Overclock.net banner
6,361 - 6,380 of 7,348 Posts
I know nothing about it so I cant help. Just seen it talked about.
Thanks for trying to help, i need to find a 10th gen CPU, my friend has his one mounted on his gaming rig and i don't want to bother him with that, gonna try 10th gen & if it doesn't work gonna buy the EVC2SX...
 
  • Rep+
Reactions: bscool
Ok, undervolting an 11700K is not ideal at all. The silicon is not as good as it’s 11900K counter parts. So, auto voltage is voltage the the silicon is requesting based on its quality. Rocket lake does just fine with light to medium loads at 1.500-1.600V. So don’t worry too much about this.

Also, if you type in a manual voltage the. The voltage will not lower down when the CPU is idle on the desktop.


Also, even when you type a manual voltage it won’t go but so low, because it’s feeding the CPU the minimum it needs or requires.

Most 11900K’s need 1.250V to 1.275V for 4.8Ghz all cores. So an 11700K at 5.0Ghz with 1.315-1.325V is not terrible at all. It’s not gonna go much lower than that, if it’s even stable under AVX loads.
I see, thanks. I was thinking msi like others overvolt too much for no reason, this was with default turbo mode..so it doesn't then?
Or does that PL1 state override it too and raises further, because the mobo asked me what cooler and I selected water which makes it 4096w.


I just wanted it to behave like old 4770K, put in max turbo and max adaptive allowed voltage and be done, at balanced it would downclock and down volt. That one did 4.7ghz 1.284v and cache 4.2ghz 1.16v. Was quite good chip :D

I now set all core to 50x, thinking it's just that. Or do I need that Turbo mode where it shows all cores and all at 50. Will it then act like the old days?


Also 1.30v in bios seems to be the limit for the quick tests like Cinebench15, 3dmark firestrike, timespy extreme cpu test, but in all these scenarios it used up to 200w, a bit high for 1.30v, cant imagine if I left it at default 1.47v.

Rectangle Font Screenshot Parallel Terrestrial plant


and it still raised a little, I think becuase of AVX in this cpu profile test and by Timespy extreme AVX instructions?


speaking of adaptive voltage and default 1.47v. I set for example 1.35v in bios and it still read as 1.47v after boot in bios next to my manually input 1.35v, why does it ignore this?


When I manually undervolted this 1.47v it was ok up to -0.160 (so this 1.30v limit), but in idle I saw it as low as 0.56v I think?. And still saw some 1.5v spikes I think, or is this HWmonitor a bit wrong? it reported some of the cores with 0C "bug", but only for a brief second and only after I enabled extra c-states thinking that might help. Does speed shift help too and lower further or it has no effect if I set all core 50x? I know you said it's bad on 11700kf, but this batch I got x140k555 (e4) seems to be "better", I remember first reviewers couldnt even boot at 5ghz and or needed 1.45v+.


I also tested gear1 at those 5Ghz and anything but 3600mhz works, as soon as I change this to 3733mhz it fails; both SA and IO2 at 1.30v, for example first time I booted mobo auto set 1.40v for 3600mhz.. On cometlake (my old 10500) it set to 1.30 and IO 1.25v. I seem to be ok at 1.25v for both at default 3600mhz now, but I've set 1.30v atm to stabilize this beast.


The best latency I got at trfc 300 is ~ 46ns, down from 52ns default, do I need to tweak extra tfaw, twcl, trrd_L S,.. This gskill ram did cl17-19-19-40 4500mhz @ 1.47v, would that tweaked trfc 300 Gear2 benefit more or be somewhat the same + extra ram bandwidth? I play mostly Destiny2, and when I tweaked this trfc and whent from 3960 to 4237 (102.50mhz blck oc on 10500) I got a more responsive game with better min fps and more consistent fps. But it also lowered latency down to ~38.5ns there from 41ns.


I hope I didnt ask too much, would really appreciate your thoughts on my questions and ideas. :)
 
I see, thanks. I was thinking msi like others overvolt too much for no reason, this was with default turbo mode..so it doesn't then?
Or does that PL1 state override it too and raises further, because the mobo asked me what cooler and I selected water which makes it 4096w.


I just wanted it to behave like old 4770K, put in max turbo and max adaptive allowed voltage and be done, at balanced it would downclock and down volt. That one did 4.7ghz 1.284v and cache 4.2ghz 1.16v. Was quite good chip :D

I now set all core to 50x, thinking it's just that. Or do I need that Turbo mode where it shows all cores and all at 50. Will it then act like the old days?


Also 1.30v in bios seems to be the limit for the quick tests like Cinebench15, 3dmark firestrike, timespy extreme cpu test, but in all these scenarios it used up to 200w, a bit high for 1.30v, cant imagine if I left it at default 1.47v.

View attachment 2539986

and it still raised a little, I think becuase of AVX in this cpu profile test and by Timespy extreme AVX instructions?


speaking of adaptive voltage and default 1.47v. I set for example 1.35v in bios and it still read as 1.47v after boot in bios next to my manually input 1.35v, why does it ignore this?


When I manually undervolted this 1.47v it was ok up to -0.160 (so this 1.30v limit), but in idle I saw it as low as 0.56v I think?. And still saw some 1.5v spikes I think, or is this HWmonitor a bit wrong? it reported some of the cores with 0C "bug", but only for a brief second and only after I enabled extra c-states thinking that might help. Does speed shift help too and lower further or it has no effect if I set all core 50x? I know you said it's bad on 11700kf, but this batch I got x140k555 (e4) seems to be "better", I remember first reviewers couldnt even boot at 5ghz and or needed 1.45v+.


I also tested gear1 at those 5Ghz and anything but 3600mhz works, as soon as I change this to 3733mhz it fails; both SA and IO2 at 1.30v, for example first time I booted mobo auto set 1.40v for 3600mhz.. On cometlake (my old 10500) it set to 1.30 and IO 1.25v. I seem to be ok at 1.25v for both at default 3600mhz now, but I've set 1.30v atm to stabilize this beast.


The best latency I got at trfc 300 is ~ 46ns, down from 52ns default, do I need to tweak extra tfaw, twcl, trrd_L S,.. This gskill ram did cl17-19-19-40 4500mhz @ 1.47v, would that tweaked trfc 300 Gear2 benefit more or be somewhat the same + extra ram bandwidth? I play mostly Destiny2, and when I tweaked this trfc and whent from 3960 to 4237 (102.50mhz blck oc on 10500) I got a more responsive game with better min fps and more consistent fps. But it also lowered latency down to ~38.5ns there from 41ns.


I hope I didnt ask too much, would really appreciate your thoughts on my questions and ideas. :)


Adaptive voltage may not work properly on a Z490 motherboard. But to answer your question, you can run low voltage to these chips. They just love high voltage. I have two 11900K’s, (With auto voltage) and 4.8Ghz all cores, one Cala for 1.275V and the other uses around 1.195-1.199V underload through R15/R23 type testing. I can run my newest 11900K at 5.3Ghz with 1.330V (It’s an incredibly good chip though) my other 11900K would need 1.465V and heavy load line just for 5.2Ghz. Just find the best balance In performance VS power for your 11700KF.

(Gear 1) memory at a slower speed with the lower latency, will always smash (Gear 2) in games. So, I’d be after gear (1) 3600 all day.

Also, try 1.500V on the VCCSA
and use up to 1.450V for VCCIO2 or VCCIO AUX (Whatever your bios calls it) these voltages would be required, especially for any memory speeds beyond 3600 (Gear 1) from what I understand the 11700K/KF does not have as good an IMC as 11900K models, but who knows, you could have a good IMC.


Chip variation is fairly large with 10th Gen, and 11th Gen CPU’s. Or just like any CPU really.

That 11700KF is not bad at all though. I would be after maximum overclock for (8) cores, and maximum clock speed for (4) cores, and maximum gear 1 memory speed. This will provide the absolute mind boggling gaming performance. Like, it really won’t get much better if any better from any platform in 2021/2022

I would personally just set a fixed static voltage, and not worry at all if it goes down during idle. Just stability test, and keep reducing the voltage until you can get it as low as possible for 5Ghz all cores, or 5.1Ghz all cores. You also want to run 3600-3733 (Gear 1) with really tight timings, and secondary timings. This will provide extremely fast performance.

Gear 2 may looks nice in a memory bandwidth benchmark. But, it doesn’t hold up for gaming.


See, properly tuned 11th Gen can absolutely smash in games. There is newer microcode, and bios update since launch. And well tuned 11th Gen is a absolute animal in gaming.

Also, I would highly recommend “Dual rank” memory sticks preferably 2x16GB= double sided ram.. It is much faster than 2x8GB.

2x16GB in Gear 1 at 3600 is plenty sufficient.


Also, make sure to overclock your ring/cache too. Rocketlake is pretty picky at 42-43 range depending on the CPU, I run 46 ring/cache daily, but that’s my chip, yours may be different. And, it cache/ring directly feeds off the CPU V-Core too. So, more cache/ring OC need more cpu Vcore voltage to keep it stable.

If your trying to post for maximum memory OC. I would set your VCCSA to 1.500V then set VCCIO2 to 1.400V. then I would apply 3600 Gear 1, then slowly bump the BCLK until it won’t post to the bios anymore. Then you’ll find the max gear 1 memory speed. Then back off a tad, or adjust timings for optimal latency and bandwidth. Test stability etc. etc.

If you have Samsung B-Die, I would go for 1.500-1.600V on the memory. Depending on how good your ram is.


I don’t go too crazy with stability testing, Rocketlake is very picky. if it’ll run through a lot of simple benchmarks and games, like R23, profile test, firestrike physics test, chances are it’s almost or already is stable. Now, memory is another thing. HCi memtest for about 4-6 hours should be good, and I call it stable.

No need to cook/hammer your CPU with prime 95 small fft’s. That is really degrading on a processor. If it’s stable, it’s stable lol. Meaning, it never crashes while using it for what you are doing daily.
 
Intel I9-10900K || Asus ROG Maximus XIII Extreme Glacial || G.Skill Trident Z Royal 32GB 4400MHz CL17

Ofcourse i have very EXTREME Custom Loop (2x XE480 1x CE420) So temps are not problem. Only dont know what volts are this new platform..

Asking few question. :) What volts are max Z590: VCCIO, DRAM, VCSSA and VCORE with these system @ 24/7?
I know exactly how overclock Z490 Platform, but not this new Extreme Glacial Z590.

Z490 Apex was run volts max:
Dram 1.56v
Vccio 1.35v
Vccsa 1.40v
Vcore 1.375v

Thanks for answers.

At moment running Linpack / TM5 Extreme stable:
All Core 5.1GHz / Cache 4700MHz
V/F curve volt -0.060 offset LLC6
Dram 1.54v
Vccio 1.34v
Vccsa 1.4v (Stress 1.376v VCCSA LLC Auto (can change 1-3))

Cant tight any mem timings come Error in TM5 Extreme but Linpack goes NICE Resis.. What i do wrong?
Font Electronic instrument Audio equipment Gadget Rectangle
Font Parallel Electric blue Pattern Rectangle
Computer Personal computer Gadget Computer monitor accessory Font
 
Adaptive voltage may not work properly on a Z490 motherboard. But to answer your question, you can run low voltage to these chips. They just love high voltage. I have two 11900K’s, (With auto voltage) and 4.8Ghz all cores, one Cala for 1.275V and the other uses around 1.195-1.199V underload through R15/R23 type testing. I can run my newest 11900K at 5.3Ghz with 1.330V (It’s an incredibly good chip though) my other 11900K would need 1.465V and heavy load line just for 5.2Ghz. Just find the best balance In performance VS power for your 11700KF.
Hi, thanks now I really got the hang of it :)

I managed to use adaptive voltage with offset, -0.140v was on the edge, -0.130v is fully stable even in cinebench20, which makes cpu VID 1.34v..

So today I wanted to test lower offset again -0.140v and it Bsod watchdog in CB20 halfway trough, is this bsod usually a LLC issue? I was using LLC5, now LLC4 is fully stable, but I noticed more heat, tighter adv ram timings couldn't do that right?

Also yesterday I frist tested Gear2 4266Mhz CL16-16-16-36 @1.50v and compared to somewhat super tight Gear1 3600 CL14-14-14-34 1.45v, everything ran the same or better on 4266mhz.


But somehow I can't get lower latency on 3600mhz, 45ns was the best I could do. On Gear2 4266 the lowest I saw was 47.2ns, not as terrible as I first thought it would be, with extra bandwidth.

Gear1
Font Screenshot Electronic device Technology Multimedia



Gear2
Font Rectangle Screenshot Software Number
Font Screenshot Software Technology Electronic device



Although on Gear2 I can't pass 4266MHz. 4400 or 4600 both fail by boot, even at CL19 in all cases I used 1.50v. Maybe i used too little SA and IO2? I think I used up to SA 1.30V and IO2 1.40V.

atm using SA1.22V and IO2 1.35V for 4266 CL16, with slightly tweaked 2nd timings. IOL/RTL, Slew? and one more are all set to dynamic.

What would be optimal for 4400 or even 4600, or are these b-dies at limit? Can higher cache limits it as well? Im at 4300MHz atm, seen a few using only 4000 or even 3900MHz,

btw, what is this IA, LLC/ring voltage @ 1.198V? Is it cache voltage? I seen it go lower when I lowered offset more.
Rectangle Font Screenshot Parallel Number



Thanks again :)
 
@TheHunter Highly unlikely you can get 2x16 4266+ stable in gear 2 on z490 Unify. Even 2 dim z590 boards 4400c17 will be doing good on DR b die gear2.

z490 MB are just not ideal for mem oc with 11th gen cpu. Can it work, yes. But you will never get # like those on z590 MB can. Latency and copy will always be worse on z490 v z590 using 11th gen.

No to discourage you ,just so you have realistic expectations if you compare your z490 unify # to z590 users yours will be behind by a bit when using 11th gen cpu.

With tuned subtiming/cache you should be able to get into low 40ns latency in gear 1 on the Unify with 11th gen. On z590 Hero as a comparison will be in 38ns range with similar subs and cahce speed.

Here is a screenshot I saved of a z590 Hero build I did from someone with XMP 3600c14-14-14

Most of your subtimings can be tightened quite a bit. Just some feedback as I ran z490 Unify with 10th and 11thgen cpu and did some mem ocing.

Also ppd on z490 unify is reversed. PPD1 lowers latency. Or it did back when I used 11th gen on it so try that. either way try both 1 and 0 and you will see which is better.
 

Attachments

@TheHunter Highly unlikely you can get 2x16 4266+ stable in gear 2 on z490 Unify. Even 2 dim z590 boards 4400c17 will be doing good on DR b die gear2.

z490 MB are just not ideal for mem oc with 11th gen cpu. Can it work, yes. But you will never get # like those on z590 MB can. Latency and copy will always be worse on z490 v z590 using 11th gen.

No to discourage you ,just so you have realistic expectations if you compare your z490 unify # to z590 users yours will be behind by a bit when using 11th gen cpu.

With tuned subtiming/cache you should be able to get into low 40ns latency in gear 1 on the Unify with 11th gen. On z590 Hero as a comparison will be in 38ns range with similar subs and cahce speed.

Here is a screenshot I saved of a z590 Hero build I did from someone with XMP 3600c14-14-14

Most of your subtimings can be tightened quite a bit. Just some feedback as I ran z490 Unify with 10th and 11thgen cpu and did some mem ocing.

Also ppd on z490 unify is reversed. PPD1 lowers latency. Or it did back when I used 11th gen on it so try that. either way try both 1 and 0 and you will see which is better.
HI,

It's a SR 2x8GB kit 3600 CL16-16-16-36

Gear2 at same timings so 4266mhz CL16-16-16-36 is giving me some really hard time, I tested it yesterday more and it became unstable, then I tested memtest64 to see what's up and to my surprise it started spitting errors.

I played with trfc a little and SA and IO2 and the more i tweaked the sooner it became unstable.

This is how it looks atm, I kept all at auto and just adjusted trfc and trefi
Font Screenshot Technology Parallel Number



Was about to test with lower cache, atm its 43x, but then when I saw your post I said im going to test this PPD, auto was 1, now testing 0.

I also lowered both SA and IO2, someone else where on tomahawk z490 had more luck with lower volts then higher.

Memtest64 became unstable either in 1st loop or sometimes in 9th loop when I tweaked voltages a bit.. Most tests always mention older cpus so now Im kind of lost, some say low SA and IO, now with RL it should be more ok with higher..


So lowering cache to 42x or even 40x won't do much here? Im kinda ok with those 45-47ns latencies atm. Just wanted this thing stable at atleast 4266mhz, am I too optimistic with them CL16-16-16 @ 4266?

EDIT: PPD 0 is slower, went form 48 back to 52ns.
 
@TheHunter just forget about gear 2, since its latency is too high and FPS is lower because of this.

You need to enable Round Trip Latency for RAM, to improve RTL timings, because RTL 62 and 64 is too high for 3600 MHz.
I have 61 and 61 for 4000 MHz CL14 gear 1.

Also only 1.35 V SA is not much.
So this could be one of the reasons why you can't do more than 3600 MHz.
Setting SA to 1.45, 1.475 or 1.5 V is fine.

And also what is the exact model of your RAM?
 
  • Rep+
Reactions: TheHunter
EDIT:
what would be a "ideal" 2x16GB kit DR kit, is 4600-4800mhz even possible then with different e.g. Micron kit?

@TheHunter just forget about gear 2, since its latency is too high and FPS is lower because of this.

You need to enable Round Trip Latency for RAM, to improve RTL timings, because RTL 62 and 64 is too high for 3600 MHz.
I have 61 and 61 for 4000 MHz CL14 gear 1.

Also only 1.35 V SA is not much.
So this could be one of the reasons why you can't do more than 3600 MHz.
Setting SA to 1.45, 1.475 or 1.5 V is fine.

And also what is the exact model of your RAM?
HI,

I would at least like to stabilize it without memory errors if nothing else. I can run 3600 CL14 no problem in games, but not 4266..

It's this ram

Font Screenshot Number Parallel Pattern




Font Screenshot Number Document
Font Screenshot Parallel Number Rectangle



EDIT: Im now at 3733MHZ, SA 1.40v did the trick, what about IO2 do i need to raise that more too? Atm its 1.30v.
 
@TheHunter 3600 MHz 16-16-16-36 1.35 V XMP is fine.

You can probably do 3866 MHz with SA close to 1.5 V or less.
1.4 V SA is not much.
1.35 IO is fine. But I think that (much) less than 1.35 V IO should be enough.
 
  • Rep+
Reactions: TheHunter
@TheHunter 3600 MHz 16-16-16-36 1.35 V XMP is fine.

You can probably do 3866 MHz with SA close to 1.5 V or less.
1.4 V SA is not much.
1.35 IO is fine. But I think that (much) less than 1.35 V IO should be enough.
Ok thanks,

I've tested more, SA up to 1.50v, IO2 up to 1.45v and nothing helped, guess Im stuck at max 3733Mhz, at least something.. At first I thought I could do only 3600.

Font Screenshot Electronic device Multimedia Software


About RTL, I set mobo to dynamic and it uses them so high, idk. In Gear2 they're more tight, both around 25-40.

Does it still apply that lower cpu/cache frequency helps by higher ram OC?
I have it now all core 5Ghz @ 1.33V and cache 4.3Ghz, and I didnt try anything else, only lowered cache back to 4Ghz to see if it might boot over 3800Mhz.. but no luck.

This is with somewhat tweaked settings, and I think final too., @bscool is latency ~43.5ns ok for a 3733Mhz?

Light Font Screenshot Red Software

5Ghz 1.33V vid, cache 4.3Ghz, noticed with 4Ghz cache vid drops to ~1.28V, might use that instead.
SA 1.40v
IO2 1.32v,


I can get another 2x8GB for later since they're SR and just keep 3733MHz @1.37v., but will it work ok in 4x8GB mode for interleave? Maybe even 4266 could be possible properly then?
 
Ok thanks,

I've tested more, SA up to 1.50v, IO2 up to 1.45v and nothing helped, guess Im stuck at max 3733Mhz, at least something.. At first I thought I could do only 3600.

View attachment 2540377

About RTL, I set mobo to dynamic and it uses them so high, idk. In Gear2 they're more tight, both around 25-40.

Does it still apply that lower cpu/cache frequency helps by higher ram OC?
I have it now all core 5Ghz @ 1.33V and cache 4.3Ghz, and I didnt try anything else, only lowered cache back to 4Ghz to see if it might boot over 3800Mhz.. but no luck.

This is with somewhat tweaked settings, and I think final too., @bscool is latency ~43.5ns ok for a 3733Mhz?

View attachment 2540428
5Ghz 1.33V vid, cache 4.3Ghz, noticed with 4Ghz cache vid drops to ~1.28V, might use that instead.
SA 1.40v
IO2 1.32v,


I can get another 2x8GB for later since they're SR and just keep 3733MHz @1.37v., but will it work ok in 4x8GB mode for interleave? Maybe even 4266 could be possible properly then?
Then 3733 MHz is the limit for you.
See by how much you can lower SA and IO for 3733 MHz.

Z490 boards have very very basic support for 11th gen - it just works, with relatively high latency (vs. Z590) and almost no BIOS updates with newer microcode and no BIOS finetuning for 11th gen.

Forget gear 2 and 4 sticks. Both are much worsе than gear 1 and 2 sticks.
 
  • Rep+
Reactions: bscool
  • Rep+
Reactions: owikh84
  • Rep+
Reactions: satinghostrider
Sup guys.
I have troubles with my CPU overclocking.

i7 11700K
Z590 Aorus Master
G.SKILL F4-4000C19D-16GTZKW

My 11700K can't pass OCCT AVX2 at these settings (screenshot) on any reasonable voltage. I've tried 5.0Ghz @ 1.45v (LLC 6) and 4.9Ghz all cores @ 1.4v (LLC 6). It will get error in 2-10 min.
Same happens in Prime95 during "Blend" test.
Even 4.8 Ghz @ 1.38v (LLC 6) will get error after ~3 hours.
CPU can sustain regular AVX load longer with same settings but will error anyway too.
OCCT gives no errors for ~20 min with AVX2 load @ 5.0Ghz and almost 1.5v (LLC 6) but I find this is too high anyway and stopped testing.
Is this means my sample so bad or I missing something?
P.S. no errors @ CPU stock settings (at least it will pass 8 hours test with any load).
P.S.S. forgot to mention Power limit is disabled in BIOS by default (4095W both).

Font Line Screenshot Software Technology
 
Sup guys.
I have troubles with my CPU overclocking.

i7 11700K
Z590 Aorus Master
G.SKILL F4-4000C19D-16GTZKW

My 11700K can't pass OCCT AVX2 at these settings (screenshot) on any reasonable voltage. I've tried 5.0Ghz @ 1.45v (LLC 6) and 4.9Ghz all cores @ 1.4v (LLC 6). It will get error in 2-10 min.
Same happens in Prime95 during "Blend" test.
Even 4.8 Ghz @ 1.38v (LLC 6) will get error after ~3 hours.
CPU can sustain regular AVX load longer with same settings but will error anyway too.
OCCT gives no errors for ~20 min with AVX2 load @ 5.0Ghz and almost 1.5v (LLC 6) but I find this is too high anyway and stopped testing.
Is this means my sample so bad or I missing something?
P.S. no errors @ CPU stock settings (at least it will pass 8 hours test with any load).
P.S.S. forgot to mention Power limit is disabled in BIOS by default (4095W both).

View attachment 2540488
This could just be your 11700K sample. These chips are an incredible value for the money, running them stock, or with a good Gear (1) optimized setup for daily work/gaming.

But there is a reason these chips are not 11900K’s.

I’m gonna say it’s probably your Silicon quality. That’s a very power hungry 11700K.
 
This is 5.4@1.385V in bios. (5.4Ghz on all cores)

Load max VID is 1.423V per HWinfo.

265 watts through R20. This is such a good sample. I love this chip!

This is how all 11900K’s should be.

Photograph Light Screenshot Font Rectangle
 
Ok I tweaked some more and this is my best latency so far.
SA 1.375v
IO2 1.27v
ram 1.41v

Can I lower it further? I also noticed at these timings it heats more , but Im not sure what part?

Font Screenshot Technology Software Number


and what timing would be to improve Write speeds? Because this 3600 CL14 try had better write speeds, but with worse timings
Font Screenshot Electronic device Technology Multimedia



I could run tTRDWR's at 10, but then I tweaked one more timing and whole thing broke down, had to build it up again for that 3733mhz oc (lol cost me 2hrs).

what about tRRD's and tWTR's are they ok by 3733mhz? and tCKE could it be lower with such timings or what is the lowest safe, 5?
 
6,361 - 6,380 of 7,348 Posts