Testing single core overclocks

[Lord_Ralex]

I am looking at how to best overclock my i7-8700k for both a mix of workload related tasks and for gaming.

 

It's already a good CPU, but I'm wanting to really test how well it overclocks for a gaming load vs for a bunch of VMs/apps running without compromising stability or heat.

 

A lot of the overclocking testing that I see hits all cores at once, so the features that my CPU has which will let it overclock higher or lower depending on the number of active cores is harder to validate. (As in, it by default clocks to 4.7 when it's one active core, 4.6 or something when it's 2, and so on, so the max clock speed when all cores are in use is lower than when one a few are).

 

I already have my i7 running at 4.6GHz when all cores are in use, but I want to have it run at 5GHz when only 1-3 of the cores are in use, and drop back to roughly 4.6-4.7 when all cores are maxed, to control heat and stability.

 

How does one test the overclock stability for when only a select few cores are active? As in, how can I test that when I run the 5GHz during gaming, it would be stable and heat be reasonable, compared to when all cores are maxed and it throttles down a little more.

 

Is there any software to natively do this, or am I looking at disabling cores to "mimic" this behavior, then using AIDA64 to validate?

[Ashiella]

18 minutes ago, Lord_Ralex said:

I already have my i7 running at 4.6GHz when all cores are in use, but I want to have it run at 5GHz when only 1-3 of the cores are in use, and drop back to roughly 4.6-4.7 when all cores are maxed, to control heat and stability.

No.

 

 

Cinebench r15 > custom thread count

userbenchmark.com

[Lord_Ralex]

3 hours ago, Ashiella said:

No.

 

 

Cinebench r15 > custom thread count

userbenchmark.com

 I fail to see how the "No" related to the quote you selected.

 

Also, not remotely sure what you are saying with Cinebench R15 + userbenchmark at all related to my question, considering I am not asking about benchmarks.

[Jumper118]

Best way to oc is to delid then run all cores at same speed while temps are 20c cooler than they were at pre delid with the cooler in an ice bucket. 

[Lord_Ralex]

1 hour ago, Jumper118 said:

Best way to oc is to delid then run all cores at same speed while temps are 20c cooler than they were at pre delid with the cooler in an ice bucket. 

Rather than delidding, which is definitely not something that is just a "yeah, let's just do this!" for everyone, this again, does not answer the question I really want to know. Way riskier, not something for the faint of heart, and yeah... Be nice, but ain't going to learn how to delid with my only Coffee Lake CPU, and I got no other cheap chips to play with. 

 

The point is not "how high can I get it". It is playing and tuning one of the features of the CPU so that I can have benefit from it, and being able to test said changes.

[Ashiella]

3 hours ago, Lord_Ralex said:

 I fail to see how the "No" related to the quote you selected.

 

Also, not remotely sure what you are saying with Cinebench R15 + userbenchmark at all related to my question, considering I am not asking about benchmarks.

 

7 hours ago, Lord_Ralex said:

A lot of the overclocking testing that I see hits all cores at once, so the features that my CPU has which will let it overclock higher or lower depending on the number of active cores is harder to validate. (As in, it by default clocks to 4.7 when it's one active core, 4.6 or something when it's 2, and so on, so the max clock speed when all cores are in use is lower than when one a few are).

 

I already have my i7 running at 4.6GHz when all cores are in use, but I want to have it run at 5GHz when only 1-3 of the cores are in use, and drop back to roughly 4.6-4.7 when all cores are maxed, to control heat and stability.

 

How does one test the overclock stability for when only a select few cores are active? As in, how can I test that when I run the 5GHz during gaming, it would be stable and heat be reasonable, compared to when all cores are maxed and it throttles down a little more.

 

Is there any software to natively do this, or am I looking at disabling cores to "mimic" this behavior, then using AIDA64 to validate?

"I already have my i7 running at 4.6GHz when all cores are in use, but I want to have it run at 5GHz when only 1-3 of the cores are in use, and drop back to roughly 4.6-4.7 when all cores are maxed, to control heat and stability." <--- This is stupid.

 

"How does one test the overclock stability for when only a select few cores are active? As in, how can I test that when I run the 5GHz during gaming, it would be stable and heat be reasonable, compared to when all cores are maxed and it throttles down a little more." <-- Run Cinebench r15 and check thermals.

[Lord_Ralex]

53 minutes ago, Ashiella said:

"I already have my i7 running at 4.6GHz when all cores are in use, but I want to have it run at 5GHz when only 1-3 of the cores are in use, and drop back to roughly 4.6-4.7 when all cores are maxed, to control heat and stability." <--- This is stupid.

Explain this one, because Intel must be stupid as well, as they do this exact thing. The only difference I want to increase said numbers.

 

54 minutes ago, Ashiella said:

"How does one test the overclock stability for when only a select few cores are active? As in, how can I test that when I run the 5GHz during gaming, it would be stable and heat be reasonable, compared to when all cores are maxed and it throttles down a little more." <-- Run Cinebench r15 and check thermals.

That makes a little more sense.

[Ashiella]

1 hour ago, Lord_Ralex said:

Explain this one, because Intel must be stupid as well, as they do this exact thing. The only difference I want to increase said numbers.

 

That makes a little more sense.

1: Might as well have the same frequency for all cores, it really dont change stability, and heat would only really change 1-6c (which as long as its below 90c its fine)

 

2: yeah, Cinebench you can set a certain number of threads for it to process on and you can check thermals while it works, and its a real-world application.

[Jumper118]

5 hours ago, Lord_Ralex said:

Rather than delidding, which is definitely not something that is just a "yeah, let's just do this!" for everyone, this again, does not answer the question I really want to know. Way riskier, not something for the faint of heart, and yeah... Be nice, but ain't going to learn how to delid with my only Coffee Lake CPU, and I got no other cheap chips to play with. 

 

The point is not "how high can I get it". It is playing and tuning one of the features of the CPU so that I can have benefit from it, and being able to test said changes.

It is turning a screw then applying thermal paste and gluing it back together. A 4 year old does these things at nursery on a regular basis. 

 

Also yes Intel aren't dumb. They have brainwashed you into thing not getting the full value out of your cpu is a feature by not having it run at it maximum speed all the time therefore loosing free extra performance .