Can I Get Some Help Overclocking My RAM?

[jordan1794]

I'm sorry to post a thread like this, I feel like I'm asking a lot...but I'd really like to overclock my RAM...
I did some research, and thought I knew what I was doing...but apparently not...

I spent 3 days working on frequencies and timings and was able to pass Memtest86+
But as soon as I booted up a game I froze...And the same thing happened with any of the settings that I passed Memtest86+ with.

So apparently that was 3 days wasted...


This is the RAM I have:
http://www.newegg.com/Product/Produ...-na-_-na&cm_sp=&AID=10446076&PID=3938566&SID=

With Memtest86+ I was able to boot and pass with 1.65 DRAM voltage, +.150 to my system agent voltage(ring voltage is 1.05, so the offset made my system voltage 1.2) with a frequency all the way up to 2600 MHz (and timings of 14-14-14-36)
I tested 2200 and 2400 first, so when I got to 2600 I became suspicious and tried to run a game(and Prime95) and both froze...that was when I realized that Memtest86+ was not stressing my system the way I wanted.



So what did I do wrong? Are there additional settings I need to change, or did I change something incorrectly?
Or does Memtest86+ just not work for stressing RAM? When I changed timings I would either fail to boot, or I would pass...nothing would create errors EXCEPT if my CAS was too low, the other timings would just fail to boot if they were too low...


I'm sorry I'm a mess, I was so frustrated after this happened I just dropped it for a whole month...now I want to try to tackle it again...but I obviously need more information, because something isn't working.

(default ram voltage is 1.5, so surely 1.65 should get me to AT LEAST 2200 or 2400 MHz, or at bare minimal some reduced timings on 2133 MHz...right?)

 

[jordan1794]

Other info...
CPU (i7 4790K) overclock settings:
Cpu Multiplier 48
Vcore 1.265
Cpu uncore 44
Vring 1.05 (stock, surprisingly)

How do I test RAM for stability, if not with Memtest86+?
What settings should I put Prime95 on, and how long do I need to run it?

So many questions, I am sorry!

 

[ZenZimZaliben]

Overclocking RAM can be a frustrating time. There are so many settings and you have to test in smaller increments. Jumping 200Mhz at a time is to big of a leap and you will need to losen timings as your go up in MHz, and not just CAS. As an example...

• DDR3 2133 at 11-13-13-31 /1.5v

• DDR3 2200 at 11-15-15-35 /1.52v

• DDR3 2300 at 12-16-16-39 /1.55v
  
  The crap part of all this is even after gaining an additional 20% in Mhz increase you have lowered your latencies so much that the real world gains are like +3%.
  
  Back in the DDR/DDR2 days overclocking ram did yield some big performance improvements. With DDR3/DDR4 the bandwidth has become so large that a small gain in Mhz with the resulting loss in latency to maintain stability tends to make overclocking modern RAM a rather pointless trial of frustration..
  
  Since you have a K chip, with full multiplier control, there really is no reason to push the Ram.

 

[st2000]

better test 1866 cl9(or lower if possible)
i think result will suprprise you

 

[jordan1794]

Overclocking RAM can be a frustrating time. There are so many settings and you have to test in smaller increments. Jumping 200Mhz at a time is to big of a leap and you will need to losen timings as your go up in MHz, and not just CAS. As an example...

• DDR3 2133 at 11-13-13-31 /1.5v

• DDR3 2200 at 11-15-15-35 /1.52v

• DDR3 2300 at 12-16-16-39 /1.55v
  
  The crap part of all this is even after gaining an additional 20% in Mhz increase you have lowered your latencies so much that the real world gains are like +3%.
  
  Back in the DDR/DDR2 days overclocking ram did yield some big performance improvements. With DDR3/DDR4 the bandwidth has become so large that a small gain in Mhz with the resulting loss in latency to maintain stability tends to make overclocking modern RAM a rather pointless trial of frustration..
  
  Since you have a K chip, with full multiplier control, there really is no reason to push the Ram.

Well, when I had it running at 2600 MHz I saw vast improvements in my desktop loading speed(Icons popped up immediately, whereas with 2133 they take up to 2-3 seconds.
These are my starting attempts/best attemps while passing Memtest86+ (obviously doesn't mean much now)
2133 @ 11-13-13-31 (stock) / 2133 @ 10-11-11-31
2200 @ 10-12-13-35 / 2200 @ 10-11-12-31
2400 @ 11-15-15-35 / 2400 @ 10-12-13-35
2600@ 14-14-14-36 / (never attempted to tighten timings)

With the best attempts on the right I could boot, run memtest, get into windows without issues...but if any of the timings were lower I would not boot.


Also, I tried 1866 with a CAS of 9 and I got no boot.
It appears 10 is the absolute lowest CAS I can have at any frequency and still boot.

 

[jordan1794]

I think my biggest problem is that I don't know how to test for stability...Can I use Prime95 for this, and if so...what settings should I use?

 

[st2000]

It appears 10 is the absolute lowest CAS I can have at any frequency and still boot.

it's sad

Well, when I had it running at 2600 MHz I saw vast improvements in my desktop loading speed(Icons popped up immediately, whereas with 2133 they take up to 2-3 seconds.

if so - you got your benefits
just relax, any hardware got it's limits

 

[jordan1794]

I get that, and I accept it fully.
It's just that I prioritized RAM with a voltage of 1.5 specifically so that I could have room to overclock it...I am baffled that even an increase to the max suggested of 1.65 did not gain me anything...
:/

 

[buildzoid]

Well, when I had it running at 2600 MHz I saw vast improvements in my desktop loading speed(Icons popped up immediately, whereas with 2133 they take up to 2-3 seconds.
These are my starting attempts/best attemps while passing Memtest86+ (obviously doesn't mean much now)
2133 @ 11-13-13-31 (stock) / 2133 @ 10-11-11-31
2200 @ 10-12-13-35 / 2200 @ 10-11-12-31
2400 @ 11-15-15-35 / 2400 @ 10-12-13-35
2600@ 14-14-14-36 / (never attempted to tighten timings)

With the best attempts on the right I could boot, run memtest, get into windows without issues...but if any of the timings were lower I would not boot.


Also, I tried 1866 with a CAS of 9 and I got no boot.
It appears 10 is the absolute lowest CAS I can have at any frequency and still boot.

I think that for 2600 you should be looking at 13-15-15-36 maybe 14-15-15-36. 14-14-14-36 doesn't make sense if you did 2400 on 11-15-15-35

AFAIK the second and third primary timings don't scale with voltage only CL does. You might also want to look into fine tuning your secondaries and tertiaries

 

[jordan1794]

See it comes back to the problem of how to test it...
I thought for sure that Memtest86+ was all I needed, but apparently not...

If I can't test for stability, then I am playing with fire...and RAM instability can very easily lead to corruption of my games files and what-not...which would lead to even more head-aches.

 

[buildzoid]

See it comes back to the problem of how to test it...
I thought for sure that Memtest86+ was all I needed, but apparently not...

If I can't test for stability, then I am playing with fire...and RAM instability can very easily lead to corruption of my games files and what-not...which would lead to even more head-aches.

I test RAM by running IBT set to use all the available RAM and a couple cores(not all of them) for several hours. If that doesn't get an error I just go with it as stable and so far haven't had issues with this method.

 

[jordan1794]

Also, I extrapolated the data to predict the lowest stable timings.
Based on what timings would/would not boot I came up with the following chart:
(Did not bother with the 4th timing for this part)
Observed:
2133 @ 9-11-11
2200 @ 10-11-12
2400 @ 10-12-13
Predicted, based on the above
2600 @ 14-14-14
2666 @ 15-15-15
2800 @ 19-18-16
2933 @ 21-21-17
3000 @ 24-24-18

Obviously I did not expect to hit anything above 2600 though.


And I'll check into using IBT for my stability, thanks for the suggestion.

EDIT: I have 8 core (well, 4...depending on how you look at it)
With IBT what would you suggest to set it to use, 2 or 4?

 

[jordan1794]

Ok so running IBT on standard with my current overclock and just regular XMP enabled I failed about 3 seconds in haha.
CPU temperature only got to 55-60 before my computer shut-off and restarted.

I'm thinking perhaps my vring is too low...I did not have to increase it to get to 4.4 GHz on my uncore, but that was passing Prime95 Small FFT's. Now that I think about it, the FFT's might not even touch the uncore stability.

Huh, maybe my RAM CAN run 2600 MHz,and it's my uncore overclock that is causing the issue?

Time to test!

 

[ZenZimZaliben]

Well, when I had it running at 2600 MHz I saw vast improvements in my desktop loading speed(Icons popped up immediately, whereas with 2133 they take up to 2-3 seconds.
These are my starting attempts/best attemps while passing Memtest86+ (obviously doesn't mean much now)
2133 @ 11-13-13-31 (stock) / 2133 @ 10-11-11-31
2200 @ 10-12-13-35 / 2200 @ 10-11-12-31
2400 @ 11-15-15-35 / 2400 @ 10-12-13-35
2600@ 14-14-14-36 / (never attempted to tighten timings)

With the best attempts on the right I could boot, run memtest, get into windows without issues...but if any of the timings were lower I would not boot.


Also, I tried 1866 with a CAS of 9 and I got no boot.
It appears 10 is the absolute lowest CAS I can have at any frequency and still boot.

Benchmark your system as you normally run it. Then once you get ram OC'd and stable bench again. You will at most see a 3%-5% increase in games/apps. So in games you might go from 30FPS to 32FPS. There are benchmarks all over the web that prove this. Just google "DDR3 Benchmark results" - http://www.tomshardware.com/reviews/dram-benchmark-fluctuations,4080.html

 

[jordan1794]

So an update...
I can't run IBT on standard even with my uncore set at stock...
But temperatures again never reached 60 C before my computer shut-down...

Maybe my cpu overclock is not stable? I don't know at this point!

 

[jordan1794]

Ok, so I must be doing something wrong...
I crash/shut down about 5 seconds into IBT even on stock settings...
Temps never exceed 60C before hand...

What's going on? I'm so lost DX

 

[buildzoid]

Ok, so I must be doing something wrong...
I crash/shut down about 5 seconds into IBT even on stock settings...
Temps never exceed 60C before hand...

What's going on? I'm so lost DX

Try lower RAM usage the OS might not be OK with IBT eating all the RAM.

 

[jordan1794]

I was using standard, which displayed 1024 MB as the RAM...but it was grayed out so I dunno...

Let me try manually setting it, and I'll also set it to use 2 cores as well.

 

[jordan1794]

Ok so manually setting 1024 MB of RAM and setting it to use 2 cores it did not freeze.
I guess now I'll test 4 cores, then 8 to see how my temps and stability are with IBT compared to Prime
(With Prime my temperatures get to 79 on my hottest core)

 

[newtekie1]

I think my biggest problem is that I don't know how to test for stability...Can I use Prime95 for this, and if so...what settings should I use?

Memtest86, but you have to let it run for a long while. Just letting it go through one pass is not good enough. Minimum of 6 hours, I prefer 12-24.

 

[jordan1794]

Ok so I am hitting shut-down temperature with IBT, it was just happening so fast that RealTemp wouldn't display it before it shut-down.
Running with 2 cores I had temperatures around 75
4 core it went to around 82
6 core i saw it hit 85 before the computer shut-down.

I'm not sure where to go from here though, my OC is rock-solid for everything else (including Prime95)
I'm lost

 

[newtekie1]

IBT beats the processor up like no real load ever will. Still, if you are hitting shutdown temp with it your pushing your cooler too much. Time to back down voltages and maybe clocks.

 

[jordan1794]

Honestly, I'm not too worried about my IBT temperatures...Prime95 I only get to 79 and real world playing games like Battlefield and doing long video encoding(1 hour+) I never see temperature rise above 60...even with 100% load and running many programs simultaneously.

Guess I'll probably just abandon my RAM overclocking endeavor...
Thanks for the suggestions and help though, at least I have more knowledge than when I started

 

[buildzoid]

Honestly, I'm not too worried about my IBT temperatures...Prime95 I only get to 79 and real world playing games like Battlefield and doing long video encoding(1 hour+) I never see temperature rise above 60...even with 100% load and running many programs simultaneously.

Guess I'll probably just abandon my RAM overclocking endeavor...
Thanks for the suggestions and help though, at least I have more knowledge than when I started

That's why I said not to run IBT on all cores when you using it to mem test. Run it on 1 or 2 cores but using most of your RAM. That way it won't overheat but will still check your RAM and IMC stability.

 

[MxPhenom 216]

@cadaveca is the one to ask for this sort of stuff, but depending on the quality of your chip, you may have to drop your CPU clock a bit to run a higher stable memory clock.