Delidded 7700K 4.9 from OCUK - issues

covenantuk · 1 Apr 2017 at 16:51

Ok, so I've built my first system for 5 years using a 4.9ghz 7700K delidded from OCUK

I had some doubts when it arrived as "4.8" was written in biro on the box lid, but I sent support a web note and they said it might just be reused box and to try it.

Setup is pretty much what 8pack uses for his overclocked bundles:

Asus ROG Strix 270G
BeQuiet 240 AIO
16GB corsair Vengeance 3200
Superflower 750 Gold

OCUK page says the following should work:
Settings for 4.9GHz:
- Vcore 1.32-1.39v load typically 1.35v needed at most.
- VCCSA 1.05 for 3200mhz mems Upto 1.25 for DDR4000mhz +
- VCCIO 1.0 for 3200mhz mems. upto 1.25 For DDR4000mhz +

I'm having issues getting it stable at 4.8, let alone 4.9 but to be fair to them I want to make sure I give the CPU a fair go - the BIOS is a fair bit more complex than my last one (and I'm pretty rusty)

So far I've updated BIOS, loaded optimized defaults and just set it for all cores at 48 adaptive. RAM is on spec at 1600mhz 16-18-18-36-2T

vCore gets up to about 1.36 at 4.8ghz. Running RealBench temps max out at 71C with Cores1/3 about 10C higher than Core 2/4. Realbench runs fine but starting up anything else and it's not stable. Pretty sure it's on good but think it's worth reseating the waterblock?

If I try increasing multi to 49 adaptive pushes voltage above 1.41v which I'm not really comfortable with, even if it was stable at that.

Is there a BIOS profile somewhere I can try or a thread giving some more detailed BIOS settings?. Or can someone help me on some settings that "should" work at 4.9ghz before I phone OCUK and ask for a replacement?

8 Pack · 1 Apr 2017 at 17:51

Vccsa auto vccio auto. Memory xmp with 1.4v. Test.....

Then try also 1.2Sa 1.15io

Doug2507 · 1 Apr 2017 at 17:56

Set voltage to manual, llc 5/6.

10 deg delta isn't anything to loose sleep over but may be due to crap re seal.

covenantuk · 1 Apr 2017 at 18:52

8 Pack said:
Vccsa auto vccio auto. Memory xmp with 1.4v. Test.....

Then try also 1.2Sa 1.15io

VCCSA and VCCIO on auto
With P95 and XTU running not stable under RealBench.

SA and IO set to above
Thos vales weren't much difference from the "Auto" values but I set them manually.
At 4.9 adaptive: CPU: 1.41v
Better, stable under P95 and XTU intially but eventually RealBench crashes. Cores 0/2 max out at 69C/70C but Cores1/3 very hot, reaching 82C and 89C

Ev0 · 1 Apr 2017 at 19:45

Sounds like similar issues others have been having with these chips, might be worth an RMA.

I had one core a good 10 to 15 degrees higher than the rest, wasn't stable at anything above 4.6.

New chip seems better, will do 5Gz with about 1.38/1.39 vCore, am running in a small itx case though so peak temps then are nearing 80 but might just keep it at that as not doing any harm

Dervious · 1 Apr 2017 at 21:10

Quality control seems to be lacking.

Sending 4.9Ghz CPUs in boxes marked 4.8 is asking for trouble.

gavinh87 · 1 Apr 2017 at 21:54

Not the first person to have issues with these delidded binned chips. I hope they are looking into the process in order to prevent this.

covenantuk · 1 Apr 2017 at 22:31

Ok, unless 8pack has any other suggestion I think it's an RMA. I reseated the waterblock and got it stable at 4.9ghz 1.42v vCore... until I fired up Mozilla at the same time, then I got this:

After this all the other stress programs crashed one after the other also. As you can see there's a 16C difference between Cores 0/2 and 1/3

Not only not stable but seems the voltage required is out of spec What you think?

zia · 1 Apr 2017 at 23:03

Return and just buy retail sample.most will do 4.9 easily unless your really unlucky with silicon lottery.Not worth the premium.

Zia

whiterabbit · 2 Apr 2017 at 01:04

covenantuk said:
Ok, unless 8pack has any other suggestion I think it's an RMA. I reseated the waterblock and got it stable at 4.9ghz 1.42v vCore... until I fired up Mozilla at the same time, then I got this:

After this all the other stress programs crashed one after the other also. As you can see there's a 16C difference between Cores 0/2 and 1/3

Not only not stable but seems the voltage required is out of spec What you think?

Are you running those stress tests simultaneously??!!

covenantuk · 2 Apr 2017 at 07:50

No, run Realbench on it's own, because it has AVX in it.

gavinh87 · 2 Apr 2017 at 07:54

Luxmark is a GPU test. I had issues with this in my ryzen build. Use DDU to uninstall display drivers and reinstall. I'm guessing you are using nvidia? Myself an others have had issue with nvidia using open CL

covenantuk · 2 Apr 2017 at 07:57

RealBench is not the only error I get, it just happened to be the one this time.

Fact remains, it's not even stable at 4.9ghz over 1.41v and 2 cores are 16C hotter than the others. At 4.8ghz it runs fine at 1.36v.
I'm thinking that "4.8" was written on the box for a reason.

wedrum · 2 Apr 2017 at 08:02

Is 4.9 from 4.8 really worth it ?

My delided 4790k needs 1.38v for 4.9ghz and 1.22v for 4.8ghz

I run it at 4.8ghz

covenantuk · 2 Apr 2017 at 08:13

If my CPU did that I'd be happy with it and I'd probably run at 4.8 too.

I'm not fussed about 0.1ghz, but I paid a premium for this CPU. Partly for speed guarantee, but also for lower temps from delidding and some headroom as I may move to Spain where the ambient temps are going to be a lot higher.

gavinh87 · 2 Apr 2017 at 08:31

Scotland to Spain, you will feel like you in the earths core!
Seriously I'm sure OCuk will sort this for you.

covenantuk · 2 Apr 2017 at 09:33

gavinh87 said:
Scotland to Spain, you will feel like you in the earths core!
Seriously I'm sure OCuk will sort this for you.

Heh, I'm ok with it, we have a flat there already.

Not had any issue with OCUK before, I just wanted to make sure it wasn't my rustiness with overclocking that was preventing me getting the stated speed.

8 Pack · 2 Apr 2017 at 11:50

Dervious said:
Quality control seems to be lacking.

Sending 4.9Ghz CPUs in boxes marked 4.8 is asking for trouble.

We are working very hard to improve this now. Did u buy one?? U have a lot to say.

We literally sold hundreds of the 5G and 4.9 bin. Percentage overall RMA is low although anything is too much. We recognize this and are acting with robust retesting after delid now in place.

Did the op actually check stock with xmp in place. Maybe a memory issue exists??

The option to RMA is no problem.

covenantuk · 2 Apr 2017 at 12:42

Cheers 8 Pack

Yes I've benched at stock no issues with P95 on Blend. I'll run MemTest today and RMA if the memory checks out all clear.

Ev0 · 2 Apr 2017 at 12:46

On my 4.9 chip, clocked to 4.9, I'm currently seeing vcore at around 1.36v, with ambient temp of 21 degrees it's idling around 29/30 in an NCase M1 and C14 cooler.

Running Cinebench as a quick test it peaks at around 69/70 degrees.

Clocking to 5.0 using 1.39v and running Cinebench a few times it peaks at around 75 degrees.

Running Prime95 with this clock for a quick 10 min run the temps are in the 80s, peaking at 85 on one core so far and hwinfo shows voltage peaking at 1.4v.

Given that airflow is more than likely not the best in such a small case to the above numbers sound reasonable and safe enough?