New Build for P3D V3.2- March 2016

[Guest]

My 2 cents (assumption is cost is not relevant):

 

1.  980Ti (SLI not very effective in P3D V3.x but I need to test more with V3.2 and latest nVidia drivers)

2.  Quad channel motherboard and CPU (Asus X99 Extreme boards work well for OC and Asus R&D socket design to optimize performance) ... memory bandwidth is very important and quad channel setups are typically over 60 GB/s

3.  Quad channel RAM (3000Mhz CL15 or better) - 3200 or 3400 might require a lower BCLK.

4.  Good water cooling solution for heat extraction.

5.  Higher BCLK OC seems to work well for me 125 X 35, 36, 37 

6.  PSU ... 850W or higher - try to keep loads around 50% of rated (my setup draws about 450-600 Watts under load, my PSU is 1200 Watt) and most importantly try to obtain output results under rapid load switching (this is where many PSUs fail causing high fluctuations in current)

7.  M.2 Samsungs 512GB drive is much faster than standard SATA6 SSD (motherboard needs to be equipped with M.2 slot(s).

8.  4K monitor (I like Sony, so long as it can do 30hz and 60Hz and look/operate as good at both refresh rates)

9.  5820K 5930K or 5960X 

 

I don't believe in "silicon lottery" ... you can make any CPU work at high OC with some tuning and efficient heat extraction.  If you are concerned about OC I suggest you obtain Intel's replacement guarantee (no questions asked).

 

I also don't believe in "selected" RAM modules at exotic prices ... these so called "performance" bins are done with internal testing equipment that is not representative of what one will encounter on any brand X motherboard using a different PSU, different GPU(s)/components, electrical outlet, PC location, etc. etc. etc.  ... it passed a very specific "controlled" environment -- and one's home environment will be entirely different.

 

Not a fan of Skylake, cache too small and wasted graphics unit just taking up space.

 

OC can help you hit that min FPS to match your monitor's refresh rate ... for example Vsync On, 30Hz refresh 4.3Ghz ... if I drop down to 28 FPS I get a stutter or two, at 4.6Ghz I don't drop below 30 FPS (that 2 FPS makes a difference), no stutters and all smooth.

 

Just my opinion of what has worked for me and I'm certainly NOT suggesting anything beyond that and no "golden" rules.

 

Cheers, Rob.

[max2770 75]

 

 

Not a fan of Skylake, cache too small and wasted graphics unit just taking up space.

I love this argument.  I agree the integrated graphics are just taking up space - but the rest is very interesting.  It's a point I debated a lot before I made the investment.  It came down to wanting the best bang for my buck, considering cost was an object, and the Canadian dollar is quite low, my system came down to roughly the same as if I bought it in USD.  

 

Although the 6700k's cost was around the same as the 5820k, the motherboard would have been 40% more expensive, and that with all considered the 6700k has more IPC than the 5820k, which is what counts more and more for modern software.  6 cores will help for heavy video editing (which I don't do) and extreme multitasking, which is above my needs.  I consider myself a power user, but I don't have nearly the needs of a guy like you, Rob (I do however drool at your system and the videos you post).  Besides, I want to run SLIx2 but that's about where it stops.  

 

In the end, I don't have enough free time to be on the computer to justify a system worth more than $2000.

[cavaricooper 2,615]

Thank-you all (and especially you Rob)... I have lots to mull over... and shall.

[martin-w 3,235]

4. Good water cooling solution for heat extraction.

 

 

 

The Noctua NH-D15/15S cools perfectly adequately for high overclocks. No need for pumps that will ultimately fail and potential water leaks. Yes, a fancy full blown loop is nice if you happen to be enthusiastic about such things, but it's not a requirement and not necessarily the right solution for all.

 

 

I don't believe in "silicon lottery" ... you can make any CPU work at high OC with some tuning and efficient heat extraction.  If you are concerned about OC I suggest you obtain Intel's replacement guarantee (no questions asked).

 

 

 

Well it kind of defies the laws of physics to say you "don't believe in the silicon lottery" to be honest, given that it's a very real phenomenon. And it depends on how you define "high overclocks". It's most definitely true that if the silicone lottery lands you with a very bad chip, that no amount of "efficient cooling" and "fine tuning" will magically render it anything different. Some very bad chips don't even respond to super high voltage and remain unstable. And if your dodgy chip does respond to super high voltage, then accelerated degradation ensues. High voltage degrades a CPU despite having the thermals in control. Current and amperage are highly significant in terms of a CPU's lifespan. 

 

CPU manufacturers dope the silicone in such a way that the required energy to for an electron to tunnel is huge, so under normal circumstances, as a result of this energy barrier, the electron does as it's told, stays where it's supposed to. However, as we increase voltage to render our bad overclocker stable, the electrons have more energy and therefore a higher probability that they will overcome the energy barrier and turn a switch that should be off, on! Ultimately, a path forms and the electrons pass through with ease, leading ultimately to CPU failure. 

 

 

 

Not a fan of Skylake, cache too small and wasted graphics unit just taking up space.

 

 

 

Yes I wish the integrated graphics wasn't there too. However... that's no reason not to be a fan of Skylake, same for your cache argument. We can hardly say the cache is "too small" when the forums are replete with individuals thoroughly impressed by Skylake's performance. The proof of the pudding is in the eating. 

[westman 257]

I prefer a unused IGP before a hole bunch of voltage regulators as in the Haswell chips ,

One of the reasons wy Haswell produce a lot of heat.

The silcon lotteria is real , Rob your 5960x is not a good one not mine , but i have test some good ones.

5ghz at 1.32v 8c/8t CB15 stable

Have one Haswell-E 5960x with asus MoBo ( V Extreme) but a prefer the Skylake up to SLI but if you run triple or Quad SLI with apps that support it ok the 5960x shine.

Its my personal opinion , the 5960x runs 4.7 3400mhz cl 14 mems and not close to the 5ghz 6700k with 4000mhz cl16 mems on a asus impact both with single 980ti sim P3D 3.1 not uppgraderar 3.2 yet settings after Robs recomendations same with both systems , thanks Rob.

[martin-w 3,235]

I prefer a unused IGP before a hole bunch of voltage regulators as in the Haswell chips ,

One of the reasons wy Haswell produce a lot of heat.

 

 

Good point, I hadn't considered that. 

[Guest]

The silcon lotteria is real , Rob your 5960x is not a good one not mine , but i have test some good ones.

5ghz at 1.32v 8c/8t CB15 stable

 

Need to know more than just Core voltage ... that's one very small element of the equation to successful stable OC.  I have operated at 4.8 and 5.0Ghz using lower BCLK (100 rather than 125) and adjusting RAM timings but that caused higher latency and less bandwidth, but more importantly I more frequent long frames at extreme graphics settings and add-ons at 4.8 and 5.0Ghz.  So I've elected to operate at around 4.6Ghz HT OFF.

 

But I'll maintain the lottery is NOT real.  What is real is the "guess work" that goes into OC ... it's far too "try this and see what happens" ... but when one has close to 100,000+ permutations of BIOS/EFI settings to work with, it's not surprising their is less "science" involved and more "try this".  I'm sure there are more permutations I could have tried to come up with higher BCLK and lower latency (at 4.8 or 5.0Ghz), but I just didn't find the right combination ... just a matter of how much time one wants to put into it and not so much a "lottery".

 

Cheers, Rob.

[westman 257]

if a CPU do 4.4 and another do 5.0ghz on the same MoBo and bios settings only swap CPUs .

For example Caseking have binned 6700k on 1000 tested is 50 good 2-3 really good.

Asus RD binned tons of CPUs intel binned CPUs for computex and even CPUs for rewiews Its common in the industri.

Its no lotteria when you bin CPUs but if you buy one Its lotteria.

[Guest]

I've taken the very same CPU out and then back into the same Motherboard and was not able to OC it has high ... there are just too many other variables to come to the conclusion it's a CPU lottery.  

 

1.  socket contamination (micro-spec of dust between a contact and/or other contaminants) on each insert

2.  cooling block tension 

3.  application of thermal paste

4.  socket contact depressions (from prior insertions)

5.  and the list goes on and on...

 

The motherboard variance is likely to be the HIGHEST factor in variance in OC's over the 100+ builds I've done over the decades.

 

I would also NOT trust anything Intel portrait to the public ... Intel have no desktop competition, so they have no incentive to provide CPUs with higher clocks but less profit.

 

Cheers, Rob

[westman 257]

Ok ,your ar the pro

Elmor and the guys at Asus RD is ?????

[martin-w 3,235]

But I'll maintain the lottery is NOT real.

 

 

With respect, this is one of the strangest statements I've seen on this forum. The variance in CPU quality and capability, isn't something the PC enthusiast community invented... it's actually fundamental to the way CPU's are manufactured!

 

http://www.cpushack.com/MakingWafers.html

 

Here's the important bit, and I quote...

 

"Once the wafer full of chips is made each chip is tested while still on the wafer. If a bad one is found it is marked so that it is not used. Most bad chips tend to be around the edge of the wafer. The best chips are in the center and are sometimes selected for extended temperature testing for military or industrial use."

 

If the military accept that the silicone lottery is real... why cant you? You don't find the military buying any old crap CPU and just tweaking it in the BIOS.

 

 

As you can see, there is a considerable variance in CPU quality defendant in part on where on the wafer the CPU is located. Manufacturers don't scrap all but the best "chip" on the wafer, they use the majority, thus some are more capable than others, and some serve as lesser variants.

 

Thus, to claim the silicone lottery isn't real is contrary to fact!

 

If you are saying that the variance in CPU quality on the wafer is real, but that you, as an individual, have the talent to fully compensate for a fundamental deficiency as a result of the way CPU's are manufactured, with the advanced knowledge of  "BIOS tweaking" you possess... then you are way more talented than the experts at Intel, AMD, Asus, EVGA etc. And moreover, posses greater knowledge than they do. 

 

 

but when one has close to 100,000+ permutations of BIOS/EFI settings to work with

 

 

I'm not quite sure how you got such a high number, but the overwhelming majority of those permutations do almost nothing to enhance an overclock to any appreciable degree. 

 

 

there are just too many other variables to come to the conclusion it's a CPU lottery.  

 

 

 

it's nothing to do with those other variables when we know how a CPU is manufactured, and know the variance in quality, dependent of where on the wafer the CPU was located, is a fact, a fact fundamental the the CPU's manufacture. 

 

 

 

1.  socket contamination (micro-spec of dust between a contact and/or other contaminants) on each insert

2.  cooling block tension 

3.  application of thermal paste

4.  socket contact depressions (from prior insertions)

 

 

 

You are vastly overstating your case to back up your argument.

 

Cooling block tension doesn't make a huge difference in temp, nowhere near enough to thermally compromise an overclock to an appreciable degree, the temp difference is small as long as a person isn't idiotic enough to vastly under tighten. The vast majority of enthusiasts know how to apply TIM. And minor differences in TIM application make a very minor difference to CPU temperature.

 

 

The motherboard variance is likely to be the HIGHEST factor in variance in OC's over the 100+ builds I've done over the decades.

 

 

 

The CPU variance is likely to be the HIGHEST factor in variance in OC's over the 100+ builds I've done over the decades. And in the experience of the technical experts at Intel, Asus EVGA etc. Not to mention thousands of PC enthusiasts out there.

[westman 257]

I stop posting on Avsim , completly aggree with you Martin .

Its no hardware forum with a moderator that no nothing about Overclocking , tuning etc.

Give the comunity false info, Its time for the Staff to do someting..

 

Thanks for the years here at Avsim

 

Hasse

[Ted Striker 338]

No need for anyone to go home. Can't we just agree to respectfully disagree and appreciate what each persons opinions brings to the forums?

 

Ted

[Guest]

 

 

Cooling block tension doesn't make a huge difference in temp,

 

I didn't actually say it did?  Uniformity of tension does ... hence don't vertical mount a CPU and your cooler.

 

Your link covers manufacturing process ... was there some other point to it?  It's does go over cause of rejected CPUs.

 

But think about this ... what is random?  A lottery indicates some type "randomness" ... none of you have explain what that "randomness" is?  Why would increasing voltage in various circuits make a difference?  It's all about contact at a molecular level ... that's not a "lottery" ... Intel's manufacturing process may make it appear to be a lottery ... I don't believe in "randomness", only flaws in manufacturing process of which there are solutions.

 

So lower the definition of acceptable manufacturing?  That's the lottery?

 

 

 

Give the comunity false info, Its time for the Staff to do someting..

 

You are more than welcome to contact AVSIM staff if you have a problem with my post and think it violates ToS.  But in most forums/moderated or not people will disagree.  You clearly disagree - which is fine, but what's the emotional attachment to that?  You really haven't provided any technical discussion on what you believe the "lottery" is about?

 

Cheers, Rob.

[martin-w 3,235]

Intel's manufacturing process may make it appear to be a lottery ... I don't believe in "randomness", only flaws in manufacturing process of which there are solutions.

 

 

 

And those "flaws in the manufacturing process" result in it being a lottery for you and I when we attempt to overclock our CPU's!!! 

 

With only 56 atoms separating transistors in some cases, of course there will be variability in chip quality. We are dealing with manufacture at the smallest scale and cutting edge technology, attempting to squeeze more and more out of our silicone devices. The manufacturing process can't be totally uniform, perfect in every way, that would be impossible given the complexity of the process. We are dealing with structures so small that just a single atom out of place can have undesirable consequences. 

 

And NO, there are no magical solutions that Intel could employ. Do you really think Intel would be so silly, so stupid that they would compromise their own profitability by neglecting to manufacture chips to the best of their ability? Do you think they are throiwing away chips from the edge of the wafer for fun?

 

Perhaps you should get in touch with Intel and tell them of your ingenious "solution" to the variability issue with CPU manufacture. I bet they would pay you millions for your advanced knowledge.

I didn't actually say it did?  Uniformity of tension does ... hence don't vertical mount a CPU and your cooler.

 

 

 

Uniformity of tension doesn't make a big enough difference to thermally compromise an overclock to an appreciable degree. Variations in uniformity of tension make a minor difference. Unless someone is dumb enough to cock it up significantly.

 

"Don't vertical mount a CPU and your cooler"... what on Earth are you saying? Almost all of us do, without issue. If the back plate is properly designed, and the static load sufficient, no issue.

 

 

Your link covers manufacturing process ... was there some other point to it? 

 

 

Well obviously it pointed out to you that variability in CPU quality is inherent to the manufacturing process.

 

 

But think about this ... what is random?  A lottery indicates some type "randomness" ... none of you have explain what that "randomness" is? 

 

 

 

The randomness is clearly structural variation in the CPU as a result of the complexity of the manufacturing process, and the impossibility of creating a wafer of utter purity across it's entire surface area. The randomness is as a result of attempting to create a device at the ultra small level, so small that a mere 56 atoms separates components, so small that you are in danger of experiencing quantum tunnelling where you don't want quantum tunnelling.

 

In fact, it's quite common that when a new CPU architecture appears, that there is greater variability, a greater number of rejected and lower quality chips on the wafer. As the architecture matures, and the manufacturing process improves, fewer sub standard chips result and less variability.

 

To expect Intel, AMD or any other CPU manufacturing company, to provide you with perfection, an utterly uniform wafer, is beyond unreasonable.

 

 

This stuff is fact! Not open to debate! And the nice thing about fact, is that it remains a fact regardless of any bizarre notions you or I might have.