I5-12600KF Undervolting won't work

[agent_x007]

Well, you need base value for AC Loadline for this to work.
In my case I had "130" (shown on right on my screenshot), but your ASUS doesn't show it, which means you have to "guesstimate" what's needed, check if it's the same as Auto (or close to it), and go from there. You could check what was mentioned ("170" for 12600k), but better read some more about this. CPU my sister uses (and this was her PC), is 12100F (and not 12600k though).

According to @Solid State Brain, you have to adjust it down (keeping in mind that for each "10" in here, vCore may change by ~30mV), from default value for undervolt under load to take effect. Less than stock is undervolt.

 

[Solid State Brain]

Offset in bios does not work, it tanks performance if I lower more than 10mv. I found this however:

Performance probably tanks because of "IA CEP Enable". CEP stands for "Current Excursion Protection" and its function appears to prevent system instability due to insufficient voltage during high CPU loads, at the cost of performance. Try setting it to "Disable" instead of "Auto".

Other than that, to undervolt the CPU you could try using different "IA AC Loadline" values than the automatic setting. You could either start from a high value like 110 (corresponding 1.1 mOhm) and go down from there, or start from the minimum value of 1 (0.01 mOhm) and test stability from that, increasing it by 5 or 10 every time if the system isn't stable during stress testing. For the 12400F, the maximum AC Loadline value should not be higher than 170 (1.7 mOhm), but at that level your CPU will probably already run quite overvolted.

There's a slim chance that tweaking voltages through the AC Loadline setting will not work at all on your 12400F. Apparently Intel prevented this function from working on locked 13th gen processors, but I'm not sure if that's the case for locked 12th gen ones.

 

[Cowboystrekk]

Performance probably tanks because of "IA CEP Enable". CEP stands for "Current Excursion Protection" and its function appears to prevent system instability due to insufficient voltage during high CPU loads, at the cost of performance. Try setting it to "Disable" instead of "Auto".

Other than that, to undervolt the CPU you could try using different "IA AC Loadline" values than the automatic setting. You could either start from a high value like 110 (corresponding 1.1 mOhm) and go down from there, or start from the minimum value of 1 (0.01 mOhm) and test stability from that, increasing it by 5 or 10 every time if the system isn't stable during stress testing. For the 12400F, the maximum AC Loadline value should not be higher than 170 (1.7 mOhm), but at that level your CPU will probably already run quite overvolted.

There's a slim chance that tweaking voltages through the AC Loadline setting will not work at all on your 12400F. Apparently Intel prevented this function from working on locked 13th gen processors, but I'm not sure if that's the case for locked 12th gen ones.

CEP disable does not wrok, changing IA AC and DC loadline works, but as soon as I lower them too much I get the clock-streching issue like with the traditional UV using core offset. More than 10mv UV performance starts to degrade :\ Seems I need a unlocked CPU to be able to UV properly.

 

[Deleted member 185158]

CEP disable does not wrok, changing IA AC and DC loadline works, but as soon as I lower them too much I get the clock-streching issue like with the traditional UV using core offset. More than 10mv UV performance starts to degrade :\ Seems I need a unlocked CPU to be able to UV properly.

12400F has no benefit from undervolt.

1.2v and 4.6ghz is where you should be with that chip.

 

[agent_x007]

He could try to drop max. multi by one or two, and gain stability for lower voltage that way.
Also, it would be best to downclock all cores to the same frequency when downvolting (preventing highest single multi from crashing system on lower voltage).

 

[Deleted member 185158]

Oh NVM.

It's a ddr4 board and no bclk.

 

[mkdr]

Offset in bios does not work, it tanks performance if I lower more than 10mv. I found this however:

You might have undervolt protection activated in your bios, disable it and try offset again. Read this:

https://www.reddit.com/r/intel/comments/10dcjd9
"There are settings to specify the negative voltage offsets, but if you specify even a tiny offset (30-40mV, for example) and have the Undervolting Protection enabled, the system will be very slow."

https://www.reddit.com/r/intel/comments/10raw6a/_/j6w60a9

 

[Cowboystrekk]

You might have undervolt protection activated in your bios, disable it and try offset again. Read this:

https://www.reddit.com/r/intel/comments/10dcjd9
"There are settings to specify the negative voltage offsets, but if you specify even a tiny offset (30-40mV, for example) and have the Undervolting Protection enabled, the system will be very slow."

https://www.reddit.com/r/intel/comments/10raw6a/_/j6w60a9

I tried but it does not work, if I UV more than -10mv then performance drops. I have tried 7 different bioses and result is the same :/

 

[WhenMusicAttacks]

Hello, i am working on some information about locked Alder Lake undervolt (without clock stretching, where you actually crash if you type -0.200 ), i have found MSI to be working quite good on their b660 lineup, but also had success with some Gigabyte boards lately - there is no clear pattern!

Is there a single main thread to discuss this, or should one be created?

 

[Deleted member 185158]

Hello, i am working on some information about locked Alder Lake undervolt (without clock stretching, where you actually crash if you type -0.200 ), i have found MSI to be working quite good on their b660 lineup, but also had success with some Gigabyte boards lately - there is no clear pattern!

Is there a single main thread to discuss this, or should one be created?

The only 2 things locked on those chips is System Agent and cpu multiplier.

There's nothing to discuss. A 12400F has gotta be the coolest 6 core chip I've ever run, can't imagine it needs an under-volt in any way....

 

[WhenMusicAttacks]

you clearly don't understand the topic, no need to comment then - clock stretching is a huge issue you are unaware of, and i would like to discuss how to push the undervolt on my chip, as i'm not on a normie forum, but the throttlestop one

 

[Deleted member 185158]

you clearly don't understand the topic, no need to comment then - clock stretching is a huge issue you are unaware of, and i would like to discuss how to push the undervolt on my chip, as i'm not on a normie forum, but the throttlestop one

If you need to under-volt, then you have a cooling issue, not a clock stretching issue.

If you under volt and experience clock stretching, see the first sentence above.

What's with the interest of an under-volt?
Why do you feel the need to under-volt?

 

[WhenMusicAttacks]

If you need to under-volt, then you have a cooling issue, not a clock stretching issue.

If you under volt and experience clock stretching, see the first sentence above.

What's with the interest of an under-volt?
Why do you feel the need to under-volt?

If you need to eat chocolate, you have some lack of love. Why do you need to eat chocolate? You can get by with bread just fine.

Please, i am interested in the topic of this thread, i don't need your opinion. I just see now your comment count, i guess that's why you insist with your nonsense.

 

[Deleted member 185158]

If you need to eat chocolate, you have some lack of love. Why do you need to eat chocolate? You can get by with bread just fine.

Please, i am interested in the topic of this thread, i don't need your opinion. I just see now your comment count, i guess that's why you insist with your nonsense.

Your opinions are not validated...

"Because I want to under-volt, but expect better performance" is a null and void statement.

"I'd like to under-volt to reduce temps because I want too" is a better statement.

But since my opinion is nonsense to you, though I've been through these ropes many times... under voting has always reduced performance. This is not a new thing. But perhaps the approach for you and the misunderstanding why transistors need voltage is beyond your comprehension.

In short, there's a reason why Intel is restricting the under-volting. Perhaps asking why would be a better approach than writing up something you obviously don't understand.

Peace.

 

[Cowboystrekk]

The only 2 things locked on those chips is System Agent and cpu multiplier.

There's nothing to discuss. A 12400F has gotta be the coolest 6 core chip I've ever run, can't imagine it needs an under-volt in any way....

12400F is quite cool, but if you OC ram SA voltage stability is quite temps-sensitive. My 12400F was only stable at 3500 G1 with stock cooler, with a basic towercooler it is no stable at 3700 G1. At load the basic tower keeps temps at 50-55C vs 65-70C with stock cooler. If I UV more than 10mv I get clockstretching. 100mv UV is probably possible if it was unlocked, that would probably lower temps by 5C more makibg 3800 G1 possible

Your opinions are not validated...

"Because I want to under-volt, but expect better performance" is a null and void statement.

"I'd like to under-volt to reduce temps because I want too" is a better statement.

But since my opinion is nonsense to you, though I've been through these ropes many times... under voting has always reduced performance. This is not a new thing. But perhaps the approach for you and the misunderstanding why transistors need voltage is beyond your comprehension.

In short, there's a reason why Intel is restricting the under-volting. Perhaps asking why would be a better approach than writing up something you obviously don't understand.

Peace.

I have to disagree a bit. From a safetyperspective UV has historically has issues (plundervolt) that can be a real concern. Atm on unlocked CPUs, the is potential to UV by 50-150mv without causing stability issues.on notebooks lack of UV will impact performance, noise etc. It makes no sense that you can UV by 10-20mv before clockstretching kicks in as UV is basically available, just with overly aggresdive stabilitymeasures.

 

[WhenMusicAttacks]

Your opinions are not validated...

"Because I want to under-volt, but expect better performance" is a null and void statement.

"I'd like to under-volt to reduce temps because I want too" is a better statement.

But since my opinion is nonsense to you, though I've been through these ropes many times... under voting has always reduced performance. This is not a new thing. But perhaps the approach for you and the misunderstanding why transistors need voltage is beyond your comprehension.

In short, there's a reason why Intel is restricting the under-volting. Perhaps asking why would be a better approach than writing up something you obviously don't understand.

Peace

On every board where undervolt works, i got some performance increase in everu test i run. That's even on desktop, not to mention on power/temp constrained scenario like notebooks - we talk double digits gains there.
I have run hours of tests and keep doing so on every build i deliver, wich are about 2 every week, maybe you tried on a couple machines for you and your friends and have no clue.

The reason to lock underolt on the software side was to address known and future potential vulnerabilities, and segmentate the market, wich is why they actually block stuff like perfectly working cores on high-yelds process to make a 12100 i3.

12400F is quite cool, but if you OC ram SA voltage stability is quite temps-sensitive. My 12400F was only stable at 3500 G1 with stock cooler, with a basic towercooler it is no stable at 3700 G1. At load the basic tower keeps temps at 50-55C vs 65-70C with stock cooler. If I UV more than 10mv I get clockstretching. 100mv UV is probably possible if it was unlocked, that would probably lower temps by 5C more makibg 3800 G1 possible


I have to disagree a bit. From a safetyperspective UV has historically has issues (plundervolt) that can be a real concern. Atm on unlocked CPUs, the is potential to UV by 50-150mv without causing stability issues.on notebooks lack of UV will impact performance, noise etc. It makes no sense that you can UV by 10-20mv before clockstretching kicks in as UV is basically available, just with overly aggresdive stabilitymeasures.

Yes, that is another good reason to undervolt, keeping the SA cooler and thus having higher stability at 3500mhz ddr4 gear 1 / 5800 ddr5 gear 2.
3800ddr4 gear1 is totally impossible under Linpack on any condition and cooling though. Even with Peerless Assassin on MSI boards with lite load and -50mv i was not able to get 3600 stable under all testing conditions.

Plundervolt is an issue only if they can access the registers on the software side, wich is something they can easily prevent while leaving the controls open in the bios, like 9th to 11th gen intel cpus on mobile and desktop platforms.

I got 150 and 165 in the first two gigabyte boards where i had full undervolt access last week, i was testing expecting the usual 10-15mv windows before performance regression and found it to be totally open. When i got a crash booting at -200 i was super happy.

 

[Deleted member 185158]

12400F is quite cool, but if you OC ram SA voltage stability is quite temps-sensitive. My 12400F was only stable at 3500 G1 with stock cooler, with a basic towercooler it is no stable at 3700 G1. At load the basic tower keeps temps at 50-55C vs 65-70C with stock cooler. If I UV more than 10mv I get clockstretching. 100mv UV is probably possible if it was unlocked, that would probably lower temps by 5C more makibg 3800 G1 possible

A few things for enlightenment, and I'm only speaking from experience here. Please don't take offence.

If you wanted all the 12th gen performance, you wouldn't have gotten a DDR4 mother board. K, so G1 3800mhz means nothing again no offence.

The 12400F doesn't care it's 55c. Or 98c.
If you're working on the memory, try the memory voltage settings instead. Maybe a bump in IMC voltage. Right? Lot of memory settings on my motherboard, what about yours??

V-core. This term clock stretching.
Does anyone ever see the clocks actually "stretch"??
What is the definition of stretch and why does it apply here?

Well what's actually happening to these chips, some of those billions of transistors stop working if there's not enough voltage. That's why when studied millions of times sustaining a consistent clock, but lowering the v-core, performance (benchmark numbers) go lower. You literally benched with less active transistors. Could be 100s could be thousands. That is if you can sustain stability with this testing.

#1 thing with All Intel chips, past to present, you over-volt to obtain performance and adjust the cooling accordingly.

AMD, is similar. But more so that these chips are voltage leakers. This means when run cooler, the cpu can sustain similar clocks, even higher clocks sometimes when run cool. I've done this testing with a 350$ new 2700x. The chips haven't changed much, it's still a leaky Ryzen core. Cool it, it boosts better even without user intervention.

Anyhow, you can go for gear 1 on ddr4. 3800mhz is a laughable frequency even on ddr4 rigs running decent b-die kits. But it's always 1:1:1, or the no you shouldn't goon squad will come out to get me.

Performance comes with frequency first. Always.
Look at ALL gpu overclocking. Your in the right forum....
More cooling, More Power Tool (for AMD cards) More Performance.
Never the other way around.

I can't get by benchmarking with an under-volt. Not sure how this is even a thing. Used to be people inquiring for the best cooling. Now we inquire how to under-volt to avoid cooling.

On every board where undervolt works, i got some performance increase in everu test i run. That's even on desktop, not to mention on power/temp constrained scenario like notebooks - we talk double digits gains there.
I have run hours of tests and keep doing so on every build i deliver, wich are about 2 every week, maybe you tried on a couple machines for you and your friends and have no clue.

The reason to lock underolt on the software side was to address known and future potential vulnerabilities, and segmentate the market, wich is why they actually block stuff like perfectly working cores on high-yelds process to make a 12100 i3.




Yes, that is another good reason to undervolt, keeping the SA cooler and thus having higher stability at 3500mhz ddr4 gear 1 / 5800 ddr5 gear 2.
3800ddr4 gear1 is totally impossible under Linpack on any condition and cooling though. Even with Peerless Assassin on MSI boards with lite load and -50mv i was not able to get 3600 stable under all testing conditions.

Plundervolt is an issue only if they can access the registers on the software side, wich is something they can easily prevent while leaving the controls open in the bios, like 9th to 11th gen intel cpus on mobile and desktop platforms.

I got 150 and 165 in the first two gigabyte boards where i had full undervolt access last week, i was testing expecting the usual 10-15mv windows before performance regression and found it to be totally open. When i got a crash booting at -200 i was super happy.

All of Intels life, a crash means the cpu requires more v-core.
Why the heck all that for the minor performance figures of gear 1 at such low memory frequencies.

Ddr5, run 6000mhz? I do that on early released memory modules that where reasonably priced and happen to be Samsung sticks, easily run cas 36..... funny enough their native frequency xmp of 5200 and defaults very unsatisfactory.

Ddr4, 4000mhz 14-14-14-34-68 1.50v B-Die is like a standard since 8th gen. If you can't pull that off on a 12th gen, then perhaps overclocking and tweaking that easily isn't for everyone.

I can't imagine trying to under-volt my 13700K. It doesn't seem to care about 90c. I dunno. My 12400F was a happy camper at 4.6ghz. An under-volt didn't accomplish that.

But it's a happy chip 1.2v and 4600mhz however!

 

[WhenMusicAttacks]

A few things for enlightenment, and I'm only speaking from experience here. Please don't take offence.

If you wanted all the 12th gen performance, you wouldn't have gotten a DDR4 mother board. K, so G1 3800mhz means nothing again no offence.

The 12400F doesn't care it's 55c. Or 98c.
If you're working on the memory, try the memory voltage settings instead. Maybe a bump in IMC voltage. Right? Lot of memory settings on my motherboard, what about yours??

V-core. This term clock stretching.
Does anyone ever see the clocks actually "stretch"??
What is the definition of stretch and why does it apply here?

Well what's actually happening to these chips, some of those billions of transistors stop working if there's not enough voltage. That's why when studied millions of times sustaining a consistent clock, but lowering the v-core, performance (benchmark numbers) go lower. You literally benched with less active transistors. Could be 100s could be thousands. That is if you can sustain stability with this testing.

#1 thing with All Intel chips, past to present, you over-volt to obtain performance and adjust the cooling accordingly.

AMD, is similar. But more so that these chips are voltage leakers. This means when run cooler, the cpu can sustain similar clocks, even higher clocks sometimes when run cool. I've done this testing with a 350$ new 2700x. The chips haven't changed much, it's still a leaky Ryzen core. Cool it, it boosts better even without user intervention.

Anyhow, you can go for gear 1 on ddr4. 3800mhz is a laughable frequency even on ddr4 rigs running decent b-die kits. But it's always 1:1:1, or the no you shouldn't goon squad will come out to get me.

Performance comes with frequency first. Always.
Look at ALL gpu overclocking. Your in the right forum....
More cooling, More Power Tool (for AMD cards) More Performance.
Never the other way around.

I can't get by benchmarking with an under-volt. Not sure how this is even a thing. Used to be people inquiring for the best cooling. Now we inquire how to under-volt to avoid cooling.



All of Intels life, a crash means the cpu requires more v-core.
Why the heck all that for the minor performance figures of gear 1 at such low memory frequencies.

Ddr5, run 6000mhz? I do that on early released memory modules that where reasonably priced and happen to be Samsung sticks, easily run cas 36..... funny enough their native frequency xmp of 5200 and defaults very unsatisfactory.

Ddr4, 4000mhz 14-14-14-34-68 1.50v B-Die is like a standard since 8th gen. If you can't pull that off on a 12th gen, then perhaps overclocking and tweaking that easily isn't for everyone.

I can't imagine trying to under-volt my 13700K. It doesn't seem to care about 90c. I dunno. My 12400F was a happy camper at 4.6ghz. An under-volt didn't accomplish that.

But it's a happy chip 1.2v and 4600mhz however!

View attachment 289008

If you wanted performance you could have spent a lot more money and energy? Really? What a surprise.

All the things you wrote are just so wrong and uninformed, you should really give up and maybe go doing some gardening.

On 12400f, IMC voltage (SA) doesn't work and is locked at around 0.95 You can change it on the board and that would do nothing.

Micro-Star International Co., Ltd. MS-7D97 - Geekbench

Benchmark results for a Micro-Star International Co., Ltd. MS-7D97 with an Intel Core i5-12500 processor.

browser.geekbench.com

Now this is i5 12500 with undervolt (lite load 1) AND overclock, with stock cooler and 3466mhz memory (can run geekbench at 3600 for sure, but not keep linpack or ycruncher). Running just so much better than any stock configuration you can imagine.

The fact that you don't know what clock stretching is means you should stop talking about stuff you don't understand - just like everything else you wrote there.

 

[Cowboystrekk]

A few things for enlightenment, and I'm only speaking from experience here. Please don't take offence.

If you wanted all the 12th gen performance, you wouldn't have gotten a DDR4 mother board. K, so G1 3800mhz means nothing again no offence.

The 12400F doesn't care it's 55c. Or 98c.
If you're working on the memory, try the memory voltage settings instead. Maybe a bump in IMC voltage. Right? Lot of memory settings on my motherboard, what about yours??

V-core. This term clock stretching.
Does anyone ever see the clocks actually "stretch"??
What is the definition of stretch and why does it apply here?

Well what's actually happening to these chips, some of those billions of transistors stop working if there's not enough voltage. That's why when studied millions of times sustaining a consistent clock, but lowering the v-core, performance (benchmark numbers) go lower. You literally benched with less active transistors. Could be 100s could be thousands. That is if you can sustain stability with this testing.

#1 thing with All Intel chips, past to present, you over-volt to obtain performance and adjust the cooling accordingly.

AMD, is similar. But more so that these chips are voltage leakers. This means when run cooler, the cpu can sustain similar clocks, even higher clocks sometimes when run cool. I've done this testing with a 350$ new 2700x. The chips haven't changed much, it's still a leaky Ryzen core. Cool it, it boosts better even without user intervention.

Anyhow, you can go for gear 1 on ddr4. 3800mhz is a laughable frequency even on ddr4 rigs running decent b-die kits. But it's always 1:1:1, or the no you shouldn't goon squad will come out to get me.

Performance comes with frequency first. Always.
Look at ALL gpu overclocking. Your in the right forum....
More cooling, More Power Tool (for AMD cards) More Performance.
Never the other way around.

I can't get by benchmarking with an under-volt. Not sure how this is even a thing. Used to be people inquiring for the best cooling. Now we inquire how to under-volt to avoid cooling.



All of Intels life, a crash means the cpu requires more v-core.
Why the heck all that for the minor performance figures of gear 1 at such low memory frequencies.

Ddr5, run 6000mhz? I do that on early released memory modules that where reasonably priced and happen to be Samsung sticks, easily run cas 36..... funny enough their native frequency xmp of 5200 and defaults very unsatisfactory.

Ddr4, 4000mhz 14-14-14-34-68 1.50v B-Die is like a standard since 8th gen. If you can't pull that off on a 12th gen, then perhaps overclocking and tweaking that easily isn't for everyone.

I can't imagine trying to under-volt my 13700K. It doesn't seem to care about 90c. I dunno. My 12400F was a happy camper at 4.6ghz. An under-volt didn't accomplish that.

But it's a happy chip 1.2v and 4600mhz however!

View attachment 289008

On locked 12th and 13th gens SA voltage is locked at about 0.95v, even a Z-board cant unlock. This means getting over 3400-3800 in G1 is impossible. On a K-sku 4000-4300 G1 is doable.

Clockstretching is when reported clockspeed and actual clockspeed misalign, can be observed i hwmonitor64.

Temp has major impact on SA-voltage stability. Higher temps makes the locked 0.95v SA voltage more prone to instability.

Undervolting does not reduce performance if minimum required voltage is met. When Intel, AMD etc bins CPUs/GPUs they overvolt a bit since quality of die varies. Also the will degrade slightly over time so if 1.0v at 4GHz was required a launch, 3 years later the same CPU may need 1.03v to run 4GHz. The producers approach is to make all CPUs run 4GHz at 1.15v to accound for all quality of CPUs. Some may operate 4GHz at 0.9v, others may need 1.1v, it varies a great deal. My 5600X runs fine with -30 CO allcore, single core performs the same, multicore is higher if I use stock PPT. My nephews 5600X has major issues with CO, only one core can do -30, other -5 to -15, good example of different binning.

 

[Deleted member 185158]

On locked 12th and 13th gens SA voltage is locked at about 0.95v, even a Z-board cant unlock. This means getting over 3400-3800 in G1 is impossible. On a K-sku 4000-4300 G1 is doable.

Clockstretching is when reported clockspeed and actual clockspeed misalign, can be observed i hwmonitor64.

Temp has major impact on SA-voltage stability. Higher temps makes the locked 0.95v SA voltage more prone to instability.

Undervolting does not reduce performance if minimum required voltage is met. When Intel, AMD etc bins CPUs/GPUs they overvolt a bit since quality of die varies. Also the will degrade slightly over time so if 1.0v at 4GHz was required a launch, 3 years later the same CPU may need 1.03v to run 4GHz. The producers approach is to make all CPUs run 4GHz at 1.15v to accound for all quality of CPUs. Some may operate 4GHz at 0.9v, others may need 1.1v, it varies a great deal. My 5600X runs fine with -30 CO allcore, single core performs the same, multicore is higher if I use stock PPT. My nephews 5600X has major issues with CO, only one core can do -30, other -5 to -15, good example of different binning.

The misalignment is from the reference clock drooping from low voltage.

2 5600X chips that vary slightly from under-volt isn't exactly binned by AMD. That's being binned by you.

If the 5600X had passed the testing to be clocked as high as a 5950X, that's where that silicon would have went. To a higher frequency and v-core.

System Agent being locked kind of sucks. I'll admit that. But again, frequency over gear. Gear 2 higher frequency will net better performance.

Being actually overclocked some locked chips, posting above 5600mhz gear 1 can be difficult, but not necessary. The performance gained from the additional frequency will make up for this.

Neither here nor there I guess. I mean to actually run a locked chip, on a locked board and expect performance of small differences from stock. G1 and G2 maybe a percent worth of frustration.

I suppose performance of locked chips is in the eye of a beholder.

2 cores, locked system agent.
This screen shot is a happy example of performance.
C-states are disabled. Pretty sure running gear 2. Locked SA.
No v-droop with an over-volt, not an under-volt.
The board, with an external clock gen, locked multiplier.
Asus ROG B660-G = under 300$ (US).
I did my research for this build specifically to gain performance with locked chips.
Next best board for adventures like this would be a Z690 Hero.

If you wanted performance you could have spent a lot more money and energy? Really? What a surprise.

All the things you wrote are just so wrong and uninformed, you should really give up and maybe go doing some gardening.

On 12400f, IMC voltage (SA) doesn't work and is locked at around 0.95 You can change it on the board and that would do nothing.

Micro-Star International Co., Ltd. MS-7D97 - Geekbench

Benchmark results for a Micro-Star International Co., Ltd. MS-7D97 with an Intel Core i5-12500 processor.

browser.geekbench.com

Now this is i5 12500 with undervolt (lite load 1) AND overclock, with stock cooler and 3466mhz memory (can run geekbench at 3600 for sure, but not keep linpack or ycruncher). Running just so much better than any stock configuration you can imagine.

The fact that you don't know what clock stretching is means you should stop talking about stuff you don't understand - just like everything else you wrote there.

Right, see the above post.
Locked chip, board was not much more expensive, is a b660 chipset and it's miles ahead of anything you've posted yet so far.

I overclock. You.... tweak, I guess the word fits, locked hardware in hopes of some performance uplift, which to me, looks like very little to none.

Never does overclockers under-volt to make a frequency gain. This is only a blanket pulled over gamers eyes. You don't under-volt a 5600X to accomplish 4.8 to 5ghz.

Y-cruncher stable... gimme a fkn break! Lol

 

[Cowboystrekk]

The misalignment is from the reference clock drooping from low voltage.

2 5600X chips that vary slightly from under-volt isn't exactly binned by AMD. That's being binned by you.

If the 5600X had passed the testing to be clocked as high as a 5950X, that's where that silicon would have went. To a higher frequency and v-core.

System Agent being locked kind of sucks. I'll admit that. But again, frequency over gear. Gear 2 higher frequency will net better performance.

Being actually overclocked some locked chips, posting above 5600mhz gear 1 can be difficult, but not necessary. The performance gained from the additional frequency will make up for this.

Neither here nor there I guess. I mean to actually run a locked chip, on a locked board and expect performance of small differences from stock. G1 and G2 maybe a percent worth of frustration.

I suppose performance of locked chips is in the eye of a beholder.

2 cores, locked system agent.
This screen shot is a happy example of performance.
C-states are disabled. Pretty sure running gear 2. Locked SA.
No v-droop with an over-volt, not an under-volt.
The board, with an external clock gen, locked multiplier.
Asus ROG B660-G = under 300$ (US).
I did my research for this build specifically to gain performance with locked chips.
Next best board for adventures like this would be a Z690 Hero.

View attachment 289040



Right, see the above post.
Locked chip, board was not much more expensive, is a b660 chipset and it's miles ahead of anything you've posted yet so far.

I overclock. You.... tweak, I guess the word fits, locked hardware in hopes of some performance uplift, which to me, looks like very little to none.

Never does overclockers under-volt to make a frequency gain. This is only a blanket pulled over gamers eyes. You don't under-volt a 5600X to accomplish 4.8 to 5ghz.

Y-cruncher stable... gimme a fkn break! Lol

No, clockstretching is UV-protection kicking in. Others have successfully removed it on K-skus and certain B660-boards. With clockstretching it will throttle down at 10-20mv, without UV protect it can be 100% stable at -100mv with no performanceloss, maybe even performance gain if cooler is limited on K-skus.

A good 5600X that can reach 4.95GHz stock on all cores is binned as 5900X due to 2 defect og very low binned cores. My 5600X can do 5.1-5.3GHz at 1.3v according to hydro on 4 of the cores, but 1 core do max 4.9 and one do 4.85. The 2 latter cores made it just 5600X and not 5900X.

As for Gear 2 it will be better in BW-scenarios, but latency is a major issue. I compared my setup with a guy with same MB running Gear 2 at 5333. In some games I won, in some he did.

 

[Deleted member 185158]

No, clockstretching is UV-protection kicking in. Others have successfully removed it on K-skus and certain B660-boards. With clockstretching it will throttle down at 10-20mv, without UV protect it can be 100% stable at -100mv with no performanceloss, maybe even performance gain if cooler is limited on K-skus.

A good 5600X that can reach 4.95GHz stock on all cores is binned as 5900X due to 2 defect og very low binned cores. My 5600X can do 5.1-5.3GHz at 1.3v according to hydro on 4 of the cores, but 1 core do max 4.9 and one do 4.85. The 2 latter cores made it just 5600X and not 5900X.

As for Gear 2 it will be better in BW-scenarios, but latency is a major issue. I compared my setup with a guy with same MB running Gear 2 at 5333. In some games I won, in some he did.

My primary hobby goal is obtaining numbers. And I try to do this as cost effectively as possible.

All systems react differently. I agree, use case, certain games ect ect, some tweaks are better than others.

But honestly, I feel your pain big time when it comes to overclocking. It does start to cost significant dollars.

Seeking performance is interesting. I'm only here to reply because of that interest.

I just don't agree with under-volting as a viable solution when there's better out there. It doesn't have to be expensive though.

Just imagine running that 12400F on my board. 4.6ghz is what the stock cooler will handle. With its cute little copper plug and all, doesn't even cover 50% of the IHS plate. Good cooling it's a 5.2ghz chip all day. I found it very cost effective to get 16gb of memory, a decent board with a clock gen, 269$ and a 12400F. Much better than being stuck at stock with an under-volt.

 

[Cowboystrekk]

My primary hobby goal is obtaining numbers. And I try to do this as cost effectively as possible.

All systems react differently. I agree, use case, certain games ect ect, some tweaks are better than others.

But honestly, I feel your pain big time when it comes to overclocking. It does start to cost significant dollars.

Seeking performance is interesting. I'm only here to reply because of that interest.

I just don't agree with under-volting as a viable solution when there's better out there. It doesn't have to be expensive though.

Just imagine running that 12400F on my board. 4.6ghz is what the stock cooler will handle. With its cute little copper plug and all, doesn't even cover 50% of the IHS plate. Good cooling it's a 5.2ghz chip all day. I found it very cost effective to get 16gb of memory, a decent board with a clock gen, 269$ and a 12400F. Much better than being stuck at stock with an under-volt.

Undervolting is viable running stock, if you find table floor it will lower temp, consumption and make SA-voltage more stable if you OC ram. On notebooks I would say UV is essential as it can increase performance. You always have headroom, but if UV protection is on which is impossible to unlock on locked CPUs it seems, then performance drops at minor UV. What is better than UV if you runs clocks stock?

 

[WhenMusicAttacks]

Undervolting is viable running stock, if you find table floor it will lower temp, consumption and make SA-voltage more stable if you OC ram. On notebooks I would say UV is essential as it can increase performance. You always have headroom, but if UV protection is on which is impossible to unlock on locked CPUs it seems, then performance drops at minor UV. What is better than UV if you runs clocks stock?

He is a broken machine, why waste more time in the discussion?
Anyway, i am tracking down the fact that at some point in 2022 a microcode with unlocked undervolt was released. I have a b760m gaming x with f1 bios (release november 2022) that does undervolt properly.
I have the video up on youtube and currently purchasing more boards to test. Sadly most b760 boards were released later in 2022 with earliest bios being dec.
I did test with a b660m gaming x but only with f4 and f23, i need to try f21 wich is from the same time window as the working ones.

At the same time, z690ud had no undervolt in f4 bios, but has it on march 2023 f23a one!

A few things for enlightenment, and I'm only speaking from experience here. Please don't take offence.

If you wanted all the 12th gen performance, you wouldn't have gotten a DDR4 mother board. K, so G1 3800mhz means nothing again no offence.

The 12400F doesn't care it's 55c. Or 98c.
If you're working on the memory, try the memory voltage settings instead. Maybe a bump in IMC voltage. Right? Lot of memory settings on my motherboard, what about yours??

V-core. This term clock stretching.
Does anyone ever see the clocks actually "stretch"??
What is the definition of stretch and why does it apply here?

Well what's actually happening to these chips, some of those billions of transistors stop working if there's not enough voltage. That's why when studied millions of times sustaining a consistent clock, but lowering the v-core, performance (benchmark numbers) go lower. You literally benched with less active transistors. Could be 100s could be thousands. That is if you can sustain stability with this testing.

#1 thing with All Intel chips, past to present, you over-volt to obtain performance and adjust the cooling accordingly.

AMD, is similar. But more so that these chips are voltage leakers. This means when run cooler, the cpu can sustain similar clocks, even higher clocks sometimes when run cool. I've done this testing with a 350$ new 2700x. The chips haven't changed much, it's still a leaky Ryzen core. Cool it, it boosts better even without user intervention.

Anyhow, you can go for gear 1 on ddr4. 3800mhz is a laughable frequency even on ddr4 rigs running decent b-die kits. But it's always 1:1:1, or the no you shouldn't goon squad will come out to get me.

Performance comes with frequency first. Always.
Look at ALL gpu overclocking. Your in the right forum....
More cooling, More Power Tool (for AMD cards) More Performance.
Never the other way around.

I can't get by benchmarking with an under-volt. Not sure how this is even a thing. Used to be people inquiring for the best cooling. Now we inquire how to under-volt to avoid cooling.



All of Intels life, a crash means the cpu requires more v-core.
Why the heck all that for the minor performance figures of gear 1 at such low memory frequencies.

Ddr5, run 6000mhz? I do that on early released memory modules that where reasonably priced and happen to be Samsung sticks, easily run cas 36..... funny enough their native frequency xmp of 5200 and defaults very unsatisfactory.

Ddr4, 4000mhz 14-14-14-34-68 1.50v B-Die is like a standard since 8th gen. If you can't pull that off on a 12th gen, then perhaps overclocking and tweaking that easily isn't for everyone.

I can't imagine trying to under-volt my 13700K. It doesn't seem to care about 90c. I dunno. My 12400F was a happy camper at 4.6ghz. An under-volt didn't accomplish that.

But it's a happy chip 1.2v and 4600mhz however!

View attachment 289008

oh, i just took the time to randomly take out the B660m mortar and unsurprisingly, your score with r20 and 12400f is trash

and i'm sure my geekbench is going to be at least 10% better even if you are on ddr5 and i am on d4

I can do very nice cinebench scores on a 10980xe, but i would still pick undervolted 12400f with tuned ddr5 any day of the week for my personal rig. Geekbench single core is a much better indication of real world performance.

 

[Deleted member 185158]

Undervolting is viable running stock, if you find table floor it will lower temp, consumption and make SA-voltage more stable if you OC ram. On notebooks I would say UV is essential as it can increase performance. You always have headroom, but if UV protection is on which is impossible to unlock on locked CPUs it seems, then performance drops at minor UV. What is better than UV if you runs clocks stock?

I guess under-volt is great for these applications. I've already admitted that.

He is a broken machine, why waste more time in the discussion?
Anyway, i am tracking down the fact that at some point in 2022 a microcode with unlocked undervolt was released. I have a b760m gaming x with f1 bios (release november 2022) that does undervolt properly.
I have the video up on youtube and currently purchasing more boards to test. Sadly most b760 boards were released later in 2022 with earliest bios being dec.
I did test with a b660m gaming x but only with f4 and f23, i need to try f21 wich is from the same time window as the working ones.

At the same time, z690ud had no undervolt in f4 bios, but has it on march 2023 f23a one!


oh, i just took the time to randomly take out the B660m mortar and unsurprisingly, your score with r20 and 12400f is trash

View attachment 289104

and i'm sure my geekbench is going to be at least 10% better even if you are on ddr5 and i am on d4

I can do very nice cinebench scores on a 10980xe, but i would still pick undervolted 12400f with tuned ddr5 any day of the week for my personal rig. Geekbench single core is a much better indication of real world performance.

Yeah, it's not the greatest score, I'll admit that. But at least I ran it properly, wrapped with Benchmate and have proof with cpu-z to validate with.
That's way high for a default frequency. I suppose that's why you didn't run Benchmate or care to take the 10 seconds to open cpu-z a couple times.
What's benchmate? It makes sure you can't skew the clocks. Been a rule since W8 released, this is due to the RTC.

Anyhow, on the list, the next best R20 time/score that's better than mine is done at 5000mhz.

Cinebench - R20 overclocking records @ HWBOT

Overclocking records

hwbot.org

Please submit and properly. Get in where you fit in, cause to me, looks like the cpu is running at least 4.8ghz. CBR20 is not very memory sensitive as you make it out to seem.
After you accomplished a proper submission, maybe I'll pull out my 12400F and put the water block on it instead of the stock air cooler, which is what I used for the submission.