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

Speaking of fan header, anybody know the max power output of stock fan header on those Nvidia FE card? Always wanted to hook some extra noctua 80mm fans to cool my GPU from that single header. And can I make use of stock LED header, like convert them into fan header with no PWM?

 

Nice fan controller. Really like it. The aircraft quality engineering is unmistakable. Overkill truly is the only way to fly for everything. I have never been disappointed with using the overkill approach that I can remember. Bigger, heavier, stronger, faster... always awesome. The only down side is it makes it way more difficult to put up with something less in the future.

 

New personal best in Cinebench R11.5: http://hwbot.org/submission/3997507_

 

New personal best in wPrime 1024M: http://hwbot.org/submission/3997536_

 

Was that from your bus speed tweak? I did the same on my CPU a few weeks back, increased CPU speed from 4.6Ghz up to 4.65Ghz (I used to be on 4.7Ghz, but was seeing a small amount of voltage related CPU degredation after running it for a year like that). As an added bonus it pushes your RAM speed up a tiny amount as well, a small notch that gives you more fine tuning than just setting CPU & RAM multipliers.

 

No, I don't think so. When I put the X299 Dark motherboard in I used Kryonaut and noticed my CPU was getting about 15°-18°C hotter than is was with the Rampage VI Apex, even with the water chiller. I had Conductonaut on it before. So, I took it apart again this evening and put Conductonaut back on between the water block and IHS. Temps are back where they should be. I think the CPU is just running more efficiently because it is so much cooler. Both of those benchmarks immediately improved after applying the Conductonaut. The BIOS settings were the same before, using saved profiles.

 

Oh right, wow, 15-18 degC temperature difference between Kryonaut & Conductonaut! Temperature differences normally aren't that large for switching to liquid metal for an IHS to CPU cooler interface, but given that you have a water chiller and huge watts of heat to dissipate it's probably made that interface more of a limiting factor compared with other less extreme systems. For example, I don't think I'd see 15 degC temp drop by using it on my system, because I'm only dissipating 80W in the day to day for gaming and only 110W during the most extreme benchmarks/stress apps (it was 148W when I used to run it at 4.7Ghz though). You've done some bus tweaking in the past though, as you bus clock is slightly above 100Mhz?

 

It was about 5-6°C cooler with the 8700K. I suspect you would see a similar improvement. I think you nailed it though. Pulling between 900-1100W it needs all the help it can get moving the heat off the IHS to the water block as quickly and with the least resistance possible, and the Conductonaut does a far superior job of that than Kryonaut or other normal thermal compounds.

 

5-6 degC cooler in my case then maybe, I don't think I'll do it though as my CPU on the hottest core for a max temperature during a gaming session is only 55-60 degC depending on room temperature, I'm using Kryonaut. Ha, if I put a 9900K in here then I'd do it, but I'd need a modified BIOS for my Z170A board, plus I'd have to actually buy the 9900K!

 

I always run with BCLK over 100 because I hate it when my clock speeds are less than the multiplier. That bugs the crap out of me, LOL. EVGA does not have as granular stepping on the BCLK for some reason. I could set the Rampage VI Apex BCLK in increments of 0.01MHz and the X299 Dark only changes in increments of 0.20MHz. So, it's actually not tweaking as much as not having as much granular control as the ASUS motherboard. The X299 Dark also lacks some other settings like the ability to disable CPU BCLK spread spectrum, RAM spread spectrum and RAM power limits and auxiliary memory VRM and power limit tuning, and CPU and memory VRM frequency response/latency. Doesn't even have LLC adjustments. Supposedly, EVGA says their BIOS is advanced enough to adjust those settings automatically based on the clock speeds set. That could be true, but I prefer to manually set all of that stuff and not have to rely on their firmware to do it.

If I decide to keep this motherboard if/when I get the Rampage back from ASUS, I may see if @Prema is willing to have a go at exposing some hidden menus for those things. For the life of me I cannot understand why EVGA would not have those settings available on their flagship extreme overclocking motherboard. Seems kind of stupid that they are not accessible.

 

Have you tried fine tuning your overclock using BCLK though? It's still finer tuning than CPU & RAM multipliers right? Although 0.2 steps are large aren't they, too much for fine tuning? (My motherboard can do it in 0.01 steps, that's MSI Krait Gaming x3, which is not a top tier board, but not budget either)

 

Not yet. I'm still just getting started with the tuning on this mobo. I was gone on a business trip all week, and I will be on another business trip next week.

Yes, I am puzzled by some of the things with the EVGA firmware not having the robust settings I am used to. This is clearly an excellent motherboard, but I conceptually abhor the idea of the firmware "AI" adjusting things automatically. That just rubs me the wrong way. And, I actually believe that I am capable of doing a better job of it on my own, without any "help" from firmware "AI" LOL.

I almost have to believe that wizards like Kingpin must have a "special" BIOS for his X299 Dark, with all of the settings ASUS makes available on the Rampage, and the peasants like me are expected to settle for the "almost good enough" firmware to go with the amazing EVGA hardware. I am really leaning toward keeping the Rampage if/when ASUS replaces it solely for this reason.

 

Managed to hit 2280 CB at 5.1 ghz (HT on) @ 1.335v (can't go lower on vcore or WHEA/CPU L0 errors) on BGA Killer.
Had to tighten RAM timings to command rate 1T (surprisingly worked on 2x16GB Gskill RGB TridentZ B-die) and tRFC=270 and tREFI=32767.
Thermals are almost out of control however. 30 minute realbench stress test was reaching the 90's.

BGAkiller can do 5.2 ghz @ 1.335v with HT off (haven't tried lower on vcore yet, that was a best guess).

Pretty happy about 5.2 ghz @ 1.335v with no HT. Turns the 9900K into a very highly binned 9700K.

 

What are you using to test stability again? WHEA errors are hardware event error reports aren't they? I'm fairly certain you can get corruption of OS/data before you get WHEA errors - in other words just using the computer 'normally' for your everyday tasks at your new overclocks and not seeing any WHEA errors isn't a guarantee that no corruption is occurring. That's my intuition/hunch on it - I always go overboard with stress testing my overclocks, so I've never relied on creating a borderline stable overclock & then looking for WHEA errors in the event log, so don't really know if that's a reliable way of going about it, my intuition is that it's not quite stringent enough.

EDIT: I see you mentioned Realbench in your post, an 8hr pass of Realbench is supposed to be a pretty good test of stability according to the folks in this Skylake Overclocking Guide/Thread ( https://www.overclock.net/forum/5-intel-cpus/1570313-skylake-overclocking-guide-statistics.html). They put it on par with x264 testing, which is how I validated my latest overclock, an overnight run of x264.

 

http://hwbot.org/submission/3998097_

http://hwbot.org/submission/3998131_

http://hwbot.org/submission/3998137_

 

Sorry mistake

 

Again

 

https://valid.x86.fr/bench/2arb6b/16

 

I'm still waiting on Performance PC to actually ship my MO-RA3 420 Pro and a crap load of fans. I also picked up the delid tool for 2066. Going to be a cooollld winter.

 

I'm happy that you like it. It's a very durable controller and I've been happy with it.

 

Not many people would have a real world encounter with 30 minutes of 100% full-tilt CPU load. If it takes that long to reach 90°C with the CPU running wide open with 100% load, then you can probably bench and game the crap out of it and never get nearly that hot.

When you says WHEA errors are you talking about the subtle type that show with Throttlestop or HWiNFO64, or a BSOD WHEA Uncontrollable Error (0x124)? That generally means voltage is too low. Try increasing cache voltage first if the cache and cores are not already in sync. It may be the cache voltage too low rather than core voltage, but either one can cause it. So can having the VCCIO or DMI voltage too low.

 

The ones that appear in HWinfo64/event viewer.
I'm actually testing 5.1 ghz fixed size 1344K FFT AVX right now@ 1.135v, LLC turbo. Not sure if "PWM Power Phase Control" matters or not but I set that to "extreme performance".
Not a power virus test either. Then disabling AVX and doing the same thing (realbench/cinebench is stable of course).

 

Did you try what I suggested? I had the same WHEA errors showing up with Throttlestop benchmark and HWiNFO64 sensors running the 8700K with stock cache voltage when the CPU cores were overclocked. Increasing the cache voltage stopped the WHEA errors in Windows. Changing the power phase control to extreme and disabling power phase protection options has no effect on this. Those settings are only for pushing things harder without interference from power-saving crap.

 

There was no need since I'm stable at 1.335v, so I'm happy here.
1.325v wasn't stable at all. Even cinebench gave WHEA errors, as well as realbench. So i raised it up to 1.335v. That did the trick.
I was just bored and decided to test 1.335v again for OCD reasons (it was already saved in a profile).

 

He's running a Noctua NH-D15 air cooler which is almost the same as my NH-D14, if his case cooling is efficient then max CPU temps will stabalise within 10 minutes easily - I've found that my CPU reaches a max stable temperature very quickly. So in that situation he would be seeing 90 degC within a 10min time frame rather than the extended 30min. If case cooling is inefficient though, (not enough air flow), then it could take longer to reach steady state, as you have to wait for heat to build up in the case in addition to heating up the air cooler radiator - of course that steady state CPU temperature would be higher in this situation too. (With your water cooling you have to heat up the whole loop of water as well as the radiator, but with air cooling it's just the radiator that has to heat up to reach steady state if case air flow is good.)

EDIT: yep, just tested mine to confirm what I remember: took 8mins to stabalise at max 58 degC on hottest core during Prime95 load.

 

Yes, this is true. But, I still doubt he would see those temperatures in most real-world workload scenarios. There are not many things most of us do that put the CPU under full-tilt load for such a long time. Running programs like Prime95, AIDA64 stress test, OCCT, etc. for 30 minutes or longer far surpasses the typical workload stress scenario and could reveal some levels of instability that might never be seen under ordinary circumstances like benching and playing games, or less taxing tasks like web browsing and email.

 

That's true too, but a lot of people like to allow for a margin of error above a "knife edge stable overclock". If you just test an overclock with everyday tasks & gaming then there could be circumstances where you get a build up of undetectable corruption that could cause widespread software issues/errors over a period of weeks/months that could end up in loss of data and need for reinstallation & retrieval of back up files. By testing an overclock with a slightly or significantly harsher workload than you normally use, then you're just finding out if you have errors or not within an 8hr time frame rather than an 8 week time frame (when it might be too late + hassle). And on a practical level there's not much to be gained from running a knife edge stable overclock vs a rock solid overclock, you're only talking 50-100Mhz (0.05-0.1Ghz) difference on a CPU or instead an extra 25-50mv (more like 25 rather 50!).

 

Forgot how fun it was to have a extreme CPU. 4.6Ghz:

 

That's good, not far away from a stock desktop 6700K! And with your CPU being a mobile CPU as well as few generations older than the 6700K, then that's a good result. Although did Mr. Fox have his running near 5Ghz!? Yours is good though, for sure.

 

He had 4.7Ghz I believe.

I can also reach 4.7Ghz but I need AC cooling to reduce temps. I should also install windows 7 on a separate esata drive so my score is even higher.


I think @Mr. Fox will agree, this was a time when mobile CPUs were considered good and clocked very well like desktops chips.

 

Yeah, absolutely. The 2920XM, 3920XM and 4930MX were all awesome mobile processors. I benched them all at 4.7-4.8GHz on AC cooling.

http://hwbot.org/submission/2345039_mr._fox_cpu_frequency_core_i7_2920xm_4906.1_mhz

http://hwbot.org/submission/2390090_mr._fox_cpu_frequency_core_i7_3920xm_4905.57_mhz

http://hwbot.org/submission/2578547_mr._fox_cpu_frequency_core_i7_4930mx_4997.04_mhz

 

On a practical level, no. On a competitive level, that small change can make a real difference in points and ranking.

 

cmon, i wanna see 5k+ cb15 points! u can do it! hey @Mr. Fox maybe you can lend him your 4200 RAM to eke out the necessary points haha

 

Got it!

 

awesome! so the RAM did it indeed! nice

 

The Ice Bear is coming! Thank goodness I picked up some quick disconnects, otherwise this build would be very difficult to move.

 

Very nice! I bet now you're glad you didn't trade me that CPU for the 7960X.

I wanna see some pix of it.

If it is difficult to move, then chances are good that it must be pretty awesome. Portability doesn't always mean mediocrity, but it's generally a pretty good indicator of compromise.

 

Are you home yet? I don't want to do a team viewer just yet but I would like to talk to you about something. Can I call you and if so what's a good time?

 

Hi. Yes, I am home. Leaving on another business trip this weekend. You can call me any time.

 

Not even delidded yet and I jumped up into the top 50 in the TS Extreme CPU HOF - https://www.3dmark.com/spy/5319977

I have much more tweaking to do, so stay tuned!

 

If you like, show some comparison pics of your setup before & after the various changes/modifications you make/made, as well as comparisons showing temperature & power differences combined with differences in 'scores'. Tell a story of it if you like.

 

Well, looks like 2x GTX 1080 Ti (especially the K|NGP|N variant) still easily destroys a single RTX 2080 Ti. Used these examples submitted by K|NGP|N himself to keep the comparison somewhat more consistent

https://www.3dmark.com/compare/spy/3345848/spy/4443441/spy/2746610/spy/2520758#



And, this should prove even more interesting. Who knows how long it will take for the 2080 Ti K|INGP|N to become available, or what kind of vital organs will need to be sold in order to pay for one of them.

EVGA releases teaser RTX 2080 Ti Kingpin

 

Did you get your Rampage back yet Bro?

 

When they teased the 1080 TI KP cards, it was a couple of months after when they launched. The K|NGP|N premium over EVGA's best general market card is normally $300 to $350, so I'm expecting the card to be between $1500 to $1549. Will they be worth the price of admission? Yes, if you are going to OC the crap out of them with water at least. I do have a 2080 TI FTW3 card here (on loan) and with the right bios (the ones that K|NGP|N or the EVGA community has), you can push this card very far.

 

Not yet. Status says repairing. I seriously doubt it can be repaired, but they're probably still looking at it. I did get an email from the engineer that called me. He wanted to let me know he had the board and would be examining and testing it. That was Thursday.

Yeah, K|INGP|N cards are definitely the way to go. FTW3 are also awesome. I don't even want a 20 series GPU at all unless it is an EVGA 2080 Ti FTW3 or K|INGP|N. Compared to what I have already with my modded SC2, I would consider anything else to be a total waste of money for such a modest performance increase. The only reason I would buy any expensive GPU is to overclock the crap out of it. Otherwise, I would go totally cheap and buy some piece of crap GPU from AMD, like an RX 590 for $279. That's good enough for playing games.