*Official* NBR Desktop Overclocker's Lounge [laptop owners welcome, too]

I'm not sure yet what fans I'll go... maybe nf-12 from noctua.

 

Shunt mod is looking promising. I did it slightly different this time. Will share more after I am done testing and confirm it, but it looks like there is a chance that cancer firmware is/was the problem... and if so, even the K|NGP|N firmware is not enough magic to fix NVIDIA's bullfeces.

 

Really looking forward to your results, was disappointed reading your initial results this morning.

How high can you crank the Samsung memory? I figure the cold temps should let it fly.

 

I haven't tried higher than 1400. I have been devoting all of my attention to maximizing the core overclock because that is more important in benching (biggest impact on scores). But, it passed 1400 without artifacts.

The temps are not the problem. Power and Voltage limits kicking in and crashing above 2175 on core were the issue. But, it's looking about 90% certain at this point that the shunt mod fixed that. Still testing to confirm. Seems like I am still having a power limit flag, so I may need to change the ohms resistance a little bit. I experimented using a .010 ohm shunt instead of 8 and that might not be lying enough to kill the power limit. .008 ohms was perfect before and .005 ohms was too much. Using the .010 ohm resistor has dramatically improved it and now it doesn't crash at 2200 on core. It just doesn't stay there due to the power limit flag. Also going to try older drivers. The latest driver is installed and it may be feces.

 

Don't sell yourself short for using soft tubing. It's far more practical and you're going to save yourself a lot of time while losing zero performance. Before finishing up your order, check if your motherboard has the older 12V RGB header (3 pins for separate R, G, and B channels + ground). Despite the fresh release of EK-Vector waterblocks, EK stuck to their old 12V LED strips, so if your motherboard doesn't have the appropriate header, you're going to want add the newer, addressable 5V RGB LED strip to your order.

What prompted you to RMA your fans? Coil whine? I have 3x EVO 120ER fans and I think their spin up is a bit loud. Not sure if that's normal behaviour or not, so I might test the fans one by one soon. The front Noiseblocker fans are a lot smoother and quieter. I would be interested in trying their eLoop fans in the future, but hopefully they've released a revised version by then.

Were your DDC pump 3.25 models (sold as EK-DDC 3.2 PWM)? Hope I don't have a ticking bomb and I thought they were all Laing pumps underneath. I built mine into a non-EK heatsink/top/reservoir combo though. My rads and fittings are Hardware Labs and Bitspower respectively.

 

Shunt mod definitely moved the needle the right direction. This is with the KPXOC vBIOS. Not my best score, but didn't crash at 2200 core clock like it was doing before the shunt mod. Seems like the voltage isn't holding under load, so I might be better off using the Aorus Extreme vBIOS. At least I can lock that at 1.093V and the KPXOC vBIOS drops below that under extreme load. Really a shame we cannot use the voltage curve tool to lock the voltage at like 1.125V with the KPXOC vBIOS. It would be so much better if we could.



@Rage Set @Johnksss

 

@Talon - 1500 on memory passed... showing just shy of 900W coming from the wall. I think I am going to flash the Aorus Extreme vBIOS after testing Fire Strike with the KPXOC.

 

Ah yes it sure is haha. It's so much wider than a reference design that I almost went with a vertical mount.

Nice updates.

For sure.

Well said.

Nice check list. I like to use primochill tubing (stays very clear) and ek cyrofuel (clear)... For the fellow starting out, I'd avoid colors for now and just stick with clear cyrofuel concentrate mixed with 1 gallon (3.78L) of water.

Yea even on my card the core isn't as strong as the memory. 2175mhz so far with 2190mhz peaks... Once I put it on the chiller it may have a little more ceiling to it... We'll see.

As for the .010 ohm's very interesting... Now I'm really curious.

That's true. There's nothing wrong with soft tuning and one of the major benefits of it is the ability to use quick disconnects. Hard tubing is nice, but imo mostly for show with a lot more drawbacks in many ways... Especially for enthusiasts.

Very nice and good to know. 1.125v was the max so far on my end as well... And yea it stinks that we can't make adjustments on the xoc vbios...

I've been having to compensate with the strong memory for the weaker core.... Totally understand.

 

New personal best and a close second to my previous personal best.

https://hwbot.org/submission/4152801_

https://www.3dmark.com/fs/19306613

https://www.3dmark.com/compare/fs/19306613/fs/19306690

 

The fans simply stopped working (EK-Vardar mostly and a couple EK-Vardar EVO's). The DDC pumps were the EK-XTOP DDC 3.2 PWM Elite. Now I did purchase the dual DDC pump for my X299 build and that's only because they gave me a good discount.

Their EVO fans are supposed to be quieter but like any fan, high RPM's are going to make noise. I won't buy anymore EK fans though, strictly Noctua fans for me. I still have several that are two to three years old that run as if they are new.

 

That's what I want to see brother!!! Give me something to aim for.

 

Looking real nice! What's the highest core with the KP again on the chiller? 2230mhz was it? Not far off at all...

Each of us all tackling different cpu class... Love the diversity. The gpu scores and the capability of the FTW3 card is impressive.

 

No, I can't get it to 2235 with the .010 ohm shunts. It might get there with .080 ohm. I might change the resistors late tonight before going to bed.

 

It'd be interesting to see, however, the fact that the FTW3 card is hitting 2220/2125 already is very impressive. Which reminds me, let me go check the mail to see if the little unleashers arrived.

 

Well fellas, hope everyone is enjoying their Saturday. I just finished up with some yard work and now about to swap out the PSU (1000w for the 1600w), drain the loop and install the KP card. As noted previously I'm going to run the KP card in the same conditions as the FTW3 to compare as accurately as possible. Taking the FTW3 card out will also give me a chance to shunt mod it so the timing is perfect.

@Mr. Fox how's it coming along on your end? @Rage Set are you able to get any benching in this weekend?

 

Here are some photos...



This was the experiment with the 0.010 ohm resistors. I actually like how they installed better. I had these in spare parts already, but now I am tempted to order more like this in 0.080 ohm instead of installing those that I already have simply because of how much easier they are to tack on. The resistor under the back plate also has more clearance this way. I had to insulate it before with a piece of tape to prevent shorting against the back plate.

 

I got a little benching done.... https://www.3dmark.com/spy/7157623

 

Look awesome Brother. Yea I can see how it'd be much easier to manage the clearance and install with that type of resistor. Hmm.. Make sense. So with the .008 ohm you'd use a .002 (wire style) for when testing .010 ohm? Ah and as a reminder, I'll send you a bunch of .008 ohms.

Very nice! Always great to see SLI action....

 

I really wish EVGA had a waterblock for the KP's out by now. I see that their other blocks are on sale. @Mr. Fox did you catch the sale price on your block?

 

Same here... can't wait until they release it... nice sale currently going on too...

PSU swapped... now drain time...

 

Well, it appears the .008 ohm resistor produces better results. It seems to be the sweet spot, as .005 is too much and .010 ohms is not enough. I tested these only because I had them and wanted to see what it would do with .010 ohms. I would likely not add another resistor, just take these off and replace them with .008 ohm.

Thanks for sending me some spares. I appreciate it.

Yup... amazing. I love SLI. I'll probably never have it again due to the cost.

 

I did. It was $20 off when I bought mine. You need it. If the K|INGP|N Hybrid cooler is anything like the FTW3 Hybrid cooler it's not anywhere even remotely close to enough cooling for overclocking. Just barely adequate for stock. I cannot imagine how hot they would get using an air cooler.

In a little while I will post a comparison of temperatures using Hybrid versus Hydro Copper running Time Spy and Fire Strike stress tests. I am running them now with the Hydro Copper.

 

Absolutely Brother my plesaure.

Looking forward to it!

One thing with full cover blocks have you noticed any condensation under the block side between the block and pcb at all with the chiiller?

 

I have not. But, two things bear mention about that, since my experience may not be a good representation of what would be normal for most people. (1) I live in a desert with extremely low relative humidity; and, (2) I check the numbers to calculate the dew point and keep an eye on things. When I see evidence of condensation starting, I turn off the chiller.

When I am running the chilled water through the radiators, it does not get cold enough to cause condensation. The radiators are constantly warming the water. Although, it still gets A LOT colder with the chiller even running the water through the radiators. For anyone that lives in a humid area, leaving the chilled water circulating through the radiators is probably going to be the safest way to combat condensation.

The down side to being in an extremely dry climate (and there are very few down sides to it) is static electricity is crazy bad... you can't touch anything without getting zapped on some days.

 

I am going DICE during the summer when I have a week or two off. I may get a chiller during the next round of X299 proc's.

 

It would be fun to try that some day. I have never done anything like DICE or LN2 overclocking.

The nice thing about the chiller for me--and I think you would appreciate it as well--is the ease of use. I don't have to take anything apart, just flip a switch and I'm ready to go. Worst case scenario, I move a couple of hoses with quick disconnect fittings if I need to get the temps down near 0°C... no fuss, no muss.

 

Great stuff and info. The dew point is definitely good to know.

Sounds great...I'm eagerly waiting for my chiller to arrive as I think I've exhausted all the thresholds at my current conditions. We should be seeing you in the top 20 with that monster of a CPU you have.

The On-Demand is nice for sure... one thing that I like about it a lot.

 

In progress...

Out with the current... (mod time)


....in with the new. (KP)



* @Mr. Fox wait do we have the same screw driver? ROFL... just noticed that.

 

It certainly looks like it.

 

OK... here are the Fire Strike and Time Spy stress tests comparing GPU core temps with the Hybrid versus Hydro Copper.
Stock FTW3 vBIOS versus K|INGP|IN XOC vBIOS. Stock clocks, no chiller and ambient temperature is 73°F in my office.

The test loops 20 times and runs for around 10 minutes. To pass the test, your system's frame rate stability must be at least 97% and all loops must be completed.



@Johnksss @Robbo99999 @jaybee83 @Talon @Papusan @Rage Set @iunlock @JoeT44

 

I don't remember... are those using 240MM radiators or a little 120MM like the FTW3 Hybrid?

 

240mm rad.

 

That's good. I'd think that would work quite a lot better than the 120MM radiator the FTW3 Hybrid uses.

 

Excellent data thanks for providing this Brother. The test bench looks awesome too.

Yup 240's... I had to reroute all the tubing, but now about to fire it up... that took forever lol... I was looking for the extra power cable for the KP so that took a while...

 

Another personal best...

https://hwbot.org/submission/4153089_ | https://benchmark.unigine.com/results/rid_1dc5c3b0737440eeb84e3badee5b10f3

 

Looking good! I'm benching at the moment...trying to dial in the KP... I can't wait for the chiller....

Excellent SP score...

 

Yeah, but why lower the voltage of the RAM when you find a certain Mhz is stable, instead why not continue to increase Mhz - it sounds like you might be missing out on higher Mhz clocks because you're backing down on the voltage in the name of finely tuning lower Mhz overclocks?

 

Don't feel bad, you did him a solid by selling him a "like new" card for probably a great price, and he knew you looked after it - it's probably a great thing to buy hardware from other forum members, I might do that sometime.

Excellent testing there, that's the difference between a custom loop and a hybrid then!

Is that your personal best of both your previous card & the one you have now, so like the highest score you've ever achieved? I hope so, makes your GPU switch more worthwhile if that's the case. Good to see you've been able to gradually get more performance out of the card by doing the shunt mods & changing it to custom loop cooling/chilled water.

 

see step 7) rinse and repeat.

once i got tightened timings and lowest stable voltages set for a specific clockspeed, i start over again @step 1. thus, voltages back up, loosen timings and try my luck with higher clocks.

not missing out on anything, no worries just dont like to run voltages any higher than they need to be

this lets me know in what way and at what clockspeed the timings and voltages really need to be higher, so i can make out trends and better predict stable settings for the next higher clocks.

as i said before, getting the first error free boot is the most difficult part cuz its based on educated guesswork.

Sent from my Xiaomi Mi Max 2 (Oxygen) using Tapatalk

 

Ah, I see, that takes a long time to do that testing, I'd probably take a shortcut rather than confirm the tightest timings stable for each 'step' of Mhz RAM overclock. My approach would be to keep XMP stock timings and increase frequency, if not stable then increase RAM voltage, and if not stable loosen timings at the higher overclock. I'd then compare that in performance to stock Mhz at stable tightened timings to make a decision between the two. But I see your logic of your approach now, downside is more work, but maybe you get a better result in the end as it sounds like you'd theoretically be comparing performance of optimised overclocks (in terms of tightened timings & voltage) at multiple different Mhz overclocks.

 

Thanks. Yes, and that is why I did it. I had every intention of switching out the hybrid kit and it was inconvenient to do it, but it was useful for me and others to have something concrete to look at rather than going on assumption. I am pretty confident the hybrid is better than air, but I did not have any way to test that. It definitely was with the 1080 Ti, but I never committed that to data that could be shared.

Yes, it was the best. My previous best was just short of 11K.

I still need to redo the shunt mod with .008 ohm resistors. Inconvenient, but at least now we know for a fact that .008 works better than .010 ohm.

I am still puzzled that the KPXOC vBIOS produces a voltage limit performance cap with higher voltage. Will be interesting to see what .008 ohm resistors do with that.

It is a sin and a crime how absurd NVIDIA'S firmware has become, with idiotic and unnecessary behavioral algorithms that should not exist. There is no good excuse for it and the people responsible for it need to be beaten severely, then fired. Firmware should just work, not interfere, and let the product buyer decide how to tweak and tune things. The only limitation there should be is a thermal limit at the thermal damage threshold.

 

Nice score Bro. Congrats on the new card.

 

Hmm, my GPU (GTX 1070) is no longer downclocking at idle, any ideas? I was recently messing with using a screensaver rather than letting my monitor go to sleep, and it's then that I noticed that my GPU wasn't downclocking at idle or during low loads like internet browsing like it was before. There was also an NVidia driver update in the last few days which I upgraded to, but don't know if that's the cause or not. I've tried disabling screensaver, but still no GPU downclock at idle. My next experiment would be reinstalling the NVidia drivers or downgrading to last NVidia driver, but I don't want to do this if I'm missing something obvious and easier, any ideas? I'm also wondering if experimenting with the screensaver has tripped a Windows 10 bug of some sort (regardless of whether screensaver is currently configured as active or not). (didn't think this was worthy of creating a new thread just yet, and we have a lot of knowledge in this thread)

EDIT: reinstalled NVidia graphics drivers and now it's downclocking. Not sure what caused it to get stuck before, probably a borked NVidia driver install given that reinstall fixed it (I always do a clean install option, so strange that it was borked before).

 

I have some ideas for you to look into. What typically causes that is a 3D process running in the background. Running a high refresh rate monitor can also cause it. Check your advanced settings in Chrome and turn off graphics acceleration and disable the ability to allow Chrome apps to continue running in the background.

 

Mission accomplished... .008 ohm resistors are in place and the GPU behaves appropriately... better than .010 ohm. Benchmarks will follow, but the power limit flags with both vBIOS are gone now.

 

Man, you fly through these mods! Good news.

(I fixed my previous GPU downclocking issue, I just had to reinstall the NVidia graphics driver, I always do a clean install so it's strange that it seems to have been an initial borked NVidia driver install).

 

It takes longer to remove the screws than it does to solder those resistors in place, LOL. It's even faster than replacing thermal paste if you clean up the old paste first.

 

Cool stuff

 

What driver version you re-installed?

 

A few months ago I re-soldered a dodgy connection in my AKG K702 headphones (a known issue with the model), and I felt like a hero, sure beat scrapping a pair of 114 quid headphones, for the cost of a 16 quid soldering iron and some youtube instructional videos!

It's the latest one, 430.64 whql. I've never had downclocking issues before, so I figured initially it was due to some bug in Windows 10 in relation to messing about with the screensaver (I've never touched the screen saver before you see). But I've got the screensaver activated now, and a reinstall of the NVidia drivers fixed the issue, the same clean install that I always do (selecting 'clean install' option when installing the drivers).