn't pass. 

 

26 minutes ago, Wildgg said:

I'm finding out fast how time consuming ram overclocking is, assuming changing one setting at the time is the way to go.

One thing to help you out, look around for a program called "ASUS MemTweakIt", it should be available on the HWBot forums and some other places (unfortunately not from ASUS directly). It lets you change almost every memory timing right in Windows instantly, so you can save a lot of time on reboots. Generally it cuts the time it takes me to dial in an overclock from about 4-6 hours for a basic one before fighting through stability testing to about 1 hour before fighting through stability testing since it's so much easier to find the limits of timings and set them in the BIOS rather than having to reboot every other minute. MSI Dragonball also does this if you prefer the GUI of that program instead, though MemTweakIt does give a bit more timings so I personally prefer using it instead. 

Link to comment
Share on other sites

Link to post
Share on other sites

8 minutes ago, RONOTHAN## said:

One thing to help you out, look around for a program called "ASUS MemTweakIt"

Thank you VERY much for answering all that, i really appreciate it! 

quick few questions on your answers:
1. so when i use asus memtweakit, lets say i find my self in a situation where my pc locks up or crashes, since it's software based it's just a simple reboot and try again right? doing everything in bios got me a bit ''worried'' as i'm probably have to reset cmos a few times when failing overclocks.
2. you mention at question/answer 4: MHz over timings, so currently i'm testing 16-18-18-36 3600MHz, should i abandon cl16 and go for cl18 timings and higher mhz into timing tuning? if Yes what would be a good start for cl18? 18-22-22-42? 
3.so right now I'm not sure what i should do, do i keep upping the MHz at cl16, or do i start somewhere else like cl18 and get possible higher clocks?

Recent build: Fractal Design - Torrent reviewMeshify C / The 1080TI Strix Noctua modDefine S X58 Xeon build  / Specs: i7-14700KF 5.8Ghz - ASUS TUF RTX 4080 super - G.Skill Ripjaws 32GB 4000mhz CL18 -  Gigabyte Z690 Gaming X d4 - Torrent Fractal Design white - EVGA 850W Supernova G2 80+ Gold - Noctua D15

Link to comment
Share on other sites

Link to post
Share on other sites

27 minutes ago, Wildgg said:

1. so when i use asus memtweakit, lets say i find my self in a situation where my pc locks up or crashes, since it's software based it's just a simple reboot and try again right? doing everything in bios got me a bit ''worried'' as i'm probably have to reset cmos a few times when failing overclocks.

Pretty much. Just reboot and all your timings will be reset. You will want to eventually set them in the BIOS, but it's a quick way to do it. Just a warning that there a couple timings that MemTweakIt can't adjust, mainly tCL (you can adjust it but it's completely bugged out and won't actually apply), tCWL, command rate, and the RTLs (though the RTLs should only really be messed with for the advanced overclockers, not on a first time as they're super annoying), so those you have to do in the BIOS, but everything else can be done in the OS. 

 

28 minutes ago, Wildgg said:

2. you mention at question/answer 4: MHz over timings, so currently i'm testing 16-18-18-36 3600MHz, should i abandon cl16 and go for cl18 timings and higher mhz into timing tuning? if Yes what would be a good start for cl18? 18-22-22-42? 

Yeah, abandon CL16 and go for as high as you can. I usually set something like CL20-30-30-50 when trying to get the max frequency my memory controller can handle, except with B die where I do 20-20-20-28. Worry about the timings later, just try and get as good a frequency as you can that's somewhat stable and work from there. You might get CL16 or CL17 in the end, but 

Link to comment
Share on other sites

Link to post
Share on other sites

8 minutes ago, RONOTHAN## said:

Pretty much. Just reboot and all your timings will be reset. You will want to eventually set them in the BIOS, but it's a quick way to do it. Just a warning that there a couple timings that MemTweakIt can't adjust, mainly tCL (you can adjust it but it's completely bugged out and won't actually apply), tCWL, command rate, and the RTLs (though the RTLs should only really be messed with for the advanced overclockers, not on a first time as they're super annoying), so those you have to do in the BIOS, but everything else can be done in the OS. 

 

Yeah, abandon CL16 and go for as high as you can. I usually set something like CL20-30-30-50 when trying to get the max frequency my memory controller can handle, except with B die where I do 20-20-20-28. Worry about the timings later, just try and get as good a frequency as you can that's somewhat stable and work from there. You might get CL16 or CL17 in the end, but 

Thank you very much again!

one more question, i assume i should lock/set the timings of (cl20-30-30-50) in bios and use that as the 'baseline' correct? so when ever i find stability issues in MSI Dragonball it will reboot back to the timings that were set in bios.

(i assume leaving timings on auto in bios will cause issues)

Recent build: Fractal Design - Torrent reviewMeshify C / The 1080TI Strix Noctua modDefine S X58 Xeon build  / Specs: i7-14700KF 5.8Ghz - ASUS TUF RTX 4080 super - G.Skill Ripjaws 32GB 4000mhz CL18 -  Gigabyte Z690 Gaming X d4 - Torrent Fractal Design white - EVGA 850W Supernova G2 80+ Gold - Noctua D15

Link to comment
Share on other sites

Link to post
Share on other sites

6 minutes ago, Wildgg said:

one more question, i assume i should lock/set the timings of (cl20-30-30-50) in bios and use that as the 'baseline' correct? so when ever i find stability issues in MSI Dragonball it will reboot back to the timings that were set in bios.

Pretty much. 

 

7 minutes ago, Wildgg said:

(i assume leaving timings on auto in bios will cause issues)

No, it shouldn't. I do tend to at least set the primaries, but it should train them loose automatically and not really matter one way or another. 

Link to comment
Share on other sites

Link to post
Share on other sites

2 minutes ago, RONOTHAN## said:

Pretty much. 

 

No, it shouldn't. I do tend to at least set the primaries, but it should train them loose automatically and not really matter one way or another. 

gotcha thx!

I'll report back when ever i find something stable!

Recent build: Fractal Design - Torrent reviewMeshify C / The 1080TI Strix Noctua modDefine S X58 Xeon build  / Specs: i7-14700KF 5.8Ghz - ASUS TUF RTX 4080 super - G.Skill Ripjaws 32GB 4000mhz CL18 -  Gigabyte Z690 Gaming X d4 - Torrent Fractal Design white - EVGA 850W Supernova G2 80+ Gold - Noctua D15

Link to comment
Share on other sites

Link to post
Share on other sites

On 1/18/2023 at 11:58 PM, RONOTHAN## said:

Pretty much. 

 

No, it shouldn't. I do tend to at least set the primaries, but it should train them loose automatically and not really matter one way or another. 

I spend a allot of time trying many options and long story short, i can run the timings my kit came with at 3600mhz instead of 3200, decent boost.
I tried your way of finding out the controler limit bacially and any time i try 3800mhz it will not do it on gear 1, even with high timings.

In the end I'm still happy with a free 400mhz boost.

thanks again for your help.

Recent build: Fractal Design - Torrent reviewMeshify C / The 1080TI Strix Noctua modDefine S X58 Xeon build  / Specs: i7-14700KF 5.8Ghz - ASUS TUF RTX 4080 super - G.Skill Ripjaws 32GB 4000mhz CL18 -  Gigabyte Z690 Gaming X d4 - Torrent Fractal Design white - EVGA 850W Supernova G2 80+ Gold - Noctua D15

Link to comment
Share on other sites

Link to post
Share on other sites

  • 2 weeks later...

Little update for those who were following this.

Since my last post i couldn't get 3800mhz on gear 1 running, turns out dram voltages weren't applying correctly + new bios update was helpfull.
4100mhz was only possible on Gear 2 (+high timings) and tbh i didn't want to give up Gear 1.

Currently I'm stable at 3800mhz cl16-18-18-36, voltages: VDDQ TX & VCCSA 1.39v , dram voltage 1.4v.
Voltages above 1.4 on VDDQ seems to cause instability? to me more voltage means more stability but i'm not that experienced with ram overclocking, maybe @RONOTHAN## can comment on that.

Recent build: Fractal Design - Torrent reviewMeshify C / The 1080TI Strix Noctua modDefine S X58 Xeon build  / Specs: i7-14700KF 5.8Ghz - ASUS TUF RTX 4080 super - G.Skill Ripjaws 32GB 4000mhz CL18 -  Gigabyte Z690 Gaming X d4 - Torrent Fractal Design white - EVGA 850W Supernova G2 80+ Gold - Noctua D15

Link to comment
Share on other sites

Link to post
Share on other sites

1 hour ago, Wildgg said:

Voltages above 1.4 on VDDQ seems to cause instability? to me more voltage means more stability but i'm not that experienced with ram overclocking, maybe @RONOTHAN## can comment on that.

That's not uncommon, a lot of memory voltages, especially on LGA 1700, sweet spot, and can sweet spot depending on frequency. On my Unify-X, for instance, if I have VDDQ TX at 1.425V I can't POST at 5600MT/s, but at 7600MT/s it's the optimal value and going above or below causes me to fail Y cruncher VST. When I messed around with a B660M-A Pro for a bit, I couldn't run DRAM voltages above 1.45V in general, even on sticks rated for 1.5V because it would either fail to train or be less stable than 1.45V. 

 

Basically, more volts =/= always better. It's usually better when temps aren't too high, but especially on memory signals where you need as precise a signal as possible raising voltage can cause some unexpected behavior. 

Link to comment
Share on other sites

Link to post
Share on other sites

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now

×