[SOLVED] 5600x lower than average performance for single core

[Cookiezaddict]

Hi guys,


I have just general question, out of curiousity, maybe it is not even an issue but well…. i will ask the questions anyway

my system are as follow;


CPU: Ryzen 5 5600X

GPU: TUF RX 6800 XT

Motherboard: Strix B550f GamingWifi2

PSU: 850w EVGA G5

CPU Cooler: Be Quiet! Pure Rock 2

Main Storage: 1 TB Samsung 980 M.2

Memory: 16GB 4000mhz cl16 Corsair LPX @ 3800mhz fclk mclk uclk 1:1:1 1900mhz.

Windows 11 Pro

The case is basically more towards single core performance, which could be normal, but slightly lower than what I expected.

So when running stock setting, cinebench r23 will score around 10.9-11.1k on multi core, 1490-1520 on single core. In this situation you can tell the MC result is normal, but the SC is lower than most 5600x

So move to PBO, no boost, ppt up to 100-120w doesn’t matter, tdc 75-90 and edc up towards 120-160, all will basically give me MC results around 11.9-12.1k for cinebench r23, even when CO only set at -5 all cores, no need to go extreme and for that? I am impressed. Without CO, it can only boost to 4.59ghz all cores and score will be slightly lower, around 11.5-11.7k on MC, again it’s normal normal, quiet good.

However, the Single core tho, if not boosted it will never go above 1540, if boosted 100mhz, it can reach 1570, if boosted 200mhz it can reach 1590-1600 on Cinebench R23.

Does anyone else think it’s a low Single Core performance for 5600x? Because online, I see the results of single core for this cpu would be 3-4% stronger than what I could get today. Multi core isn’t my concern here.

I tested on Geekbench, R20 Cinebench, all the same, my results is slightly lower than we can expect from 5600x

no thermal throttle, this set up only runs up to 84c on the most hardcore Prime95 even after 9 hours of stable test. Peak around 69-71’C on Cinebench r23 multicore, and only peak at around 51-53’C on single core test.

My timespy score for cpu test is around 8800 when boosted 100mhz, stays the same when boosted 200mhz, or no boost at all, because apparently it’s kind of multi core test

my 3dMark CPU profile test put me in great position for multi core results, and acceptable for single core result.

I would like to know, not about your 5600x score, but can you spot anything that may cause this?

Does anyone think it is because Win11? If it is win11, what can potentially cause this? Because even after fresh install, bios flash, I tried it all, all the same.

I do not turned my core integrity virtualisation no, so I am very curious what’s causing this

tried all LLC, tried vCore offset, tweak all I can possibly tweak changing global c state, and some things on bios, for weeks trying to get what most online reviewer claim, 1560-1590 single core R23 for 5600x, I can’t just can’t get it. Lol

thanks for checking and reading all the way.

hope I could get someone help to solve the mystery

 

[Darkbreeze]

3-4% can be anything. Different motherboard. Different memory configuration. Clean vs not-so-clean Windows installation and environment.

C-state settings won't have anything to do with CPU performance, only power savings.

You have latest BIOS version?

You've gone to the product page for your motherboard and downloaded, then installed, the latest chipset, audio and network adapter drivers?

Tried a clean install of Windows to ensure it's not related to a cluttered OS installation?

To be honest, I don't think you even HAVE a problem, but making sure you've done all those things and that you have XMP enabled for your memory in the BIOS, are the only things that are really going to make any difference. The rest could be as simple as not the best CPU sample.

And while you might not be seeing actual "throttling", having better CPU cooling will allow boost times to remain higher and boost further, so comparisons between your system with a relatively lower end CPU cooler and other systems with better cooling are probably not really relevant since that alone could be the difference you are seeing.

 

[Cookiezaddict]

3-4% can be anything. Different motherboard. Different memory configuration. Clean vs not-so-clean Windows installation and environment.

C-state settings won't have anything to do with CPU performance, only power savings.

You have latest BIOS version?

You've gone to the product page for your motherboard and downloaded, then installed, the latest chipset, audio and network adapter drivers?

Tried a clean install of Windows to ensure it's not related to a cluttered OS installation?

To be honest, I don't think you even HAVE a problem, but making sure you've done all those things and that you have XMP enabled for your memory in the BIOS, are the only things that are really going to make any difference. The rest could be as simple as not the best CPU sample.

And while you might not be seeing actual "throttling", having better CPU cooling will allow boost times to remain higher and boost further, so comparisons between your system with a relatively lower end CPU cooler and other systems with better cooling are probably not really relevant since that alone could be the difference you are seeing.

Thanks for the reply man…

yes…. Done most thing I can tell, apart from reseating cpu cooler and re paste, but not sure if needed since everything works alright.


I cleaned installed windows like 4 times for the first 4 weeks of owning this lol. Tried this drivers that drivers and yes I flashed my bios, updated last Agesa first thing first before even installing Steam lol oh and I flashed it like twice because so curious. I even found that led app like Armory Crate ruins cpu performance by up to 5% depends on what kind of rgb effects we like to apply

I changed my ram from 16gb 3600mhz, to 32gb, back to 16gb that I currently use all within 2-3 weeks lol. And no I don’t see much difference. The 16gb 3600 was actually the weakest tho, but like 1-2% difference, almost nothing, 16gb 3800 cl16 provide No difference ti my last 32gb 3600 cl 18.

and about cooling problem I am certainly sure nothing like that because I even run 30min Cinebench multi thread, PBO, all cores runs 100%, peak at 74’c slightly more than 10 mins run where it peaks at 71’c, however score relatively the same around 11.9-12k

only single core that confuse me and my cpu runs so cold on single core.

funny that even on 3dMark cpu profile, for 100mhz boost, my result somekind of impressive, again, apart from single core test which seemed to be more towards the average side, so it’s kinda confusing for me

people tend to have problems on all cores boost, struggle to get multi core good results, and here I am confuse why my single core running so bad at stock? 1520 stock is bad while most people got 1560 stock r23 cinebench single. Yes it is not the end of the world and yes it’s like 3% difference I can live with that.

however at the same time, I kinda want to know pretty badly, what’s the cause? And you mentioning cooling got me thinking about that but I am not gonna spend $200 on AIO for 2% single core improvements lol

if what you said about cooling relevant to my case,, maybe the algorithm about boosting single core is different than multi and maybe temperature too measured differently.

Because I do see a movement between fastest core to second fastest core to execute tasks like they kinda ‘rest’ and not able to boost long enough so another fast core needed to take over. Different to multi core boost, it’s never rest, it’s just go 100% on HWinfo, all the way, no drop, that’s why my single core is higher than what most people share online. I get 12k with only -5 or -7 all core CO, and I tried -10 all core and they can run 4.67ghz all cores LOL, gave me 12.1k or 12.2k at cinebench r23 I forgot it was a while ago and it gave me error on prime95 after 7hrs test with hyperthreading hence I just go easy undervolt now around -7 all cores and -5 or -4 on fastest cores I believe.

Also maybe, because of that maybe it creates some sort of delay in execution, which results in say 3% lower performance compared to those with higher end cooling system.

 

[Karadjgne]

There's also 2 different types of ppl. There's those chasing bragging rights and those who don't. Many of the top number posts are the former, they want their score #1, so everything they setup is geared towards absolute performance. That means grade A cooling, hours spent maximizing ram timings, running their 4000 MT/s ram at 3966 just to get the 1933 fclock, disable Antivirus and any other stay resident application or service that could possibly interfere with cpu calculated performance numbers etc etc etc.

Your assumption is that you are missing 3-4% single core based on online results. Reality is that 3-4% only exists because those online results are not real, they are cherry picked and specialty tweaked, just to get the score.

This is a game of throwing hand-grenades. Your throw does not have to be perfect, anywhere even close is perfectly good enough.

A 3-4% performance difference generally equates to 1-3fps at absolute best, so in CSGO if your buddy with a tweaked pc is showing 300fps on average, and you are getting 298fps on average, that's a game changer, right? You'll be able to see the difference?

 

[CountMike]

Some questions. What is highest frequency on single core benchmark ? It's directly tied to it.
How many passes ? On a single pass without waiting few minutes for temps and system to stabilize, +/- 3% is just statistical error, It also may not be running every time on same core or on best core. You could run Core cycler https://github.com/sp00n/corecycler with Ryzen Master running to see which core is best,
General thing about any benchmark. it's thankless to compare with other systems when results are close as you never know which condition that other system is. It's enough for one of systems to have one more or less thread running to skew result by that much.

 

[Cookiezaddict]

There's also 2 different types of ppl. There's those chasing bragging rights and those who don't. Many of the top number posts are the former, they want their score #1, so everything they setup is geared towards absolute performance. That means grade A cooling, hours spent maximizing ram timings, running their 4000 MT/s ram at 3966 just to get the 1933 fclock, disable Antivirus and any other stay resident application or service that could possibly interfere with cpu calculated performance numbers etc etc etc.

Your assumption is that you are missing 3-4% single core based on online results. Reality is that 3-4% only exists because those online results are not real, they are cherry picked and specialty tweaked, just to get the score.

This is a game of throwing hand-grenades. Your throw does not have to be perfect, anywhere even close is perfectly good enough.

A 3-4% performance difference generally equates to 1-3fps at absolute best, so in CSGO if your buddy with a tweaked pc is showing 300fps on average, and you are getting 298fps on average, that's a game changer, right? You'll be able to see the difference?

make sense 😂 and yes mostly that I see are either overclockers or benchmarker and sometimes I heard about setting apps priprity HIGH that thing I never do it because I want ‘pure results’ the only thing I did to CHEAT the score is turning off my LED, then now I replace all my fans to be quiet all blacks and leaving LED only on 2 exhaust fans up the top. 😂

I somehow become excessively obsessed with this single core number man because ffs…

since I saw online that 5600x score 1590 on stock At cinebench r23, I started to see everything can be wrong with my system, while it may just actually perfect and it could be as simple as I chose win11 over 10, and I can’t go 10, cuz I started with 11 but people say 10 actually faster 😂

Mixed review Maybe comparing stuffs online especially benchmark is a bad idea, sometimes

 

[Cookiezaddict]

Some questions. What is highest frequency on single core benchmark ? It's directly tied to it.
How many passes ? On a single pass without waiting few minutes for temps and system to stabilize, +/- 3% is just statistical error, It also may not be running every time on same core or on best core. You could run Core cycler https://github.com/sp00n/corecycler with Ryzen Master running to see which core is best,
General thing about any benchmark. it's thankless to compare with other systems when results are close as you never know which condition that other system is. It's enough for one of systems to have one more or less thread running to skew result by that much.

I only boost 100mhz auto OC, and fastest core are #1 and #3 and they both hit 4750 all the way, but after several second, one core take over the job for quick moment, and the fastest core take it back for a while… and repeat… between those two they keep throwing the tasked every now and then but they both boost all the way to 4750mhz

I did try to boost them 4850mhz, but at current Curve Optimiser I run, -7 all, -5 for 2 fastest cores, they can only hit as high as 4820-4840mhz and I need to be more aggressive, can do it at -12-15 some cores and -9 for 2 fastest cores but even then the score only 1610, max max I reached on single core 200mhz boost, someone who bought my ram once, shared his single core reached 1640mhz on 5600x 200mhz boost lol

So maybe cooling, maybe

one thing I am curious is, does single core run suppose to be shared by 2 cores like this below, or just one core all the way? And what the reason they keep throwing job between each other? Temp? It’s only 51’c tho

both are running single core r23 cinebench

in this clip #1 do the main job

https://youtube.com/shorts/hg0AetPwNnM?feature=share

In this clip somehow task executed by core #3 ,

https://youtube.com/shorts/tfHTB8pt39Y?feature=share

However, if checking HW info, in general whatever the pattern of execution the cpu does, effective core clock will generally stays the same around say 700-900mhz fluctuating like that and score are always the same too, for the single core which is normal as it is lightly threaded so effective clock tend to be much lower than the ‘clock’

 

[Darkbreeze]

LEDs don't have ANYTHING, AT ALL, to do with system performance. Nothing.

 

[Cookiezaddict]

LEDs don't have ANYTHING, AT ALL, to do with system performance. Nothing.

lol. Apparently yes it does. I test different aura effect on multiple benchmark they are consistent with cpu performance drop, super small, max at 5%, in average 2%. Depends on the effect.

View: https://youtu.be/NzAeAFudylI


same with icue and all other bloatware apps. I do believe some people love seeing their pc look great and don’t care much about 2% performance drop, which eventually nothing for high end rig, even for me too.

It’s just, I feel curious, if I miss something because there are so many people sharing their 5600x literally score 1560-1590 bone stock 😂 on single core test R23 cinebench. Me, 1520-1540, and even boosted 100mhz, only 1570, and need 200mhz single core boost to even touch 1590-1600.

View: https://youtu.be/WDPGSj7UtA0

 

[Cookiezaddict]

Again, I am here discussing very small margin of performance, for those who may be interested or could help. And there will be dozens of article and contents that show how argb / rgb / led effect , especially the heavier effect that requires more cpu utilisation, will just drop cou performance, by tiny amount

 

[kerberos_20]

5600x has max boost clock 4.6GHz
and yes ram will affect cpu performance aswell (when ap doesnt fit in cpu cache and has to move data to ram, it causes cpu delays)
can you post userbenchmark score?

anyway
View: https://www.reddit.com/r/Amd/comments/jtaokr/5600x_cinebench_r23_multi_11540_single_1571/

here people sayin +200 mhz PBO = around 1600 single core cb 23

 

[Cookiezaddict]

5600x has max boost clock 4.6GHz
and yes ram will affect cpu performance aswell (when ap doesnt fit in cpu cache and has to move data to ram, it causes cpu delays)
can you post userbenchmark score?

anyway
View: https://www.reddit.com/r/Amd/comments/jtaokr/5600x_cinebench_r23_multi_11540_single_1571/

here people sayin +200 mhz PBO = around 1600 single core cb 23

https://imagizer.imageshack.com/img923/3808/rTKE51.jpg

saw that reddit post since forever and that’s one thing that got me wonder, I assume that OP was stock, and he got 1570 lol. About userbenchmark I always got it between 96-100 percentile, depends on certain stuffs like if I got things running in the background or not. But yeah I could get higher score there, but I don’t think their cpu test is good.

 

[Darkbreeze]

Sorry man, but no, it doesn't. Because if it DID we'd be hearing about it front page on every tech review site for the past five years. And, we haven't. In fact, it's not even possible. And I AM talking about the hardware side of the LEDs and ARGB lighting.

The SOFTWARE on the other hand, if you are using an installed desktop utility, especially any of the major ones like CAM and iCue, then sure, like any application or utility that is running all the time there will be some resource usage but it would generally be fairly minimal.

On the hardware side however, the lighting itself has zero effect on system performance. Many systems don't require the use of a desktop utility to configure your lighting the way you want, and additionally there are VERY lightweight open source replacement control utilities for people who feel they MUST have control from the desktop for most of these systems.

And if you want to look at it like that, then you are simply trying to chase bragging rights because the average person isn't going to just remove all the things that make their PC useful or desirable to them just to gain a few percentage points worth of performance. I mean, sure, the whole thing will be faster if you simply install Windows, AND NOTHING ELSE, but what is the point of that? There is no point, unless you are in competition or chasing bragging rights. Simple as that.

Otherwise, like I said before, get better cooling so your cores will boost to higher clocks and sustain them longer, maybe tighten your memory timings a bit and then just use the system for whatever it is you use it for. The rest is just, well, not really what we are here to try and help with.

 

[Unolocogringo]

Out of the box my 5600x did 4.65 on 1-2 core boost. All are different but you are guaranteed 4.6
After installing a Pure Rock 2 and many hours of Curve optimizer tweeking she will now do 4.65 1-4 core boost and 4.5 all core boost on super heavy loads and 4.6 on lighter loads.
I can get 100mhz more out of it but that require too much voltage for 24/7/365 usage.
Moral of the story is if you are running stock settings expect your scores to be lower if comparing against online scores.

 

[Unolocogringo]

So for grins and giggles I ran R23 to check scores.
Single core 1527
Multi core 11581
MP ratio 7.59
Within a few percentage points of your system.
Better cooling makes a huge difference in all core clock speeds.

This is with the browser still open and all of my normal programs running.
Folding but paused. AI suite for fan controll. MSI afterburner, Malware Bytes.
I can probably increase that a few percent by closing down all unneeded services and all programs.
But this is a Folding/gaming rig, not a highest score at all cost rig.
Plus my systems run folding @ home 24/7/365 so Stability/reliability/performance are my top 3 concerns.
Not a synthetic benchmark score.

You might can get a little more out of your memory @1933 as stated above. Still runs 1/1/1.
Probably get a little more out of the processor with a lot of trial and error running Curve optimizer and testing one core and one setting at a time.
I wound up with a -29 value for all cores. Your chip will most likely be different.
Increasing LLC slightly can also help stability at higher boost on lower voltages.
You just have to play with your system to find its limits.

For running benchmarks I can get another 200mhz out of her ,but she is not 100% stable at that point but will run and complete almost all benchmarks .

 

[Darkbreeze]

Better cooling makes a huge difference in all core clock speeds.

Exactly.

 

[Cookiezaddict]

Sorry man, but no, it doesn't. Because if it DID we'd be hearing about it front page on every tech review site for the past five years. And, we haven't. In fact, it's not even possible. And I AM talking about the hardware side of the LEDs and ARGB lighting.

The SOFTWARE on the other hand, if you are using an installed desktop utility, especially any of the major ones like CAM and iCue, then sure, like any application or utility that is running all the time there will be some resource usage but it would generally be fairly minimal.

On the hardware side however, the lighting itself has zero effect on system performance. Many systems don't require the use of a desktop utility to configure your lighting the way you want, and additionally there are VERY lightweight open source replacement control utilities for people who feel they MUST have control from the desktop for most of these systems.

And if you want to look at it like that, then you are simply trying to chase bragging rights because the average person isn't going to just remove all the things that make their PC useful or desirable to them just to gain a few percentage points worth of performance. I mean, sure, the whole thing will be faster if you simply install Windows, AND NOTHING ELSE, but what is the point of that? There is no point, unless you are in competition or chasing bragging rights. Simple as that.

Otherwise, like I said before, get better cooling so your cores will boost to higher clocks and sustain them longer, maybe tighten your memory timings a bit and then just use the system for whatever it is you use it for. The rest is just, well, not really what we are here to try and help with.

well I don’t know about the bragging but I feels like I am some kind of entitled to that stock ‘listed benchmark’ lol. It’s just all that, that’s why I don’t boost it to 200mhz Auto OC because it is very very easy for me to do and I did it, my cou can run all core at 4.67-4.68 and single core boosted all the way to 4850mhz, by only doing -15 up to -20 on some cores, -9 and -12 fastest cores, voltage offset to make sure it gets at least 1.35v for Prime95 stable, done. I did that, the first week of owning the PC, get 12.2k all cores, or maybe 12.1k I forgot, with ARGB lights all on set by Armory Crate, single core I scored I believe 1610, same, ARGB sync and on.

However, I see no points of doing that, because when cpu loads up, the best noise from fans, to be temperature wise, I found running 4.5-4.6ghz all core is best, can maintain temp without aggressive fans setting.

and for the gaming side, I tried 200mhz vs 100mhz actually it provides minimum real world difference maybe because my cooling. For bench mark, it does provide me bit of a jump like 3-4% just 100mhz different but for gaming, I barely notice 5fps jump, maybe 2-3fps, but almost nothing.

again, instead of calling it bragging side, I feel like being simply curious, why some people claim they score 1590 stock on single core? Countless of people say they score 1570 single core, stock stock…. While me, if all stock, best of the best, I may get 1540 and that’s something, it’s like, I lost 2% not knowing what caused it while at the same time, I can ‘BRAG’ about my multicore being super good, stable, score higher than those benchmarkers claim. Most of them I saw scored 1570-1590 pbo no boost, scored 11.5-11.9k multi cores, I never hit below 11.7k and I need to run internet explorer, armory sync, to get it as low as 11.7k.

saying that? I am kind of happy to learn, eager to learn if there is way to sacrifice this multi core performance, I’ll sacrifice 4% multi to get extra 2% single core, just to learn something that I can learn. Because I believe if I speak with their engineer now they’d know what need to be done. 😂 I need to solve this mystery, badly

 

[Cookiezaddict]

Exactly.

my all core clock speed can go as high as 4.67-4.69 at cinebench R23, it can be stable at prime too I actually tested it and it is said to be the best we can go without throwing too much on cooling and tweaking too much on voltage.

i have ZERO, literally Zero for my multicore performance, if any, it’s something I can brag about 😂 my confusion is for my single core, it feels like S* seeing my stock benchmark score like 2-3% lower than what people claim they score stock.

 

[Cookiezaddict]

So for grins and giggles I ran R23 to check scores.
Single core 1527
Multi core 11581
MP ratio 7.59
Within a few percentage points of your system.
Better cooling makes a huge difference in all core clock speeds.

This is with the browser still open and all of my normal programs running.
Folding but paused. AI suite for fan controll. MSI afterburner, Malware Bytes.
I can probably increase that a few percent by closing down all unneeded services and all programs.
But this is a Folding/gaming rig, not a highest score at all cost rig.
Plus my systems run folding @ home 24/7/365 so Stability/reliability/performance are my top 3 concerns.
Not a synthetic benchmark score.

You might can get a little more out of your memory @1933 as stated above. Still runs 1/1/1.
Probably get a little more out of the processor with a lot of trial and error running Curve optimizer and testing one core and one setting at a time.
I wound up with a -29 value for all cores. Your chip will most likely be different.
Increasing LLC slightly can also help stability at higher boost on lower voltages.
You just have to play with your system to find its limits.

For running benchmarks I can get another 200mhz out of her ,but she is not 100% stable at that point but will run and complete almost all benchmarks .

thanks for the explanation man…

I don’t go aggressive on CO, I found that when I use -22 or even -24 on some cores, and like -12 or so on fastest cores, my system can be ‘stable’ on Prime without AVX and Without Hyperthreading. But with Hyperthreading and all hardcore test, I figured I can choose either running aggressive CO but provide VCore offset, or, just do minimum CO, and I choose minimum CO. Only around -10 all cores -7 and -5 for each fastest cores, that’s for 200mhz boost good, but for 100mhz boost I go -7 all cores and around -4 or -5 on fastest cores.

the difference is marginal really, very tiny and all more towards self satisfaction; curiousity… nothing about bragging or comparing my results much towards others. But I just feel like,

do people really score 1570-1590 stock no boost no auto oc on single core!? I am dead keen to know that and you just shered 1527 score which makes me feel my cpu could be just normal now 😂

 

[Unolocogringo]

thanks for the explanation man…

I don’t go aggressive on CO, I found that when I use -22 or even -24 on some cores, and like -12 or so on fastest cores, my system can be ‘stable’ on Prime without AVX and Without Hyperthreading. But with Hyperthreading and all hardcore test, I figured I can choose either running aggressive CO but provide VCore offset, or, just do minimum CO, and I choose minimum CO. Only around -10 all cores -7 and -5 for each fastest cores, that’s for 200mhz boost good, but for 100mhz boost I go -7 all cores and around -4 or -5 on fastest cores.

the difference is marginal really, very tiny and all more towards self satisfaction; curiousity… nothing about bragging or comparing my results much towards others. But I just feel like,

do people really score 1570-1590 stock no boost no auto oc on single core!? I am dead keen to know that and you just shered 1527 score which makes me feel my cpu could be just normal now 😂

If your plan is high scores, you need to do a minimal windows install,a windows 10 lite edition.
This has all of the unnecessary windows services and gadgets removed. And can improve scores a few more percent.
Also setting core affinity to your fastest core/or dissabling all other cores. so that windows does not swap cores during the single core test can add another percent or 2.
So yes the higher scores are obtainable, IF you are willing to make the hours of tweeking happen. And run a stripped down version of windows to make it happen.
There are lots of false/deceiving information on the internet.

 

[Unolocogringo]

With CO you are telling the processor to boost to higher frequency's while using less Voltage.
Which in turn produces less heat/amperage/wattage at the same clock speed or more speed at the same heat/amperage/wattage draw.
So aggressive CO is your friend if you take the hours needed to fine tune it.

 

[Cookiezaddict]

If your plan is high scores, you need to do a minimal windows install,a windows 10 lite edition.
This has all of the unnecessary windows services and gadgets removed. And can improve scores a few more percent.
Also setting core affinity to your fastest core/or dissabling all other cores. so that windows does not swap cores during the single core test can add another percent or 2.
So yes the higher scores are obtainable, IF you are willing to make the hours of tweeking happen. And run a stripped down version of windows to make it happen.
There are lots of false/deceiving information on the internet.

well I am running Win 11 pro too now, lots of blowtware and ‘bloat’ features there maybe that explains. 😂

 

[Cookiezaddict]

With CO you are telling the processor to boost to higher frequency's while using less Voltage.
Which in turn produces less heat/amperage/wattage at the same clock speed or more speed at the same heat/amperage/wattage draw.
So aggressive CO is your friend if you take the hours needed to fine tune it.

I know how the CO works but trust me most aggressive CO give impression in term of benchmarking performance but sadly it will fail Prime95 with all the test include hyperthreading in no time. And All I aim is normal single core score around 1690 stock or 100mhz boost, not like 1640 single core score boosted.

 

[geofelt]

Single core benchmarks will vary, depending on the instruction set used.
Such instructions may have little resemblance to your games or whatever the single thread performance is used to achieve.
Best to not obsess about it unless something is really off.

As another test, try the cpu-Z bench and look at the single thread performance rating.
It should be about 643:
https://valid.x86.fr/bench/rsf5p1/1

Again, realize that the cpu-z test submissions may reflect higher than normal performance.

 

[Darkbreeze]

Let's make one thing clear. ANY and ALL "scores" that you are seeing where people have submitted the results of their benchmarks for the same CPU you have DO have PBO and boosting enabled. If you don't, you are never going to have the same scores, and again, even if you DO, you might not depending on how much memory you have installed, how fast your memory is, how tight your timings are, what motherboard you have (Yes, JUST the motherboard can have up to a 5% different in performance with all other things being equal), what other hardware you have installed and exactly what software you have installed. So, by all means, make every effort to eek out every ounce of performance you can for your own system, but base your results against your own system, not somebody else's. You'll just be trying to catch smoke.