Overclocking Cache Ratio, How can you tell if its stable?

Overclocking your ring ratio can be an extremely frustrating and tricky process, this is what i've been doing for a couple years now to save myself time, I call it, the "frame chasers special"if you guys want more of these tips and tricks to help you in your frame chasing journey, join the community below

Join the community over on facebook gaming, twitch and discord to talk tech and play some games with fellow frame chasers.

[hemulen]

That Fortnite cache/memory check is brilliant. Thanks.

[TheKelz]

Man, this video needs more views. Much more informative and clear than other videos, kudos.

[dt]

One of the best OC computer ideas I have ran into on YouTube👍

[toonnut]

Something really interesting happened today. Yesterday I copied your video and raised the cache to 48 had no problems. Today I took my 3070 back to the store because I didn't think 8gb of video ram will be enough in future. So I started my computer with no gpu and it wouldn't boot. I eventually found out that it was the cache it was too high. So it basically worked fine with my 3070 but didn't boot with the Intel graphics. Hopefully this might help someone with the same problem 👍

[ribeack]

This dude saved my life. Even if I am playing with i7 7700k in 2022 this video has saved me . Dude you are a genius!!!

[noxxphox]

Insane test technique. I am overclocking my eam at the moment, after I saw your Warzone ram speed video. Started at 16-16-16-36 TRFC 631 3600Mhz 1, 35V and I am now at 16-16-16-33 TRFC 300 4000Mhz 1, 45V with my Samsung B-Dies, still testing. First run TM5 was successfull.
Will try this in the next days. First I try to boot 4200 CL17, that didn't work before. Maybe I soften some timings for first tests if the Kit / CPU can achieve the speed.

[jericosangbaan]

damn this is one of the best overclocking videos here in youtube. i am currently on an i5 7600k and was getting crashes when my cache is at 4.7gh - lowered it down to 4.5 and it is now stable ❤

[brovid-]

Madlad is benching with fuckin fortnite

Subbed

[jasonwatson]

OMG, been working on sqeezing out my max uncore OC, and this is such a fast and reliable way to test.

[dt]

Very smart Ideas!!! Thanks so much for these tips much appreciated and subscribed for the way you explain things very clearly.... 💯😉👍

[stuartsloan]

You don’t know how much this helped me! I had what I thought was a stable OC at core 50 and ring ratio 47 but when I would play games it would stutter and feel delayed. I went crazy trying to figure out what the issue was. I even went back to stock and was having these same issues. It wasn’t until I saw this video that I tried lowering my ring ratio that I knew what the issue was. I had to lower my ring ratio to 42. I tested using fortnite lobby screen and BOOM stable frame rates in the high 300’s! I noticed if I tighten the timings of my ram frames start dropping again. I wonder is ring ration/cpu cache frequencies are linked to ram timings? Anyway thank you so much for your help on this!

[Frenchstud]

@15:56 still laughing at his "Tongue Twister" and he STILL got it wrong omg epic, even when he says "OOF! That was Rough.." its like ok i F it up lets move on lol
Edit: had to come back and type it "...so you know WHAT you did last WAS... WHAT... WAS... the thing ya did wrong". FYI I think he meant "so you know the last thing you did was wrong".

[ilovehotdogs]

This video saved my ass back in the day

[ZombieRommel]

Epic capped the lobby FPS to 120 now, so this test no longer works.

EDIT: I think you can still achieve the same thing by loading into an exploration creative map and just standing still.

[BryanAM]

Ty to you I managed to get an old 7700k at estable 5.1Ghz with cache at 4.7 and Voltaje at 1.340 ty

[mattcoles]

This is really helpful. I'm enjoying (mostly) slowly understanding how these things work.. Thanks again man

[GodKitty]

Prime 95 small FFTs will tell you if the cache is stable. With cache you never want to run anything if you are unstable. Its an easy way to get lots of data corruption very quickly. It can take days before you start to see issues, by then you will have to redownload half your game because of data corruption.

[ashryver]

I used this tip to check my cache on my 4790k recently. My fps could only go to 400-422 but definitely maintained mid to high 300s with the odd high 265ish and very rare flicker of 150ish but never below that.

One thing that interested me - and I'm still new to ocing- is that when my ram was unstable from overclocking 1866 10 11 10 30 to 2200 10 12 12 34 (got 1 memory error in hci memtest at 150% on one thread, while the rest cleared 300%+)... I lowered my cache down to 4.0 from 4.1. I used Aida memory bench and Cinebench to quickly see if anything noticeable changed because Uncore does extremely little apparently on Devils Canyon, and observed lower reads and higher latency with a wide variance, and cinebench scores were lower avg and wide variance compared to pre ram oc - and when initially going back to 4.0 stock and auto volts, I bsod'd in Cinebench which never happened before at stock, the only difference was RAM OC.

Upping the cache voltage, under the impression it was secretly unstable like the Fortnite tip reveals, from 1.2, 1.21, 1.22 yielded increased avgs, tighter variance, and less latency. By 1.22 5 passes of Cinebench were all 5200+, and Aida was very close (albeit lower - 31.6 to 32.2 53.5 latency vs 31.9 to 32.X 49-51 latency) to the original 4.1 uncore @ above ram speeds.

I haven't gone back to test HCI yet, but it seemed that with the ram overclock, 4.1 uncore 1.21v was secretly unstable, and 4.0 1.21 was still unstable and needed 1.22 now (and could no longer run Auto). Fortnite also became noticeably smoother gameplay!


Sorry for tldr, but I hope I help someone digging into OCing cache and ram, especially for Haswell/DC/DDR3!

I also noticed remarkable improvements to cinebench score and latency by doing VCCIO 1.22 and VCCSA 1.064 (ish) vs 1.152 on each. That's where I finally saw significant and consistent improvements from nudging my already oc'd 2133mhz ram to 2200mhz. (Reads were higher as it should be, but quite varied, sometimes lowering down to original 2133. After those voltage adjustments was when I got 31.9-32.X every aida64 memory benchmark and 49-51 latency. This was still at 4.1 uncore 1.21v and was when it was actually unstable in memtest too, prior to the uncore dropping and volt increase. So it's possible that if I fed 4.1 uncore like 1.25 or 1.26v, under the teachings of THIS video, that it may be slightly higher or tighter variance! But I get +1C degree just jumping from 1.21 to 1.22

... my main takeaway or point here I guess in hopes someone might find it useful in the future is that if there's memtest ram instability.. maybe check out your uncore and voltage. It might not fix the instability, but it appears ram OC impacts cache/cache voltage. It MAY also explain why I had a random watchdog bsod gaming.

[EldenLord.]

Tested this with my 13900K. Seems to be stable at 5Ghz cause my "low" FPS never dip below 900 while my high FPS is also around 1k. Thanks a lor for this video.

[NotReallyLaraCroft]

One funny thing I found about cache oc: if I run intel xtu stress test it doesn't crash or freeze, but after playing some games for a while and doing xtu test again my pc bsods almost immediately.Wtf is actually going on