300 point difference in Cinebench scores on two 8700k. Why?

[LaboonTheWhale]

Hey guys

 

So I decided to test me and my gf's builds out in Cinebench. 

 

I have an 8700k OC'd to 5Ghz @ 1.35v and she also has the same CPU + OC. 

 

I have mine in a Maximus X Hero and she has hers in a STRIX Z370 itx. We both have 16GB of 3200Mhz trident Z ram. Mine is CL16 and hers is CL14. 

 

I am averaging about 1560 in Cinebench CPU test but she is sitting at roughly 200-300 less points around 1216. I have ran this test 5 times now and it all comes out the same. 

 

Her's is being cooled by a Dark Rock TF w/ fresh air as intake and I have a Noctua U14s cooling mine. Hers is rated for 220W TDP and mine is rated for 160. both of our temps get hammered but it seems mine always performs better. Is it because its an ITX board vs my Maximus board? I can't really find a difference in why it performs so differently when they're both OC'd the same w/ same speed ram. 

 

Edit. I know it's not a big deal. Just curious on what's causing the difference

 

Edit 2: 1.35v. not 1.5v

[Cookybiscuit]

Are you monitoring clock speeds across all cores to make sure they are staying at what they are supposed to be? 

 

Also I'm sure you already know, but 1.5V is extremely high, I have a very hard time believing a NHU14 is keeping a 8700K at 5GHz 1.5V under TJMax unless you live in Antarctica. Also probably not particularly healthy for your motherboard VRMs.

[LaboonTheWhale]

Just now, Cookybiscuit said:

Are you monitoring clock speeds across all cores to make sure they are staying at what they are supposed to be? 

 

Also I'm sure you already know, but 1.5V is extremely high, I have a very hard time believing a NHU14 is keeping a 8700K at 5GHz 1.5V under TJMax unless you live in Antarctica. Also probably not particularly healthy for your motherboard VRMs.

Ahhh I forgot to edit! its 1.35v. oopsies. the 3 got dropped. 

[Jurrunio]

Hyperthreading in her chip turned off?

[LaboonTheWhale]

Just now, Jurrunio said:

Hyperthreading in her chip turned off?

Had no idea that was a thing

[pas008]

Multi core enhancement?

[LaboonTheWhale]

4 minutes ago, pas008 said:

Multi core enhancement?

set to auto in bios 

[VicciVicVic]

7 minutes ago, pas008 said:

Multi core enhancement?

^^^^^^^^^^^^^^^^^^^^

[LaboonTheWhale]

14 minutes ago, VicciVicVic said:

^^^^^^^^^^^^^^^^^^^^

its set to auto in bios. can only be auto or disabled. 

[scottyseng]

2 minutes ago, LaboonTheWhale said:

its set to auto in bios. can only be auto or disabled. 

auto on both motherboards?

 

XMP profiles enabled for both motherboards?

 

Yeah, I'm not sure what the cause of the difference is...

[LaboonTheWhale]

1 minute ago, scottyseng said:

auto on both motherboards?

 

XMP profiles enabled for both motherboards?

 

Yeah, I'm not sure what the cause of the difference is...

yes and yes. Power delivery of itx board maybe cant keep up? i dunno. 

[OnionRings]

As the one person suggested, run both pcs again while watching clock speeds, look for dips.

[scottyseng]

Just now, LaboonTheWhale said:

yes and yes. Power delivery of itx board maybe cant keep up? i dunno. 

Eh, they aren't crazily overclocked so I'd be shocked if the power delivery couldn't keep up on the smaller one.

 

But yeah, I would track clock speeds during a synthetic CPU load test and look for dips as mentioned above.

[LaboonTheWhale]

1 minute ago, scottyseng said:

Eh, they aren't crazily overclocked so I'd be shocked if the power delivery couldn't keep up on the smaller one.

 

But yeah, I would track clock speeds during a synthetic CPU load test and look for dips as mentioned above.

 

2 minutes ago, OnionRings said:

As the one person suggested, run both pcs again while watching clock speeds, look for dips.

If there were dips, would it be caused because the CPU can't handle it? or the mobo? or do i need to add more voltage? 

[OnionRings]

1 minute ago, LaboonTheWhale said:

 

If there were dips, would it be caused because the CPU can't handle it? or the mobo? or do i need to add more voltage? 

Honestly I haven't got that far, owning a 6700 myself Ive never OCd before but I do know anything above 1.35v is probably too much. Find if it is dipping first, if it is, then maybe you can narrow the searching and find out why.

[scottyseng]

7 minutes ago, LaboonTheWhale said:

If there were dips, would it be caused because the CPU can't handle it? or the mobo? or do i need to add more voltage? 

Usually thermal throttling in my experience. If it weren't enough power, it would just bluescreen.

 

Is the vcore set to auto as well?

[LaboonTheWhale]

14 minutes ago, scottyseng said:

Usually thermal throttling in my experience. If it weren't enough power, it would just bluescreen.

 

Is the vcore set to auto as well?

I have to go check. also whats the best software to make sure I can confirm its all going the correct speeds? HWInfo? 

[scottyseng]

1 minute ago, LaboonTheWhale said:

I have to go check. also whats the best software to make sure I can confirm its all going the correct speeds? HWInfo? 

Yeah, I use HWinfo / HWmonitor or CPU-Z (any works)

 

As for CPU stress tests, AIDA64 is pretty mild. Intel Burn Test is usually what I use for OC, but is very intense (temps you see on this you will never see in normal use, even at 100% rendering load). Though any PC that survived Intel Burn Test was a rock solid stable OC for me.

[Blackie Sheen]

It can be due to a combination of factors like thermal throttling, maybe she got a shit sample too. Because from what I could figure out most people get in the high 1300,low 1400 when the cpu is at stock so 1216 is straight up awful. Also maybe its not fully stable and causes to perform poorly but that would surprising since it would most likely cause it to crash.

Firstb thing though is to find if there is throttling happening because that is most likely it

[LaboonTheWhale]

Just now, Blackie Sheen said:

It can be due to a combination of factors like thermal throttling, maybe she got a shit sample too. Because from what I could figure out most people get in the high 1300,low 1400 when the cpu is at stock so 1216 is straight up awful. Also maybe its not fully stable and causes to perform poorly but that would surprising since it would most likely cause it to crash.

Firstb thing though is to find if there is throttling happening because that is most likely it

Maybe I should try reseating the thermal compound as well with the cooler. so many things could be contributing. 

[Blackie Sheen]

2 minutes ago, LaboonTheWhale said:

Maybe I should try reseating the thermal compound as well with the cooler. so many things could be contributing. 

Yeah, honestly thats what I would do but first just do a stress test for like 15-20 or until the temps plateaus and observe if it is throttling. If not, Its likely not the issue. But if you dont mind the hassle of removing, cleaning the thermal compound, replacing it and reseating the cooler, it sure cant be a bad idea to do it

[Jurrunio]

1 hour ago, LaboonTheWhale said:

Had no idea that was a thing

just a guess. 8700k at 5GHz could score around 200 in single core score, which means with hyperthreading disabled, the all core score will be around 1200

[Cookybiscuit]

1 hour ago, LaboonTheWhale said:

Maybe I should try reseating the thermal compound as well with the cooler. so many things could be contributing. 

You just need to look at the clock speeds, simple as. Run the test and make sure no cores on either CPU hit 100C (the point at which they will downclock) and that they are all sitting at 5GHz, if you've accidentally set up a turbo overclock it'll change the clockspeed based on how many cores are in use.

 

My bet is inadequate cooling is causing it to hit 100C and downclock. I'm delidded and on a much bigger air cooler and still hit 90C+ in Cinebench at 5GHz 1.35ishV, but there's too many variables to say for sure without testing. Also worth mentioning that if it is thermal throttling any stability testing you've done is invalid for obvious reasons.

[johnukguy]

Cinebench scores are just about meaningless. For one thing, Cinebench is only really useful for testing whether a CPU overclock is nominally stable. For another, the scores change all the time based on a whole host of variables.

[Emberstone]

2 hours ago, johnukguy said:

For another, the scores change all the time based on a whole host of variables.

^^^^^

Simply having certain background programs open can cause your score to drop 100+ points, I've noticed. When running Cinebench, the first thing you should do is close everything (end processes in Task Manager if you have to) to ensure that you're getting a close to accurate score.

 

If you ever wondered why tech YouTubers often end up with Cinebench scores lower than yours, this is why. They often don't close background programs and leave things such as CPU-Z, HWiNFO, and whatnot open during the tests. I see results all the time like 1300 multi/145 single on a 4.0 GHz R5-1600, or 1490/190 on a 5.0 GHz 8700K, and it baffles me why they don't think these scores are weird.

 

For context, my R5-1600X at 4.0 GHz got 1350/165 and my 8700K at 4.7 GHz (all cores) gets 1557/209. Close ya damn background programs, peeps.