My i7 7700k Has Arrived - Insights, Benchies & Overclocks Inside (Now With Delid!)

[superstition]

"apparently"?

The Stilt said Vishera chips should probably be limited to 1.475 and yet people have pumped 1.55 – 1.6+ into them.

The other question is how long does it take to degrade a chip. I see review sites pump lots of voltage into CPUs and write their articles as if they did this prior to stock testing and lower-voltage testing.

I was surprised to learn that idling at high voltage is just as damaging as running it full-tilt in Prime, in terms of damage from excessive voltage. It also makes me wonder about voltage spike transients because I saw a big difference between the maximum voltage in an ASUS Crosshair board's second sensor set versus the first sensor set. The first sensor set didn't show the highs of the second. This could lead a person into thinking their voltage range is safe when it's not. If most boards are using lax sensors then that's another issue — what voltage is actually going into the chip?

If a CPU gets degraded quickly by excessive voltage then it becomes useless as a model for normal performance.

 

[RichUK]

"apparently"?

The Stilt said Vishera chips should probably be limited to 1.475 and yet people have pumped 1.55 – 1.6+ into them.

The other question is how long does it take to degrade a chip. I see review sites pump lots of voltage into CPUs and write their articles as if they did this prior to stock testing and lower-voltage testing.

I was surprised to learn that idling at high voltage is just as damaging as running it full-tilt in Prime, in terms of damage from excessive voltage. It also makes me wonder about voltage spike transients because I saw a big difference between the maximum voltage in an ASUS Crosshair board's second sensor set versus the first sensor set. The first sensor set didn't show the highs of the second. This could lead a person into thinking their voltage range is safe when it's not. If most boards are using lax sensors then that's another issue — what voltage is actually going into the chip?

If a CPU gets degraded quickly by excessive voltage then it becomes useless as a model for normal performance.

The question is, is it voltage or current that knackers a chip, and how does LLC factor into this?

Increasing voltage and LLC exponentially increases risk exposure.

 

[superstition]

The question is, is it voltage or current that knackers a chip, and how does LLC factor into this?

Increasing voltage and LLC exponentially increases risk exposure.

Unless the droop is so big without LLC that one ends up putting in higher voltage than with it, eh?

 

[bbhaag]

I think it's very true. Zen is literally around the corner, and will probably be a much better purchase than a quad core Intel. At the very least it might affect prices. Buying CPU's when they are just about to be discontinued is always a mistake IMO (7700K wil have higher resale value in the future, most people see bigger numbers as better, even if Kabylake is just an overclocked Skylake).

AMDs track record over the past several years doesn't inspire confidence in me. While Zen maybe a very nice cpu I just don't have the faith that you seem to have in it. I also don't buy cpus for their resale value.

 

[RichUK]

Looks like 5Ghz is a no go - 1.42v and Prime 95 small FFTs fails on 3 cores after a few seconds, temps are ~90 degrees even with the AIO in performance mode. A de-lid may give me more temperature headroom, but i'm guessing it'll require closer to 1.5v to stabilize. I can pass OCCT (AVX Linkpack ticked) and IBT (20 mins) no issue at 1.36v, however.

4.9Ghz when stressed with Prime 95 small FFTs is chugging along just fine at 1.34v (only tested 20 mins before stopping). I need to tweek the memory speed and timings before i leave for a prolonged blend test.

That's a shame.

I'll still be deliding to get the temps lower, as even with this AIO it still hovers around ~82 degrees during prime.

And 5.4Ghz, only briefly stable in windows:

 

[Dave2150]

Disappointing for sure. So far we've seen a few 7700K's all fail to reach 5.0 stable, on air and AIO water cooling.

Obviously for anyone buying now, a 7700k is still the way to go, but Intel's hype about the '14nm+' process is truly nonsense. A 100 or 200Mhz gain in average overclockable clock speed over the original 14nm process is simply not worth shouting about. It should have been an unmentioned improvement, just as it has been with previous processes development.

Maybe Intel know something we don't - with Zen coming very soon, I'd have expected Intel to push to bar if Zen was going to threaten them. I guess Zen will be a massive disappointment too, or Intel will rush forward Cannon/Coffeelake/Skylake-E.

 

[RichUK]

Disappointing for sure. So far we've seen a few 7700K's all fail to reach 5.0 stable, on air and AIO water cooling.

Obviously for anyone buying now, a 7700k is still the way to go, but Intel's hype about the '14nm+' process is truly nonsense. A 100 or 200Mhz gain in average overclockable clock speed over the original 14nm process is simply not worth shouting about. It should have been an unmentioned improvement, just as it has been with previous processes development.

Maybe Intel know something we don't - with Zen coming very soon, I'd have expected Intel to push to bar if Zen was going to threaten them. I guess Zen will be a massive disappointment too, or Intel will rush forward Cannon/Coffeelake/Skylake-E.

Yeah, surprised this 14nm+ qualifies as a new generation processor. I reckon I may still need to bump up the voltage to achieve 4.9ghz to be fully stable over 12 hours at least.

It makes no sense upgrading from a skylake quad to kaby lake, but if you are looking for an upgrade from an older generation, kaby lake is worth waiting for given its release is just around the corner.

I'm disappointed as I genuinely thought I'd achieve 5ghz with 1.4v. Never mind, I didn't really need to do this upgrade but just felt like spending some fun coupons.

I'm also a little disappointed in this AIO cooler - just stick with a Noctua D15s.

Now to focus on my next purchase in q2 2017, a new BMW M4 Competition Pack

 

[.vodka]

Hm, this is similar to Bloomfield C0 -> D0 stepping change... only difference is that it now qualifies as a new generation due to the improvements on the GPU side. Oh well. That, and Z270 bringing Optane support. That could make for an interesting platform.

Zen, your turn.

 

[RichUK]

Hm, this is similar to Bloomfield C0 -> D0 stepping change... only difference is that it now qualifies as a new generation due to the improvements on the GPU side. Oh well. That, and Z270 bringing Optane support. That could make for an interesting platform.

Zen, your turn.

I guess kaby lake only really benefits mobile, given increased clocks at the same voltage and improved igpu.

Not sure when the skylake 6600k / 6700k processors were released, but it'll essentially be 2 years between true generation releases ( skylake to kaby Lake successor), or when the next true step change will occur.

That's what happens with a lack of competition.

 

[Justinbaileyman]

Wow what a bummer... I Hope not all the chips are gonna act like this. I want my 5.0Ghz 7700k for 24/7 use. Did you try upping the LLC along with the vcore? Is it stable at 5.0GHz with 1.45v or is that even a no go? Thanks buy the way for posting and sharing all the info you've work for so far!!

 

[RichUK]

Wow what a bummer... I Hope not all the chips are gonna act like this. I want my 5.0Ghz 7700k for 24/7 use. Did you try upping the LLC along with the vcore? Is it stable at 5.0GHz with 1.45v or is that even a no go? Thanks buy the way for posting and sharing all the info you've work for so far!!

LLC was set at level 2 - all LLC does is limit v droop. I haven't tried 1.45v because the temps were getting silly - any higher and you'll be very close to tjmax (or 100 degrees), which I believe is when it'll start to throttle. There's no way I'd stress it for 12 hours with temps getting that high. I'll perhaps give it a try once I delid, but anything over 1.4v is a bit too high in my opinion.

 

[VirtualLarry]

but anything over 1.4v is a bit too high in my opinion.

Agreed. RS posted that some people were running their 6700K CPUs at like 1.48V, and I was like, "are you crazy!?!?".

Temps get pretty high for my i5-6400 @ 1.410V, 165.0 BCLK, 4.455Ghz, 86C under OCCT:CPU load, with a Zalman CNPS5x 92mm tower heatsink. But at least, it seems stable.

 

[RichUK]

I'll update the OP with the detailed outcomes once I've achieved 12 hours prime stable. I'm hopeful for 4.9ghz, but it could end up being 4.8ghz.

Stability is key. I could leave it at 5ghz if IBT and OCCT doesn't fail an extended test, but that's not truly stable in my book especially when Prime 95 quickly fails (and doesn't fail at stock, as you'd expect).

 

[RichUK]

Wow what a bummer... I Hope not all the chips are gonna act like this. I want my 5.0Ghz 7700k for 24/7 use. Did you try upping the LLC along with the vcore? Is it stable at 5.0GHz with 1.45v or is that even a no go? Thanks buy the way for posting and sharing all the info you've work for so far!!

1.45v hits 99 degrees on 2 cores - not sure if the AIO is under performing as I'd expect it to be doing a better job, though I am still using the original paste so will try changing it to AS5.

Speedstep is enabled hence why the screenshot doesn't show 5ghz, but you'll see prime 95 fails rather quickly from the utilisation graphs (for 3 cores, afterwards i stopped the test on all cores):

 

[RichUK]

Can someone tell me whether these temps are normal on a H110i in performance mode (ignore the max temps as they were from the 1.45v stress test)?

To clarify, I changed the multi to 48 and reduced the voltage offset via the Asrock tool in windows as this would be more comparable with the 6700k - temps just seem to high, I thought they'd be lower.

 

[Dave2150]

Noctua high end coolers usually outperform the cheap corsair AIO's, in my experience at least. Much quieter too of course.

I replaced my H110I GT (280mm) with a Noctua UH-12S (single tower) and got similar temperatures but absolute silence (6700k@ 4.7Ghz, 1.3v).

 

[Justinbaileyman]

Can someone tell me whether these temps are normal on a H110i in performance mode (ignore the max temps as they were from the 1.45v stress test)?

To clarify, I changed the multi to 48 and reduced the voltage offset via the Asrock tool in windows as this would be more comparable with the 6700k - temps just seem to high, I thought they'd be lower.

Your temps seem awfully high to me... are you sure you have your pump plugged in and its running?Or are your fans running?? Something is off and I would double check everything over real good cause I was running a single 120mm NZXT x31 AIO setup and had a FX-8350 overclocked to 5.0Ghz at 1.5v and it never went over 70c even maxed out. Aren't these intel CPU's supposed to run cooler even while overclocking? I understand that they can run up to 80-90c but they are not supposed to if you can help it. If that were me I would be Pooping bricks right now casue I would never never never let my temps hit 80c.

 

[Majcric]

Just want to give many thanks to RichUk and the findings. I was glad to see you throughly test with prime95 small FFTs included the mix

 

[LTC8K6]

Have you tried Intel XTU for temp reporting?

 

[CHADBOGA]

Just want to give many thanks to RichUk and the findings. I was glad to see you throughly test with prime95 small FFTs included the mix

Majcric,
Is that you in your avatar?

I'm guessing it is probably your Dad or Grandfather, but it would be nifty if you really were an elderly gentleman that has the level of interest that you do in CPU's.

 

[Excessi0n]

Agreed. RS posted that some people were running their 6700K CPUs at like 1.48V, and I was like, "are you crazy!?!?".

I've been running mine at 1.5v since I delidded it. 4.9 GHz.

 

[RichUK]

Having never used an AIO before, it's a bit of a learning curve - I discovered the VRMs overheating yesterday which kept throttling voltage and CPU speed (every 2 seconds it would flip between 4.2ghz and 4.9ghz), which is a consequence of not having air moving over the VRMs as I did with the tower cooler.

Edit - this happened under IBT load as well as prime 95. For comparison, 5ghz with IBT load never dipped below 5Ghz when I was using the tower cooler (which was actively cooling the VRMs).

Question I need to ask myself is, do I try the Kraken X62 and hope for reduced temps (can't imagine I'd see a huge difference) and still have to resolve the vrm overheating issue, or cancel it and place an order for a DH15s.

Hmmmm.

 

[RichUK]

Have you tried Intel XTU for temp reporting?

I will try it.

EDIT: - Doesn't load, shows an error message suggesting a conflict with another performance tuning app, but no other apps are open.

HWinfo showed the temps hit 99 degrees and you could see the speed in CPU-z throttle back to 4.2ghz when it hit this threshold.

 

[RichUK]

Your temps seem awfully high to me... are you sure you have your pump plugged in and its running?Or are your fans running?? Something is off and I would double check everything over real good cause I was running a single 120mm NZXT x31 AIO setup and had a FX-8350 overclocked to 5.0Ghz at 1.5v and it never went over 70c even maxed out. Aren't these intel CPU's supposed to run cooler even while overclocking? I understand that they can run up to 80-90c but they are not supposed to if you can help it. If that were me I would be Pooping bricks right now casue I would never never never let my temps hit 80c.

Yes, everything was working as expected. At stock clocks the temperature was low 70s when stressed with prime.

I'm only testing, so the short burst of high temps won't do any harm.

 

[RichUK]

Apologies for all the images, not sure if they take a while to load (it doesn't for me).

As an aside, I've also applied AS5 instead of the standard paste on the H110i which doesn't seem to make any difference to temps.

The issue here is VRMs overheating when using an AIO.

So, problem:

Tactical solution:

Outcome (solved):

Accepting 4.8Ghz and using a Noctua D15s would be the 'sensible' option (the next 100Mhz could be expensive and untidy in terms of cooling [X62 + vrm fan], assuming it's stable at 4.9Ghz), but....

There is also the delid which may negate the need for the X62, but will still need active cooling on the VRMs if I stick with the h110i. Hmmm

EDIT: Given I still need to up the VCCIO and VCCSA volts to get a decent memory overclock, I think I'll continue with the X62 which should be here tomorrow. I'll be installing this into a Corsair Obsidian 450D, the case in the pics is an old case for testing purposes.

De-lid tool should be here on Thursday.

The journey continues....