Overclock.net banner

Overclocking i7-10700k results, comparisons & discussions

351K views 889 replies 110 participants last post by  GuruSR  
#1 · (Edited)
I saw the i9 version of the thread so I thought I'd start the one dedicated to the i7.

So what are your results with this chip? ofcourse, proper results with aida64/rb/p95

msi unify z490 + nzxt x73 = currently sitting at 5.2ghz(48 cache) 1.295v LLC3

i guess it's a good chip, seen a lot of people struggling at higher voltages.

So what about you guys?
 
#2 ·
Just getting into it. Currently 5.0ghz(44 cache) 1.275v LLC4. I have been crashing in OCCT at 5.1 With as much as 1.31v. Your Chip seems better than mine.
 
#3 · (Edited)
Thanks for getting this thread up and running. Just got the 10700K with the Gigabyte Aorus Ultra. Running the Arctic Freezer II 360mm AIO. Just started overclocking on this board and have an adaptive overclock at 5GHz all core, 47 cache and LLC at Performance with a max VID at 1.336v with a vcore at max 1.345v.
Temps in Cinebench R20 80 degrees and Prime95 non AVX just under 60 degrees.

Prefer adaptive to fixed so a bit more work required to get the vcore down as on the Gigabyte board adaptive seems a bit more complex...Any help on settings appreciated and I will update further on the best I can get...

Just did a bit more tuning and have the max vcore under the adaptive mode down to 1.322v and the max VID at 1.297 stable in both Prime95 and Cinebench R20 with max temp hitting 82 degrees on the hottest core. Fingers crossed should be able to lower the vcore to around a 1.3v or maybe a fraction lower.
 
#4 ·
Thanks for getting this thread up and running. Just got the 10700K with the Gigabyte Aorus Ultra. Running the Arctic Freezer II 360mm AIO. Just started overclocking on this board and have an adaptive overclock at 5GHz all core, 47 cache and LLC at Performance with a max VID at 1.336v with a vcore at max 1.345v.
Temps in Cinebench R20 80 degrees and Prime95 non AVX just under 60 degrees.

Prefer adaptive to fixed so a bit more work required to get the vcore down as on the Gigabyte board adaptive seems a bit more complex...Any help on settings appreciated and I will update further on the best I can get...
LLC "Performance" is NOT a LLC setting!
Please do not confuse LLC (Loadline Calibration) with CPU Internal ACDC Load Line. They are NOT the same and CPU Internal AC DC Load Line affects the CPU's power supply (AC Loadline) when in offset or adaptive modes, not in fixed vcore mode. CPU Internal AC DC Loadline has nothing to do with vcore vdroop.

Vdroop (Loadline calibration) is controlled only by Loadline calibration.

If you want to see your "VR VOUT", use the tool I posted in the Gigabyte z390 thread. You cannot run this tool and hwinfo64 at the same time. VR VOUT functionality will be added to HWinfo64 at a future time.
The Vcore readings in CPU-Z and in HWinfo, etc are NOT accurate. They're okay for idle but way off at load. The tool I posted is accurate to 10mv.
 
  • Rep+
Reactions: Pinnacle Fit
#6 ·
I got a 10700K and MXIIF. I delided it it and threw it in my custom loop. I'm coming from a 3770K and MVF so I basically feel like I have no clue what I'm doing on these new platforms.

Quick question: in terms of what is "safe", am I worried about the set voltage or the load voltage? IE: is setting 1.4v safe if the load is closer to 1.3v? (using llc4)
 
#11 ·
Okay, I have gone the fixed vcore route as adaptive just seems to give me to many headaches...I am at a 1.290v for a 5GHz all core overclock though in HarwareInfo the vcore peaks at 1.308v during Cinebench and that is with no AVX ofset...It is a little lower in Prime95 at 1.295v.

Temps actually seem better with Cinebench R20 hitting a max 74 degrees on one core, all the other cores are between 66 and 74 degrees. Also it does downlock the CPU speed to a minimum of 800MHz when idle which makes me feel better.

Have not tried 5.1GHz yet but will try that shortly after a I do a long Prime95 run and Cinebench loop to test long term stability...
 
#12 ·
Do your vcore tests again for me and use this tool and tell me the minimum vcore at heavy load at 5 ghz.
If you ran hwinfo64 before, please REBOOT before running this tool or you will get 0's. Do not run this tool in the same windows session as HWinfo64.

Select 0 for vcore/VR Loop 1

Courtesy of Asus.
 

Attachments

#13 ·
Here you go. I used the first one:
 

Attachments

#14 ·
Ok your minimum accurate (die sense) vcore at 5 ghz at max load is 1.235v, which is accurate within 5mv.

Reducing voltage any more in BIOS lower is unstable?
 
#15 ·
Yes it becomes unstable.
 
#17 ·
Thank you. Need to compare results with others.
 
#16 ·
But those figures are better than I get in HardwareInfo....
 
#18 ·
No, thank you sir!...Just a quick follow up, is it okay to run this overclock and fixed vcore for my normal day to day use?
 
#20 · (Edited)
Yes you are fine. Did you have cpu-z or anything else running though? Your amps draw was not shown properly.
Yes, hwinfo64 and other programs are using the Bios "Socket sense" reading (MLCC caps behind the socket), and the second reading in HWinfo64 is from Super I/O. These are not going to be accurate because of impedance across the power plane causing a voltage drop (which means the sensors will read MORE voltage than what you are using).

This tool probes the VRM directly. This is the same reading you got as "VR VOUT" on the Z390 boards.
HWinfo's programmer is working on implementing VR VOUT for the Gigabyte Z490 boards but he isn't sure how to communicate with the chip yet, through PMBUS, but it uses a certain protocol for it.

Try rebooting and running NO monitoring programs at all, and having NO monitoring programs auto-start with windows, Just use this tool. See if you can get IOUT and POUT to register properly. IOUT should be around 100 to 150 amps and POUT probably 150 watts.
 
#21 ·
No nothing else running other than background like Bitdefender, Corsair iCue and Soundblaster Command. Also I ran Cinebench R20 during the vcore test.
 
#22 ·
Disable bitdefender and iCue. If you can. But you need to reboot after and keep them disabled.
 
#23 ·
Done and here you go:
 

Attachments

#24 ·
Hmm. Thank you for trying. Its still not reading amps and watts properly. At least vcore is working though!
 
#25 ·
Yep and thanks again as it is more than appreciated. Feel much better now that I can accurately gauge the vcore...Not sure why on the Amps side but will try to look into it...
 
#26 ·
This tool is a test tool released by Asus for reading the VRM on Gigabyte boards. Consider it as alpha software, basically.
We were trying to get proper voltages from the VRM because the onboard sensor measurements are just terribly inaccurate.

Once hwinfo64 programmer figures out how to access the Intersil VRM,, you will have much more reliable operation.
 
#27 · (Edited)
Here's my results from some preliminary tuning on the Aorus Pro AX board.

The stock settings on the board with 'normal' llc was putting the voltage closer to 1.29v under load out of the box at stock all-core.

At 4.7Ghz stock all-core testing using cb15 to find the voltage floor, it finally hard locked down at 1.045v measured in hwinfo.
I'm using 1.10v for 4.7 currently, seems to be happy there and runs much cooler.

I got up to 5Ghz all-core, 4.7Ghz cache at 1.23v load in Prime95 small FFT.

CPU load is at 82c in this condition after only a few minutes, waiting on some new fans to see if I can push it further.
Running an EK block, D5 pump and 360mm radiator. Normal gaming load or something like cb15 stays cool enough.

These chips seem to run great, and don't use much power if the clocks are kept down.
Power usage at 5ghz / 1.23v is pretty close to my 3770k was at 4.8 / 1.392v.
Idle power draw at the wall is significantly lower as well.

RAM tuning has been an issue, it took me hours to get the board to boot at 3600mhz with this g.skill/hynix 32GB kit.
The XMP profile would not boot, had to manually train it and bump the VCCIO and sys agent up to 1.2v. Finally got it running stable at the XMP 3600/cl16 spec with some sub timing configuration.
The ram issues seem to be immature bios to me, I tried another 3600mhz cl16 e-die kit I had laying around and it still took a lot of time to get it to boot higher than 3000mhz.
 
#28 · (Edited)
I cant get 4133mhz Ram to boot with either XMP 4133 or XMP2 4000 even with VCCIO and VCCSA at 1.300. Any ideas? 10700k + MSI Gaming Edge WIFI . Ram is 16gb 4133 Patriot viper, B-die. Got the CPU running at 5.1 @ 1.280v, cache at 4.6. Currently Running the Ram at 3600 14-16-16-36 but The subtimings are on the 4000mhz XMP AUTO, not sure how to improve subtimings, afraid of setting timings that dont work well together.
 
#29 ·
@SmackHisFace

There were also two advanced voltage settings I changed, the 'DRAM training voltage' to 1.45v and 'DRAM termination' to 0.680v. I believe the termination is supposed to be half of the dram set voltage.

I don't think sub timings are very important until you get the frequency you want to boot. My bios was setting a low tRFC value for my kit, and setting it manually to the appropriate value seemed to help.
There's several good articles and forum threads on sub timings around the net, they're very time consuming to setup.

Other than that, patience and letting the bios do its training thing. Cold booting after bios changes helped as well.
 
#30 · (Edited)
Here are some of my settings. Cant seem to locate either of the settings you mentioned. Possible they have another name on MSI bios? EDIT: What is the recommended way to monitor VCore in windows? I dont see the option in HWinfo Beta so Im using CPUZ, is CPUZ accurate?
 

Attachments

#37 ·
Okay i pushed my 10700K to 5.1GHz all core on a fixed 1.3v in HWinfo the vcore hits a max of 1.320v but VID can get to 1.4v...Temps are all okay...Is this okay?
 
#38 ·
Please install the newest beta version of hwinfo64 and screenshot your VR VOUT while under max load, please, and post it (or upload it to imgur and post a link here).
 
#44 ·
I can do 5 GHz with 1.25v mannual set on bios, and 47 ring LLC2 on MSI what it keep the voltaje linear.
But what is very rare that I know a guy with same #batch as mine, and he can do the 5 GHz with only 1.10v I cant even get in windows with that voltage.


Here:
 
#45 ·
I can do 5 GHz with 1.25v mannual set on bios, and 47 ring LLC2 on MSI what it keep the voltaje linear.
But what is very rare that I know a guy with same #batch as mine, and he can do the 5 GHz with only 1.10v I cant even get in windows with that voltage.


Here: https://www.youtube.com/watch?v=sBEk9nAUlpw
5 ghz @ 1.10v load voltage for a 10700k? Good chip for sure. Probably within 10900k SP levels of quality. I can do that easily if I disable 2 cores on my 10900k, but all 10 cores requires 1.128-1.143v.
 
#54 ·
Just lowered vcore to 1.285v and ran test loop on CinebenchR20 with HarwareInfo beta stats below:
 

Attachments

#63 ·
Hi,
Memory presets are pretty wild lol timings and voltages.
 
#71 ·
So true....

No matter. I've had nothing but good experiences here for the last 13 years. There's always bound to be a jackass or 2 floating around. At least I know who to ignore now.,[/QUOTE]