Waisted some cash 6700k\gtx1070

[gamble 76]

Some results that I can aim to beat... until you get that new cpu going ;)

 

Probably won't beat the graphics score but if I run it with the cpu to max I might be able to give you a decent challenge

[Dasa 288]

you have about a month till i get back to get that gpu on water :)

 

unfortunately the day i spent testing the 3770k is probably wasted since i will be away for so long all the games will probably receive updates in that time potentially affect performance

 

oh and in 3dm11

first two tests were ~2000-2074 boost next two were ~2050-2088 combined ran 2114 with +125 boost

it scored better without evga software open to keep an eye on clocks while it ran so i ran it later to see what it was doing

Edited August 6, 2016 by Dasa

[Dasa 288]

it seems the vrm on this card may get a little toasty at least with low fan speeds

http://www.guru3d.com/articles_pages/evga_geforce_gtx_1070_sc_superclocked_gaming_review,10.html

 

the r33 chip at the top there and the first two black chips to the right of it are apparently the memory vrm

then each of the gpu r22 chips has one vrm to its right for themselves

when i put it on water i think i will glue some hs to the heatplate and maybe put some thermal tape between the pcb and the back plate

https://www.techpowerup.com/reviews/EVGA/GTX_1070_SC/4.html

Edited August 7, 2016 by Dasa

[gamble 76]

Could you plz screenshot AB with Heaven few loops maybe 3 or 4. Just want to see how your core responds when OC around 2000mhz. Mine is kinda all over the place but within a range though, definitely not flat. Then if you do water would be curious if it flattens more, that is if yours jumps around like mine.

[gamble 76]

Here is BF4 for a good long run. My core is usually around 1974 to 2012 with +150. It will boost to 2050 maybe once in a great while, if Im guessing perfect conditions. I ran the fan curve up to 80% to get keep under 70c.

 

If you notice that the core drops are nearly perfectly in tune with the voltage.

 

https://drive.google.com/open?id=0BwBp8FSaZ50sb3JJVjRwSWVWNm8

[Dasa 288]

Could you plz screenshot AB with Heaven few loops maybe 3 or 4. Just want to see how your core responds when OC around 2000mhz. Mine is kinda all over the place but within a range though, definitely not flat. Then if you do water would be curious if it flattens more, that is if yours jumps around like mine.

ok im back

stock clocks drop ~100mhz once the card warms to 76c

oc clocks drop ~25mhz with fans at 100% once the card warms up

Edited September 3, 2016 by Dasa

[gamble 76]

Thanks for those screenshots..... I can't beat your scores even before you move to the new cpu/mb :(

[Dasa 288]

can i see your best score?

 

hmm those afterburner screenshots are crap quality

when i say the amount the fps drop as it heats up thats at the same point in the test next time round

there is far more variance than that across the test

lets try this

stock

oc

Edited September 7, 2016 by Dasa

[Dasa 288]

Ok rebuilt the system today its all on water 6700k is in xmp 3200c14 is working no problem

 

the gpu on water is staying under 31c in heaven and at stock its running 1962-2000mhz

total system power use peaked at 229w with a 1050w psu :\

need to grab a camera for a better picture

swapped to the ek gpu block to lower temps but unlike the old swiftech block it cares about flow direction so it doesnt look as nice with the inlet is on the wrong side

i had a bunch of old copper heatsinks all lapped up ready to go onto the gpu heatplate over the vrm only to find one of my old tubes of arctic adhesive has dried up

 

Edit:

spent some time overclocking seems fairly stable at 4.7ghz core & cache 1.376v typical load sometimes spikes close to 1.4v

with the ram at 1.4v (vccsa & vccio 1.2v) 3466 14-14 t1 tRFC270

Edited September 9, 2016 by Dasa

[gamble 76]

System looking great there ^

 

Sure, here is a few results. I have a few newer but they are lower scores or near these scores:

 

http://www.3dmark.com/3dm11/11457517

http://www.3dmark.com/fs/9507883

http://www.3dmark.com/spy/152149

 

Ill be rerunning with cpu @ 5ghz for timespy and FS. But, mostly was just checking the graphics score at the time I was running. Look forward to see what you get out of the new cpu/mb :)

[Dasa 288]

been able to push the gpu to +150 in firestrike now its on water

these are the scores so far

22121

17764

 

squeezed a little more out of the ram this morning got it up to 4600 with 1.45v barely posts with that but it passed 5min of realbench

couldnt do 3466 c13 at 1.45v

tried 4.9ghz 1.45v but no luck so i may stop pushing for now

 

system is now fanless at idle using ~44w just a slight hum from the pump at 1200rpm

Edited September 9, 2016 by Dasa

[Dasa 288]

gave arma3\fallout4 a quick run at the stable oc core\cache 4.7\4.7ghz 3466c14 ram

there running ~45-50% faster than the 3770k at its max stable oc

arma3 82fps avg

fallout4 74fps avg

 

rainbow6 193.5fps avg ~0.4% faster

3dmark said the gpu score was up 0.4% vs same clocks on the 3770k probably due to slightly more stable boost clocks with the gtx1070 on water

 

 

edit

dropping the ram to spd 2133 has taken ~20fps of the arma3\fallout4 tests above

xmp3200c14 is somewhere in the middle

 

3466 proved to be unstable

but 4000c17 t2 seems stable and performs about the same in game

3866c16 t1 also performs much the same although this needs fast boot turned off otherwise it fails to post on reboot so 4000mhz may be the 24\7 speed unless another hidden gremlin pops up

 

t1 wont post without xmp enabled

4000 wont post without xmp disabled

Edited September 12, 2016 by Dasa

[Dasa 288]

3770k --- 6700k
Asus Maximus V Gene Z77--- Asus Maximus VIII Gene Z170
Samsung oem DDR3-1600c11 8GB 2x4GB --- G.Skill DDR4-3200c14 16GB 2x8GB
EVGA GTX1070 SC
64g Samsung 830, 1tb Sandisk ultra II, 3tb WD green
Seasonic XP 1050 <tad overkill for the current system which has topped out at 306w from the wall
Silverstone TJ08-EW
Swiftech Apogee drive II
200mm Phobya rad
EK VGA HF

 

 

ok here we have the first graph with 3770k vs 6700k and memory scaling

maybe i should change the graph order from min fps to cpu clock speed then memory speed?

also skylake allows clocking the cache separately as it doesnt get far on some chips so its 4.7 core/4.7 or 4.2 cache

edit: ok and here is arma

edit

here is r6 your typical gpu bottlnecked game used in most ram reviews as mentioned earlier the ~0.4% increase vs 3770k is probably just due to more stable boost clocks now that the gtx1070 is on water

 

also have intel xtu and aida64 tests at each speed for the 6700k

took a sharper image of the rig and just realized some of the above cropped screenshots may not fit 1080 displays all that well?

Edited September 13, 2016 by Dasa

[gamble 76]

wow nice job with all those graphs. Im a bit surprised by how the difference with 4.7 4000c17 t2 is compared to 4.7 2133c15 in Fallout... Also, I would have guessed that you would have had better results with 3866c t1 than 4000c17 t2. Looks like 4000c17 t2 is giving great results so I would stick with it.

 

As for the 3770 vs 6700, looks like a larger upgrade than I would have thought too. Sure it depends on the game, I knew it was going to be a bit higher but those results are encouraging. Could you do one run with 4.5\4.5 4000c17 t2 in Fallout? Just to get an apples to apples compare. Also, was that run with FS @ 4.7, looks like 4.8?

[Dasa 288]

yeah top firestrike test was at 4.8ghz i kept to stable clocks for the game tests except for 3000c12 which wasn't stable

for some reason i cant get the system to post again at 4000mhz ram there must be a sub timing it auto configured correctly last time that isn't quiet right this time or something but since it passed all stability tests and i saved its profile im a bit baffled

 

the increase from the higher ram speeds doesnt really surprise me its ~20% which isnt much more than i have seen with slow vs fast ddr3

but i am a bit surprised at how little difference overclocking the cpu makes and how much difference single vs dual channel made

and like you i would have thought that 3866c16 t1 would have the edge over 4000c17 t2 its very close though other than the fallout 4 average which could be a bit out im afraid

 

there is a little more variance in fallout4 results than i would like

with such a short run low fps at the start and high at the end i only need to start or stop the fps a second out and it can throw the average by a few points

so focus more on its min fps as they are fairly consistent unlike arma where i left them out

 

running at the 6700k 4.5GHz 4000c17 may drop off ~1-2fps vs 4.7GHz

Edited September 13, 2016 by Dasa

[gamble 76]

maybe toggle the xmp on and off and back to your profile? I have to do a cmos reset it seem now if I run 5ghz... even if I load the 4.7ghz profile it will keep the 5ghz profile. I have to reset and then load the 4.7ghz profile then its fine... Just a guess since you have probably been playing with the bios alot :)

 

These are the best I can do with 5ghz running:

 

cpu @ 5ghz - gpu core is all over the place depending on the part of a bench... But the core is around 1974 - 2025ghz with a few higher and lower... Memory is at 4750 and starts to artifact around ~4800.

16251 - http://www.3dmark.com/fs/10145773

6165 - http://www.3dmark.com/spy/436959

 

Oh and this is interesting:

Edited September 13, 2016 by gamble

[Dasa 288]

lol 1 point short in timespy so dam close! i bet you could tweak some more out of your ram sub timings

timespy sees artifacts much sooner than firestrike for me i had to back off the gpu\mem clocks a lot for it

firestrike can run +825 with maybe the odd minor artifact but i run +750 when gaming

1974 - 2025 sounds like a fairly typical range for the most part sounds like your gpu really wants a waterblock though ;)

 

got 4000 running again after loading the profile this time i noticed that the ram was reporting 2400 and that was the max speed available till if poked at it a few times in the uefi

its not perfectly stable at the moment though need to tweak it again

[gamble 76]

I know right!? 1pt and ran 3 times and restarted once, would have to come from timings or slight blck bump... I think your time spy was 4.5ghz so you'll see more probably with 4.7ghz?

[Jeruselem 788]

Say, can you try the new geekbench 4 on that rig?

[Dasa 288]

here you go at 4.7ghz 4000

sc 6337 mc 21305

https://browser.geekbench.com/v4/cpu/306559

 

just hit 6000 write speed 5200 copy in aida64 then still over 37ms at 4000 though

 

since going on water at stock clocks the gtx1070 stays at 1999mhz in heaven

increasing gpu memory speed also reduces core clocks when overclocking due to increased power use but its worth it as the performance still goes up

Edited September 13, 2016 by Dasa

[Dasa 288]

just did the power mod gamble linked above with liquid ultra

must say i was a bit worried since the last time i fiddled with a card i killed my 7970 by having the diy vrm slide due to bad glue and short the vrm

thankfully it posted ok but then failed to boot sure made my hear skip a beat but it was just the mb had decided to give me a scare by changing the boot order

 

it seems to have worked! heaven was running a solid 1139 with 140+boost at 1050mv instead of dropping back to ~1000mv due to power limits

reported power use does go over 100% so it may still restrict clocks a little under some circumstances

 

firs run of firestrike at stock clocks reported some inconstancy error but timespy was fine

need to do further testing

 

also need to get some new arctic adhesive to stick those hs on over the vrm plate so they dont cook with the extra v

[gamble 76]

just did the power mod gamble linked above with liquid ultra

must say i was a bit worried since the last time i fiddled with a card i killed my 7970 by having the diy vrm slide due to bad glue and short the vrm

thankfully it posted ok but then failed to boot sure made my hear skip a beat but it was just the mb had decided to give me a scare by changing the boot order

 

it seems to have worked! heaven was running a solid 1139 with 140+boost at 1050mv instead of dropping back to ~1000mv due to power limits

reported power use does go over 100% so it may still restrict clocks a little under some circumstances

 

firs run of firestrike at stock clocks reported some inconstancy error but timespy was fine

need to do further testing

 

also need to get some new arctic adhesive to stick those hs on over the vrm plate so they dont cook with the extra v

 

 

That was pretty ballsy mod there ;) I need to get under water so I can do this mod... nice job.

[Dasa 288]

final stable speeds for now...

Dasa--i76700K @4.7/4.6 1.376-1.4v---3866Mhz-C16-16-16-34-2T----1.425v---SA 1.25v---IO 1.2v

idle power use 40-50w peak 315w

Edited October 13, 2016 by Dasa

[gamble 76]

So, I have been looking over your charts... Age old mhz vs latency, but with ddr4... It looks like mhz wins slightly in most? Holding out for ryzen before I pick next build but thought about buying ram before and a few other pieces before then.

[Dasa 288]

provided you can keep the final latency close to the same with mhz vs timings then higher freq gains the edge through increased bandwidth

some software cares more about bandwidth some more about latency so depending on what your running it can swing more one way than the other

games typically seem to lean towards final latency but some do like a bit of bandwidth as well

Edited February 9, 2017 by Dasa