5600X Overclocking advice

[pendragon1]

Just so people don't think I'm a moron here. There is a recent thread on reddit about this ram being available and people who bought them having issues with AMD.
https://www.reddit.com/r/bapcsalesc...ingston_fury_renegade_32gb_3600mhz_ddr4_cl16/

notice how it says XMP, not A-XMP or EXPO? thats intel's spec, not AMDs, hence it doesnt always play nice. try manually setting 3600 18-22-22-(and up the last one 2 points) with 1.4v.

 

[trick0502]

Profile 1 16 20 20 should work no issue. 15 17 17 I’m not so sure.

 

[riev90]

Just so people don't think I'm a moron here. There is a recent thread on reddit about this ram being available and people who bought them having issues with AMD.
https://www.reddit.com/r/bapcsalesc...ingston_fury_renegade_32gb_3600mhz_ddr4_cl16/

Like pendragon1 said, it's XMP profile and I've mentioned before that DJR's subtiming is optimized for intel platform, so consider this as learning time for optimizing new dram kit you have

Profile 1 16 20 20 should work no issue. 15 17 17 I’m not so sure.

that 15-17-17 is doable as it's just 3000mhz of frequency.
at 3200mhz, DJR can do 14-17-8-19 with vdimm ~1.5v, and 16-17-8-19 with vdimm ~1.36v.

 

[SpongeBob]

Like pendragon1 said, it's XMP profile and I've mentioned before that DJR's subtiming is optimized for intel platform, so consider this as learning time for optimizing new dram kit you have


that 15-17-17 is doable as it's just 3000mhz of frequency.
at 3200mhz, DJR can do 14-17-8-19 with vdimm ~1.5v, and 16-17-8-19 with vdimm ~1.36v.

What's with the 8? I feel like I haven't see a timing under 10 setting in years?

If the 5600X can't make the cl16 3600 speeds work I might try this for fun since gamers nexus has a real thing for cl14 3200. How sketchy is 1.5v though? I certainly don't have an extreme gamer 8 fan rig anymore. Granted the heat spreaders on these are beefy

 

[pendragon1]

What's with the 8? I feel like I haven't see a timing under 10 setting in years?

i was wondering about that too. should be 17 or 18

 

[riev90]

What's with the 8? I feel like I haven't see a timing under 10 setting in years?

If the 5600X can't make the cl16 3600 speeds work I might try this for fun since gamers nexus has a real thing for cl14 3200. How sketchy is 1.5v though? I certainly don't have an extreme gamer 8 fan rig anymore. Granted the heat spreaders on these are beefy

What's with the 8? I feel like I haven't see a timing under 10 setting in years?

If the 5600X can't make the cl16 3600 speeds work I might try this for fun since gamers nexus has a real thing for cl14 3200. How sketchy is 1.5v though? I certainly don't have an extreme gamer 8 fan rig anymore. Granted the heat spreaders on these are beefy

This is the best I can do on my ex ryzen 3600 + MSI B540M Pro-M2 Max + Gskill Trident Z Neo 3600c16 1.35v.
Tuning dram to 3800 14-14-8-14 trfc 252 1T 1.52v with gdm disabled:

Doing the tRCDWR to 8 on AM4 with mature agesa is not hard these days even on chipset B450, increase smoothness and low / 1% fps.

This is the example of another tRCDWR I managed to do fully stable using Ryzen 5600G + ASRock B550M Pro4 + Hynix CJR kit xmp rated 3600c18 1.35v:
Tuning dram to 4267 18-21-8-21 trfc 546 1T 1.45v with gdm enabled:

 

[crazycrave]

thats the kids pc. i was able to do 3600 on a x470 with a 3700x. i would just but the soc to 1.1 and put the dram at 1.37-40. it should work.
https://valid.x86.fr/kv8sft

thats the kids pc. i was able to do 3600 on a x470 with a 3700x. i would just but the soc to 1.1 and put the dram at 1.37-40. it should work.
https://valid.x86.fr/kv8sft

651 with new drivers https://valid.x86.fr/bench/ly4n4q/1

 

[jlbenedict]

View attachment 549261
notice how it says XMP, not A-XMP or EXPO? thats intel's spec, not AMDs, hence it doesnt always play nice. try manually setting 3600 18-22-22-(and up the last one 2 points) with 1.4v.

So does this mean on AMD, it may not still play nice with reading from SPD, settings that are set by the memory... if picking one of the XMP profiles?

 

[pendragon1]

So does this mean on AMD, it may not still play nice with reading from SPD, settings that are set by the memory... if picking one of the XMP profiles?

yes, possibly. some just work, others take a bunch of fiddling.

 

[SpongeBob]

This is the best I can do on my ex ryzen 3600 + MSI B540M Pro-M2 Max + Gskill Trident Z Neo 3600c16 1.35v.
Tuning dram to 3800 14-14-8-14 trfc 252 1T 1.52v with gdm disabled:

View attachment 549516

Doing the tRCDWR to 8 on AM4 with mature agesa is not hard these days even on chipset B450, increase smoothness and low / 1% fps.

This is the example of another tRCDWR I managed to do fully stable using Ryzen 5600G + ASRock B550M Pro4 + Hynix CJR kit xmp rated 3600c18 1.35v:
Tuning dram to 4267 18-21-8-21 trfc 546 1T 1.45v with gdm enabled:
View attachment 549517

If you tried it what was your best 1.4V cl14/16, with 1T, GDM disabled timings/speed? I'm going to be totally honest with you I really don't have the time to tinker this much so I'm going to set my main timings, voltage, speed, save. Then done with it. I'm not going to waste 2 weeks OCing my ram again for a whopping 1-5 extra FPS.

So does this mean on AMD, it may not still play nice with reading from SPD, settings that are set by the memory... if picking one of the XMP profiles?

So in my case with the ram I'm going to fiddle with it was a crap shoot (I realized this going in but didn't think it would be this big of a gong show since my others have never been). However I would assume if the XMP speed is shown on your QVL it should work. You would think. For example a QVL kit I punched in on the manufacturers website.

 

[riev90]

If you tried it what was your best 1.4V cl14/16, with 1T, GDM disabled timings/speed? I'm going to be totally honest with you I really don't have the time to tinker this much so I'm going to set my main timings, voltage, speed, save. Then done with it. I'm not going to waste 2 weeks OCing my ram again for a whopping 1-5 extra FPS.

For 3600mhz at 18-22-22-22, it can do vdimm 1.36~1.4v for daily.
To ensure the stability, don't forget to test the settings using TM5 config usmus25.

 

[SpongeBob]

For 3600mhz at 18-22-22-22, it can do vdimm 1.36~1.4v for daily.
To ensure the stability, don't forget to test the settings using TM5 config usmus25.

Well tried this with my 3800X tonight and!!!!!!!!!! Blue screen. Even tried raising the voltage to 1.45 and raising SOC voltage max I did was 1.1 on the SOC. All I know is my other kit 3200Mhz kit can OC to 3600. I ended up settling on 3200 16-18-18-18-38 1.39v. Passes tombraider benchmark I'm going to play some games. I'll swap processors tomorrow (Saturday) finally have some time to do this and I'll revist the 3600Mhz as Legendary Gamer has suggested swapping CPU's may do the trick. However if these sticks are wonky with the 3600Mhz I'm straight up ordering another set and returning these. I'm pretty surprised these can't run at 18-22-22. My only theory is that these are dual rank and maybe that has something to do with it if these sticks are in reality 100% fine.

So I ran Windows Memory Diagnostics tool. No idea if it's good or poop but this ram passed it with no issues. So anyone with experience does that clear any potential for these having defects?

Total side note, why does everything like websites and such seem to load significantly faster with PBO off and autoboost on? Well on my rig anyway I noticed this.

 

[SpongeBob]

Well...... I spent like 5 hours today doing some upgrades added a m.2, Noctua 140mm, and the 5600X. Then tweaking the ram. So long story short I tried all the settings you folks suggested even the easy 18-22-22 etc and just nothing but unstable. Just not worth the hassle anymore I spent way too much time trying to make this kit work and it was a no go with both the 3800x/5600X timings, speeds, SOC increases etc. I popped the 16GB 3200Mhz kit back in OCed it to 3600Mhz with easy timings 18-22-22-42, 1.4v, increased SOC cause why not and and been stable all evening like it was before. So can confirm everything is fine just that ram is a headache. Dealing with the retailer for a return, I have a ticket with Kingston so worst case scenario maybe they can set me out a QVL kit.

So in the mean time what should I do with this the chip?

I have PBO off, Allcore boosting to 4.65Ghz.

 

[riev90]

Well tried this with my 3800X tonight and!!!!!!!!!! Blue screen. Even tried raising the voltage to 1.45 and raising SOC voltage max I did was 1.1 on the SOC. All I know is my other kit 3200Mhz kit can OC to 3600. I ended up settling on 3200 16-18-18-18-38 1.39v. Passes tombraider benchmark I'm going to play some games. I'll swap processors tomorrow (Saturday) finally have some time to do this and I'll revist the 3600Mhz as Legendary Gamer has suggested swapping CPU's may do the trick. However if these sticks are wonky with the 3600Mhz I'm straight up ordering another set and returning these. I'm pretty surprised these can't run at 18-22-22. My only theory is that these are dual rank and maybe that has something to do with it if these sticks are in reality 100% fine.

Well after take a look at your thaiphoon screenshot, your kingston memory is indeed dual rank.
It affects the overclockability of your ram, giving more stress towards your IMC.

So I ran Windows Memory Diagnostics tool. No idea if it's good or poop but this ram passed it with no issues. So anyone with experience does that clear any potential for these having defects?

I thing the dram is not having defects, it's just not optimized for AMD / AM4.
Usually single rank Hynix DJR can still be configured easier than say Micron chips.

Total side note, why does everything like websites and such seem to load significantly faster with PBO off and autoboost on? Well on my rig anyway I noticed this.

For ryzen 5000, be it 5600G or 5600X, I only applied all core OC rather than PBO + Curve Optimizer.
It's far more stable even though it needs the vcore to be fixed number.

 

[Legendary Gamer]

Well congrats on the upgrade and

Well...... I spent like 5 hours today doing some upgrades added a m.2, Noctua 140mm, and the 5600X. Then tweaking the ram. So long story short I tried all the settings you folks suggested even the easy 18-22-22 etc and just nothing but unstable. Just not worth the hassle anymore I spent way too much time trying to make this kit work and it was a no go with both the 3800x/5600X timings, speeds, SOC increases etc. I popped the 16GB 3200Mhz kit back in OCed it to 3600Mhz with easy timings 18-22-22-42, 1.4v, increased SOC cause why not and and been stable all evening like it was before. So can confirm everything is fine just that ram is a headache. Dealing with the retailer for a return, I have a ticket with Kingston so worst case scenario maybe they can set me out a QVL kit.

So in the mean time what should I do with this the chip?

I have PBO off, Allcore boosting to 4.65Ghz.

Now you know the RAM is absolute garbage. It happens, even if it's rare. If you are OC'ing your 3200Mhz Stuff to 3600 ... Might want to just keep that stuff and try to tighten your timings. The performance difference between CL18 and CL15 is gonna be negligible. You have the sweet spot frequency for Ryzen already dialed in.

 

[SpongeBob]

Well congrats on the upgrade and

Now you know the RAM is absolute garbage. It happens, even if it's rare. If you are OC'ing your 3200Mhz Stuff to 3600 ... Might want to just keep that stuff and try to tighten your timings. The performance difference between CL18 and CL15 is gonna be negligible. You have the sweet spot frequency for Ryzen already dialed in.

Oh I'm still going for the upgrade to 32GB. My 3200 kit you mentioned above are a random set of 16GB Oloy's my buddy got with the purchase of his 3080 and didn't want and lol good luck trying to find a matching kit of these. I'm loving the 5600X so far it's significantly faster than the 3800X in certain games and runs cooler. So I enabled PBO but disabled allcore just set it to per core and left it at auto. Seems like it performs pretty darn good? Am I missing out on anything by not running it all core? I feel like pushing allcore seems pointless for gaming if not all cores are being used or if certain cores can boost higher if I'm keeping it cooler. I could be misunderstanding everything which is likely since I've kinda skipped a couple generations of hardware.

Should I try to run it with a voltage offset? I'm using my 3800X's stock cooler (Wraith Prism RGB) not the little cooler a 5600X comes with so it's keeping it around 67 degrees while gaming in Zero Build Fortnite. Seems pretty decent in terms of temps for being in a NZXT 510 case.

 

[Legendary Gamer]

You seems like you're doing fine. The 5600X is significantly faster than the 3000 series in most metrics. It was like a 25% uplift. You would push all core if the games you're using consistently utilize more than 2 cores on a regular basis. Otherwise, the 5000 series is pretty good at dynamically boosting clocks on the Cores being utilized for workloads. The Wraith Prism is fine, it's as good as you're gonna get from AMD coolers. AIO or bigger Air coolers might get you a couple percent increase but it's not going to be a night and day difference.

Maybe explore the Undervolting and Curve optimizer a bit to see if you can squeeze anything else out of the processor without actually forcing an overclock yourself. But sounds like you are doing quite well.

 

[applegrcoug]

So how I tune ryzen 5000's is by using PBO and then core cycler https://github.com/sp00n/corecycler to work through the offsets.

Start each offset with -30 in the bios. Then I start running the thing. It methodically goes through each core of the CPU testing it with Prime95. After it finds an error, it moves to the next core. At the end of the run, I go to the cores with an error and increase the offset...it usually takes several runs. But I don't have to be there.

 

[Digital Viper-X-]

I have some samsung B-Die 4266(2x16g) that ran at 4000 on my 5xxx series at one point, my latest 5800x had it at 3800 CL14. Interested?!?!
(they are XMP, but you have to hand tune the timings to get 4000 stable using the Ryzen memory tool)

 

[SpongeBob]

Under the FCH BCLK Settings, FCH Base clock what do you guys have yours set for MHZ or do you just leave it at Auto?

 

[pendragon1]

Under the FCH BCLK Settings, FCH Base clock what do you guys have yours set for MHZ or do you just leave it at Auto?

the setting name doesnt ring a bell so i assume mines on auto.

 

[riev90]

Under the FCH BCLK Settings, FCH Base clock what do you guys have yours set for MHZ or do you just leave it at Auto?

That's for BCLK overclocking.
Better leave it auto or 100mhz.
Once you set it above 100mhz, it will affect your MCLK clock and your core clock overall, thus will require more burn-in test to verify it's stability

 

[SpongeBob]

the setting name doesnt ring a bell so i assume mines on auto.

Not sure if other mobos use different terms but just an example screenshot I found on Google.

 

[riev90]

Not sure if other mobos use different terms but just an example screenshot I found on Google.

Almost MSI AM4 bios have same layout / interface, so I assume your MSI MPG B550 Gaming Plus has the same interface with that screenshot.
You can always try to squeeze a bit of performance by increasing BCLK / FCH clock beyond auto (in this case, 100mhz), but it will take another time and harder to do.

Sometimes in TM5 or any other ram stress test, it can be stable (due to increase of MCLK clock going up following any increase to BLCK).
But in other application, even in idle state, suddenly there would be WHEA error appeared, means that whole system is not fully stable.

 

[SpongeBob]

Almost MSI AM4 bios have same layout / interface, so I assume your MSI MPG B550 Gaming Plus has the same interface with that screenshot.
You can always try to squeeze a bit of performance by increasing BCLK / FCH clock beyond auto (in this case, 100mhz), but it will take another time and harder to do.

Sometimes in TM5 or any other ram stress test, it can be stable (due to increase of MCLK clock going up following any increase to BLCK).
But in other application, even in idle state, suddenly there would be WHEA error appeared, means that whole system is not fully stable.

Is 100 overclocking it or what's default if not auto? I honestly don't know what this does. Also you are correct about the mobo shot picture.

 

[pitingres]

The default BCLK is 100, and that's what I'd expect auto to get you as well.

 

[riev90]

Is 100 overclocking it or what's default if not auto? I honestly don't know what this does. Also you are correct about the mobo shot picture.

100 is the auto value.
for ASRock AM4, sometimes exactly 100, sometime 99.8mhz or 100.2 mhz so around ± 0.2%.

 

[njs2k]

Thought I'd share my experience on getting things stable as I've been tweaking for optimal voltages.

I’m currently at 1866FCLK with 150MHz PBO boost and per core CO between -17 to -30.
I’ve found 1866MHz FCLK to be stable at 1.10625v VSOC on my 5600x with two single-rank sticks. From what I’ve read, 1.10v VSOC seems to be the sweet spot for finding your max daily FCLK on the 5000 series. I was able to boot into windows at 2000MHz FCLK @1.1125v but it was not stable beyond 1866MHz FCLK on my previous BIOS (Asus B550-F 2806). I’ve yet to test this on the latest BIOS v3002.

I spent some time on finding the lowest voltages for VSOC, CLDO VDDP, and VDDG CCD/IOD so that I can maximize power efficiency. After a lot of reading around I’ve found that in some cases lower voltages perform better. I’ve learnt that there should be a minimum 75mv delta between VSOC and VDDG IOD, and that a lower VDDP voltage between 0.80v - 0.90v is better to stability.

Initially, the lowest voltages I found where I was able to pass TM5 (Extreme1 @anta777) with OCCT VRAM stress test (without CPU OC/Boost/PBO) were:
VSOC: 1.10625v (Increased from 1.100v to fix apps freezing and minimise Vdroop)
VDDP: 0.85v
VDDG CCD: 0.80v
VDDG IOD: 0.90v

(Taken after also tightening RAM timings)

But I found two other problems:

• VDDG IOD too low was causing audio corruption and web video playback issues.
• CDDG CCD too low was giving me idle reboots when CPU overclocking. And CCD too high was limiting the maximum negative limits for the curve optimizer (-15 vs -17 for the worst core)

So, after enabling CPU OC+PBO+CO I’ve settled with these system stable values:
VSOC: 1.10625v
VDDP: 0.85v
VDDG CCD: 0.938v
VDDG IOD: 0.988v

My CPU PBO limits are now:
PPT: 90 (This keeps load temps below ~80c)
TDC: 64
EDC: 114
Scalar: 1x
Boost: 150Mhz

I decided to settle with 150Mhz boost as opposed to 200Mhz as this lowered overall power usage and temps when idle/browsing/working.
My approach to PBO and CO was to start with PPT limit only. I found this limit when stress testing with y-cruncher and monitoring max temps. Then after applying initial CO values, monitor peak TDC and EDC when stress testing (with y-cruncher & p95) and put those as limits afterwards. In my case EDC peak was 118 but I used 114 as this was the lower high when stress testing with p95 Large FFTs/y-cruncher. Then continue long term stress testing to make minor adjustments to the Curve Optimizer.

So far it’s been stable in CoreCycler for 5hrs+, p95 Small FFTs 2hrs+, and OCCT core cycler with CO at: -29, -20, -30, -30, -17, -27.
CB r20 scores 4435 with all cores hovering around 4.5GHz.

 

[SpongeBob]

It seems like some folks can go -30 on their negative offset. What would you say is average? I believe mine is rated to be bronze but it boosts well enough. -20 safe for all cores do you think?

 

[SpongeBob]

Well here we go again. So last time my whole cl16 3600 dual rank fiasco still hasn't officially come to an end yet I'm still waiting for my friggin money back from the retailer however Amazon had kits of these on sale for $50 CMK16GX4M2D3600C18 so I snagged 2 of them. So I'll have 4x8GB of cl18 3600. HOWEVER these are on my QVL list and show 4 sticks should be fine. So fingers crossed I can actually run them in XMP and not deal with another horror show. In terms of memory that's as much time as I'm willing to sink it to it at this point.

If this is stable I'll look at maybe doing a small offset on my 5600X and tweak my RX 6750 XT.

If I run thaiphoon burner with all 4 modules in there can it read all 4 sticks? I don't use this software much.

I swear though if QVL ram won't even work with this I'm turning this into my new server and ordering an intel processor and mobo next.

 

[pendragon1]

xmp, not expo? dont be surprised if you still have to manually config them, xmp is intels spec....

 

[SpongeBob]

xmp, not expo? dont be surprised if you still have to manually config them, xmp is intels spec....

I'm just going off of what's in my BIOS. I have XMP settings at the top of my screen, XMP option in my ram settings.
Not my computer just screen shot from the internet but this is what it looks like at the top of our BIOS. Lists XMP profiles you can choose top left. I mean if they don't work that's fine I mean I'm not going to complain about putting in 18-22-22-42 3600 as long as it works.
LOL so let's hope. Compatibility: AMD 300 Series, AMD 400 Series, AMD 500 Series, AMD X570. I've been searching for ram for months and I don't know if I've even seen ram that's expo on it. Seems like a real shortcoming of AMD if their mobo manufacturers with QVL's are basically Bullcrap.

 

[pitingres]

EXPO is only for DDR5, isn't it?

As for the Corsair kits, they ought to work, but who knows. I've never attempted corsair sticks and complaints about them seem moderately common. Either they have a somewhat higher than normal fall-short rate, or maybe corsair buyers just have a higher than normal whinge factor. Anyway, best of luck.

 

[pendragon1]

EXPO is only for DDR5, isn't it?

might be, doesnt change that xmp is intels spec and doesnt always play nice. hence the warning. and as long as it run at the correct spec when set manually does it really matter?

 

[SpongeBob]

EXPO is only for DDR5, isn't it?

As for the Corsair kits, they ought to work, but who knows. I've never attempted corsair sticks and complaints about them seem moderately common. Either they have a somewhat higher than normal fall-short rate, or maybe corsair buyers just have a higher than normal whinge factor. Anyway, best of luck.

Well as nice as the Kingston heatspreaders were, dual rank, but not functional. A whole lot of other people not just myself who jumped on them on a deals subbreddit all found out they wouldn't run in our systems. So I'll take my chances with these kits that are actually on my QVL list, 20% cheaper than the Kingston to get me to 32GB, and have a 83% 5 star rating on Amazon. I'm doing my best to hedge my bets here.

 

[SpongeBob]

Okay.......... So after one wasted Saturday and having to reinstall windows everything seems to be good now. Plus I have functioning 32GB of ram wooo!

Has anyone ever had Thaiphoon burner not open for them or do you guys know a different program I could run to get all the specs off of my ram. I saw a way around it but if I don't have to do a reg edit I'll just use another program.

In Aida64 It's latency stock is better than my previous ram overclocked plus it runs on lower voltage, but the AIDA64 speed/latency test the speed a bit slower. Not that I think I'm missing much but if I can tighten the timings a bit I can try.. What do you guys think Speed vs. Latency?

From there I'll work off CPU offsets to try and reduce some heat when boosting.