5820K Overclocking, need some help please

[ebilcake]

I have just put a system together with 5820K, Asus X99 S motherboard, Corsair DDR4 LPX 2666Mhz memory, H100i cooler.

But I'm new to Haswell overclocking and seem to hitting wall at 4.2GHz which is disappointing.

Currently 42x100 with 1.350VC which seems a lot, temps are around 70 - 75C

Any help would be appreciated.

Current settings

42x100
Using XMS Profile
Fully manual Enabled
Core Voltage 1.350v
CPU Cache Voltage 1.250v
CPU SVID Disabled
CPU Input Voltage 1.9v
Spread Spectrum Disabled
Turbo Enabled (Can't seem to disable it anyway)
LLC Level 6
Speedstep Disabled

Everything else is pretty much set to auto.

 

[mickyflinn]

I would use manual timings and voltage on your ram and loosen the timings.

 

[ebilcake]

I would use manual timings and voltage on your ram and loosen the timings.

Thanks for suggestion Micky but how will that help with achieving a higher CPU overclock?

I don't think the memory is limiting me, I would be surprised if it was as when stress testing I decrease the memory speed so can rule it out as the cause of any issues.

 

[mickyflinn]

Sometimes when overclocking your cpu it affects the ram, adding more voltage to compensate or loosening the timings may help to achieve higher overclock.

 

[ALXAndy]

Usually the trick is to just enable XMP on your ram, then work on the multi. That's the easy way and ram should not affect your overclock.

I can get my 3970x to 4.9ghz using that method, so ram definitely is not the limiting factor. I may try and push harder some time soon.

Ram will only be a factor if you use the strap, as it ups the FSB.

 

[ebilcake]

Lets ignore memory for now, any comments on the settings above or any other settings I could try changing?

1.350v seems like a lot of voltage for 4.2 from what I've seen, so either I've got a lemon or I'm doing something wrong.

 

[ALXAndy]

Lets ignore memory for now, any comments on the settings above or any other settings I could try changing?

1.350v seems like a lot of voltage for 4.2 from what I've seen, so either I've got a lemon or I'm doing something wrong.

Dude without being funny all 5820ks were lemons, that's why they're not in a 5930k box.

4.2ghz is about what to expect, 4.4 if you were lucky and 4.6+ if you are god.

If you were told different then some one wasn't being honest with you.

1.35v on a die that small is a lot of voltage. I've seen people achieve 4.6 with 1.26v so obviously yours is just a poor clocker.

Even with all of the bravado certain members who could clock them higher ended up settling with 4.2ghz because of heat and insane voltage requirements.

 

[ebilcake]

Dude without being funny all 5820ks were lemons, that's why they're not in a 5930k box.

4.2ghz is about what to expect, 4.4 if you were lucky and 4.6+ if you are god.

If you were told different then some one wasn't being honest with you.

1.35v on a die that small is a lot of voltage. I've seen people achieve 4.6 with 1.26v so obviously yours is just a poor clocker.

Even with all of the bravado certain members who could clock them higher ended up settling with 4.2ghz because of heat and insane voltage requirements.

Well that's certainly possible but I've gone through a lot of 5820K threads and I haven't seen any examples of users needing 1.350V for 4.2GHZ.

I know users do tend to be less than honest at times when reporting results but you would have expected a few to post something similar.

I suppose I won't know until I try another 5820k.

 

[ALXAndy]

Well that's certainly possible but I've gone through a lot of 5820K threads and I haven't seen any examples of users needing 1.350V for 4.2GHZ.

I know users do tend to be less than honest at times when reporting results but you would have expected a few to post something similar.

I suppose I won't know until I try another 5820k.

I had a GTX 670 that would bomb with anything over 15mhz. I looked around to find any as bad as mine and couldn't lol.

It could be your bios too man. Early bioses on new techs are always ropey. Excuse me if you said, but what board are you using? it's probably better to look for bad board info rather than bad CPU info.

But you might simply have lost the lottery, ripped up your ticket and weed on it like I did with my 670

 

[ebilcake]

I had a GTX 670 that would bomb with anything over 15mhz. I looked around to find any as bad as mine and couldn't lol.

It could be your bios too man. Early bioses on new techs are always ropey. Excuse me if you said, but what board are you using? it's probably better to look for bad board info rather than bad CPU info.

But you might simply have lost the lottery, ripped up your ticket and weed on it like I did with my 670

It's the Asus X99-S and it's using the latest bios, I can get pretty much everything replaced if I need to at no extra cost, that's not an issue.

Just inconvenient.

 

[Boomstick777]

Dude without being funny all 5820ks were lemons, that's why they're not in a 5930k box.

4.2ghz is about what to expect, 4.4 if you were lucky and 4.6+ if you are god.

4.6Ghz here with 5820K..

My best multi score is 1303. With 4.6Ghz / 1.35v overclock and default 1066mhz DDR4 memory speed.

P18646 with NVIDIA GeForce GTX 980(1x) and Intel Core i7-5820K Processor

Graphics Score
20892

Physics Score
14488

Combined Score
13554

http://www.3dmark.com/3dm11/8770484

 

[ginjaninja]

Dude without being funny all 5820ks were lemons, that's why they're not in a 5930k box.

Hi Andy do you have evidence for this position?

its is just as reasonable to suggest that it is not commercially efficient for intel to bin more than the bare minimum to meet the their specs.

If they could make money out of finer granular binning then wouldnt they have more granular products.

it could be that intel marketing suggest they need X% 5960s, Y% 5930s and Z% 5820s per production run...once they have found enough 5960 they find enough of the next tier then the next...resulting in a much smaller likely of worse clocking due to not being binned higher and the larger factor being the quality of the overall production.


of course you could be right and intel run every chip at its highest and work down until they find its limit...but i suspect they make much more money out of 5820s than 5930s, so as long as enough people think they are more likely to perform higher, the market for the more expensive wont collapse.

so it would be good to know the background to your assertion.
or from anyone who knows how cpus are binned.

 

[ALXAndy]

Hi Andy do you have evidence for this position?

its is just as reasonable to suggest that it is not commercially efficient for intel to bin more than the bare minimum to meet the their specs.

It's just good olde binning.For example, the 5820k could have had defective lanes, poorer clocks etc etc.

With the 3930k it was pretty much just one chip, one model. Well, apart from the higher binned 3960x and 3970x (I've had 4.9ghz out of my 3970x @ 1.46v on a H110 no custom loops).

However with the 5820k there are two models, and one is a hell of a load cheaper so you can bet Intel are binning.

 

[ginjaninja]

However with the 5820k there are two models, and one is a hell of a load cheaper so you can bet Intel are binning.

oh..you learn something every day...i had always assumed, that there was single manufacturing process for all three chips, and as they came off the 1st stage of the production line, they were tested for 5960 spec, then 5930 spec, then 5820 spec, and then at a '2nd' stage some minor microcode/pcb adjustment made to make the designation permanent. Which licensed motherboard manufacturers then adhered to..

 

[ALXAndy]

ooooo ! ya got close there Boom but still your chip can't keep the pace with a chip two generations older.



http://www.3dmark.com/3dm11/8947535

I've not even touched anything else yet.

 

[555BUK]

I see binning as similar to moving the Business Class curtain on British Airways flight's. All CPU's are made to be top-end but only a percentage of buyers are willing to pay top prices. As such, many CPU's that are good enough to be 5830K's are simply relegated to lesser parts by disabling cache, PCIE lanes, cores etc. On a flight they simply disable the lounge, metal cutlery and unlimited booze options.

Of course not all parts will meet ultra high-end specs but Intel tends to leave very high margin for error. Very few recent Intel CPU's overclock by less than 20% when given a little encouragement and many "lesser parts" can greatly exceed high-end parts after tweaking. It is slightly different with AMD currently, because they have to clock top parts close to the red line to come anywhere close Intel offerings.

 

[ALXAndy]

oh..you learn something every day...i had always assumed, that there was single manufacturing process for all three chips, and as they came off the 1st stage of the production line, they were tested for 5960 spec, then 5930 spec, then 5820 spec, and then at a '2nd' stage some minor microcode/pcb adjustment made to make the designation permanent. Which licensed motherboard manufacturers then adhered to..

Of course they're all from the same manufacturing line. Then they're binned for quality, speed and functionality, then they're packaged.

Now sometimes when that happens decent chips are boxed up to make the numbers.

However from all I have seen 5930ks always tend to clock better than 5820ks. And that's most certainly because they were of inferior quality and so Intel just sell them as 5820k.

Every other X820k has been quad w/ht. The fact Intel decided to make the 5820k a hex is pretty cool but they seem to be no match for the 5930k.

And, contrary to Boom's excitement when they launched (saying they easily beat a 4960x) the fact is they're no better than Sandy or Ivy because they clock lower. And, in some cases you'll find a 3970x that can clock to 5ghz (still not played with anything more than 4.9 as I've been busy today) which actually beat the 5820k, cost the same for the CPU, less for the board and less for the ram. Plus I've got all of my lanes intact, more cache etc.

Anand were the only reviewers not to get excited and grow a Poncchio nose, and actually be honest about it.

 

[ginjaninja]

I see binning as similar to moving the Business Class curtain on British Airways flight's. All CPU's are made to be top-end but only a percentage of buyers are willing to pay top prices. As such, many CPU's that are good enough to be 5830K's are simply relegated to lesser parts by disabling cache, PCIE lanes, cores etc. On a flight they simply disable the lounge, metal cutlery and unlimited booze options.

this is what i had assumed; it would be good to see some cold light of day numbers with more statistical significance.

the fact that there are few data points in the haswell-e owners thread suggests people are not reaching these levels often or not many people like contributing for another reason...

 

[jimbaw]

even if binning plays a large part, this is ridiculous! get the intel tuning plan and swap it.

and axl, his score was using a single gpu . you're using two titan blacks

 

[Boomstick777]

ooooo ! ya got close there Boom but still your chip can't keep the pace with a chip two generations older.



I've not even touched anything else yet.

Like the way you dodged the comment, you said

Dude without being funny all 5820ks were lemons, that's why they're not in a 5930k box.

4.2ghz is about what to expect, 4.4 if you were lucky and 4.6+ if you are god.

I got 4.6Ghz with ease.