[Placekicker19]

You mean on intel or AMD ?

Yes the intel

 

[dr.Rafi]

Having not yet figured out adaptive voltage or offsets, I am trying to decide which way to go for a 24/7 OC. I am on a 10900K (SP 75) with an Asus Z490-E Gaming (0707 bios) and X73 AIO (360mm).

I am stable at 5.1 with LLC 5 and bios vCore set at 1.395. Temps at 30 minutes P95 Small FFT No AVX and 8 Hours Real Bench are 82-87. Under load, vCore is 1.350 and VID is 1.359. Running games, the vCore is 1.376, VID 1.325 and temps sit in the high 40s.

If possible, I would like to not have the CPU idling at this voltage. So, I figured I would try just setting the core ratio to 51 and leaving voltage and LLC to auto. Two minutes into P95 (Small FFT No AVX) temps were 91 to 95. vCore under load was 1.430 and VID was 1.395. While gaming, vCore was 1.483, VID was 1.490, and temps were in the mid 50s.

Obviously, I am not going to be stress testing my PC all the time, so I do not think the temps on P95 are a huge concern on auto. My rig is used for light productivity (MS Office) and gaming. When it comes to longevity, do you think I should go with the manual voltage at 1.395, or the auto voltage? Manual is cooler with lower voltage under load, but it sits at 1.395 all the time. Auto is hotter and higher voltage when gaming, but obviously the voltage drops off considerably while idle. For what its worth, I kept my last machine for 7 years, and would like to try to do the same with this one.

Any thoughts would be appreciated.

Same thing here still not sure how to set offcet or adaptive voltage

 

[dr.Rafi]

Yes the intel

cash 48x with 51 cpu and ram [email protected] 16 16 16 36 370 2t and not any tweaking with secondary and tertiary , but graphic is rtx 3080 watercooled , shunted to 525 max real power which is boostong 2130-2115 during whole bench of gear5

 

[dr.Rafi]

I use -50% vdroop setting, its equivalent to asus llc6, gigabyte llc turbo . Asus reports actual die voltage, Msi and gigabyte report vrout, which is the same as die voltage. Evga doesn't read die voltage, so voltage will read higher than the way other board manufacturers read voltages.

I know on the evga z390 dark setting -50% vdroop with 1.25v resulted in a actual die voltage of 1.213. If the z490 is anything like the z390 my actual die voltage on 1.27 set it bios would be 1.223v. I wish the dark read die voltage that way you could compare it to other boards. Im not stable at 1.27, it was just my minimum set voltage it took to pass 3 b2b runs of r20 without error.

You seem to have gotten a nice sample of a 10850k, have you attempted higher clocks yet?

Its 10900kf , @5.2 ghz with asus z490 formula, vcore set to 1.28 v bios fixed,ll7, ram to 4000 16161636 , cash 48x ,real bench stable 15min,max temp 78c but mostly 68 to 73c stable, prime 95 72c max no avx load, aida test any combination of fpu, cash, memory,cpu is stable. load voltage drop to 1.24 in heavy loads, 1.254v gaming,its Sp63 chip , and asus prediction is not correct, showing 53x vcore should be 1.5v vwhich is not true at all

 

[LesPaulLover]

Quick question for yall: How bad are 10850K in terms of binning compared to the full 10900k CPUs?

 

[munternet]

Quick question for yall: How bad are 10850K in terms of binning compared to the full 10900k CPUs?

No guarantee of 5GHz from all accounts

 

[Placekicker19]

Its 10900kf , @5.2 ghz with asus z490 formula, vcore set to 1.28 v bios fixed,ll7, ram to 4000 16161636 , cash 48x ,real bench stable 15min,max temp 78c but mostly 68 to 73c stable, prime 95 72c max no avx load, aida test any combination of fpu, cash, memory,cpu is stable. load voltage drop to 1.24 in heavy loads, 1.254v gaming,its Sp63 chip , and asus prediction is not correct, showing 53x vcore should be 1.5v vwhich is not true at all

Nice, how much voltage does 5.3 with 5 cache require ? Ive seen bad sp63 chips and awesome ones. I bought 2 chips and the one that had quite a bit lower vid could not pass cinebench r15 @ 5.3ghz without L0 cache errors, tested up to 1.45vs. The chip with the higher vid scaled much better with voltage. They both could boot 4800mhz cl19 ddr4.

 

[ExDarkxH]

I have a 10900kf and did direct die with liquid metal on a optimus foundation block

temps are phenomenal
Currently doing 5.2Ghz @ 1.375v with LLC set to 4
with cinebench r15 im topping out at 56c

HOWEVER, cinebench crashes when i try 5.3ghz. its definetly not the temps i think because LLC is at 4 its not getting enough power to handle 5.3

I tried up to 5.4Ghz on games without any issues though
should i just ignore cinebench and simply run this at 5.4?

Running an Apex mobo with 4200 CL 16 and left ring at 49 if that makes a difference

 

[ThrashZone]

Hi,
Cinebench 15/ 15 extreme/ r20/ r23 are the easiest benchmark to get through.
Might be corrupt maybe restore a system image memory oc often scraps an oc pretty quickly.

 

[Placekicker19]

I have a 10900kf and did direct die with liquid metal on a optimus foundation block

temps are phenomenal
Currently doing 5.2Ghz @ 1.375v with LLC set to 4
with cinebench r15 im topping out at 56c

HOWEVER, cinebench crashes when i try 5.3ghz. its definetly not the temps i think because LLC is at 4 its not getting enough power to handle 5.3

I tried up to 5.4Ghz on games without any issues though
should i just ignore cinebench and simply run this at 5.4?

Running an Apex mobo with 4200 CL 16 and left ring at 49 if that makes a difference

Use load line 6 and try again. If you have apex LLC 6 is the minimum I would use, llc7 is safe too. Your temps are great so you have plenty of voltage headroom. Go by the load voltages not bios voltages. Load voltage is the actual voltage the cpu is receiving. Setting 1.375 llc4 is probably only like 1.2 actual volts. If your crashing in cinebench your cpu is very unstable. Run cinebench r20 like 5xs and check hwinfo64 for hardware errors. If it passes r20 then stress test real bench 2.56 for atleast 2 hours. If your stable in realbench you should be fine for gaming.

 

[dr.Rafi]

Nice, how much voltage does 5.3 with 5 cache require ? Ive seen bad sp63 chips and awesome ones. I bought 2 chips and the one that had quite a bit lower vid could not pass cinebench r15 @ 5.3ghz without L0 cache errors, tested up to 1.45vs. The chip with the higher vid scaled much better with voltage. They both could boot 4800mhz cl19 ddr4.

Will try today going higher on cash but for my experince decreasing ram frequency or loosening ram timing make it easier to go higher with cash ,if you keep the whole voltages constant (vcore,ca,io,ddr voltage),not sure i have to find the max cash overclock on stock ram setting ,then go higher on ram ?

 

[dr.Rafi]

Also thinking not to delid the cpu but lapping the whole ihs until i expose the solder to die, so will have only the sides of the ihs left then clean the die solder and remove the square frame left of the ihs clean the silicon underneath, so after cleaning the silicon, recheck the ihs on cpu and sanding the ihs until i have couple micrometer of die higher than ihs frame, so i can use it as direct die frame to protect the die from damage with water block mounted , what do you thing guys any suggestion will be highly appretiated.
Simply dont want to pay for delid and direct frame and wait for 2 weeks shippings.

 

[Placekicker19]

Will try today going higher on cash but for my experince decreasing ram frequency or loosening ram timing make it easier to go higher with cash ,if you keep the whole voltages constant (vcore,ca,io,ddr voltage),not sure i have to find the max cash overclock on stock ram setting ,then go higher on ram ?

I would max out my core and cache before working on ram. Leave ram at stock while dialing in your overclocks

 

[Dvnsta]

It would be interesting to see what a de-lidded 10850k can do with direct die cooling. Do you think Intel bin their chips with the IHS on? As my SP63 10900k improved massively after de-lid and direct die cooling.

I delidded and i'm direct die with mine. Got 5.1 all core 1.38 vcore. llc6 maximus hero XII .

 

[dr.Rafi]

Interisting finding, i sanded the ihs from 2.6 mm to 1.4 mm,its really painfull process by hand, refit the cpu found that water block touching the intel mounting mechanism of the motherboard before touching the cpu so removed the mechanism and keep only its back metal plate with screws, so now the cpu is holded to the socket only by waterblock run some tests and overclocks, found i cant go higher than 49 on cash, then after couple hours 49 become unstable have to go 48 and after while i have to clock the cash on stock but the ram and core clock overclocked have no issue, so i thought really i degrade my my cash very quickly finally even stock setting start freezing, it turn the cpu start getting lose and no enough pressure is applied to socket was the problem and my water block flimsy metal mounting is not applying enough pressure,tightned more ande everything went normal, so definitly need delidding and direct die kit for best result because applying high pressure on motherboard pcb is not good for long term use .

 

[ExDarkxH]

So i made some adjustments and I'm running 5.3Ghz all core 10900kf @ 1.35v (bios) LLC6
moved cash to 50x and adjusted ram to 4266 cl16

Tested cinebench and still not touching 60c on any of the cores so i have more thermal headroom, but I'm actually happy with this setting. Did aida64 and prime 95 stress tests without errors

 

[ExDarkxH]

also, not sure how much stock to put into sp rating. My card is a 63 and hasnt really hindered me doing anything. I was gaming at 5.4Ghz for a few days with 1.375v llc4.
I set my vccio to 1.18 and SA to 1.23 mainly for my memory and it takes very well

I want to try gaming at 5.5Ghz. will report back how much voltage was needed. Regardless of the result though i will probably switch back to 5.3 as im quite content

 

[munternet]

I fitted my air cooled 3080 Tuf OC in place of my water cooled 1080 ti a few days back and since then I have had 2 bsod while running realbench 2.56. Something which never happened before. Been running this setup for a few months now with no problems at all.
I just now turned the GPU voltage limiter up from 100 to 110% in afterburner which seems to have solved the bluescreen problem but I'm not really sure why.
Having removed the GPU from the water loop I have a temporary soft section and I seem to get more restriction now as the coolant can't take the easier route through the GPU waterblock and must pass through the CPU block.
The block is very old and I'm wondering if I should bin it and get a new one? The pump really struggles to turn the flow wheel now and I'm wondering if the early EK blocks were any good?.
I was testing to see if I need to water cool the GPU and I don't think I will bother at this stage as the frames in BFV are plentiful.
Any suggestions appreciated as always
Original loop and temporary loop pics

 

[dr.Rafi]

I fitted my air cooled 3080 Tuf OC in place of my water cooled 1080 ti a few days back and since then I have had 2 bsod while running realbench 2.56. Something which never happened before. Been running this setup for a few months now with no problems at all.
I just now turned the GPU voltage limiter up from 100 to 110% in afterburner which seems to have solved the bluescreen problem but I'm not really sure why.
Having removed the GPU from the water loop I have a temporary soft section and I seem to get more restriction now as the coolant can't take the easier route through the GPU waterblock and must pass through the CPU block.
The block is very old and I'm wondering if I should bin it and get a new one? The pump really struggles to turn the flow wheel now and I'm wondering if the early EK blocks were any good?.
I was testing to see if I need to water cool the GPU and I don't think I will bother at this stage as the frames in BFV are plentiful.
Any suggestions appreciated as always
Original loop and temporary loop pics
View attachment 2466687 View attachment 2466688 View attachment 2466689

I just feel (not sure)your old water loop have issue cooling both the cpu and gpu to max, putting 2 identical blocks like 2 gpus in pareller work well because they have same resistant to fluid flow and the fluid will pass equally in and out of both but having cpu/gpu blocks in parellel,if one of them is more resistant (ristricted flow) the coolant will pass through the other one faster and you lose the cooling effecincy in more resistant one, only what i am thinking, and what istudied back in school, but never tested, i always do cpu gpu in serial passing first through the gpu because want it better temp to perform well in gaming, then pass through cpu which usually run hotter in games than the gpu if both are on water.

 

[munternet]

I just feel (not sure)your old water loop have issue cooling both the cpu and gpu to max, putting 2 identical blocks like 2 gpus in pareller work well because they have same resistant to fluid flow and the fluid will pass equally in and out of both but having cpu/gpu blocks in parellel,if one of them is more resistant (ristricted flow) the coolant will pass through the other one faster and you lose the cooling effecincy in more resistant one, only what i am thinking, and what istudied back in school, but never tested, i always do cpu gpu in serial passing first through the gpu because want it better temp to perform well in gaming, then pass through cpu which usually run hotter in games than the gpu if both are on water.

Hi dr.Rafi
I see what you are saying but I never had any overheating problems before and I think the extra flow actually helped bleed all the air from the top of system really well
I'm probably more interested in the reasons for the blue screens and whether I need to get a new waterblock tomorrow because it's black Friday
I have turned the pump up from 45% to 80% and flow is slower than before