Did I lose the Silicon Lottery or am I missing something (i5-9600k)

[JoshOohAh]

I bought myself an i5 9600K with the intention of OCing to that magical 5GHz that everyone else seems to be able to hit. I've seen a lot of people saying they easily hit 5GHz or higher while <1.4V. I'm reaching out for help because I'm hoping that I'm doing something wrong as it's been over 10 years since I last OC'd anything. So far I can't seem to get anything beyond 4.6GHz that's stable.

 

This PC is primarily for gaming but more specifically, sim racing (iRacing) in VR. I play competitively and I'm doing some longer stability tests to make sure I don't run into any issues mid-game. iRacing has a neat feature where it can tell you the time your CPU spent rendering and the time your GPU spent rendering (in mS) and I'm actually much more CPU limited as this game is heavily reliant on single core performance. I also believe iRacing (or SteamVR) uses an AVX instruction set because whenever I set an AVX offset, my clock speed drops as soon as the game boots up.

 

My methodology has been pretty similar to this thread:

1. Prime 95 Small FFT no AVX for ~15 minutes
2. Real Bench Stress Test 8 Hour 16GB
3. Prime 95 Blend w. AVX (8 Hours)
4. Increase Ring and repeat 3

If the tests fail at any point, I would adjust the voltage as needed. Based off my temperatures for a long term OC, I would like to stay at or below 1.36V which would give me a maximum of 90degC after hours of Prime95 Blend w. AVX and a maximum of <75degC when actually gaming. I'd consider getting some better thermal paste or even deliding the CPU to push it a few extra mV but my CPU does not seem happy at anything above 4.8GHz even temporarily.

What I've tried:

• 5GHz up to 1.43V (just to see) fails test 1 immediately, thermals were <90degC okay when it BSOD.
• 4.9 GHz up to 1.4V (just to see) fails test 1 immediately, don't remember the thermals.
• 4.8 GHz up to 1.36V after several hours on test 3, a couple cores error on prime95 but no BSOD. Peak of 90degC, equilibrium is in the mid 80s.
• 4.7 GHZ up to 1.36V after several hours on test 3, a couple cores error on prime95 but no BSOD. Peak of 90degC, equilibrium is in the mid 80s.
• 4.6 GHZ passes all at 1.32V. Peak of high 80s, equilibrium low 70s. Currently trying lower voltages.

For all my OC I had the following settings:

• XMP: On
• Intel C-State: Disabled
• Enhanced Turbo: Disabled
• LLC: "Mode 3"
• Ring: Auto (4.3GHZ)
• AVX offset 0 (unless specified)

The rest of the rig:

• i5 9600K
• MSI MPG Z390M Gaming Edge AC
• be quiet! Dark Rock Pro 4
• Corsair RM650x 80+ Gold
• Corsair Vengeance DDR4 3200MHz C16
• Fractal Meshify C Mini w.
  • 2x 140mm in front, 1x 120mm top, 1x 120mm rear
• Ambient Air ~20degC
• 2070 Super

Am I missing something?

[Bombastinator]

Might be neither.  One thing I noticed was that while there were many claims that my 4770k should do a much higher OC than the one I got,  I also had fairly stringent stability standards.  Part of the issue with OC is it seems that people have different opinions on what constitutes “rock solid stable”.  If they’re higher it seems the capacity of an overclock goes down sometimes. 

[JoshOohAh]

14 minutes ago, Bombastinator said:

Might be neither.  One thing I noticed was that while there were many claims that my 4770k should do a much higher OC than the one I got,  I also had fairly stringent stability standards.  Part of the issue with OC is it seems that people have different opinions on what constitutes “rock solid stable”.  If they’re higher it seems the capacity of an overclock goes down sometimes. 

 

I hope that's the case. Part of my reason for posting this thread was to try and see if I'm being unnecessarily strict on my definition of stable or if I just got bad luck

[xg32]

5 minutes ago, JoshOohAh said:

 

.

irc the lowest bin was around 4.7 avx, alot of 5ghz claims were bs, those are top 30%

[JoshOohAh]

22 minutes ago, xg32 said:

irc the lowest bin was around 4.7 avx, alot of 5ghz claims were bs, those are top 30%

Good to hear, I thought there might be something wrong with my methodology because I really haven't found many results below 5GHz and I wasn't sure the variance between the different bins.

 

Just bad luck it seems like ...

[Bombastinator]

37 minutes ago, JoshOohAh said:

 

I hope that's the case. Part of my reason for posting this thread was to try and see if I'm being unnecessarily strict on my definition of stable or if I just got bad luck

You have more stringent standards than I did which turned out to be more stringent than a lot of people claiming rock solid stability. I’m not sure your standards are stringent for you though. 
 

You have an unusual situation as a competition player.  The thing not crashing is a livelihood. Conversely if it’s not running at the speed of your competitors that is equally unacceptable.  Which is more important?  It may be that it might be worth rmaing the chip just to see if you get a better draw in the silicon lottery. 

[papajo]

OCing due to so many contributing factors is more like an empirical art than a strict set of rules you can apply in any situation. 

 

 

You mentioned all the settings you used but you did not mention two important factors. 

 

 

Time (as in duration) and thermals during the duration of those tests. 

 

Like when exactly did it crash @ 4.8GHz 1.36v at step 3? and where the temps gradually raising? or imidiate peaking? at what max temp etc? 

 

 

Also what's the airflow on your case? 

 

 

 

[JoshOohAh]

1 hour ago, papajo said:

OCing due to so many contributing factors is more like an empirical art than a strict set of rules you can apply in any situation. 

 

 

You mentioned all the settings you used but you did not mention two important factors. 

 

 

Time (as in duration) and thermals during the duration of those tests. 

 

Like when exactly did it crash @ 4.8GHz 1.36v at step 3? and where the temps gradually raising? or imidiate peaking? at what max temp etc? 

 

 

Also what's the airflow on your case? 

 

 

 

 

Sorry, sort of indirectly mentioned the thermals but not per test. I've updated the main thread accordingly.

 

• 5GHz up to 1.43V (just to see) fails test 1 immediately, thermals were <70degC and rising when it BSOD.
• 4.9 GHz up to 1.4V (just to see) fails test 1 immediately, don't remember the thermals.
• 4.8 GHz up to 1.36V after several hours on test 3, a couple cores error on prime95 but no BSOD. Peak of 90degC, equilibrium is in the mid 80s
• 4.7 GHZ up to 1.3V currently running 3 and seems fine after an hour with peak temps >80degC and equilibrium mid 70s.

Airflow seems fine, like mentioned in the thread, there's 2x intake 140mm in the front, 1x exhaust 120mm on the top, and 1x exhaust 120mm on the back.

[papajo]

13 minutes ago, JoshOohAh said:

• 4.8 GHz up to 1.36V after several hours on test 3, a couple cores error on prime95 but no BSOD. Peak of 90degC, equilibrium is in the mid 80s
• 4.7 GHZ up to 1.3V currently running 3 and seems fine after an hour with peak temps >80degC and equilibrium mid 70s.

Sounds like you could become more stable by deliding your CPU use conformal coating on the entire surface other than the core chip (also make sure its a thin coating with no exposed areas or air bubles by reflecting some intense light on the chip to be sure everything is coated last but not least before proceeding leave the coating to cure for a few hours  ) use liquid metal and then youll be fine, if possibe buy a cooper ihs as well.

[Mateyyy]

3 minutes ago, papajo said:

Sounds like you could become more stable by deliding your CPU use conformal coating on the entire surface other than the core chip (also make sure its a think coating with no exposed areas or air bubles by reflecting some intense light on the chip to be sure everything is coated last but not least before proceeding leave the coating to cure for a few hours  ) use liquid metal and then youll be fine, if possibe buy a cooper ihs as well.

OP's temps are acceptable at 1.36V 4.8GHz. Assuming they can get it stable at around 1.375V, those temps should still be in check.

I don't see how it'd be worth delidding the chip in this case, especially considering it's soldered, as opposed to using paste between the die and IHS.

[JoshOohAh]

8 minutes ago, Mateyyy said:

OP's temps are acceptable at 1.36V 4.8GHz. Assuming they can get it stable at around 1.375V, those temps should still be in check.

I don't see how it'd be worth delidding the chip in this case, especially considering it's soldered, as opposed to using paste between the die and IHS.

That's my thought as well. If I could get it to run at a higher clock speed, I would consider it, but it doesn't seem worth the risk/trouble.

[papajo]

2 hours ago, Mateyyy said:

OP's temps are acceptable at 1.36V 4.8GHz. Assuming they can get it stable at around 1.375V, those temps should still be in check.

I don't see how it'd be worth delidding the chip in this case, especially considering it's soldered, as opposed to using paste between the die and IHS.

There are improvements to be done especially if he changes the IHS to a better cooper one that has been made for that specific reason by small tool-shops (sold on ebay or elsewhere) 

 

In this video we can see that by keeping the factory IHS the temperature drop was about 5C

 

The reason I proposed deliding (which ok is not for the faint of heart but OP expressed the desire to OC the cpu and along with his description which shows that he spent considerable time I think he could have the determination to do such a task by his own risk of course) 

 

Is that the average temp is low but specific core temps are high, this more often than not indicates bad contact which even soldered chips could have minuscule difference in the surface between the core and IHS (as for example solder density or tiny gas/air bubbles created during manufacturing) can contribute to that. 

 

OP is almost there at 4.8 GHz running stable for at least several hours as he claims having only a few cores reaching 90ish degrees a single -5 drop can do wonders here (especially if he changes IHS as well which will contribute to a bigger delta)  In fact I wouldnt be surprised if he manages to hit 5GHz after that at under 1.4 volts (besides deliding though a nice application of a good thermal paste between the IHS and the cooler is needed as well preferably a kryonaut level of paste or better. ) 

 

Whenever my temps are low but specific core temps were high deliding helped the case. 

 

[Mateyyy]

26 minutes ago, papajo said:

Is that the average temp is low but specific core temps are high, this more often than not indicates bad contact which even soldered chips could have minuscule difference in the surface between the core and IHS (as for example solder density or tiny gas/air bubbles created during manufacturing) can contribute to that. 

That's true, but an uneven IHS could also do that, where lapping would be useful.

Still, would I recommend that for a 9600K with very average overclocking potential by the looks of things? Probably not.

27 minutes ago, papajo said:

OP is almost there at 4.8 GHz running stable for at least several hours as he claims having only a few cores reaching 90ish degrees a single -5 drop can do wonders here (especially if he changes IHS as well which will contribute to a bigger delta)  In fact I wouldnt be surprised if he manages to hit 5GHz after that at under 1.4 volts

90C does not impact stability. These chips start throttling at 100C by default.

[Bombastinator]

29 minutes ago, Mateyyy said:

That's true, but an uneven IHS could also do that, where lapping would be useful.

Still, would I recommend that for a 9600K with very average overclocking potential by the looks of things? Probably not.

90C does not impact stability. These chips start throttling at 100C by default.

There are many reasons. Just to throw another in all cores are not created equal.  It appears that the way binning is done is some cores just fail.  Other cores work but not well enough to pass and are turned off anyway, others are still weak but good enough to pass minimums.  These might be left on.  Several, or even one weak core, could mean a lead chip.  A gold mid grade chip could still have several bad cores that are turned off but the ones that are on are good ones.  This is why the 10700 has a poor rep.  It’s a 10900 with 2 cores turned off.  A flawed 10900. So it’s a chip with problems but not enough problems to make a 10600 out of it.   The 10600 has a better rep because it’s a 10900 with its worst 4 cores turned off. This seems to mean that the 6 cores left are often pretty decent ones.  A lead 10700 would be a chip that narrowly escaped being a 10600.

[JoshOohAh]

Thanks all for the input!

 

I just got home and noticed that I had a couple cores failed test 3 at 4.7GHz 1.3V, max temp got to 90 on one of the cores but rest maxed out around the mid 80s. I have a bit more headroom to play with here and I'll try 4.7GHz, 1.32V tonight and see. Maybe 4.7GHz is going to be unobtainable...

[cried22]

9 hours ago, JoshOohAh said:

I bought myself an i5 9600K with the intention of OCing to that magical 5GHz that everyone else seems to be able to hit. I've seen a lot of people saying they easily hit 5GHz or higher while <1.4V. I'm reaching out for help because I'm hoping that I'm doing something wrong as it's been over 10 years since I last OC'd anything. So far I can't seem to get anything beyond 4.7GHz that's stable.

 

This PC is primarily for gaming but more specifically, sim racing (iRacing) in VR. I play competitively and I'm doing some longer stability tests to make sure I don't run into any issues mid-game. iRacing has a neat feature where it can tell you the time your CPU spent rendering and the time your GPU spent rendering (in mS) and I'm actually much more CPU limited as this game is heavily reliant on single core performance. I also believe iRacing (or SteamVR) uses an AVX instruction set because whenever I set an AVX offset, my clock speed drops as soon as the game boots up.

 

My methodology has been pretty similar to this thread:

1. Prime 95 Small FFT no AVX for ~15 minutes
2. Real Bench Stress Test 8 Hour 16GB
3. Prime 95 Blend w. AVX (8 Hours)
4. Increase Ring and repeat 3

If the tests fail at any point, I would adjust the voltage as needed. Based off my temperatures for a long term OC, I would like to stay at or below 1.36V which would give me a maximum of 90degC after hours of Prime95 Blend w. AVX and a maximum of <75degC when actually gaming. I'd consider getting some better thermal paste or even deliding the CPU to push it a few extra mV but my CPU does not seem happy at anything above 4.8GHz even temporarily.

What I've tried:

• 5GHz up to 1.43V (just to see) fails test 1 immediately, thermals were <90degC okay when it BSOD.
• 4.9 GHz up to 1.4V (just to see) fails test 1 immediately, don't remember the thermals.
• 4.8 GHz up to 1.36V after several hours on test 3, a couple cores error on prime95 but no BSOD. Peak of 90degC, equilibrium is in the mid 80s.
• 4.7 GHZ up to 1.3V currently running test 3 and it seems fine after an hour with peak temps in the mid 80s and equilibrium mid 70s. I'll check it when I get home but I feel I at least have more voltage headroom if needed.

For all my OC I had the following settings:

• XMP: On
• Intel C-State: Disabled
• Enhanced Turbo: Disabled
• LLC: "Mode 3"
• Ring: Auto (4.3GHZ)
• AVX offset 0 (unless specified)

The rest of the rig:

• i5 9600K
• MSI MPG Z390M Gaming Edge AC
• be quiet! Dark Rock Pro 4
• Corsair RM650x 80+ Gold
• Corsair Vengeance DDR4 3200MHz C16
• Fractal Meshify C Mini w.
  • 2x 140mm in front, 1x 120mm top, 1x 120mm rear
• Ambient Air ~20degC
• 2070 Super

Am I missing something?

I built a system for my GF using this chip and cannot get over 4.8. Seems pretty normal. If you bought it from amazon you can return and try for a better chip.

[JoshOohAh]

So doesn't look like 4.7GHz is going to pass Prime 95 Blend test at any voltage. A couple cores kept erroring out during the test.

 

I ran 8 hours of 4.6GHz at 1.32V overnight and it seems stable. I'm just tweaking to see how low I can get the voltage now. I'm pretty disappointed with these results overall.

[Mateyyy]

6 minutes ago, JoshOohAh said:

So doesn't look like 4.7GHz is going to pass Prime 95 Blend test at any voltage. A couple cores kept erroring out during the test.

 

I ran 8 hours of 4.6GHz at 1.32V overnight and it seems stable. I'm just tweaking to see how low I can get the voltage now. I'm pretty disappointed with these results overall.

Eh, I mean, it's passable. The stock all-core boost is 4.3GHz on that chip so it definitely could've been worse.

I need to shove a lot of voltage into my 9900K even for just 4.9GHz, compared to others I've seen on the internet, but it is what it is.