Have I bought the most unlucky Haswell-E CPUs in a row?

S

sblantipodi

2[H]4U
Joined
Aug 29, 2010
Messages
3,765
Hi,
I have three 5930K in my home running Asus X99 deluxe now.
All rigs has same hardware, Intel 5930K, Asus X99 Deluxe, Corsair H80i GT, 16GB of Corsair LPX 2800MHz and Corsair AX860i.

My three CPUs are not able to go over 4.2GHz and they require from 1.26V to 1.28V to get it stable with RealBench for 8 hours.

Am I the only one who is so unlucky or there are other CPU that can't do more than 4.2GHz? Have I done a guinees world record, three unlucky CPU in a row?
 
Interesting because while there are claims of getting 4.5 stable on the internet of the 4 people I know with x99 boards one only gets 4.0 and the other 3 got stuck at 4.2, with 5920k
 
zaniix said:
Interesting because while there are claims of getting 4.5 stable on the internet of the 4 people I know with x99 boards one only gets 4.0 and the other 3 got stuck at 4.2, with 5920k
Click to expand...

yes, there are lot of people saying they are stable at 4.5GHz with less than 1.3V and this is very strange to me since my three 5930K can't do more than 4.2GHz with that voltge.
 
My 5960x konks out after 4.5GHZ at 1.4v. System becomes very unstable. I keep the system at 4.3Ghz at 1.3v.
 
The Cobra said:
My 5960x konks out after 4.5GHZ at 1.4v. System becomes very unstable. I keep the system at 4.3Ghz at 1.3v.
Click to expand...

1.4V is too much even for custom liquid solutions with this CPUs.
nice to know that there is something else who have similar CPU to mine.

Every people on the net seems able to do 4.5GHz with 1.2V but I think that there are a lot of kids that says bullshit on forums :D
 
Psyshack's 5930K can do 4.8ghz at 1.30v and KickAssCop's 5930K can do 4.6ghz at 1.32...
 
Imho...

The key to these CPU is input voltage and 125 strap.

Give it up to 1.95 input, leave vcore just below 1.3. Run 125 bstrap and adjust multiplier and ram speed accordingly. Leave everything else auto. See what u get.
 
wrangler said:
Imho...

The key to these CPU is input voltage and 125 strap.

Give it up to 1.95 input, leave vcore just below 1.3. Run 125 bstrap and adjust multiplier and ram speed accordingly. Leave everything else auto. See what u get.
Click to expand...

that's correct in fact the default input voltage for the Sabertooth X99 its 1.92v to allow better stability out of the box...
 
Every X99 system I've worked with needed 1.95-2.00 VCCIN to be stable at higher clock speeds. A lot of people want massive uncore as well, which does as close to nothing as possible besides make it hard to get stable. Basically, operator error.
 
wrangler said:
Imho...

The key to these CPU is input voltage and 125 strap.

Give it up to 1.95 input, leave vcore just below 1.3. Run 125 bstrap and adjust multiplier and ram speed accordingly. Leave everything else auto. See what u get.
Click to expand...

125 strap helps highering the memory frequency, it change pretty anything on my CPUs in terms of maximum frequency (with the same vcore/input voltage)

Araxie said:
that's correct in fact the default input voltage for the Sabertooth X99 its 1.92v to allow better stability out of the box...
Click to expand...

they use 1.92V for CPUs at stock frequency? :eek:

LigTasm said:
Every X99 system I've worked with needed 1.95-2.00 VCCIN to be stable at higher clock speeds. A lot of people want massive uncore as well, which does as close to nothing as possible besides make it hard to get stable. Basically, operator error.
Click to expand...

Please define "higher clock speeds". I need 1.92 + LLC7 for 4.2GHz, highering VCCIN more doesn't help me to get it stable at higher frequency.

Nasgul said:
What other settings are you changing besides Vc? What memory RATIO are you using?

AFAIK, Intel CPUs crap out after more than 500-600mhz above their Turbo Max rated speed. With my 5 Way Opt. Z97, my CPU goes to 4.50ghz. Sure if I do it manually it could go higher, but this is not like the old type of BIOS to fiddle with, they used to be much easier, but I'm happy with 4.50ghz.
Click to expand...

I tried 100 strap and 125 strap.
On my best CPU I highered
VCORE to 1.280V,
cache voltage offset to +0.280V,
VCCIN to 1.92V + LLC7,
System agent offset to +0.230V

and I get it stable at 4.3GHz and 3.8GHz cache if I don't use AVX software, 4.2GHz if I use AVX software. Lowering cache freq doesn't help going higher with CPU frequency in this case. This results of a daily of 4.2GHz for my best CPU.
Pretty poor for a bunch of three CPUs.
 
I never really got the impression that 4.5 was a solid average OC, more like high end or lucky, maybe I don't pay enough attention tho... 4.2 isn't that bad, probably not even a 5% difference, tho I can understand wanting to wring every last bit... :p
 
Impulse said:
I never really got the impression that 4.5 was a solid average OC, more like high end or lucky, maybe I don't pay enough attention tho... 4.2 isn't that bad, probably not even a 5% difference, tho I can understand wanting to wring every last bit... :p
Click to expand...

4.2GHz is not bad but 4.4GHz is more cool. :D
 
My 5820k does 4.3 @ 1.27vcore. Definitely not a great chip, but not horrible either. I haven't tried going further, but I will likely need 1.35 or more to get it stable at 4.4 and I don't feel like going over 1.3. It already gets uncomfortably hot during small FFTs.

I used prime95 for stability (this seems somewhat controversial but w/e) and noticed that my #1 core (thread 2) always failed first. The other cores were stable at 1.24, 1.25, and 1.26v but core #1 always failed with rounding errors. With these chips, I think it's just less likely to get all 6-8 cores that are great overclockers. Just a probability thing.
 
khaki said:
My 5820k does 4.3 @ 1.27vcore. Definitely not a great chip, but not horrible either. I haven't tried going further, but I will likely need 1.35 or more to get it stable at 4.4 and I don't feel like going over 1.3. It already gets uncomfortably hot during small FFTs.

I used prime95 for stability (this seems somewhat controversial but w/e) and noticed that my #1 core (thread 2) always failed first. The other cores were stable at 1.24, 1.25, and 1.26v but core #1 always failed with rounding errors. With these chips, I think it's just less likely to get all 6-8 cores that are great overclockers. Just a probability thing.
Click to expand...

not bad, what Input Voltage you use and what LLC?
 
I can do 4.4GHz at 1.28v on my 5960X. I have to go to 1.30v for 4.5GHz. That's really not bad. My engineering sample needs 1.35v to do 4.5GHz. I think everything but my CPU voltage and CPU input voltage is set to "auto" for both chips on most boards.

And yes, on the Sabertooth X99 the default CPU input voltage is 1.92v.
 
sblantipodi said:
Please define "higher clock speeds". I need 1.92 + LLC7 for 4.2GHz, highering VCCIN more doesn't help me to get it stable at higher frequency.
Click to expand...

I've overclocked six 5820k's and two 5930k's, all of them were in the 4.5-4.6ghz range. I haven't had a chance to get my hands on a 5960X but I would suspect it to average lower with two extra cores. Most of them with bargain boards like the MSI X99 SLI Plus or ASRock Extreme4. I've heard of a few chips getting stuck at 4.2-4.3ghz but not many, maybe you got really unlucky.
 
LigTasm said:
I've overclocked six 5820k's and two 5930k's, all of them were in the 4.5-4.6ghz range. I haven't had a chance to get my hands on a 5960X but I would suspect it to average lower with two extra cores. Most of them with bargain boards like the MSI X99 SLI Plus or ASRock Extreme4. I've heard of a few chips getting stuck at 4.2-4.3ghz but not many, maybe you got really unlucky.
Click to expand...

it is probable, I hope that this thread will help to shed some light.
 
I've got a 5820K in a box on my desk and newegg is dropping off an asrock extreme4 and 16gb of gskill 3000 today.

I plan on going 1.3v and let the cards fall where they may. 4.4 I'll be a little dissapointed, 4.5 will be a nice hit. 4.6 is what I am realistically hoping for. If I get 4.7 or better I'll be elated.
 
obiwansotti said:
I've got a 5820K in a box on my desk and newegg is dropping off an asrock extreme4 and 16gb of gskill 3000 today.

I plan on going 1.3v and let the cards fall where they may. 4.4 I'll be a little dissapointed, 4.5 will be a nice hit. 4.6 is what I am realistically hoping for. If I get 4.7 or better I'll be elated.
Click to expand...

please share your experience once you get that nice CPU. :)
 
Dan_D said:
You probably need more than 1.40v at 4.6GHz. Good luck keeping that cool.
Click to expand...

I'm hoping since 4.5 was so easy (might be stable at less voltage too), that I get 4.6 with some tweaking.

Temps were mid 70s with 1.35v @ 4.6 until it locks up.
 
obiwansotti said:
I'm hoping since 4.5 was so easy (might be stable at less voltage too), that I get 4.6 with some tweaking.

Temps were mid 70s with 1.35v @ 4.6 until it locks up.
Click to expand...

With the more recent CPU'S it seems like that last 100MHz costs a lot. You pay in voltage and heat generation.
 
Last edited:
Dan_D said:
Worn the more recent CPU'S it seems like that last 100MHz costs a lot.
Click to expand...

Yeah that's a bummer.

Using Cinebench as a reference new proc is 50% faster single threaded than my old [email protected] and 100% faster multithreaded.

I can live with that.

Not quite as good as my last upgade where I went from an opteron 180 @ 2.5ghz to the i7-920, but beggars can't be choosers.
 
Dan_D said:
With the more recent CPU'S it seems like that last 100MHz costs a lot. You pay in voltage and heat generation.
Click to expand...

So what's the consensus on the Processor Integrated VR Faults and Processor Integrated VR Efficiency Mode?

I seem to get the idea that I should turn these off as both the bios and the XTU app say that it increases headroom, but what's the downside?
 
Dan_D said:
You probably need more than 1.40v at 4.6GHz. Good luck keeping that cool.
Click to expand...

1.4 could be an insta kill voltage on this CPUs.
everthing over 1.35V IMHO should be considered a damaging voltage.

too much watt on the core.
 
sblantipodi said:
1.4 could be an insta kill voltage on this CPUs.
everthing over 1.35V IMHO should be considered a damaging voltage.

too much watt on the core.
Click to expand...

Uh that's for your pessimism I guess some of us are more [H] than others.

1.4v totally boots, is certainly not an instakill.

1.35v seems to be stable with stress test temps around 85c.
 
You must log in or register to reply here.
Back
Top