P3D upgrade CPU from 4930k 4.4 OC'd hex core to 6700k 4.2 turbo or wait for 7700k kaby lake?

[theskyisthelimit 48]

just go full custom waterloop. much more fun and works great

 

Thats tantilizing, but i figured way more costly and riskier in terms of leaks, i'm sure that may not be true though.. havent researched it much.. and i'd want to cool the 1080 ftw + cpu in such a setup i guess. aio seems so much easier.

 

I"m leaning towards just going with the $179 riing version of the 360 thermaltake to replace the h100i corsair v2.. however, i may also test out the h115 240mm while i'm at it

 

My cpu is leaning towards needing 1.375 volts (fixed) with L6 LLC + 110% + VRM, leading to a load of a round 1.376 at 4.6ghz..

 

some foot notes too..

 

4.2 ghz, i was achieving 40 fps paused on fsxmark07 (ftx, gex, utx etc) this was at 2133 ram speed though.

 

Somewhere along the lines i accidentally hit 2900 or so on ram speed (it will be 3200 before done)..

 

at 4.4 ghz, i saw 44 fps (2900 ram)

 

and at 4.6 ghz, i saw 47 fps (2900 ram).. need to scale the ram back to 2133, i expect those to drop a frame or so..

 

at any rate, you can see the scaling as you go up.. 40 (4.2), 44 (4.4), 47 (4.6).. i havent tested 4.7 just yet, as i have to wait on the better cooler for this chip, if in fact it needs 1.375 for 4.6, it will possibly need 1.4 for 4.7 (i had tried around 1.38 and it spontaneously rebooted on realbench).

 

Or i need to try a different chip

[james_2k 36]

Its expensive at first, but when you have the radiators pump etc you just need small tweaks whenever you change hardware. its daunting at first but its great!

[theskyisthelimit 48]

at any rate, you can see the scaling as you go up.. 40 (4.2), 44 (4.4), 47 (4.6).. i havent tested 4.7 just yet, as i have to wait on the better cooler for this chip, if in fact it needs 1.375 for 4.6, it will possibly need 1.4 for 4.7 (i had tried around 1.38 and it spontaneously rebooted on realbench).

 

 

Updating these stats:

 

 

Numbers shifted slightly, not sure what changed, but here they are all in one shot, ie: before I thought i saw 40 fps at 2133, but this was on the corsair vengeance ram, i've since switched to the cas15 gskill.. so maybe they faired slightly worse at 2133 for some reason even though not cas16?)  Again these are with the 1080 FTW Evga.  So that said, the change from 4930k at 2133 on the other side for DDR4, was maybe 2fps better on Skylake at 4.2 (or as high as 5), given the old test value i had on 4930k at 2400mhz ddr3 was 35fps and the new, now appears to be 37 not 40.  But watch as the numbers jump as you increase DDR4 speeds ..

 

 

DDR4 Ram at 2133 on 6700k (AM=85):

-37 fps 4.2ghz   (old 4930k hex at 4.2 was 35 with AM=1365)

-39 fps 4.4ghz

-41 fps 4.6ghz 

 

DDR4 Ram at 2933 on 6700k:

-43 fps 4.2 ghz (an INCREASE of 6 fps by going faster ram)

-45 fps 4.4ghz 

-47 fps 4.6ghz

 

**i did try 3200 (and left it at 3200), but there was NO change going from 2933 to 3200.. 

**PerformanceTest Passmark Memory Mark did NOT change when going from Cas15 to cas16:

**PerformanceTest Memory Mark was 2998 for 2133mhz Cas16 Corsair Vengeance LPX

**PerformanceTest Memory Mark was 2998 for 2133mhz Cas15 Gskill 

**PerformanceTest Memory Mark was 3667 for 3200mhz Gskill Cas15 

 

 

Back to the cpu temp/voltage issues:

Last nite i decided to remove the h100i use alcohol to clean and reapply using arctic silver 5.. either the job i did wasnt good, the arctic silver was too old or i put too much (pea size pod in middle, maybe a little bigger than a pea this time), but now running IBT all things held equal, the temps were almost immediately (30sec) at 99c, whereas before after 30sec they hit 84C..

 

I have the thermaltake 3.0 360 regular version coming, but i also have a different chip coming.. i'm going to see how the volts and thermal load are on the new chip and also try the 360mm, same mode, exhaust mode i guess too.

 

Once i determine which chip is the winner, i am going to delid it and hopefully get the 10-18c drop i saw when i did my last delid.

[theskyisthelimit 48]

 

Updating these stats:

 

 

Numbers shifted slightly, not sure what changed, but here they are all in one shot, ie: before I thought i saw 40 fps at 2133, but this was on the corsair vengeance ram, i've since switched to the cas15 gskill.. so maybe they faired slightly worse at 2133 for some reason even though not cas16?)  Again these are with the 1080 FTW Evga.  So that said, the change from 4930k at 2133 on the other side for DDR4, was maybe 2fps better on Skylake at 4.2 (or as high as 5), given the old test value i had on 4930k at 2400mhz ddr3 was 35fps and the new, now appears to be 37 not 40.  But watch as the numbers jump as you increase DDR4 speeds ..

 

 

DDR4 Ram at 2133 on 6700k (AM=85):

-37 fps 4.2ghz   (old 4930k hex at 4.2 was 35 with AM=1365)

-39 fps 4.4ghz

-41 fps 4.6ghz 

 

DDR4 Ram at 2933 on 6700k:

-43 fps 4.2 ghz (an INCREASE of 6 fps by going faster ram)

-45 fps 4.4ghz 

-47 fps 4.6ghz

 

**i did try 3200 (and left it at 3200), but there was NO change going from 2933 to 3200.. 

**PerformanceTest Passmark Memory Mark did NOT change when going from Cas15 to cas16:

**PerformanceTest Memory Mark was 2998 for 2133mhz Cas16 Corsair Vengeance LPX

**PerformanceTest Memory Mark was 2998 for 2133mhz Cas15 Gskill 

**PerformanceTest Memory Mark was 3667 for 3200mhz Gskill Cas15 

 

 

Back to the cpu temp/voltage issues:

Last nite i decided to remove the h100i use alcohol to clean and reapply using arctic silver 5.. either the job i did wasnt good, the arctic silver was too old or i put too much (pea size pod in middle, maybe a little bigger than a pea this time), but now running IBT all things held equal, the temps were almost immediately (30sec) at 99c, whereas before after 30sec they hit 84C..

 

I have the thermaltake 3.0 360 regular version coming, but i also have a different chip coming.. i'm going to see how the volts and thermal load are on the new chip and also try the 360mm, same mode, exhaust mode i guess too.

 

Once i determine which chip is the winner, i am going to delid it and hopefully get the 10-18c drop i saw when i did my last delid.

 

 

Update here, i went with a siliconlottery.com 4.7 ghz rated at 1.40 or better (realbench 1 hour).. turns out it needed more juice than that for p95 blend 12 hour stability, but i'm ok with that (bios 1.402-0.008 LLC6) with vcore 1.424-1.44 under load.. 86C p95 blend 10 hours in at least for the max.  I bought the version delided by them..

 

Super low temps compared to what i was experiencing before.. roughly 15-18C on average at the same given voltages.  I also settled on the Thermal take 360mm ultimate riing version 3.0 (after comparing directly with the h100i, the h115i and regular thermaltake 3.0 ultimate).. it was the quietest and coolest of the bunch 

 

At any rate.. 

 

Results:

 

-49 fps at 4.7ghz

[james_2k 36]

Good results!

[theskyisthelimit 48]

Good results!

Thanks..

 

just an update.. i decided to try 3333 for ram speed.. there was no change at that level.. so around 2900 + seems to be the key.

I may toy with overclocking the cache frequency to see if it does anything, so far i'm still on the default for that which is around 21x.

 

My full validated results with my siliconlottery.com 4.7 cpu are here  passed 24 hours prime95 blend with 10 minute intervals, max temp 87c.  I couldnt push it any further, i figure to hit 4.8 i need about 1.48 and thats too close for comfort.. ill stick with 1.4 with the adaptive offset and 1.44 max under load.

 

In real world testing now though.. prepar3d is making the cpu max about 68C.. 3dmark test about 60c.. gpu.. 43c around 97% in one test for p3d and 50c for the 3dmark test.. this is on the new watercooled 1080 hybrid ftw gpu which is fabulous

[westman 257]

Only 2900mhz? You have 3200gskill ? Do you enable XMP? They shall do easy 3400+ C15

All 6700k i tested have done +3500mhz

[GSalden 4,657]

Vcore 1.44 ... that is a lot .

How many months you want the processor to last ?

Intel says that 1.45 is the safe max and you are just below that.

 

If it was my cpu I would not go beyond 1.42 ..

[theskyisthelimit 48]

Only 2900mhz? You have 3200gskill ? Do you enable XMP? They shall do easy 3400+ C15

All 6700k i tested have done +3500mhz

 

I'm saying from 2900mhz and up is where fps starts to jump.. i have mine at 3200.. there was no difference in frames from 2900 all the way to 3400.

 

Actually from everything i've read 1.5 is the max limit (though yeah might be closer to 1.45) .. and with temps under 85c.. I think with haswell it was 1.35, but its higher with skylake, especially if water cooled and not stress testing all day every day.

 

being as my real world p3d temps dont exceed 68c and volts under load is 1.424.. i feel confident i'm more than fine.

 

I did realize though.. i never turned IGP off in the bios, which might help bring temps/volts down.. may retest at some point and see.

Only 2900mhz? You have 3200gskill ? Do you enable XMP? They shall do easy 3400+ C15

All 6700k i tested have done +3500mhz

curious, once you start going over the stock highest rating of 3200 for instance.. how far do you push the ram/vccio/sys agent volts? 

 

Right now i'm at 1.408 volts on ram (1.35 stock) and vccio/sys agent at 1.20 (i had them at 1.25, 1.20 worked fine for 3200)

I might retest and see if i can see a difference at 3500 or 3600 even (for curiosity sake)

[westman 257]

At 3600 c14 1.37v on ram 1.2/1.2v.

CPU multi48 /47 uncore 1.37v adaptive

For me the mems scale linear from 2133 to 4000 and scale on cl, strange that it stop at 3000mhz

[theskyisthelimit 48]

At 3600 c14 1.37v on ram 1.2/1.2v.

CPU multi48 /47 uncore 1.37v adaptive

For me the mems scale linear from 2133 to 4000 and scale on cl, strange that it stop at 3000mhz

Thanks, ill test again..

 

I need to boost my uncore cache up from 41, with my 4.7 overclock i havent tried.. did that require more vcore?  if so i may be out of luck.. i dont want to push under load past the very very max of 1.44 that i've already hit with p95.

 

In my tests it was just the paused screen for fsxmark07, noting the initial paused fps and rebooting, changing the mhz on the ram.. not hugely scientific, i think more so would be to run the full 120sec benchmark note the avg fps and switch.. i may give that a go for fun

[westman 257]

Normaly no need for more vcore , tumb of rule for Skylake minus 1-2 multi for uncore/cashe

[theskyisthelimit 48]

Normaly no need for more vcore , tumb of rule for Skylake minus 1-2 multi for uncore/cashe

 

with core at 4.7 and uncore at 4.5 (up from 4.1), at least with a quick paused test, fps didnt change any.. will do the full 120sec tests later to confirm.  (oh and it was unstable with the same vcore.. a cpu test caused a bsod)

[swiesma 625]

As we talk about RAM... Thinkijng about these:

https://www.alternate.de/G-Skill/DIMM-32-GB-DDR4-3200-Kit-Arbeitsspeicher/html/product/1283710?

or

https://www.alternate.de/G-Skill/DIMM-32-GB-DDR4-3-200-Kit-Arbeitsspeicher/html/product/1260883?

Any thoughts?

 

At the moment I use:

https://www.alternate.de/Ballistix/DIMM-16GB-DDR4-2400-Kit-Arbeitsspeicher/html/product/1184846

[westman 257]

IF you need 32gb it dosent matter take the one you think look better, same chip i both.