Dialing in the overlock on my new 12600kf, matured silicon process showing great results...

[Space Lynx]

update: I settled on a final oc 1.280v at 3.9 ecore and 5.3 pcore, great temps, stable, and same performance as 14600k but half the cost. rock on


This 12600kf is probably one of the last ones being made, so the factory process has matured, same with Ryzen 5800x3d's which if you buy one now the node has matured enough you are pretty much promised -30 undervolt oc.

Well, initially I was going to leave it at 1.260v 5.2 all core and 3.9 ecore, that was stable for me so I decided to leave it... well I got greedy.

I went up to 1.280v and 5.3 all 3.9 ecore, and got 450 more points in r23 multi before, my single r23 score was 1999 only 54 points shy of single core score of 14600k r23 single. (running it again now at 5.3 as I type this) for $149 CPU I am very happy...

So... now I am wondering should i push it further or just be happy?

also, do I need to change my ring clock in BIOS or can I just leave that on auto? been a long time since I did traditional overclocking, heh. this is 10 mins of r23 multi test below:

edit: single r23 2038 points, stock 14600k is 2054 points. I don't think I am going to push it anymore, be happy with this $149 gem

 

[Onasi]

Cool in terms of showing how silicon production improves over time. Not sure if this actually gives you anything beyond CB numbers though, any real application performance uplift is likely to be marginal to non-existent. That’s just how overclocking is nowadays. Still, there’s still headroom seemingly, so you probably could push just a bit further.

 

[Space Lynx]

Cool in terms of showing how silicon production improves over time. Not sure if this actually gives you anything beyond CB numbers though, any real application performance uplift is likely to be marginal to non-existent. That’s just how overclocking is nowadays. Still, there’s still headroom seemingly, so you probably could push just a bit further.

yeah its just for fun, if there is an extra 5-10 fps in games I'll take it, if not that's alright lol

it actually runs at a colder voltage with a manual oc like this then letting it do its own thing though, cause mobo manufacturers are dumb these days. so i am getting colder temps than stock 14600k, and same performance. meh i'll take it


edit: I want to see how low I can get the voltage at 5.2 all core... here is 1.245v score was still good. I am afraid to go any lower but i think i will try 1.235v

edit 2: well I got greedy, 5.4 finally crashed on me (though I was only doin 1.32v, really should have tried higher, but i don't want the extra heat so no ty... sweet spot seems to be 5.3 all core and 3.9 ecore at 1.280v - temps are still good, and performance is only like 1% less than a 14600k in r23

I just witcher 3 ultra in busy city area - 61-65 celsius on each core. so yeah I am happy with this, rock on

 

[ir_cow]

5.2 / 4.0 is pretty good for alder lake. You might get another 100mhz with 1.35v.

 

[Space Lynx]

5.2 / 4.0 is pretty good for alder lake. You might get another 100mhz with 1.35v.

5.3 / 3.9 is fully stable at 1.280v in override voltage mode. that and xmp only changes I made. I am happy with this.

 

[GerKNG]

I settled on a final oc 1.280v at 3.9 ecore and 5.3 pcore, great temps, stable, and same performance as 14600k but half the cost.

So... now I am wondering should i push it further or just be happy?

"Stable"
did you ran even a single stress test?

 

[phanbuey]

Cool in terms of showing how silicon production improves over time. Not sure if this actually gives you anything beyond CB numbers though, any real application performance uplift is likely to be marginal to non-existent. That’s just how overclocking is nowadays. Still, there’s still headroom seemingly, so you probably could push just a bit further.

Something as benign as power limits have a tangible performance uplift (5-15%).

Pushing to higher clocks while uncapping or locking in at stable power limits will absolutely and massively increase the performance - add in ring tuning and you're looking at huge uplifts from stock.

 

[P4-630]

I just have set all my P-Cores @ 5Ghz, good enough for me, it doesn't need to run any hotter imo.
While gaming I see around 65c max @ 5Ghz.

 

[Onasi]

Something as benign as power limits have a tangible performance uplift (5-15%).

Pushing to higher clocks while uncapping or locking in at stable power limits will absolutely and massively increase the performance - add in ring tuning and you're looking at huge uplifts from stock.

In several VERY specific workloads perhaps. I have heavy doubts that you’d see 15% improvements in, say, gaming at a modern resolution. Couple of frames here and there at best. Rendering workloads that use all cores might benefit the most, but nobody renders using CPU nowadays anyways.

 

[phanbuey]

In several VERY specific workloads perhaps. I have heavy doubts that you’d see 15% improvements in, say, gaming at a modern resolution. Couple of frames here and there at best. Rendering workloads that use all cores might benefit the most, but nobody renders using CPU nowadays anyways.

I think you would with a 7900xt - the 12600k bottlenecks pretty hard in modern games (i used to run one at 5.3 for a solid year before upgrading to the 13700k). An OC'd 12600k at 5.3 with tuned ring and ram is very roughly matching speed in games of a r7 7600. At modern resolutions with the Op's 7900xt, the difference between the 7800x3d and the 7600 is not small , especially in titles like baldurs gate, starfield, etc. so there's much more headrom for more CPU performance.

Keep in mind a stock non mce 12600k runs all core at 4.5 ghz, so a 5.3Ghz OC is absolutely massive.

 

[P4-630]

I think you would with a 7900xt - the 12600k bottlenecks pretty hard in modern games (i used to run one at 5.3 for a solid year before upgrading to the 13700k). An OC'd 12600k at 5.3 with tuned ring and ram is very roughly matching speed in games of a r7 7600. At modern resolutions with the Op's 7900xt, the difference between the 7800x3d and the 7600 is not small , especially in titles like baldurs gate, starfield, etc. so there's much more headrom for more CPU performance.

Keep in mind a stock non mce 12600k runs all core at 4.5 ghz, so a 5.3Ghz OC is absolutely massive.

A somewhat better buy for gaming might have been an 8-p core i7 12700k.

 

[Onasi]

@phanbuey
The difference between 7600 and 7800X3D has nothing to do with clocks though and everything with cache. Games nowadays, with very few exceptions, don’t scale THAT aggressively with clock speed. There is a reason why modern Intel CPUs often perform better in games at stock vs an all core OC.

Keep in mind a stock non mce 12600k runs all core at 4.5 ghz, so a 5.3Ghz OC is absolutely massive.

Yeah, but games don’t use all cores evenly all the time, so in practice the difference isn’t as huge as it seems on paper.

tl:dr Nowadays CPU clock speed is significantly less important for gaming performance past a certain point compared to other architectural factors.

 

[wNotyarD]

@phanbuey
The difference between 7600 and 7800X3D has nothing to do with clocks though and everything with cache. Games nowadays, with very few exceptions, don’t scale THAT aggressively with clock speed. There is a reason why modern Intel CPUs often perform better in games at stock vs an all core OC.


Yeah, but games don’t use all cores evenly all the time, so in practice the difference isn’t as huge as it seems on paper.

tl:dr Nowadays CPU clock speed is significantly less important for gaming performance past a certain point compared to other architectural factors.

And that's the great jump from Alder to Raptor: more cache. And the same reason the non-K RPL i5's perform no better than ADLs with same core count.

 

[phanbuey]

And that's the great jump from Alder to Raptor: more cache. And the same reason the non-K RPL i5's perform no better than ADLs with same core count.

But if that was the case, then the 14900K wouldn't take a double digit % nosedive in newer titles when 188W/253W powerlimits are applied.

That was certainly true when the 12600k was released, When I was OCING mine, i gained more from the memory OC and maybe 5% from the CPU OC - mine was at 5.3 ghz as well.

I'm not so sure anymore that's still the case - the 12600K is now getting bottlenecked... and you're going from 4.5Ghz to 5.3Ghz(18%) more performance is going to make a difference at 1440P. Ram/Ring is probably still more important, but if you're OCing a STOCK 12600K to 5.3Ghz and pumping the ring and tweaking the ram - altogether you're going to see substantial uplift, especially in the mins where you get a CPU/Ram bind.

On a 14900K - just a slight clock decrease you're already looking at drops, and this is much less than a 18% clock speed dump - you're looking at a 14900K here above 4.9Ghz from uncapped 5.6. Ignore the bottom line, that's gigabyte being gigabyte, but you can see the mins go from 106 to 94.... that's at 5.6 to 4.9 to 5.3... so 11% decrease in min fps for less than 700mhz on a 14900k at 1080P Epic settings. (epic settings used to make the cpu bind LESS not more). All the newer games are like that... if they did older games there would be 0% difference.

 

[Onasi]

@phanbuey
That is an example which, while valid, also is of a 4090 running at 1080p (in itself a questionable scenario) and in a title which is known to have atrocious CPU resources usage due to developers getting lazy with the default UE4 DX12 config. I would assume that at 1440p and especially 4K, which are far more likely to be paired with a halo card, the difference is negligible. And if one plays at 1080p he is likely to do so with a much less powerful card and be GPU bottlenecked anyway.
In short, while I find people OCing their CPUs to be a fine hobby I really don’t think there’s much practical benefit nowadays. We are far from the days of Sandy Bridge where a 2500/2600K could be pushed so far beyond stock with no real issue that it actually noticeably affected both performance and the longevity of those parts.

 

[dgianstefani]

@phanbuey
The difference between 7600 and 7800X3D has nothing to do with clocks though and everything with cache. Games nowadays, with very few exceptions, don’t scale THAT aggressively with clock speed. There is a reason why modern Intel CPUs often perform better in games at stock vs an all core OC.


Yeah, but games don’t use all cores evenly all the time, so in practice the difference isn’t as huge as it seems on paper.

tl:dr Nowadays CPU clock speed is significantly less important for gaming performance past a certain point compared to other architectural factors.

Yep. Same with the differences between 12 and 14600K. Architectural, more cache, different ring clock, better memory etc.

Clocking till you get same Cinebench unfortunately won't translate to gaming performance where a 12600K won't reach a 14600K.

It's Golden Cove cores vs Raptor Cove cores.

 

[phanbuey]

@phanbuey
That is an example which, while valid, also is of a 4090 running at 1080p (in itself a questionable scenario) and in a title which is known to have atrocious CPU resources usage due to developers getting lazy with the default UE4 DX12 config. I would assume that at 1440p and especially 4K, which are far more likely to be paired with a halo card, the difference is negligible. And if one plays at 1080p he is likely to do so with a much less powerful card and be GPU bottlenecked anyway.
In short, while I find people OCing their CPUs to be a fine hobby I really don’t think there’s much practical benefit nowadays. We are far from the days of Sandy Bridge where a 2500/2600K could be pushed so far beyond stock with no real issue that it actually noticeably affected both performance and the longevity of those parts.

Well we can always ask @Space Lynx to run a few benchies at 1440p . I generally agree with you, but the 12600K is an exception to that rule, owing to Intels aggressive downclocking to segment the i5.

My bet is still on 18% clock boost being a noticeable OC when CPU bound (which for the stock 12600K is all the time) -- when you're going from 4.9 to 5.1 or from 5.4 to 5.6 then yeah no big deal... but 18% is sandy bridge era level of clock gain - something I would consider "far beyond stock".

 

[ir_cow]

5.3 / 3.9 is fully stable at 1.280v in override voltage mode. that and xmp only changes I made. I am happy with this.

Check what your AVX offset is. That will make a difference on voltage for sure. E-Cores has a pretty hard limit as well. After the vdroop, my 1.35v is really 1.3v. It's possible for 5.3 to work at that lower voltage.

 

[Space Lynx]

hmm thanks everyone. I might trade it in for store credit and lose out on the restocking fee, I didn't realize any of this.

I might just wait until 9900x3d after all. I think I will, its only 5 months away most likely, meh not much longer to wait.

 

[wNotyarD]

hmm thanks everyone. I might trade it in for store credit and lose out on the restocking fee, I didn't realize any of this.

I might just wait until 9900x3d after all. I think I will, its only 5 months away most likely, meh not much longer to wait.

Wait, why? If it works, it works. I see nothing wrong with your 12600KF, quite the opposite.

 

[Space Lynx]

Wait, why? If it works, it works. I see nothing wrong with your 12600KF, quite the opposite.

i thought i read in this thread its holding back my 7900 xt? im kind of sad about that...

 

[wNotyarD]

i thought i read in this thread its holding back my 7900 xt? im kind of sad about that...

Is every single frame crucial to you? Check the following graphs:

Spoiler: Graphs

With the behemoth called 4090, the lack of cache of your 12600K gets apparent when it gets smacked by 13/14600K, but is still within 5% of a 5800X3D. With a more "mundane" 3080, the difference between 12th to 13th gen is way lower and it matches the 5800X3D.
Extrapolating the data, you can conclude that the stronger the card, the more a lower CPU will "hold it back". As the 7900XT is stronger than a 3080 but weaker than the 4090, using these graphs as baselines, your 12600K should still be in touching distance of the 5800X3D. 5700X3D-performance, maybe, and that's at gaming exclusively?

I use the 5700X3D as an example, because I know first-hand it doesn't allow my 7900XTX to flex its wings at its fullest potential. And there's absolutely nothing wrong with that. I knew it'd happen when I bought the CPU. Without leaving AM4, though, not even 5800X3D would be able to. So why should I bother at those "missing" frames?

 

[Space Lynx]

Is every single frame crucial to you? Check the following graphs:

Spoiler: Graphs

View attachment 345626

View attachment 345627

With the behemoth called 4090, the lack of cache of your 12600K gets apparent when it gets smacked by 13/14600K, but is still within 5% of a 5800X3D. With a more "mundane" 3080, the difference between 12th to 13th gen is way lower and it matches the 5800X3D.
Extrapolating the data, you can conclude that the stronger the card, the more a lower CPU will "hold it back". As the 7900XT is stronger than a 3080 but weaker than the 4090, using these graphs as baselines, your 12600K should still be in touching distance of the 5800X3D. 5700X3D-performance, maybe, and that's at gaming exclusively?

I use the 5700X3D as an example, because I know first-hand it doesn't allow my 7900XTX to flex its wings at its fullest potential. And there's absolutely nothing wrong with that. I knew it'd happen when I bought the CPU. Without leaving AM4, though, not even 5800X3D would be able to. So why should I bother at those "missing" frames?

i thought a 12600kf and 7900 xt was decent pairing. i was under impression you only needed top tier cpu with a 4090, its hard to tell these days since all the reviews are done with 4090 only.

your points make sense though yeah, honestly it was a good deal, so i may just keep it meh

 

[ir_cow]

It's all game and resolution dependent. Could you lose 10fps with a 12600K. Sure, will you notice between 160 and 170. No.

Also the cost to swap over to AM5 exceeds any value it will bring.

 

[wNotyarD]

i thought a 12600kf and 7900 xt was decent pairing. i was under impression you only needed top tier cpu with a 4090, its hard to tell these days since all the reviews are done with 4090 only.

your points make sense though yeah, honestly it was a good deal, so i may just keep it meh

And here's the conclusion: IT IS. May very well not be the greatest there is, but it is a very good pairing nevertheless.