Is 1.45 V too much for an air-cooled i7 7700k running at 5.2 GHz?

[trafiq]

Hello everyone,

My first post on here and I thought I'd ask for a second opinion on my system for peace of mind. I don't think I've struck any sort of silicon lottery by a longshot, but at the same time I just want to check that my Vcore won't kill my CPU in the long run.

Please see my setup below; it is my modest and first attempt at overclocking:

• i7 7700k at 5.2 GHz 1.45 V, delidded-relidded w/ Conductonaut liquid metal TIM and cooled by Noctua NH-D15, AVX offset 2
• Gigabyte AORUS Z270X Gaming 7 mobo, BIOS version F4
• LLC profile set to "High"
• 64 GB RAM (4x16GB Corsair Vengeance LED/RGB 3000MHz @ 15-17-17-35 1.35V) running at 3200 MHz and 16-19-19-36 1.35 V
• Gigabyte AORUS XTREME 1080 Ti with Conductonaut liquid metal TIM, clocked +25 MHz on GPU and +480 MHz on memory, hitting 2025-2060 MHz on the clock at about 55 degrees

I'm running stable at these clocks, with CPU temps hitting about 80-81 degrees under OCCT medium dataset with/without AVX, running for 20 minutes. Dropping voltage to 1.44 V gives a BSOD. If I try 5.3 GHz at 1.45 V under load gives a BSOD (boots into Windows fine though).

My questions are:

• Is 1.45 V Vcore too much? Do I run the risk of prematurely frying the CPU if I run this at 24/7? Is it ok if Vcore jumps to 1.46 V during load?
• Do I need to fiddle with max TDP in BIOS at all? It's currently untouched at 91 W; and
• If I want to try 5.3 GHz on air cooling, how feasible do you think it is?

Thanks in advance.

 

[Blaylock]

Personally, I wouldn't run it at that high of a voltage for a 24/7 OC. You are at the breaking point and very well could degrade the CPU prematurely. Try and keep the vCore in the 1.40-1.42v range for maximum life.

 

[EarthDog]

This ^^.

I wouldn't go much over 1.4V 24/7 either for that CPU.

5.3 Ghz isn't feasible for daily operation at the voltage you are currently at.

 

[trafiq]

Thanks for response gents, I'll dial it down to 5.1 GHz and that should bring me in the 1.40 V ballpark.

On the topic of RAM, what would be an approximate recommended DRAM voltage to run 3000 MHz rated sticks at 3600 MHz or even at 3200 MHz? I am at a loss because when I load the XMP profile, all works well at 1.35 V 3000 MHz and 15-17-17-35 timings and I can manually set the clock to 3200 MHz, which then automagically finds the timings 16-19-19-36 and boots, is stable, etc. But when I try to disable the XMP profile and manually set the clock, timings and bias myself, the mobo doesn't POST...

Any ideas?

 

[Blaylock]

For your ram you might be able to get it to run 3600 at the stock 1.35v but it could take up to 1.45v. I think the bigger question will be if your IMC can overclock 64GB of memory. I would start around these timings and voltage 3600 17-18-18-36 1.35V.

 

[trents]

I would not call running OCCT for 20 minutes without AVX a valid stress test. Try two hours with AVX or run the Realbench stress test using half your RAM amount for 2 hr. Many applications now use some AVX instruction. And personally, I would not feel comfortable with more than 1.376 vcore for 24/7.

 

[trafiq]

I would not call running OCCT for 20 minutes without AVX a valid stress test. Try two hours with AVX or run the Realbench stress test using half your RAM amount for 2 hr. Many applications now use some AVX instruction. And personally, I would not feel comfortable with more than 1.376 vcore for 24/7.

It was 20 minutes for both with and without AVX. I've got an AVX offset of 2, so it was running fine.

For your ram you might be able to get it to run 3600 at the stock 1.35v but it could take up to 1.45v. I think the bigger question will be if your IMC can overclock 64GB of memory. I would start around these timings and voltage 3600 17-18-18-36 1.35V.

Thanks, I couldn't get anything over 3333 MHz on my sticks no matter what I tried. Turns out the DRAM Calculator for Ryzen reckons that 3600 MHz is not supported for Hynix memory (SK Hynix as reported by CPU-Z), but I did manage to get it to 3333 MHz at 16-18-18-36 (unstable) and 3333 MHz 16-19-19-38 (diminishing return). I think I'll just stick to 3200 MHz 1.35 V 16-19-19-38 and be on the safe side.

For the record, I managed to get to 3333 MHz and XMP profile off by setting the memory training voltage to be matched to the operating voltage, in this case 1.35 V or 1.4 V. Its "Auto" setting was defaulting to 1.2 V.

Edit: Should I play with the max TDP values in the BIOS to keep the Vcore low and achieve stability?

Thanks a lot for your feedback. Regards

 

[trents]

20 minutes is not a long enough stress test.

 

[EarthDog]

A 7700k with 1.35V is perfectly fine.

20 minutes for 'each' stress test is not nearly enough to test for stability.

 

[Blaylock]

If it was stable for 20 mins that's good for an initial stress test, but you need more as E_D and Trents mentioned. Let it run for a full, uninterupted 2 hours now.

As far as your ram speed, you will be just fine at 3200. It's doubtful you would even notice any difference if you did get to 3600. Most improvements of this nature are only witnessed in simulated benchmarks, in games you can't tell the difference.

I'm not familiar with "Max TDP settings in BIOS" so I will wait for others to respond to this question, hopefully learning something along the way too.

 

[trafiq]

A 7700k with 1.35V is perfectly fine.

20 minutes for 'each' stress test is not nearly enough to test for stability.

Agreed. I've dropped the voltage down to 1.415 V and clock to 5.1 GHz, I've seen stability with OOCT on 20 minutes for no AVX and AVX (offset is now 1, running the AVX instructions at 5.0 GHz) as well as a CPU intensive unzipping task. I've had to up the RAM voltage from 1.35 to 1.4 V at 3200 MHz to achieve stability; now I will run longer stress tests.

Saw the power draw in Core Temp go down from 110-120 W at 5.2 GHz 1.45 V to 80-90 W at 5.1 GHz 1.415 V. Still hoping someone could shed some light on the max TDP settings in BIOS though. Cheers