[SOLVED] core vs ring ratio

LORYT699

Great
Apr 6, 2022
143
2
95
0
Hi,
The question is:
For gaming is better have increase cpu ratio or ring(cache) ratio?Does ring ratio actual give some visible benefits?
I have an i5 9600k at 4.5(actually I m about to bring it to 4.6 he can cause I tried it)(is at all core btw, I don t know why my mobo do not save oc for each core but only apply on all cores)
the cache is at 4.3 and I think is the default value.
I want to understand if is better increase by 0,1 the cpu core or try to get a 4.5 chace.
I have a voltage limit(I don t want to get more then that), is 1.23 that actually is the sweet spot for my cpu(it will have a max power of 96W).
I ve done several test with cinebench (the last version) and I can go up to 4.7(core) and 4.5 or .6 dont remember(cache) with 1.25v I think, actually I ve binned the paper where I wrote it.
I choosed a lower clock cause I would be more calm about possible stability issiue.
Anyway, for gaming is better a +0.1 on core or a +0.2 on cache?
Till I m here, I would like to talk about stress test for thermals, why cpu-z or chinebench do not make hot my cpu as msi kombustor cpu burn? they are both at 100% but kombustor make the cpu more hot
 

Ralston18

Titan
Moderator
"More influent": details and accuracy matter.

FYI:

https://cybersided.com/cpu-ring-ratio/#:~:text=We can help!,the CPU, including its cache.

There is a lot of dependency on how accurate the various temperatures etc. are being measured plus likewise determining some specific performance change.

Especially when "visible benefits" are subjective in many cases.

And again I am not convinced that generic stress testing is a good basis.

Actually play the game(s) and test. More direct and reduces the number of variables.

In any case, for any given gaming system, the results are likely to vary. There may be some general rules of thumb and/or caveats involved.

Why not do some testing on your own and present the results for others to comment on?

Key is to make only only one change (of any sort) at a time and use a consistent testing method between changes. Make note of the results: good, bad, indifferent.

I will defer to others for further comments regarding real world "cpu ratio vs ring ratio" affects on performance and/or temperatures.

E.g.:

"Yes your results show the temperatures of "X" to be too high".

Or "no, there will be no noticeable visual benefits."

Maybe - "The temperature margin of error is greater than the measured difference in temperatures."

"Problems with thermal paste and/or GPU contact would result in more prominent problems with temperature and performance."

(Etc., etc.. I have no vested interest involved one way or another.)

= = = =

Plus someone may be able to suggest other testing apps and tests that could help you find that "sweet spot" you are seeking.
 

Ralston18

Titan
Moderator
Update your post to include full system hardware specs and OS information.

Include PSU: make, model, wattage, age, condition (original to build, refurbished, used)?

Disk drive(s): make, model, capacity, how full?

The key documents are the motherboard's User Guide/Manual and the CPU's corresponding documentation.

Start with the default/recommended settings and then, as you previously appear to have done, make small changes one at a time as you test changes.

Be sure to keep notes on what you do/did and the results.

Between changes - test by actually gaming - determine if there is really any noticeable performance increases. Stop when you reach the point of no noticeable improvement.

= = = =

"Till I m here, I would like to talk about stress test for thermals, why cpu-z or chinebench do not make hot my cpu as msi kombustor cpu burn? they are both at 100% but kombustor make the cpu more hot "

As for stress tests, they can have some value but only to a point. No one wants an application stress test that creates enough stress to "end badly".....

Which would be the results of true stress testing - when does it really break?

You would need to watch the test using Task Manager, Resource Monitor, Process Explorer, etc. to discover what is actually being done. Run as admin, CPU tab, watch %'s.
 

LORYT699

Great
Apr 6, 2022
143
2
95
0
Update your post to include full system hardware specs and OS information.

Include PSU: make, model, wattage, age, condition (original to build, refurbished, used)?

Disk drive(s): make, model, capacity, how full?

The key documents are the motherboard's User Guide/Manual and the CPU's corresponding documentation.

Start with the default/recommended settings and then, as you previously appear to have done, make small changes one at a time as you test changes.

Be sure to keep notes on what you do/did and the results.

Between changes - test by actually gaming - determine if there is really any noticeable performance increases. Stop when you reach the point of no noticeable improvement.

= = = =

"Till I m here, I would like to talk about stress test for thermals, why cpu-z or chinebench do not make hot my cpu as msi kombustor cpu burn? they are both at 100% but kombustor make the cpu more hot "

As for stress tests, they can have some value but only to a point. No one wants an application stress test that creates enough stress to "end badly".....

Which would be the results of true stress testing - when does it really break?

You would need to watch the test using Task Manager, Resource Monitor, Process Explorer, etc. to discover what is actually being done. Run as admin, CPU tab, watch %'s.
Ok,
There is a misunderstanding here so let s do by steps:
1)For gaming what is more influent?rig/cache ratio, core ratio or a middle step balanced both?(ex. cache 4.6ghz core 4ghz, the inverse or 4.3 to both?)
2)Then, I have tryed this stress test:cpu-z,cinebench r23,cpu burner by msi kombustor.
The first 2 make about the same temp, cpu burner make temp a bit higher(like 2-3 C°).
3)(NEW) I have notice that 2 core of 6 hare more hot then the others by an avg of 3°, this mean that with some the difference is minimal but with others is about 5°, is this normal, could be the thermal grease or the cpu doing a not actually good contact, or maybe they are the centrals core and are wormed by the other 4?

That s all

P.S.
I got a both(cache and core ratio)(temp while gaming 60/73)(max temp on stress test 75°):
i5 9600k @4.6Ghz(core) 4.5Ghz(cache) 1.21v(96W)
Aorus ATC800 (>250W)
msi Tomahawk Z390
Sharkoon silentstorm icewind black 750
2x8 G.Skill TridentZ Royal 3200
Cooler Master TD500l
Nvidia 1060 6GB Asus Dual @ ~2Ghz (TGP 98W)
 

Ralston18

Titan
Moderator
"More influent": details and accuracy matter.

FYI:

https://cybersided.com/cpu-ring-ratio/#:~:text=We can help!,the CPU, including its cache.

There is a lot of dependency on how accurate the various temperatures etc. are being measured plus likewise determining some specific performance change.

Especially when "visible benefits" are subjective in many cases.

And again I am not convinced that generic stress testing is a good basis.

Actually play the game(s) and test. More direct and reduces the number of variables.

In any case, for any given gaming system, the results are likely to vary. There may be some general rules of thumb and/or caveats involved.

Why not do some testing on your own and present the results for others to comment on?

Key is to make only only one change (of any sort) at a time and use a consistent testing method between changes. Make note of the results: good, bad, indifferent.

I will defer to others for further comments regarding real world "cpu ratio vs ring ratio" affects on performance and/or temperatures.

E.g.:

"Yes your results show the temperatures of "X" to be too high".

Or "no, there will be no noticeable visual benefits."

Maybe - "The temperature margin of error is greater than the measured difference in temperatures."

"Problems with thermal paste and/or GPU contact would result in more prominent problems with temperature and performance."

(Etc., etc.. I have no vested interest involved one way or another.)

= = = =

Plus someone may be able to suggest other testing apps and tests that could help you find that "sweet spot" you are seeking.
 

LORYT699

Great
Apr 6, 2022
143
2
95
0
The link was actually usefeull, was exactly what I was searching for.
It say that clock speed is the most important but also the cache speed is, so make they catch together with the clock speed higher by a bit.
The other part of my question, reguarding the stress test, I think you didn t get what I was meaning.
Without talking about temperatures, too high or low, how is possible that 2 full loads(100%) make 2 thermals different, 100% is 100% so the cpu is already working at the max, no?
How can 2 loads(100%) make differents thermals?
P.S.
I have done my oc and is working perfectly, I didn t need help on the actual oc but just in theory questions.
btw, is normal that 2 core of 6 are hotter then the others?
Please simple answare

Anyway thanks for the time and help
 

Ralston18

Titan
Moderator
"Till I m here, I would like to talk about stress test for thermals, why cpu-z or chinebench do not make hot my cpu as msi kombustor cpu burn? they are both at 100% but kombustor make the cpu more hot "

Fair question. Answer not so simple. My thoughts:

What is the source and accuracy of that 100% value? From the stress test tool(s) themselves, Task Manager > CPU tab, or perhaps the Overview tab in Resource Monitor, or somewhere else.

Would you consider the stress test value valid if the end test results never went beyond 95% ? Probably not.

And again - a true stress test breaks things. And most users would not use test software that actually broke the system or proves capable of doing so by accident, misuse, improper configuration, application etc..

What differences exist between the three stress testing tools?

If you run cpu-z, chinebench, and kombuster they are all likely to be doing different things with respect to their stress test methods.

You might discover some significant differences by observing Resource Monitor's CPU tab. Monitor the % 's

Run Resource Monitor as admin. Check all four CPU tab categories: Processes, Services, etc..

Watch what each stress test is doing in each category. No harm in watching other tabs as well.

You can click the small downward pointing arrow in the CPU column headers to establish a descending or ascending sort. (Actually you can do so in any selected tab > column.)

Maybe cpu-z and chinbench are more conservative and scale back on the test as the percentage reaches/nears 100% because the developers do not really want to over stress anything.

Kombuster may be a bit less conservative and somehow force the CPU to work just a bit harder and/or longer. Result being that the CPU (in some manner) gets slightly hotter but scales back when likewise reaching or nearing 100%.

= = = =

btw, is normal that 2 core of 6 are hotter then the others?

If you google "Comparisons of CPU core temperatures" there are quite a number of links discussing the topic.

For example (and you can find other similar links using other search criteria):

https://www.thetechwire.com/one-cpu-core-hotter-than-others/

Bottom line being that there are likely natural differences in core temperatures that can be ignored but greater core temperature differences indicate a problem.
 

ASK THE COMMUNITY

TRENDING THREADS