14900k - Tuned for efficiency - Gaming power draw

[JustBenching]

Watched a few videos on your channel, good work man Interesting to see TLOU utilizing ecores quite well it seems.

Lots of games use ecores nowadays but the problem with 12th gen specifically is the cache frequency has to be 4ghz or below with ecores on, which means that sometimes turning them off and bumping the cache to 4.7+ gives you better gaming performance. That is not the case with 13th and beyond though and there is absolutely no reason to turn them off

I can try You need to level up etc to unlock harder missions. May be a few days before I can. What do you record with? Team up on a match or 2 and we can compare real time in same match if you want?

I'm using nvidia thing from Geforce experience (nvenc). Noticed that ultra settings are a lot more taxing for the CPU.

I can try You need to level up etc to unlock harder missions. May be a few days before I can. What do you record with? Team up on a match or 2 and we can compare real time in same match if you want?

Yeah sure can do that today afternoon or tommorow. I'll send you a pm

 

[iameatingjam]

Tried it yesterday, everything ultra with low resolution just to push the cpu, it's not that bad, people were going crazy about how heavy it is. Dropped to 150 when i was in a 4 man team, around 180-200 when solo, that's on a stock 12900k.I expected drops below 100 to be fair

Spoiler

UV can theoretically (very theoretically) kill your mobo - maybe - in extreme scenarios, but probably it will never happen. But killing your CPU i'm almost certain it can't happen.

Well thats the easy answer. The more difficult question, that I can't seem to get a straight answer on, is this:

Since I have b760 mobo, I cannot undervolt without first downgrading my microcode to a 13th gen version, version 0x104, after that, intel decided no more undervolting on non-Z mobos. Is it safe to do that? All answers I've got on that one are very wishy washy. "maybe" "probably" "I don't see why it would, but nobody can say for sure" yada yada. which as you can imagine, doesn't exactly fill me will confidence.

I even went to an overclocking subreddit, and people insisted on answering the obvious question, rather than the one that I bolded and underlined, in regards to the microcode. And I take that as... nobody is really sure.

 

[BoggledBeagle]

... Besides, limiting my chips to just 253w, while asking 5.5ghz all core in a gaming workload, seems rather safe in my experience. ...

253W is a lot of power. And 5.5 GHz is a high frequency too. These may not be long term safe settings and the CPU is not efficient at these settings at all.

 

[InVasMani]

The 14700K is rated for 5.6GHz turbo and 253W max. It should be fine with adequate cooling. I don't remember if b760 allows for reducing the max multipliers or not, but if it does maybe adjust the max multipliers on some of the E cores if you want to save power a lower max frequency requires less voltage so it's similar to a UV, but the MT performance is going to drop a bit where as a stable UV w/o dropping frequency is just squeezing out a bit better efficiency provided it's stable. The power limits and turbo duration is more important in terms of performance and efficiency.

The E cores also eat into the power limits you set and enforce more heavily than P cores because you've got a cluster of 4 individual cores so it's similar to dropping the multiplier across 4 P cores, but not quite since those are more efficient. The P cores are also higher clocked so not sure how it balances out, but either way a cluster of 4 E cores is roughly about the same power required as a 3 to 4 individual P cores.

Like if you want to shed some power drop the 2nd cluster of E cores multiplier down to x42 and 3rd cluster to x41 or something like 41x and x39 and you'll save a bit of power that way even w/o adjust the power limits and turbo duration. That will drop MT a bit it's better than just disabling the E cores outright which if you want to save a good bit of power do that and you'll shave off a lot of power at full load.

 

[Solaris17]

Well thats the easy answer. The more difficult question, that I can't seem to get a straight answer on, is this:

Since I have b760 mobo, I cannot undervolt without first downgrading my microcode to a 13th gen version, version 0x104, after that, intel decided no more undervolting on non-Z mobos. Is it safe to do that? All answers I've got on that one are very wishy washy. "maybe" "probably" "I don't see why it would, but nobody can say for sure" yada yada. which as you can imagine, doesn't exactly fill me will confidence.

I even went to an overclocking subreddit, and people insisted on answering the obvious question, rather than the one that I bolded and underlined, in regards to the microcode. And I take that as... nobody is really sure.

“I want to do something unsupported that is:

- CPU specific
- Board specific
- BIOS specific

Why can’t anyone just tell me?”

I mean what do you really want dude?

 

[iameatingjam]

“I want to do something unsupported that is:

- CPU specific
- Board specific
- BIOS specific

Why can’t anyone just tell me?”

I mean what do you really want dude?

I don't want anything. I was just sharing my experience which was originally, this:

I haven't had much trouble getting my new 14700kf to where I want it to be. Mind you, I only play my games at 60hz, and thats not necessarily a big ask for a chip like this. I used to undervolt too, but since my first chip died after one month of undervolting.... It could be a coincidence, but I'd rather not take the chance again. Besides, limiting my chips to just 253w, while asking 5.5ghz all core in a gaming workload, seems rather safe in my experience. Haven't had issues in the last 6 months. The only difference being... the undervolt. I know it goes against pretty much all common knowledge. So it could be coincidence, or it could just be luck of the draw. Who knows?

And its really not that specific, the question boils down to "can unsupported ucodes from a prior generation cause cpu instability/death" but I wasn't asking that question, I was just referencing it as a question I once wanted answered, that I have long since stopped caring about.

Though I suppose if I got an answer from somebody who really knew what they were talking about, I would still be intrigued. But the reality is, with the kind of work I put my pc through, its really only relevant for benchmarking scores.

 

[InVasMani]

Seems unlikely to cause death, but maybe instability. I wouldn't rule out either being plausible hypothetically in some weird bizarre instance though wouldn't expect it to be the case either. If anything you'd probably just need to reset the CMOS if it didn't work, but I'd have reservations about blindly trying it. I don't have a answer for that though since I didn't make the bios. It really seems pretty unlikely however because I don't think they'd knowingly put in a option that a user might randomly change and brick their hardware. In fact in some ways it would be kind of dumb if the bios didn't prevent that from happening if you're using hardware that's incompatible. If you have incompatible hardware with a setting that would brick it then it shouldn't even allow you to change it is how I see it.

In any case you're basically trying to get unintended behavior out of the hardware and unlikely to get any real official support with that understandably.

 

[JustBenching]

I don't want anything. I was just sharing my experience which was originally, this:



And its really not that specific, the question boils down to "can unsupported ucodes from a prior generation cause cpu instability/death" but I wasn't asking that question, I was just referencing it as a question I once wanted answered, that I have long since stopped caring about.

Though I suppose if I got an answer from somebody who really knew what they were talking about, I would still be intrigued. But the reality is, with the kind of work I put my pc through, its really only relevant for benchmarking scores.

I can't see how they could possibly cause death. If your concern is power draw in games and you don't mind the slight mt performance hit just turn off HT. VOILA

 

[iameatingjam]

I can't see how they could possibly cause death.

Well neither could I, until it happened to me. Coincidence? Possibly. But I'd still rather not risk it as if I had to do another rma, I'm certain it would not go as smoothly as last time because 1) intels RMA system is probably overloaded right now 2) this would be my second application in one year, which might set off some red flags and cause it to get delayed. Why are you guys razzin me anyway? All I know is what happened to me.

There's even a couple threads about it if you absolutely must know more.

My ecores... are causing bluescreens?

Okay I'll try my best to keep this short, I know I have a tendency to ramble... My 14700kf was working great for about a month. Then a couple days ago I started getting bluescreens, bluescreens, bluescreens constantly no matter what I was doing. Even when booting up from USB to reinstall...

www.techpowerup.com

Do you trust intel, or newegg more?

In regards to my 14700kf's dead cores (click for that story). I started getting the ball rolling with the intel rma, perhaps jumping the gun before my 12100 arrived. Man I was an idiot, I said I turned xmp off during troubleshooting totally blanking on the fact that its overclocking. Then all of...

www.techpowerup.com

Oh, and I do disable hyperthreading for applications than benefit from it.

 

[natr0n]

Have you tried disabling more cores ? Can save even more power and gain performance.

Ive done this with 5950x smt/ht off then reduced cores eventually using a 5600x with boost off smt/ht on and off dont notice anything unless memory benchmarking.

Fun to see how low you can go

I would prob use 8 cores only.

 

[JustBenching]

Have you tried disabling more cores ? Can save even more power and gain performance.

Ive done this with 5950x smt/ht off then reduced cores eventually using a 5600x with boost off smt/ht on and off dont notice anything unless memory benchmarking.

Fun to see how low you can go

Nah I don't want to neuter the CPU in MT workloads, I just wanted to maintain top performance in games + MT while sipping power, so HT was the obvious way.

 

[ShrimpBrime]

Well thats the easy answer. The more difficult question, that I can't seem to get a straight answer on, is this:

Since I have b760 mobo, I cannot undervolt without first downgrading my microcode to a 13th gen version, version 0x104, after that, intel decided no more undervolting on non-Z mobos. Is it safe to do that? All answers I've got on that one are very wishy washy. "maybe" "probably" "I don't see why it would, but nobody can say for sure" yada yada. which as you can imagine, doesn't exactly fill me will confidence.

I even went to an overclocking subreddit, and people insisted on answering the obvious question, rather than the one that I bolded and underlined, in regards to the microcode. And I take that as... nobody is really sure.

The answer is YES, using outdated micro code with updated windows MS ME software could cause instability. Possible reduction in performance for some applications, an increase in performance for other applications.

GL!

 

[iameatingjam]

The answer is YES, using outdated micro code with updated windows MS ME software could cause instability. Possible reduction in performance for some applications, an increase in performance for other applications.

GL!

I guess I should have been clearer, but everything worked fine for the first month. Then instability started ( I went into detail on the post I linked). But even after reverting back to stock, including ucode, XMP, power profiles etc, still the machine would give constant bsods unless either ecores were disabled OR 4 or less were enabled. If I enabled more than 5, it was bsod city yet again.

So, yeah, I am very hesitant to undervolt anything on my cpu now. After that happened I also don't push my ram nearly as far as I used to. Basically, it was when I realized how much I do indeed like stability ( I mean who cares about stability until it affects them? RIght XD) Used to run my ram at 1.5v 6400 cl30 with trefi at something like 60,000.

Now I run my ram at 6000 cl32 with trefi at 40,000 and voltage at... either 1.4 or 1.45 ( can't remember). That was basically as fast as I could get it without having to mess with any cpu-side voltages while keeping it stable. And lets be honest, its not like the difference between the two has much real world difference in performance, just option B is a little easier on the imc. I did decrease my cinebench and passmark scores a bit but w/e its not like its enough that I'd notice in day to day computing.

I'm pretty happy with how things are right now. And I might even agree that undervolting itself doesn't cause permanent damage but I am still not so sure if some of the things around undervolting, can cause permanent damage. Could have been the case with my hardware, or it could just be because I am one of those intel chips that are known for instability/high rma rates, even if its worse for i9s, it certainly still happens with i7s, even if they aren't overclocked. Both 13th and 14th gen are pushed way harder than they should have been.

 

[ShrimpBrime]

I guess I should have been clearer, but everything worked fine for the first month. Then instability started ( I went into detail on the post I linked). But even after reverting back to stock, including ucode, XMP, power profiles etc, still the machine would give constant bsods unless either ecores were disabled OR 4 or less were enabled. If I enabled more than 5, it was bsod city yet again.

So, yeah, I am very hesitant to undervolt anything on my cpu now. After that happened I also don't push my ram nearly as far as I used to. Basically, it was when I realized how much I do indeed like stability ( I mean who cares about stability until it affects them? RIght XD) Used to run my ram at 1.5v 6400 cl30 with trefi at something like 60,000.

Now I run my ram at 6000 cl32 with trefi at 40,000 and voltage at... either 1.4 or 1.45 ( can't remember). That was basically as fast as I could get it without having to mess with any cpu-side voltages while keeping it stable. And lets be honest, its not like the difference between the two has much real world difference in performance, just option B is a little easier on the imc. I did decrease my cinebench and passmark scores a bit but w/e its not like its enough that I'd notice in day to day computing.

I'm pretty happy with how things are right now. And I might even agree that undervolting itself doesn't cause permanent damage but I am still not so sure if some of the things around undervolting, can cause permanent damage. Could have been the case with my hardware, or it could just be because I am one of those intel chips that are known for instability/high rma rates, even if its worse for i9s, it certainly still happens with i7s, even if they aren't overclocked. Both 13th and 14th gen are pushed way harder than they should have been.

I would think e-cores can fail, without notice, perhaps without reason. At least it seems failure rates are pretty low in this aspect. You just where unlucky once.

AMD used to lock down bad cores on their Phenom II chips. 565BE or 960T as unlocked chips in example. Sometimes only unlock 1 core, sometimes stable, sometimes not when unlocked.

My 14700K runs pretty well just on 16 threads. I don't find e-cores pleasurable. Imo.

6400 CL30 is really decent. What was the SA voltage?

 

[iameatingjam]

I would think e-cores can fail, without notice, perhaps without reason. At least it seems failure rates are pretty low in this aspect. You just where unlucky once.

AMD used to lock down bad cores on their Phenom II chips. 565BE or 960T as unlocked chips in example. Sometimes only unlock 1 core, sometimes stable, sometimes not when unlocked.

My 14700K runs pretty well just on 16 threads. I don't find e-cores pleasurable. Imo.

6400 CL30 is really decent. What was the SA voltage?

Yeah its always possible it was just a bad sample, actually I think thats more likely than not, still, would rather not take the same risk again, if you know what I mean.

As for the SA, I remember it being quite low. I think I actually had to decrease it from default to get it stable believe it or not. I want to say 1.15, but that might not be right... This was 6+ months ago tbf. My new 14700kf is an entirely different beast, it doesn't like low SA whatsoever, it prefers high. I haven't been able to get memory as stable on this new one as I did on the original. But w/e, at least this one runs stable with all cores intact.

Sorry I'm not able to remember exact values for you, I don't have the best memory (the kind in my brain that is, lol).

 

[ShrimpBrime]

Yeah its always possible it was just a bad sample, actually I think thats more likely than not, still, would rather not take the same risk again, if you know what I mean.

As for the SA, I remember it being quite low. I think I actually had to decrease it from default to get it stable believe it or not. I want to say 1.15, but that might not be right... This was 6+ months ago tbf. My new 14700kf is an entirely different beast, it doesn't like low SA whatsoever, it prefers high. I haven't been able to get memory as stable on this new one as I did on the original. But w/e, at least this one runs stable with all cores intact.

Sorry I'm not able to remember exact values for you, I don't have the best memory (the kind in my brain that is, lol).

All good was just curious. My board goes red at 1.205v SA. So I use 1.250v for 6800mhz CL38. It'll run 7000mhz with some instability though, still working on that. G.Skill Z5 6000mhz CL36 kit. They don't clock too bad, but I should get a faster kit, cpu should be capable of 8000.

 

[iameatingjam]

All good was just curious. My board goes red at 1.205v SA. So I use 1.250v for 6800mhz CL38. It'll run 7000mhz with some instability though, still working on that. G.Skill Z5 6000mhz CL36 kit. They don't clock too bad, but I should get a faster kit, cpu should be capable of 8000.

I'm still hoping we will eventually get cheap single rank 64GB kits. Thats when I'll upgrade. Buuut maybe that'll be more a ddr6 thing.

 

[InVasMani]

You probably need more VCCIO or VDD.

 

[phanbuey]

All good was just curious. My board goes red at 1.205v SA. So I use 1.250v for 6800mhz CL38. It'll run 7000mhz with some instability though, still working on that. G.Skill Z5 6000mhz CL36 kit. They don't clock too bad, but I should get a faster kit, cpu should be capable of 8000.

i've been running 1.3v sa for over a year now... anything lower and the 7600mhz xmp kit isn't fully stable.

 

[JustBenching]

All good was just curious. My board goes red at 1.205v SA. So I use 1.250v for 6800mhz CL38. It'll run 7000mhz with some instability though, still working on that. G.Skill Z5 6000mhz CL36 kit. They don't clock too bad, but I should get a faster kit, cpu should be capable of 8000.

The samsung kits run 6000c30-34-34-58. Not worth pushing for 6800mhz cause you have to lax the timings a lot. Not much difference between a samsung tuned kit and my 7600c34

 

[ShrimpBrime]

The samsung kits run 6000c30-34-34-58. Not worth pushing for 6800mhz cause you have to lax the timings a lot. Not much difference between a samsung tuned kit and my 7600c34

I need to push harder!!! Lol.

Pimod 32m time is 4m 2s xxms. 6.9ghz. I'd love to get under 4 minutes, that was done on ddr4 with very lax timings. Like cl20 2200mhz if I remember right.

So now I have this MSI Z790 ddr5 board, want to try again. The problem is cresting 6ghz on ambient temps. It hits maybe 65c, but it knows it's not frozen so will BSOD at 6.1/6.2. About 1.45v-core. Tested at ring 4.6ghz only, it'll 5ghz or faster under dryice.

Because of the major cooling differences of what I'd like to accomplish, any settings under ambient may not matter for frozen, if that makes sense.

 

[chrcoluk]

Well neither could I, until it happened to me. Coincidence? Possibly. But I'd still rather not risk it as if I had to do another rma, I'm certain it would not go as smoothly as last time because 1) intels RMA system is probably overloaded right now 2) this would be my second application in one year, which might set off some red flags and cause it to get delayed. Why are you guys razzin me anyway? All I know is what happened to me.

There's even a couple threads about it if you absolutely must know more.

My ecores... are causing bluescreens?

Okay I'll try my best to keep this short, I know I have a tendency to ramble... My 14700kf was working great for about a month. Then a couple days ago I started getting bluescreens, bluescreens, bluescreens constantly no matter what I was doing. Even when booting up from USB to reinstall...

www.techpowerup.com

Do you trust intel, or newegg more?

In regards to my 14700kf's dead cores (click for that story). I started getting the ball rolling with the intel rma, perhaps jumping the gun before my 12100 arrived. Man I was an idiot, I said I turned xmp off during troubleshooting totally blanking on the fact that its overclocking. Then all of...

www.techpowerup.com

Oh, and I do disable hyperthreading for applications than benefit from it.

I agree with ir_cows and I think you had a defective CPU. That you undervolted just coincidence.

 

[JustBenching]

I need to push harder!!! Lol.

Pimod 32m time is 4m 2s xxms. 6.9ghz. I'd love to get under 4 minutes, that was done on ddr4 with very lax timings. Like cl20 2200mhz if I remember right.

So now I have this MSI Z790 ddr5 board, want to try again. The problem is cresting 6ghz on ambient temps. It hits maybe 65c, but it knows it's not frozen so will BSOD at 6.1/6.2. About 1.45v-core. Tested at ring 4.6ghz only, it'll 5ghz or faster under dryice.

Because of the major cooling differences of what I'd like to accomplish, any settings under ambient may not matter for frozen, if that makes sense.

I thought you were using the PC for gaming, seems you are doing some ungodly things with it

You might need a z790 or a 2dim z690 to go over 7k ram then.

 

[iameatingjam]

I agree with ir_cows and I think you had a defective CPU. That you undervolted just coincidence.

Well I agree that it was probably a coincidence. But considering I have to run a unsupported microcode from a previous gen to get that undervolt working, are you guys surprised I am hesitant to do it? Especially when something going wrong a second time has a chance to mean I'd be out of a computer? I can't believe this has blown up into such big thing.

 

[chrcoluk]

Well I agree that it was probably a coincidence. But considering I have to run a unsupported microcode from a previous gen to get that undervolt working, are you guys surprised I am hesitant to do it? Especially when something going wrong a second time has a chance to mean I'd be out of a computer? I can't believe this has blown up into such big thing.

I believe that was my only reply on the issue, I responded as I was notified of updates in the thread and you also linked to some other threads.

Of course you dont have to do a new undervolt, thats your choice and you clearly not comfortable doing it so dont do an undervolt. You can still save power/heat in other ways.