What is most realistic all core overclock I could get on a Ryzen 7 7700X and 7950X or 13900K/13700K with e-cores off using Noctua NH-D15

[ir_cow]

Again, temps are an issue in CBR23 or other multithhreaded workloads. In which workloads, no matter how much you throttle, your CPU will perform WAY better with ecores on.

Just for the record I believe in e-cores and what they provide. @Wolverine2349 is using a air-cooler and limited by the CPU temp. Personally I would leave them enabled with a overall lower clock.

 

[JustBenching]

Just for the record I believe in e-cores and what they provide. @Wolverine2349 is using a air-cooler and limited by the CPU temp. Personally I would leave them enabled with a overall lower clock.

Im also using an aircooler, u12a. Doesnt really matter, doesnt make a difference. Ecores dont stop your Pcores for clocking to whatever you want to clockc them at during gaming.

Of course you're free to disable E-cores for higher P-core OC

But you DONT get a higher P core oc....thats the problem

 

[GerKNG]

But you DONT get a higher P core oc....thats the problem

of course.
just the four e-cores in my 12600k gave me around a 30-40mv headroom and turned an unstable 5 Ghz to a fully (P95) stable 5.1 Ghz.

 

[JustBenching]

of course.
just the four e-cores in my 12600k gave me around a 30-40mv headroom and turned an unstable 5 Ghz to a fully (P95) stable 5.1 Ghz.

Nope. How can that be, lol. If your four ecores at stock require more voltage than your overclocked pcores, how the heck are you stable at stock?

 

[GerKNG]

Nope. How can that be, lol. If your four ecores at stock require more voltage than your overclocked pcores, how the heck are you stable at stock?

what do you mean?
the P Cores needed around 1.33V (Load Vcore) and without the E-Cores the P Cores were stable at 1.30V flat and 1.34V was stable at 5.1 Ghz instead of 5.0.
the cinebench accelerators pull current as well and they influence the P Core Headroom by quite a bit.
less cores, less vdroop, less temperature = higher OC headroom.

 

[AusWolf]

I'm not sure if the thread is still alive, but here's my two cents:

You don't overclock a 7700X, or any Zen 4 CPU. You put the biggest cooler on it that you can afford and your chassis supports, and let it do its thing. The better cooler you have, the better the CPU will perform.

 

[phanbuey]

It seems like people still want to overclock the old way - static multiplier, hand adjusted voltage, and then test with a stupid load - and turn off parts of the chip to give them a meaningless amount of headroom. When these chips have all-core turbo that can hit much higher sustained clocks and can self overclock with the right amount of volts. 99% of time you're not sitting at a cinebench load. My 12600K was sitting at 5.4Ghz during everyday use -- when I hand OC'd I would be at 5.2Ghz max.

Just adjust your turbo ratio as high as it will go before becoming unstable, let the 100C throttle do it's thing, and power limit to 225w or 190W and find a stable undervolt then let her rip. You will get better results in virtually all scenarios and have a lower cinebench temp to boot with virtually the same MT performance.

 

[Wolverine2349]

It seems like people still want to overclock the old way - static multiplier, hand adjusted voltage, and then test with a stupid load - and turn off parts of the chip to give them a meaningless amount of headroom. When these chips have all-core turbo that can hit much higher sustained clocks and can self overclock with the right amount of volts. 99% of time you're not sitting at a cinebench load. My 12600K was sitting at 5.4Ghz during everyday use -- when I hand OC'd I would be at 5.2Ghz max.

Just adjust your turbo ratio as high as it will go before becoming unstable, let the 100C throttle do it's thing, and power limit to 225w or 190W and find a stable undervolt then let her rip. You will get better results in virtually all scenarios and have a lower cinebench temp to boot with virtually the same MT performance.

For AMD yes. Intel no.

Though what is the point of an unlocked multiplier on AMD Zen 3 and 4 boards if PBO and CO give better OC results.

Of course you're free to disable E-cores for higher P-core OC, but E-cores are here to stay whether you like them or not.

The bottom line, E-cores are free cache and Intel's lifeline to staying competitive in gaming performance. Uncore portion of each Gracemont cluster includes the same 3MB L3 slice that each P-core has - remove them from the die entirely and you're down to a 24MB 8-core and it's Rocket Lake vs. Comet Lake all over again, a massive core doesn't mean much in games if it's starved for cache.

What that makes no sense???

Yes e cores help Intel scale in multi threaded workloads that scale to as many threads as possible as they do not have thermal and powrr headroom to put more than like 2 more p cores.

But e cores provide free L3 cache cach to games??? The exact same amount of L3 cache is available to the p cores whether e cores on or off.

And how is it Rocket Lake vs Comet Lake??? Golden Cove is 20% better IPC than Rocket Lake if it was on 10nm so in reality it's much much faster than 14nm crippled Rocket Lake. And that has nothing to do with e-cores. Unless you mean scaling to more than 8 cores.

Though 8 cores Intel is king as they stomp Zen 3 in IPC since Golden Cove and even still a little ahead of Zen 4 in clock normalized IPC.

 

[tabascosauz]

What that makes no sense???

I never compared RKL with RPL, I never said Raptor Cove is slow in IPC, and I never said that disabling cores affects L3. Nobody asked about Zen. You were the one that wanted a die with P-cores only:

Stop telling others what to do. Each to his and there own on what they want to do with their hardware. Some people want the P cores but not the e-cores. Does Intel have an 8 P core only 13th Gen CPU on the shelf. Nope.

If Intel had an 8 core only Raptor Cove with no e-cores and 36MB of L3 cache I would have gotten it.

You clearly know how to get all the benefits without the drawbacks (ie. E-cores disabled)......so why all the random bitching about E-cores being there?

If you hate E-cores so much that you're stubbornly waiting for a P-core with 4.5MB L3 attached, well, hurry up and wait I guess LOL.

I'll have to check on Intel, but you do lose L1 and L2 cache as you disable Zen cores. I'm looking at it right now. I assume this is the same for Intel since L1/L2 is part of the die where as L3 is shared to all cores.

Yes, but only L3 is shared and part of uncore, so the loss of L1/L2 doesn't affect other cores. Both are private in nature across both Intel and AMD

 

[ir_cow]

I'll have to check on Intel, but you do lose L1 and L2 cache as you disable Zen cores. I'm looking at it right now. I assume this is the same for Intel since L1/L2 is part of the die where as L3 is shared to all cores.

 

[AusWolf]

For AMD yes. Intel no.

Though what is the point of an unlocked multiplier on AMD Zen 3 and 4 boards if PBO and CO give better OC results.



What that makes no sense???

Yes e cores help Intel scale in multi threaded workloads that scale to as many threads as possible as they do not have thermal and powrr headroom to put more than like 2 more p cores.

But e cores provide free L3 cache cach to games??? The exact same amount of L3 cache is available to the p cores whether e cores on or off.

And how is it Rocket Lake vs Comet Lake??? Golden Cove is 20% better IPC than Rocket Lake if it was on 10nm so in reality it's much much faster than 14nm crippled Rocket Lake. And that has nothing to do with e-cores. Unless you mean scaling to more than 8 cores.

Though 8 cores Intel is king as they stomp Zen 3 in IPC since Golden Cove and even still a little ahead of Zen 4 in clock normalized IPC.

If you hate E-cores, just buy anything up to an i5-12500, or AMD.

Personally, I don't think the software environment is ready for the mass adoption of heterogenous architectures, but I've got Zen 4 in the meantime, so there's no reason to complain.

 

[ThrashZone]

Hi,
Yeah voodoo e threads
I believe most would kill/ limit as many background processes/ tasks as possible so technically waiting for the os to optimize the use of them is going to be mixed reviews

In my view more threads the better but that applies normal cores not voodoo e threads which are very limited thermally to me thermal defects so unless benchmarking I'd likely disable them to.

 

[phanbuey]

Hi,
Yeah voodoo e threads
I believe most would kill/ limit as many background processes/ tasks as possible so technically waiting for the os to optimize the use of them is going to be mixed reviews

In my view more threads the better but that applies normal cores not voodoo e threads which are very limited thermally to me thermal defects so unless benchmarking I'd likely disable them to.

Use them first, then make the call. Otherwise you're just speculating without really having any experience.

 

[ThrashZone]

Use them first, then make the call. Otherwise you're just speculating without really having any experience.

Hi,
I've have plenty of experience in cooling lots of cores I wouldn't waste time on e threads.

 

[Wolverine2349]

If you hate E-cores, just buy anything up to an i5-12500, or AMD.

Personally, I don't think the software environment is ready for the mass adoption of heterogenous architectures, but I've got Zen 4 in the meantime, so there's no reason to complain.

Then you get less than 8 P cores and much worse binning and non-k thus no overclocking ability.

I already got a 13900K and disable the e-waste cores and use the best P cores in existence.

I'll have to check on Intel, but you do lose L1 and L2 cache as you disable Zen cores. I'm looking at it right now. I assume this is the same for Intel since L1/L2 is part of the die where as L3 is shared to all cores.

Yeah I would think you lose L1 and L2 cache on the e-cores if you disable them. But the cache is only used if the e-core is used as all L1 and L2 cache is private to the core. If you do not use the cores, the cache is never used nor can be used so does not matter.

 

[AusWolf]

Then you get less than 8 P cores and much worse binning and non-k thus no overclocking ability.

I already got a 13900K and disable the e-waste cores and use the best P cores in existence.

What makes you think it's got worse binning? It's a different chip that physically doesn't have E-cores.

If you need 8 P-cores... fair enough.

 

[Wolverine2349]

What makes you think it's got worse binning? It's a different chip that physically doesn't have E-cores.

If you need 8 P-cores... fair enough.

Its not K and thus cannot be overclocked so of course it is worse binned.

And I want 8 P cores.

 

[phanbuey]

Hi,
I've have plenty of experience in cooling lots of cores I wouldn't waste time on e threads.

If you haven't used them and tested them for yourself, then it's likely most of your assumptions about voodoo threads and temperature defects are not based on anything.
In any case, if you're not ready to "waste time" on them now -- when Zen 5 uses little cores, and there isn't a high end chip that doesn't rely on them, then you can see for yourself.

And I want 8 P cores.

So get a 13700K and disable the e-cores. Then when you realize that it doesn't do anything, you can enable them. MSI motherboards allow you to configure a hotkey to turn them on/off on the fly without rebooting.

RTX 4090 & 53 Games: Core i9-13900K E-Cores Enabled vs Disabled Review - Conclusion | TechPowerUp

"We would still recommend that the E-cores be left untouched, for the simple reason that the P-cores don't appear to be gaining from the freed up power budget of the processor to sustain boost clocks better, as hypothesized in the article's introduction. Leaving the E-cores enabled benefits games, but the OS itself can dump low-priority background processes (think Antivirus, firewall, etc.,) as well as audio and networking stacks, to the E-cores (which have plenty of compute power to deal with them)."

My chip bin sucks and can still sustain 5.7Ghz without issue (needs volts tho). A delidded 13900ks could probably run well above 6ghz.

Most of the "I don't want e cores and want a static clock speed at all times" comes from back in the day when OG eist caused latency penalties and frame variation/performance drops and power savings were always shut off immediately for gaming rigs. Also the resistance to e cores is usually a psychological resistance to having "weaker" cores running, rather than anything based on benches.

I say use the combination that gets you the best temps and performance in the apps you use -- right now there are better profiles available than the classic "tune my rig for a static GHZ under stress load, but then play low-threaded games" approach.

 

[Wolverine2349]

If you haven't used them and tested them for yourself, then it's likely most of your assumptions about voodoo threads and temperature defects are not based on anything.
In any case, if you're not ready to "waste time" on them now -- when Zen 5 uses little cores, and there isn't a high end chip that doesn't rely on them, then you can see for yourself.


So get a 13700K and disable the e-cores. Then when you realize that it doesn't do anything, you can enable them. MSI motherboards allow you to configure a hotkey to turn them on/off on the fly without rebooting.

RTX 4090 & 53 Games: Core i9-13900K E-Cores Enabled vs Disabled Review - Conclusion | TechPowerUp

"We would still recommend that the E-cores be left untouched, for the simple reason that the P-cores don't appear to be gaining from the freed up power budget of the processor to sustain boost clocks better, as hypothesized in the article's introduction. Leaving the E-cores enabled benefits games, but the OS itself can dump low-priority background processes (think Antivirus, firewall, etc.,) as well as audio and networking stacks, to the E-cores (which have plenty of compute power to deal with them)."

My chip bin sucks and can still sustain 5.7Ghz without issue (needs volts tho). A delidded 13900ks could probably run well above 6ghz.

Most of the "I don't want e cores and want a static clock speed at all times" comes from back in the day when OG eist caused latency penalties and frame variation/performance drops and power savings were always shut off immediately for gaming rigs. Also the resistance to e cores is usually a psychological resistance to having "weaker" cores running, rather than anything based on benches.

I say use the combination that gets you the best temps and performance in the apps you use -- right now there are better profiles available than the classic "tune my rig for a static GHZ under stress load, but then play low-threaded games" approach.

I like extra L3 cache of 13900k.

 

[phanbuey]

I like extra L3 cache of 13900k.

cant go wrong either - binning's better too better chance of 6ghz+.

 

[AusWolf]

Its not K and thus cannot be overclocked so of course it is worse binned.

And I want 8 P cores.

I mean, the chip itself is physically a 6+0 (P+E) variant. It doesn't come in K versions at all. It's not a worse bin of higher end chips. It's a completely different chip.

 

[Outback Bronze]

It's not a worse bin

This is what goes on with these chips: Intel 13th Gen Raptor Lake binning – over 500 CPUs tested! | Part 1: i9-13900K and 13900KF | igor'sLAB (igorslab.de)

Getting the i9 has better binning.

 

[AusWolf]

This is what goes on with these chips: Intel 13th Gen Raptor Lake binning – over 500 CPUs tested! | Part 1: i9-13900K and 13900KF | igor'sLAB (igorslab.de)

Getting the i9 has better binning.

I was talking about the 6+0 (P+E) chips, that are only used for the 12500 and below.

 

[Outback Bronze]

What makes you think it's got worse binning

Because you said this.

You can get a 13900K turn off all ecores, HT and 2 Pcores and it will still clock a hell of a lot better because of the binning.

 

[AusWolf]

Because you said this.

You can get a 13900K turn off all ecores, HT and 2 Pcores and it will still clock a hell of a lot better because of the binning.

I didn't say this.

What I said was that it's a completely different chip! Just like Navi 23 is not a lower bin of Navi 21, or GA106 is not a lower bin of GA102.

Edit: I found the article:

Intel "Alder Lake-S" Comes in a 6+0 Core Die Variant

Intel's 12th Gen Core "Alder Lake-S" silicon apparently comes in two variants based on core count. The first one is a larger variant with 8 P cores and 8 E cores, while the second variant is a visibly smaller die with only 6 P cores, no E cores. This was revealed by an MSI Insider video...

www.techpowerup.com