• We’re currently investigating an issue related to the forum theme and styling that is impacting page layout and visual formatting. The problem has been identified, and we are actively working on a resolution. There is no impact to user data or functionality, this is strictly a front-end display issue. We’ll post an update once the fix has been deployed. Thanks for your patience while we get this sorted.

Intel "Coffee Lake" Builders Thread

Page 62 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.
I should clarify, for 24/7 use I personally wouldn't run much more than 1.4 to 1.45 voltage through my chips on water. But my point from above still stands, always try to find the LOWEST possible stable voltage. Whether it's at 1.25, or 1.22 or 1.2 or 1.19 or 1.08 etc. Whatever the lowest voltage you can achieve stable is what you should run. From my experience, it will probably be somewhere in the 1.2 to 1.22 range though for 4.8ghz from the sound of your chip. But you might get lucky and be able to get it into the 1.19 range as well, just gotta test it and see.

So, you’re of the opinion that running an 8700k @ 5.1 GHz with a voltage of 1.41v is probably safe as the “daily driver” config? It is delidded and will be running an Arctic Liquid Freezer 240mm AIO.
 
Ironically I have a huge case, vertical motherboard- main gear shift case. Only fits a 120 rad up top unless I mount a bigger one on the outside
 
So, you’re of the opinion that running an 8700k @ 5.1 GHz with a voltage of 1.41v is probably safe as the “daily driver” config? It is delidded and will be running an Arctic Liquid Freezer 240mm AIO.

Absolutely. De8uer has already made that statement as well, between 1.4 and 1.45v as long as temps are within check would be completely safe for 24/7 usage. I would crank up the voltage, enjoy the blazing fast speed and keep the temps in check. It's not like you will kill it, people have ran much much much more voltage through these chips during benchmarking with ln2 and dry ice etc without issue.
 
These are some numbers Ive been able to achieve with my setup, would appreciate some input as to how I can get some temps down and some possible settings to change with my setup.

I have an 8700k with the Maximus hero x wifi board. I have followed the settings proposed by der8auer. I am using an EK Water Blocks 120mm radiator liquid cooling loop, one vardar fan. 3200hmz trident z ram cl16. I am limited to the 120 cooler due to case design. (Large case, ala silverstone rz01)

Load line calibration set to 6, at 5 it would crash under these numbers.

At 5ghz, I was able to get it stable at 1.375, but 26.6 prime 95 was hitting nearly 97c.
At 4.9 ghz, 1.31 was stable, around 86c prime 95.
At 4.8ghz, 1.25 seems stable, around 77c stress test.

Do those numbers look ballpark? Any other setting I could look at to get the the temps down? Id like 5ghz, but 97c is too high. ideally Id like everything below 80c stress test, but I may need to stick to 4.8 for that.


A few updates on my overclocking journey:

1.) Removed single Vardar fan from my 120 EK water cooled radiator; added 2 Noctua NF-F12 in push/pull.
2.) Was able to get down to a stable 1.3 at 4.9ghz, with prime 95 26.6 at max temps of 84c.
3.) Previously only 1.31 was stable, but at 87c, so a slight improvement with the fans running 100%.
4.) At 4.8ghz, 1.26v seems stable and stays at 80c or under prime 95.

My testing methodology is to get at least an hour in prime, then I do realbench, cinebench, etc. If it proves stable, I'll do a 24 hour test and see how far I can take it.
 
I loaded Windows on my new rig this morning and have it running Prime with the following settings:

1. CPU multiplier: 50
2. Cache multiplier: 44
3. CPU voltage offset: .2v (Vcore is now at 1.365 during the Prime run)
4. AVX offset: -2

I’ll have a better feel for temps shortly but I have a few questions in the meantime:

1. What should I shoot for on cache multiplier?
2. I have my Arctic Liquid Freezer 240 pump connected to the CPU_OPT/WPUMP header and have that header set for a water pump running in PWM mode in the BIOS. Should I set that to DC mode to ensure the pump is always running at full speed?
3. If I opt to go to beyond the lowest LLC setting, should I set the voltage to a fixed value?

The rig is insanely quiet and I had to put my head into the case to hear the pump.
 
Last edited:
I loaded Windows on my new rig this morning and have it running Prime with the following settings:

1. CPU multiplier: 50
2. Cache multiplier: 44
3. CPU voltage offset: .2v (Vcore is now at 1.365 during the Prime run)
4. AVX offset: -2

I’ll have a better feel for temps shortly but I have a few questions in the meantime:

1. What should I shoot for on cache multiplier?

As high is as stable. I have mine at 1 less than my CPU multiplier but I'm only at 4.8ghz.
 
Nice, I don't think Newegg will do that for those of us that paid $420. I'm curious as to what would be the best delidding tool out there, truthfully it's something I'll probably only use once.
 
Nice, I don't think Newegg will do that for those of us that paid $420. I'm curious as to what would be the best delidding tool out there, truthfully it's something I'll probably only use once.

Honestly, if you’re going to only use the tool once, send your CPU to Silicon Lottery and have them do it for $30. They’ll also delid AND bin if for you for $60. That’s what I did and they binned mine as a 5.1GHz model. So far, in limited testing, it is comfortably running 5 GHz with a -2 AVX offset and roughly 1.32 volts; I can probably drop that voltage and temps have been very good.
 
Honestly, if you’re going to only use the tool once, send your CPU to Silicon Lottery and have them do it for $30. They’ll also delid AND bin if for you for $60. That’s what I did and they binned mine as a 5.1GHz model. So far, in limited testing, it is comfortably running 5 GHz with a -2 AVX offset and roughly 1.32 volts; I can probably drop that voltage and temps have been very good.

Why would you get them to bin your CPU, you can do that yourself?

Plus they probably only use basic tests to prove stability.
 
Why would you get them to bin your CPU, you can do that yourself?

Plus they probably only use basic tests to prove stability.
So how many CPUs should the average enthusiast buy to do this binning process? Then you have to sell the ones that didn't make the cut. Silicon Lottery's markup is more bearable if you weigh the odds and think through what might have to be done to produce their guaranteed result.
 
Why would you get them to bin your CPU, you can do that yourself?

Plus they probably only use basic tests to prove stability.

I hadn't actually even opened it when I sent it to them and didn't even have all my parts yet. I had them bin it because 1) I was genuinely curious how it would fare in their tests 2) I wanted to double-check their results against what I could achieve - I wanted to use their results as a baseline. I haven't built and overclocked a new Intel rig in ages and wanted a good comparison so if I was falling way short, I'd know I was doing something wrong. Plus, in their binning process, I *think* they only vary voltage, multiplier, and AVX offset, meaning that you can probably go higher than their results (or, conversely, achieve the same results with lower voltages) by changing other parameters as well.

So how many CPUs should the average enthusiast buy to do this binning process? Then you have to sell the ones that didn't make the cut. Silicon Lottery's markup is more bearable if you weigh the odds and think through what might have to be done to produce their guaranteed result.

I had intended on likely buying my 8700K from them in the first place, but they kept pushing back the date and I scored one on Newegg. I was already going to pay them to do the delid, so why not bin it while they're at it? I was really curious to see if I got a "golden" chip or not. As it turns out, I didn't get a golden one, but I got a pretty decent one.

Silicon Lottery is very clear on how they bin and how they guarantee the speeds. IIRC, they only vary voltage, multiplier, and AVX offset. To reach their guaranteed results, you must use components off their QVL. That wasn't a big deal for me because I had planned on using the components on their QVL anyway even before I saw their QVL.
 
Yesterday I noticed Micro Center dropped the price of the 8700K from $420 to $380. Went to the store today with my receipt for the 8700K I bought two weeks ago and they refunded me the $40 difference.

If they had seen your signature, they would probably ask you for 40$ more! 😀

j/k
 
I had intended on likely buying my 8700K from them in the first place, but they kept pushing back the date and I scored one on Newegg. I was already going to pay them to do the delid, so why not bin it while they're at it? I was really curious to see if I got a "golden" chip or not. As it turns out, I didn't get a golden one, but I got a pretty decent one.

How can you be sure they returned the same CPU to you if you got a golden chip? 😉
 
A few updates on my overclocking journey:

1.) Removed single Vardar fan from my 120 EK water cooled radiator; added 2 Noctua NF-F12 in push/pull.
2.) Was able to get down to a stable 1.3 at 4.9ghz, with prime 95 26.6 at max temps of 84c.
3.) Previously only 1.31 was stable, but at 87c, so a slight improvement with the fans running 100%.
4.) At 4.8ghz, 1.26v seems stable and stays at 80c or under prime 95.

My testing methodology is to get at least an hour in prime, then I do realbench, cinebench, etc. If it proves stable, I'll do a 24 hour test and see how far I can take it.


Ive been doing more testing, and I am happy with 4.9ghz at 1.29v, 42 multiplier, LLC at 5. Prime 95 26.6 stable at 79c max.

I added a second 120mm rad to my EK water cooling loop, and made both rads push pull, which helped bring down temperatures on average of about 7-10 degrees c.

5ghz was stable, but at 1.36v, and max temp of 87, so I cut it down to 4.9ghz above. I think I'd need at least a 360 rad setup for 5ghz and lower temps.

EK water cooling with 2 separate 120mm radiators, but each with 2 fans in push pull, for a total of 4 fans at 100% to achieve this.

Question: Are 2 separate 120mm rads the equivilent of 1 240mm or does it not translate like that? I Can't fit 1 240mm, but I can do 2x 120mm in different places.
 
How much are you guys paying total for the delid service? $30? Does that include shipping both ways?

ROCKIT is having a sale on their delid kit for $40 plus 20% off and also free shipping for Cyber Monday FYI. I think I am going to go with that for my 8700K.
 
How much are you guys paying total for the delid service? $30? Does that include shipping both ways?

ROCKIT is having a sale on their delid kit for $40 plus 20% off and also free shipping for Cyber Monday FYI. I think I am going to go with that for my 8700K.

$30 and it includes return shipping. You have to pay to ship it there (was roughly $6 with insurance IIRC).
 
A little bit off topic but it's kind of obnoxious that USB 3.0 was renamed to USB 3.1 gen 1. I wonder if they did this just to confuse people.... the 5.25 inch USB device is not what I thought it would be and I ended up using it for my nephew's computer and keeping my existing 2.0/3.0 one.

If anything gen 1 should be called USB 3 and gen 2 should be called USB 4.
 
So right now, I'm testing a multiplier of 50x, AVX offset of -1, Vcore offset of 100 mV, and LLC level 3. Running Prime95 using the heat generating test, all cores are in their low-mid 70s except core 0, which is in the mid-upper 70s with a high of 80. CPU voltage is reading 1.296 during this test. If I adjust AVX to -2, I'll drop 3-5 degrees. Thoughts? Idle temps for the CPU are in upper 20s/low 30s in a room which is 72 degrees F.

EDIT: System finally hard locked. I could raise the voltage to stabilize it, but I think I'm going to dial it back down to AVX -2 and leave it.
 
Last edited:
Back
Top