overclocking 5820K Disabling cores for better performance

[Curufinwe_wins]

1 minute ago, jwakeford said:

Thank you very much for your answer, you really clarified my doubts about the matter. 
I also really appreciate you taking your time to bring some actual proofs to settle the matter. 

In the end, I will go with a 5820K !  
-> Modularity 
-> Future proofing with multicore hungry games
-> Value just better paying the same as 6700k but +2 cores and double the threads all while still being comparably overclocked. 
 

Really though. The biggest thing is that you can see insanely huge differences once you start doing multi-gpu configurations (if you ever do that). 

 

 

 

This is maxed out 4k gaming on 3 of the most gpu demanding games made at the time, and there is still a huge difference.

 

Heck digital foundary saw this issue quite a bit with their fury x cf commentary.

[done12many2]

Just now, Curufinwe_wins said:

I hates you...

 

I can run 4.9 at 1.45 V on my 5820k all 6 cores, but my IMC is so bad I can't even do DDR4 2666 with any kit I've tried. I also can't push even 1 MHz out of spec on bus clocks.

You won't believe this but the cache on my 5960x does 1:1 all the way to 4.9.   I just happened to be on my 4.7/4.7 daily profile when I saw a bunch of people spitballing their theories as fact in this thread so I rebooted up'd the multiplier and took some really quick temp/voltage comparisons.

 

I hate you back for a 4.9 CPU at 1.45v.  Then again, I might actually be able to do that with only 6 cores active.  Oh boy, wouldn't that further solidify the argument.    

[jwakeford]

4 minutes ago, Curufinwe_wins said:

Really though. The biggest thing is that you can see insanely huge differences once you start doing multi-gpu configurations (if you ever do that). 

 

 

 

This is maxed out 4k gaming on 3 of the most gpu demanding games made at the time, and there is still a huge difference.

 

Heck digital foundary saw this issue quite a bit with their fury x cf commentary.

Yes I do plan on getting a 2 card setup in the future, but I have to wait and see what AMD has up its sleeve before biting the Pascal bullet in June.
 
 

[done12many2]

7 minutes ago, jwakeford said:

Thank you very much for your answer, you really clarified my doubts about the matter. 
I also really appreciate you taking your time to bring some actual proofs to settle the matter. 

In the end, I will go with a 5820K !  
-> Modularity 
-> Future proofing with multicore hungry games
-> Value just better paying the same as 6700k but +2 cores and double the threads all while still being comparably overclocked. 
 

Yeah man, no sweat, just remember that you're going to need to cool that badboy well if you expect it to keep up with a big dog like the 6700k.  A big mistake that most 5820k, 5930k and 5960x owners make is they try to keep up with the frequencies of the 6700k with all cores active and while running  a cooling setup comparable to something a 6700k would use.  It's just not going to happen.  That's why you see a lot of threads on 5820k's only hitting 4.4 to 4.5.  Cool it, give it some juice, shut down a couple of cores and let her rip.  Oh, and while I personally don't care which ones are running, you may want to take the time to see which cores OC the highest and obviously use them for your 4 core profile in BIOS.

[Moonzy]

goddam did this thread turn into a good debate

 

its still down to silicone lottery anyway ._.

 

and op might be interested in waiting for broadwell-e? 14nm and what not, might be better in the efficiency department thus causing less heat

 

interesting question, which cpu have an higher average overclock in clock speed? 5820k or 6700k?

[jwakeford]

2 hours ago, done12many2 said:

OP, there sure is a lot speculating at best going on in this thread.  With 6 cores and higher, your biggest battle and limiting factor in overclocking is going to be heat.  It is very common for those with 6 or more cores to turn off unneeded cores in order to to achieve higher clock rates.    

 

No, turning off cores wont overcome the limits set by the quality of your chip, but adding voltage while reducing/maintaining heat does within reason, therefore your overclock will in most cases increase at a lower core count.

 

My particular chip does well with 8 cores active at 4.9 GHz, however anything above that requires dropping cores.  Not so much due to heat because my cooling setup tames that well, but it just runs faster on 4 cores with more voltage than it does 8.

 

Also, generally speaking, most people with a 5820k, 5930k, or 5960x don't disable cores because most people with these setups are using their rigs for purposes other than benchmarking wars and the such.  On the occasions that I beat out a quad core, it's usually because I was only running 4 cores at the time.  That's my experience and I believe it to be directly relevant to what you're asking.

Great I was just typing a question about that but you were faster  

[Curufinwe_wins]

10 minutes ago, done12many2 said:

You won't believe this but the cache on my 5960x does 1:1 all the way to 4.9.   I just happened to be on my 4.7/4.7 daily profile when I saw a bunch of people spitballing their theories as fact in this thread so I rebooted up'd the multiplier and took some really quick temp/voltage comparisons.

 

I hate you back for a 4.9 CPU at 1.45v.  Then again, I might actually be able to do that with only 6 cores active.  Oh boy, wouldn't that further solidify the argument.    

I can hold it up to 4000 then I can't get a lick higher. I mean I'm only on a Sabertooth board, but still. I haven't tried messing with SA etc etc voltages since I tried and still couldn't get my stupid IMC to work at/above 2666. Fuck I had a hard enough time getting quad channel 2400 to work.

 

I had to do a shitload of multi step crap just to get it to register.

1: Load PC and see 3 sticks registered

2: Set speed to 2133.

3. Restart to Bios

4: Set voltage to 1.35V on each dimm

5: Restart to bios

6: Set speed to XMP 2400

7: pray to the gods.

8: If it worked I can tighten some timings at least. And from then on I'm good at my settings.

 

I've tried with other quad channel sets and my IMC just is trash. I couldn't even get the cpu to post with 4 other motherboards I tried (with more than one or two sticks). It was a fucking disaster when I bought it (I wanted to return it but I told the guys at microcenter if you can get it working with a different motherboard fine I'll take that one.)

 

But hey, I can push core clocks really nicely. My 24/7 OC is 45/40 @ 1.25 which is even a bit much on the voltage side (but hey adaptive 1.05+.2 works well).

[jwakeford]

14 minutes ago, Moonzy said:

goddam did this thread turn into a good debate

 

its still down to silicone lottery anyway ._.

 

and op might be interested in waiting for broadwell-e? 14nm and what not, might be better in the efficiency department thus causing less heat

 

interesting question, which cpu have an higher average overclock in clock speed? 5820k or 6700k?

Well it is an interesting subject I think and it touches to the future of our hardware...
Yes I might wait for broadwell-e while I'm at it waiting for pascal/polaris.
But the question and the answer is equally relevant. 
As for higher clock speed I would say the 6700K but 5820 has 12 vs 8 threads to feed those cores so the 6700K might win in speed but in real life it may be a different story. 
 

[Curufinwe_wins]

3 minutes ago, Moonzy said:

goddam did this thread turn into a good debate

 

its still down to silicone lottery anyway ._.

 

and op might be interested in waiting for broadwell-e? 14nm and what not, might be better in the efficiency department thus causing less heat

 

interesting question, which cpu have an higher average overclock in clock speed? 5820k or 6700k?

Compared to stock?

 

For the 5820k a fairly conservative 99+% possible overclock is 4.5 at 1.3V.

For the 6700k a fairly conservative 95%+ possible overclock is 4.6 at 1.35V.

 

From there the 6700k starts to vary dramatically while the 5820k has arguably the lowest "silicon lottery" variance of any cpu ever made outside of the 4790k.

 

Assuming you start messing with VCCIN and you can keep the thing cool enough almost every 5820k should be able to hit 4.7-4.8 at 1.45 (I can hit 4.9).

The 6700k also depends quite a bit more on motherboard quality.

 

But generally speaking OC to OC they reach fairly similar clock speeds at fairly similar voltages (with the 5820k being lower top end clock speeds but also slightly lower voltages).

 

Obviously this means in OC margin over stock the 5820k is king.

[done12many2]

2 minutes ago, Curufinwe_wins said:

I can hold it up to 4000 then I can't get a lick higher. I mean I'm only on a Sabertooth board, but still. I haven't tried messing with SA etc etc voltages since I tried and still couldn't get my stupid IMC to work at/above 2666. Fuck I had a hard enough time getting quad channel 2400 to work.

 

I had to do a shitload of multi step crap just to get it to register.

1: Load PC and see 3 sticks registered

2: Set voltage to 1.35V on each dimm

3: Restart to bios

4: Set speed to XMP 2400

5: pray to the gods.

 

I've tried with other quad channel sets and my IMC just is trash. I couldn't even get the cpu to post with 4 other motherboards I tried (with more than one or two sticks). It was a fucking disaster when I bought it (I wanted to return it but I told the guys at microcenter if you can get it working with a different motherboard fine I'll take that one.)

 

But hey, I can push core clocks really nicely. My 24/7 OC is 45/40 @ 1.25 which is even a bit much on the voltage side (but hey adaptive 1.05+.2 works well).

I had the same problem with my IMC initially.  Went through approximately 10 set of RAM and not matter what I did with DRAM, SA, or cache voltage, it just wouldn't OC well.  After two different rocket scientists at Asus customer support told me that I had a faulty CPU, I came to the conclusion that I was heading in the right direction when I suspected the motherboard.  Asus customer service loves to blame other parts for issues.  By the way, I really like Asus, but I hate their CS.  

 

Anyways, it was the board.  I took mine back to Micro Center and swapped it out.  Guess what, 3200MHz XMP loaded perfectly with one click.  Turns out Asus CS may have been a tad bit wrong on their assessment of my CPU being the issue. 

 

Now memory easily overclocks past 3200 even with tighter timings and the cache performance is ridiculous!

[Curufinwe_wins]

1 minute ago, jwakeford said:

Well it is an interesting subject I think and it touches to the future of our hardware...
Yes I might wait for broadwell-e while I'm at it waiting for pascal/polaris.
But the question and the answer is equally relevant. 
As for higher clock speed I would say the 6700K but 5820 has double the threads to feed those cores so the 6700K might win in speed but in real life it may be a different story. 
 

6700k has 8 threads, 5820k has 12.

 

For older games where it doesn't matter, 6700k wins.

For newer games with poor cpu optimization (single gpu), 6700k wins.

For newer games with good cpu optimization or mult-gpu configurations, 5820k wins.

For future games with ever increasing cpu optimization, you would naturally expect the 5820k to more than keep up.

 

As a side note, memory bottle-necking is becoming more and more of an issue with newer games (that use CPU's better), while for now the 6700k paired with fast DDR4 can capture most of those gains, the issue is far from being a problem for the 5820k with quad channel memory (for reads at least, for writes the x99 platform seems capped a bit lower than it should be and can be overtaken by 3600+ speed dual channel ddr4 skylake systems.)

 

http://www.corsair.com/en-us/blog/2015/september/ddr3_vs_ddr4_generational

[Curufinwe_wins]

6 minutes ago, done12many2 said:

I had the same problem with my IMC initially.  Went through approximately 10 set of RAM and not matter what I did with DRAM, SA, or cache voltage, it just wouldn't OC well.  After two different rocket scientists at Asus customer support told me that I had a faulty CPU, I came to the conclusion that I was heading in the right direction when I suspected the motherboard.  Asus customer service loves to blame other parts for issues.  By the way, I really like Asus, but I hate their CS.  

 

Anyways, it was the board.  I took mine back to Micro Center and swapped it out.  Guess what, 3200MHz XMP loaded perfectly with one click.  Turns out Asus CS may have been a tad bit wrong on their assessment of my CPU being the issue. 

 

Now memory easily overclocks past 3200 even with tighter timings and the cache performance is ridiculous!

I don't like ASUS... I didn't want to buy this board, but I wasn't going to get a blue one and I was out of time. Now it's grown on me quite a bit (esp with the insane fan support), but it still has so many stupid design flaws.

 

Also those 4 motherboards were from MSI lol (I really wanted a SLI Plus board, and then I tried one Gaming 7 as well). Got shafted really hard. (although in physical stores it's not uncommon for issues to occur in batches and being low volume all parts might be from the same batch).

 

Yea I don't even want to talk about ASUS CM. Crucial/Micron is my favorite followed far behind by EVGA then a ways back MSI then everyone else in the dust.

[Moonzy]

1 minute ago, jwakeford said:

Well it is an interesting subject I think and it touches to the future of our hardware...
Yes I might wait for broadwell-e while I'm at it waiting for pascal/polaris.
But the question and the answer is equally relevant. 
As for higher clock speed I would say the 6700K but 5820 has double the threads to feed those cores so the 6700K might win in speed but in real life it may be a different story. 

i believe the 5820k will run hotter than the 6700k even with 4c and 8 threads on both due to architecture differences and mainly transistor size differences

so broadwell-e should be able to patch that

3 minutes ago, Curufinwe_wins said:

For the 5820k a fairly conservative 99+% possible overclock is 4.5 at 1.3V.

For the 6700k a fairly conservative 95%+ possible overclock is 4.6 at 1.35V.

so if the 5820k disables 2 cores, it should be able to clock higher right? (theres no 100% in OC but yea) 

which will theoretically be better than the 6700k, even with differences in IPC

 

all in all, the 5820k looks like a better package here since its priced almost similarly to the 6700k

while having more cores and more cache

 

besides, DX12 is still iffy, there's the multi gpu support which isnt that well established yet (mixing amd and nvidia and what not)

so only time can tell whats going to happen, and whether or not if we need more pcie lanes to benefit from DX12 features

[done12many2]

13 minutes ago, Moonzy said:

its still down to silicone lottery anyway ._.

 

and op might be interested in waiting for broadwell-e? 14nm and what not, might be better in the efficiency department thus causing less heat

 

interesting question, which cpu have an higher average overclock in clock speed? 5820k or 6700k?

6700k on average because they run cooler.  It's just the way it works when you have two less cores.  Now take a 5820k and throw the appropriate amount of cooling at it along with voltage (just like you do with a 6700k) and they're all about the same.  6700k still wins clock for clock due to the IPC.

 

I do believe that the "silicon lottery" has something to do with it, but not nearly as much as one's understanding of overclocking.  Throw some voltage at something, keep it cool, and watch it go.  That's exactly how it's always worked for me.

 

Lastly, I will be the first one in line at Micro Center with Broadwell-E drops.

[jwakeford]

2 minutes ago, Moonzy said:

i believe the 5820k will run hotter than the 6700k even with 4c and 8 threads on both due to architecture differences and mainly transistor size differences

so broadwell-e should be able to patch that

so if the 5820k disables 2 cores, it should be able to clock higher right? (theres no 100% in OC but yea) 

which will theoretically be better than the 6700k, even with differences in IPC

 

all in all, the 5820k looks like a better package here since its priced almost similarly to the 6700k

while having more cores and more cache

 

besides, DX12 is still iffy, there's the multi gpu support which isnt that well established yet (mixing amd and nvidia and what not)

so only time can tell whats going to happen, and whether or not if we need more pcie lanes to benefit from DX12 features

Yes that is my point, hold on vs 6700K with current games by disabling 2 cores and a nice overclock, and as dx12and cpu optimizations become widespread I will reactivate these cores and finally destroy the 6700K. 
I am inveting 4500€ in my new setup ( X34 predator plus PC ) so yeah I want it to be future proof as I won't be changing parts for at least 3 years. 
 

[Curufinwe_wins]

3 minutes ago, Moonzy said:

i believe the 5820k will run hotter than the 6700k even with 4c and 8 threads on both due to architecture differences and mainly transistor size differences

so broadwell-e should be able to patch that

so if the 5820k disables 2 cores, it should be able to clock higher right? (theres no 100% in OC but yea) 

which will theoretically be better than the 6700k, even with differences in IPC

 

all in all, the 5820k looks like a better package here since its priced almost similarly to the 6700k

while having more cores and more cache

 

besides, DX12 is still iffy, there's the multi gpu support which isnt that well established yet (mixing amd and nvidia and what not)

so only time can tell whats going to happen, and whether or not if we need more pcie lanes to benefit from DX12 features

Actually the soldered IHS outperforms Skylake's TIM (which is much better than first gen haswell TIM but worse than haswell-refresh TIM) by quite a large margin.

 

But yea, you could generally clock a little bit higher (there isn't any real reason to think a 4 core 5820k couldn't match clocks fairly well with a 4790k at least until exotic cooling), but core per core I doubt it will ever exceed Skylake. In many cases it may match it though.

[done12many2]

16 minutes ago, jwakeford said:

Well it is an interesting subject I think and it touches to the future of our hardware...
Yes I might wait for broadwell-e while I'm at it waiting for pascal/polaris.
But the question and the answer is equally relevant. 
As for higher clock speed I would say the 6700K but 5820 has 12 vs 8 threads to feed those cores so the 6700K might win in speed but in real life it may be a different story. 
 

I personally would wait if I were you.  The increase in IPC may not be much, but it'll put you and a 6700k dead even clock for clock.  Also, when Haswell became Skylake, overclocking became more predictable and consistent.  When Haswell-E becomes Broadwell-E, I suspect the same will occur.

[jwakeford]

1 minute ago, done12many2 said:

I personally would wait if I were you.  The increase in IPC may not be much, but it'll put you and a 6700k dead even clock for clock.  Also, when Haswell became Skylake, overclocking became more predictable and consistent.  When Haswell-E becomes Broadwell-E, I suspect the same will occur.

Yes I will wait for Broadwell-E 
Any info on the specs of the lower priced one ( similar to 5820K) 

[Curufinwe_wins]

Just now, done12many2 said:

I personally would wait if I were you.  The increase in IPC may not be much, but it'll put you and a 6700k dead even clock for clock.  Also, when Haswell became Skylake, overclocking became more predictable and consistent.  When Haswell-E becomes Broadwell-E, I suspect the same will occur.

Ehh...

 

Haswell-refresh is still by far the most consistent overclocker ever made.

 

As far as overclocking consistency:

Devil's Canyon>Haswell-E>Skylake>Broadwell>Haswell (Haswell 1st gen's main issue was the truely ghastly TIM).

 

If they super over-build the FIVR on Broadwell-E like they did on Devils canyon then yes it will be more consistent, but otherwise die shrinks generally make things less consistent not more (and much more temperature variant).

 

 

[done12many2]

1 minute ago, jwakeford said:

Yes I will wait for Broadwell-E 
Any info on the specs of the lower priced one ( similar to 5820K) 

High 3's based on the last bit of speculation I read.  Last time I was at Micro Center you could pick up a 5820k for $299 and if you bought a motherboard, they knocked $20 off.  Unbelievable if you ask me.  Micro Center is my go to.  They will pricematch anything and the only thing that sucks is sales tax.  Returning items is a painless process with no questions asked.

[jwakeford]

1 minute ago, done12many2 said:

High 3's based on the last bit of speculation I read.  Last time I was at Micro Center you could pick up a 5820k for $299 and if you bought a motherboard, they knocked $20 off.  Unbelievable if you ask me.  Micro Center is my go to.  They will pricematch anything and the only thing that sucks is sales tax.  Returning items is a painless process with no questions asked.

If only I had Micro Centers in France... 
Here the cheapest 5820K you can get is 409€ which is 460$ ! 

[done12many2]

1 minute ago, Curufinwe_wins said:

Ehh...

 

Haswell-refresh is still by far the most consistent overclocker ever made.

 

As far as overclocking consistency:

Devil's Canyon>Haswell-E>Skylake>Broadwell>Haswell (Haswell 1st gen's main issue was the truely ghastly TIM).

 

If they super over-build the FIVR on Broadwell-E like they did on Devils canyon then yes it will be more consistent, but otherwise die shrinks generally make things less consistent not more (and much more temperature variant).

Okay, maybe I'm talking out of my ass because you're right  

 

Now that I'm thinking about it, I think my 4790k is a badass little chip!  That sucker will run 4.8 all day long on a simple little loop that I threw together for use as a backup PC.  It'll go higher, but I don't really use it.

[Curufinwe_wins]

3 minutes ago, jwakeford said:

If only I had Micro Centers in France... 
Here the cheapest 5820K you can get is 409€ which is 460$ ! 

I got mine for 279.99 USD. And 20 dollars off the motherboard.

 

Mind you I don't normally live near a Microcenter, but I had a summer job near one so my last week I bought and built my computer. Sadly things were not going very smoothly and it was super stressful, but the idea was good and I did save lots of money.

[done12many2]

10 minutes ago, jwakeford said:

If only I had Micro Centers in France... 
Here the cheapest 5820K you can get is 409€ which is 460$ ! 

 

5 minutes ago, Curufinwe_wins said:

I got mine for 279.99 USD. And 20 dollars off the motherboard.

 

Mind you I don't normally live near a Microcenter, but I had a summer job near one so my last week I bought and built my computer. Sadly things were not going very smoothly and it was super stressful, but the idea was good and I did save lots of money.

Gentleman, prepare to hate.  There are two within a 30 to 40 minute drive from me.  

 

I hope you guys have a great night.

[Curufinwe_wins]

Just now, done12many2 said:

 

Gentleman, prepare to hate.  There are two within a 30 to 40 minute drive from me.  

 

I hope you guys have a great night.

At least I don't have to drive across the ocean for one.