6700K fails Prime 95 blend at stock speed (Update: Resolved. Now limited by temps.)

[dank69]

I've been out of the game for 5+ years now. What are the best tools for testing a new system these days? I used to use Prime 95, Real Temp, and some other stress test app I can't remember back then.

Anyway, my new system is failing Prime 95 after a few hours on one worker out of eight. The other seven workers were stable after 8+ hours. Core temps peaked at 72C, so I don't think cooling is the problem, right?

System specs:
6700K
Asrock Z170 Extreme+
2x16GB Gskill Trident Z 2800 CAS 14
Coolermaster 212 with push/pull fans

Everything is stock. I haven't even begun changing anything in the BIOS. Anyone have any ideas what I can do to isolate the problem?

UPDATE: Getting latest version of Prime 95 resolved the issue.

 

[Burpo]

I trust IBT (Intel Burn Test) over Prime on newer processors. What power supply are you using?

 

[ShintaiDK]

What version of prime? One of them is buggy if I recall.

 

[Ken g6]

Have you run memtest to rule out bad RAM?

Try Prime95 with HT off.

 

[hectorsm]

I will also suggest checking the memory with memtest86 first. CPU failing at stock speed is very unusual. It's more likely that something else is wrong.

 

[dank69]

I trust IBT (Intel Burn Test) over Prime on newer processors. What power supply are you using?

I have a Seasonic 750 80+ gold. I'll try the memtest next as the others have suggested.

 

[Brunnis]

Before moving on to another test, I would lower the mem speed and timings and make another try. I suggest going for DDR4-2133 with timings at auto.

 

[Deders]

Mae sure you test with Prime version 28.7 or later. Previous versions report errors where there shouldn't be with FMA3/AVX loads.

 

[Flapdrol1337]

Any bios updates avialable?

could also be bad ram like others have said, so maybe worth it to test the memory modules individually.

 

[dank69]

So I ran memtest86 overnight. 2 passes completed without any errors. I left it running and will check when I get home. How many passes necessary before we can rule out RAM?

 

[Deders]

What version of prime are you using?

 

[nanaki333]

I know this is not really relevant to your situation as it was a different CPU altogether, but I got an AMD chip (FX 8120) that had half the cores fail Prime95 after about 30 minutes but would pass up until then. It had a decent cooler on it but would fail after it was going awhile at the 30 minute mark give or a take a minute.

Thought I would give it a pass and try gaming and I had games crash after playing about 30-45 minutes too so contacted AMD and they gave me an RMA. Got the new chip back and I had no issues at all. Let Prime95 run a few hours then tested it for gaming and had no issues at all.

Afterward, I boxed the machine up and gave it away for one of those silent charity auction things and to my knowledge, the owner still plays on it to this day!

Like I said, completely different chip but I wasn't about to take any chances with it being brand new. AMD got my chip and sent a new one all in less than 2 weeks. I guess they had a lot of these chips with issues and had the RMA down to an art. If it's easy, I would RMA it. Especially if you can send it back to a seller and get an immediate replacement.

Now you got me wanting to get motivated to get my i5 Skylake in its new home to make sure it works!

 

[dank69]

What version of prime are you using?

I'll have to check when I get home today. I thought I downloaded the latest version though.

 

[Deders]

I'll have to check when I get home today. I thought I downloaded the latest version though.

So did I but there was a newer version in the forums. If the one on the main page is still the one I downloaded a couple of months back then it is buggy.

 

[dank69]

So did I but there was a newer version in the forums. If the one on the main page is still the one I downloaded a couple of months back then it is buggy.

Hopefully that is the problem then. I just checked and it is 27.7. I'll have to get the latest version. Thanks!

 

[BTRY B 529th FA BN]

Use Memtest HCI

 

[Termie]

So I ran memtest86 overnight. 2 passes completed without any errors. I left it running and will check when I get home. How many passes necessary before we can rule out RAM?

You can rule out the RAM if the system fails Prime95 at DDR4-2133. If you haven't touched XMP, it's extremely unlikely it's the RAM, but running on a single stick would help you be even more certain.

Or use Intel Burn Test.

 

[dank69]

I got Prime 95 28.7 and ran blend for 24+ hours with no problems. It seems it was just the 27.7 version was faulty as Deders suggested. Now I need to find time to see what kind of overclocking I can achieve. :thumbsup:

 

[ShintaiDK]

Good it was just the faulty software

 

[dank69]

So I have begun the process of overclocking and things look great so far. I was able to validate 4.9GHz @ only 1.4v! Of course, it wasn't anywhere near stable, but is this typical for a 6700K on air? I also noticed that when I bumped the vcore to 1.45v to see if it was more stable, CPU-Z showed vcore spiking to 1.5v. Is this safe? I read that you shouldn't set it above 1.45v, which I haven't. You can see with this validation the vcore spiked to 1.488.

 

[wanderica]

So I have begun the process of overclocking and things look great so far. I was able to validate 4.9GHz @ only 1.4v! Of course, it wasn't anywhere near stable, but is this typical for a 6700K on air? I also noticed that when I bumped the vcore to 1.45v to see if it was more stable, CPU-Z showed vcore spiking to 1.5v. Is this safe? I read that you shouldn't set it above 1.45v, which I haven't. You can see with this validation the vcore spiked to 1.488.

I have been OCing mine for the last couple of days on a U14S, and Windows wouldn't boot at 4.9 GHz @ 1.4v. 4.8 @ 1.4v would boot just fine, but I get a BSOD 2 cycles into LinX. I haven't tried upping the VCore any more yet. After delidding, temps are in the low 70s running Prime95 with 4.7 GHz @ 1.4v. So, I dunno about typical, but you've managed slightly better than I have at 1.4v. My voltage doesn't seem to spike that high. I have seen it shoot up to 1.45 though.

 

[Termie]

Where are you guys finding that 1.45v is an acceptable voltage for skyline?

Anandtech's review:

http://www.anandtech.com/show/9483/intel-skylake-review-6700k-6600k-ddr4-ddr3-ipc-6th-generation/6

Seems that 1.45v is what you use for suicide runs, not 24/7 overclocks, and that voltage really doesn't help Skylake, with maximum stable oc coming at low voltage, perhaps under stock.

 

[Deders]

https://drive.google.com/file/d/0Bz2VRRbLPrZnMXBnOXRWeVlHcHM/view
​

Good samples can achieve 4.7GHz with around 1.40 Vcore fully stable. The highest voltage we recommend using is 1.42V with triple-rad water-cooling if running stress tests with AVX2 routines (lower if ambient temps are high). Those not concerned with stress testing may wish to use up to 1.45V for maximum CPU frequency.​

 

[dank69]

I have been OCing mine for the last couple of days on a U14S, and Windows wouldn't boot at 4.9 GHz @ 1.4v. 4.8 @ 1.4v would boot just fine, but I get a BSOD 2 cycles into LinX. I haven't tried upping the VCore any more yet. After delidding, temps are in the low 70s running Prime95 with 4.7 GHz @ 1.4v. So, I dunno about typical, but you've managed slightly better than I have at 1.4v. My voltage doesn't seem to spike that high. I have seen it shoot up to 1.45 though.

Did you delid yourself? It kind of pisses me off that delidding can drop temps 12-13C. Consumers shouldn't need to assemble the chips properly.

 

[Ketchup]

Did you delid yourself? It kind of pisses me off that delidding can drop temps 12-13C. Consumers shouldn't need to assemble the chips properly.

Chip manufacturers think that the safety of the chip is worth a couple degrees, and I partially agree, but they could offer a desktop model without the lid. They could even charge a premium for it.