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

Wow, I missed a lot! Including throwing in a bid for one of your customized notebooks. I envy whoever picked those up. And kudos on the awesome desktop!

Sent from my Pixel 2 XL using Tapatalk

 

Took advantage of Cyber Monday and bought an MSI codex X-069 (micro centers version) with the i7 8700k and GTX 1080ti.

Will be watching this thread for help along the way. Ran it last night and scored 22k on fire strike. That's a huge step up from my Sager NP9280 (D900F) with the 680m scoring 5,400 Lol!

Charlie

 

I think it does matter, for all CPUs that are K or X. The Maximus and Rampage ASUS boards are designed for people that are serious about CPU and memory overclocking. 'Gaming' doesn't necessarily mean the board is good for overclocking, or has components built to handle it. I have never seen so many overclocking features before. You can control everything, LOL. I think it ALWAYS make good sense to pay extra for a motherboard built around the intent of having the snot overclocked out of it by the end user. They are built with components made to handle more abuse than you can dish out. Some of the less expensive motherboards may not be as durable, and may not overclock with as much stability. @Johnksss@iBUYPOWER is the resident guru on this stuff. Maybe he can comment on whether it matters. I do not believe the core count of the CPU has much to do with whether it makes sense to pay more for the better motherboard.

Thank you for the kind words.

Awesome! Congratulations, bro.

 

My thinking behind more cores = more important to get decent motherboard, is just to be sure that the motherboard can handle the increased power it will need to dish out when overclocking those monsters.

When I bought my own motherboard for the my 6700K, I took the approach of looking up prices for Z170 motherboards, seeing which were the best deals, reading some of the features, and then trying to find specific reviews & user ratings for that particular model. I think I ended up buying one of the cheaper/better deal ones that also got good reviews. The BIOS was easy to navigate, and overclocking both CPU & RAM was easy, whether or not I could have gotten better overclocking results with another motherboard I'm not sure, but I doubt much better on the CPU side: 4.8Ghz & up are not easy for 6700K! RAM overclocking? For some reason my RAM just won't go stably above 3200Mhz even though I can tighten timings at 3200Mhz, so perhaps that's something that could be improved with a different motherboard, maybe, just postulating!

 

RAM overclocking requires excellent RAM and excellent motherboard, built by people that understand it. RAM overclocking is something ASUS excels, and some of their competitors with good reputations struggle with. The CPU and chipset also has some relevance. For example, 7700K has better memory overclocking than 6700K did. I noticed a distinct improvement with that change alone.

 

Ha, if that's the case, just as well RAM overclocking has the least impact on Intel systems, seeing as I have an MSI motherboard - one to remember for Ryzen folks then!

 

Actually, I think you give Asus a bit too much credit here. On board materials, best bar none. On firmware, it is a joke for a fair amount of the issues you get because of it. Also, during the Z170 boards, the Asrock OC Formula had the best ram OC performance, with Asus then playing catch up and releasing the Maximus VIII Impact. Even then, the Asus boards had to be binned to get the best ram overclocking, as that had as much to do with the achievement of over 3866 as finding a strong IMC. So, it isn't exactly fair to characterize Asus so highly, although they would be up there for recommended boards. With that said, they do offer, often, more granular controls than what may be found with other board manufacturers. So it is a blend.

For the conversation generally, you should look for board roundups a couple months after release to allow for firmware updates to deal with specific bugs and for the comparison in performance, as there is a difference. Starting with the Z170 time frame, Asrock has gotten better, although still chapped they only went 8+4 instead of 8+8 on their HEDT platforms.

 

Time Spy: https://www.3dmark.com/3dm/23658285 | http://hwbot.org/submission/3722356_

 

3400 1T: http://hwbot.org/submission/3722883_ | http://hwbot.org/submission/3722882_

 

3500 1T: http://hwbot.org/submission/3722902_ | http://hwbot.org/submission/3722901_


3500 1T: http://hwbot.org/submission/3722930_

 

GPUPI - http://hwbot.org/submission/3722992_

 

Have you done gpupi for cpu yet? You may need to download Intel's opencl driver.

 

I probably need to do that because I do not get the menu option to upload the score. I did run it, just cannot do anything with it. The menu option is grayed out to submit the score.

Do I have to re-enable the iGFX in the BIOS for that to work before or after installing the driver? I have the Intel iGFX disabled and only PEG enabled in the BIOS now

.

 

Enabling iGPU in the BIOS is the only way, briefly connect your display to integrated GPU's HDMI port and install the intel driver via inf or their setup file. After that simply unhook your display and connect to your 1080Ti and everything will be golden.
EDIT: You don't need anything like that, just use this package. I got this when I was fiddling OpenCL Runtimes for linux. http://registrationcenter-download.intel.com/akdlm/irc_nas/9022/opencl_runtime_16.1.1_x64_setup.msi

 

Thanks. That runtime fixed it.

GPUPI for CPU 100M: http://hwbot.org/submission/3723428_



GPUPI for CPU 1B: http://hwbot.org/submission/3723440_

 

Yeah nice bench scores @Mr. Fox

 

No, this isn't using the iGP at all. It uses the cpu to do all the work.

Grab this driver and try it:
https://software.intel.com/en-us/articles/opencl-drivers
You will need hpet enabled.

Edit: just saw I was beat. But that is the official Intel link.

 

While HPET is on, you should run HWBot prime, X265 1080 and 4k. For HWBot prime, you need java installed. But good way to get some global and world record points if not done yet.

EDIT: So HWBot prime no longer gets points, but the other two have world record and global. Wanted to clarify. Also, I cannot remember, but the HWBot realbench may also need it, IIRC. I only really turn on HPET for those benches, so I run them as a group. GPUPI/for CPU, HWBot Prime, HWBot X265 1080 and 4K, and HWBot realbench. If you can think of any I'm missing from this list of HPET benches, let me know.

 

Hey, check this out, my overclocked 6700K beats a 5Ghz 7700K in Timespy Physics CPU test!

My CPU score (6122): http://www.3dmark.com/spy/2859585

7700K 5Ghz score from Gamers Nexus: https://www.gamersnexus.net/hwreviews/3157-intel-i5-8600k-review-overclocking-vs-8700k-8400/page-3

My CPU also beats stock 8600K and stock Ryzen 6 1600X as seen in above link. This is strange because my 6700K is only at 4.7Ghz compared to 5Ghz on the 7700K - my thinking is that my cache is also overclocked to 4.7Ghz and I've got overclocked RAM tightened to 3200Mhz CL14 (Gamers Nexus use only 3200Mhz CL16 RAM) - I have an intuition that Timespy Physics test is very dependant on fast low latency RAM and possibly CPU cache frequency. My 6700K doesn't win in the other CPU tests.

 

Well, there are a few things to consider, all of which could be very relevant:

1. You know what you are doing and you have tuned your system well. So, credit to you for that. We cannot ignore this aspect of your results. Nice job, bro.
2. Professional reviewers are, for the most part, incompetent and their results should not be deemed reliable except with stock clock. In the case of stock clocks/BIOS defaults, we can expect lackluster results because defaults suck 99.9% of the time and need optimization to run right. (GamersNexus is generally better than most, and they know what they are doing most of the time, so I don't lump them all in the same slop bucket.)
3. Last, but not least, Time Spy is a pretty messy piece of crap benchmark. You can view it as "advanced" (some do) because it is very sensitive to things that other programs and benchmarks are oblivious to. I don't really consider it to be advanced as much as fragile and buggy. It is similar to some games that are only stable with stock clocks to the extent that anything that upsets its apple cart in the slightest amount causes it to crash or error out instead of recovering from the error and moving on like better programs and benchmarks do. It is one of the least stable benchmarks I have encountered in years. Perhaps the screwed up mess formally known as DX12 has something to do with that as well.

 

FINALLY, got a Cinebench R15 run on Windows 10 almost as good as my Windows 7 top score (1725). Not too shabby! It is definitely more of a pain in the ass to extract good CPU performance from that crappy " new and improved" pile of trash OS. But, since it's not better, there are no HWBOT points to be gained from it. So, "meh" for that part of the deal.

 

When your Cine-15 score & silicon both just got smoked, & Fox can't even see you anymore in his rearview mirror ... ...

 

GeForce GTX 1080 Ti At 2.5 GHz+: Overclocking On Air, Water, And LN2-Tom's hardware

 

As I reflect on my benching over the past couple of weeks, I observe the following:

• Windows 10 beats Windows 7 in AIDA64 Memory Test
• Windows 10 beats Windows 7 in Fire Strike
• Windows 7 beats Windows 10 in 3DMark 11
• Windows 7 beats Windows 10 in 3DMark Vantage
  
• Windows 7 beats Windows 10 in Sky Diver
  
• Windows 7 beats Windows 10 in Cloud Gate
  
• Windows 7 beats Windows 10 in Ice Storm
  
• Windows 7 beats Windows 10 in Ice Storm Extreme
  
• Windows 7 beats Windows 10 in HWBOT Prime
  
• Windows 7 beats Windows 10 in wPrime 32m
  
• Windows 7 beats Windows 10 in wPrime 1024M
  
• Windows 7 beats Windows 10 in Cinebench R11.5
  
• Windows 7 beats Windows 10 in Cinebench R15
  
• Windows 7 beats Windows 10 in Aquamark3
  
• Windows 7 beats Windows 10 in MaxxMEM2
  
• Windows 7 beats Windows 10 in GPUPI

Quite the legacy of failure and incompetence that the Redmond Retards are building for their worthless and pathetic selves. Maybe @Prema should start working on Windows 10 mods to at least make it come closer to matching Windows 7 performance. I know @Phoenix and @j95 have put a crap ton of effort into making Windows 10 better through tweaks. While every little bit counts and their tweaks do help, I'm pretty sure all the effort in the world might not ever be enough for Windows 10 to be as good as Windows 7. Windows 10 is just such a poopy mess, and isn't getting any better with time.

 

Cloud Gate: https://www.3dmark.com/cg/4166101

Ice Storm Extreme: https://www.3dmark.com/is/4278892

Ice Storm: https://www.3dmark.com/is/4278898

HWBOT Prime: http://hwbot.org/submission/3725893_

 

Cool, thanks for the credit about me tuning my system. One thing I've found for Timespy - it's an excellent tool for tuning in VRAM overclocks, running the Timespy Stress Test, it will crash at lower VRAM overclocks than any other program I've tried. Firestrike Graphics Test 1 on a loop I've found to be the best way to dial in a rock stable GPU core overclock, seems to be the most stringent test for the core of the GPU. That has been the methodology for me tuning my max GPU overclocks, and then gaming as the final ongoing test - it's true that for some games/benchmarks you can push higher core or VRAM overclocks, but I like the knowledge of being 100% stable, so I don't change my overclocks per game, I just use the rock stable overclock - 2100Mhz max boost frequency equating to 2050Mhz in game once the card has heated up (auto fans) and 2285Mhz (9140Mhz effective) on the VRAM. (If I increase voltage to 50% I get an extra 13Mhz on the core frequency overclock, as well as being slightly more resistant to temperature related decreases in clock frequency, so with added voltage I get 2075Mhz in game once the card has heated up, but I game most of the time without the added voltage - 2050Mhz vs 2075Mhz makes virtually no difference to a gaming experience).

 

I can't get hung up about extreme overclocking using LN2, but it is impressive to see the power circuitry of the Lightening GPU being able to handle loads of over 1000W - that's some solid manufacturing!

 

In cinebench r15 you can change the process priority from normal to Real-time and you might just break your previous records. You won't see the Image processing on the screen periodically like you used to, but the process will be very fast and everything will look like its locked up. For example, your mouse won't respond, No num lock response etc.. Just wait until you get the final score.
One more tweak I found is to cleanup .Net assembly files using Dism++ and reboot the PC. After that you can ngen update command to generate updated assemblies and runtime files optimised for your PC and guess what Win32/Classic Desktop apps startup much faster than UWP apps. Its just an estimate Win32 apps loading times have been reduced by twice.

 

What's all this, no proper gaming CPUs from Intel until 2019 earliest, best CPU we'll have gaming wise is 6 core 8700K until 2019 earliest! See: http://www.guru3d.com/news-story/ro...mini-lake-and-coffee-lake-s-(300-series).html

I'm sure there was an 8 core CPU in the works based on the same architecture as the 6 core 8700K - ie. using a ring bus & not the 'mesh' of the current Skylake X CPUs that have the many cores (e.g 8 cores+). That 'mesh' slows down their gaming performance. I was hoping to see an 8 core version with the same architecture as the 8700K. Looks like 8700K is gonna be the best gaming CPU until well into 2019.

 

My best Cinebench runs on W10 are with priority set to real time and all non-Micro$loth and non-essential Windows services disabled. I even tried running it in a RAM drive and it is still slower than W7 in spite of all that effort to match W7.

 

Even BITS and WU along with MS office 2016 background updater is stopped, correct?

 

https://overclock3d.net/news/cpu_ma...ap_has_been_leaked_revealing_cascade_lake-x/1
So, looks like I was right on sitting on cascade. No new 10nm in 2018.

In fact, instead of cannon for hedt, they are going cascade, which is coffee.

 

Yes, BITS is disabled and WU is permanently disabled. I never allow Windows Update to run... ever. It's cancer. And, I do not use Office 2016. I prefer Office 2013. I use a custom installation with none of the SharePoint, OneDrive, Skype, OneNote and telemetry/updater filth installed. Only the core programs (Access, Excel, Word, PowerPoint, Publisher and Outlook are installed.) And, I manually delete all of the registry keys for the OneDrive sync crap using NSudo with TrustedInstaller rights.

 

GPUPI for CPU @ 5.3GHz: http://hwbot.org/submission/3726695_

 

Fire Strike: https://www.3dmark.com/fs/14338069 | http://hwbot.org/submission/3726717_


Slowly, but surely, figuring out 53x6. Almost there, but not quite stable enough for Cinebench, 3DMark 11 or Vantage... yet.

 

Brownouts around the Fox neighborhood incoming

 

Cool, what are the variables you're adjusting to try & eek out 5.3 Ghz, are there any tips apart from the ones I can think of: increased core voltage (VCore) (load line calibration too maybe), toning down cache frequency, relaxing RAM overclocks/timings?

 

No, not backing off the cache much. No need to lower cache ratio with 7700K or 8700K since the max is 300MHz less than core with the same voltage as core. If you have to give the cache too much voltage, might as well up the cache ratio to use it. I am not finding anything adverse about increasing cache ratio. Before Sky Lake that was more of a stability issue than it is now, and the situation improved with Kaby Lake.

Changing a lot of other things in small increments and re-testing. Tedious process, as usual. Our enemy is heat, of course. More heat increases the need for voltage which, in turn, creates more heat. It's the same vicious cycle we always had to deal with using laptops, just not as severe of a problem with a liquid cooled desktop. I need to start saving for a water chiller to make that a non-issue on the desktop as well. I could use AC cooling, and just duct that directly into the 360mm radiator, but chilled water sounds like more fun.

Basically, tinkering with everything you see on this screen. The power limits are already maxed out, so those require no thought.

 

Yeah, you could use the AC ducting in the short term, if you ever do that would be interesting to hear what kind of a difference it makes in terms of max overclock possible for you. Yep, I remember hearing about those water chillers from yourself & JohnKSS - noise & condensation right, but definitely colder!

 

I have never used one, so I cannot comment based on experience. I suspect the relative (ambient) humidity has a role in that. I have been a desert dweller for many years and never needed to be concerned about condensation using AC. I suspect how low you turn the temperature on the water chiller makes a difference as well. The closer you get to 32°F (0°C) the more likely condensation will be. AC cooling would have the advantage of cooling everything, not only the CPU, since a water chiller does nothing for the things that are not attached to it.

 

How much higher do you reckon you're willing to go with the voltage - you're already at 1.465V. 1.45V is often deemed as the max safe voltage for normal 24/7 operations as long as it's not too hot. I suppose for short benchmark runs you could do 1.5V, but you'll get degredation of the silicon at 1.5V I think, eventually. Have you set yourself some voltage & temperature limits so as to not damage the CPU, for example 1.45V under 90 degC, 1.5V under 70 degC, 1.6V under 30 degC (I picked those voltages & temperatures fairly randomly just as an example)? My personal tolerance for voltage for 24/7 use is what I'm currently running, 1.4V; I'm relatively confident that this won't degrade the CPU, especially as it idles in the 20 degC's, games in the 40 degC's, and benches in the 50 degC's & low 60's.

 

Sort of. Silicon degradation doesn't generally happen overnight. I cannot afford to replace my CPU today, but if I am only doing that for random benching once in a while, I suspect the 8700K will be too old to be relevant and will likely already be gone before the silicon degrades to the point it needs to be replaced. Other than an XTU benchmark, I have not seen temps in the 90's. Running Fire Strike at 5.3GHz, my core max was still like 76°C, and that was using 1.520V.

Once I have taken the CPU as far as it can go, there is no point in pushing it to the max on frequent basis. Once it reaches the point where it will not yield any more ranking or HWBOT points for me, then I can just use it like a normal PC until I can afford newer/better hardware, or move to LN2 or something so I can push it harder. My appetite for spending gobs of money is starting to dwindle, so I may never go that far.

Another thing I discovered with my experimentation, ASUS built some nice over-voltage protection into this motherboard. If I try to set it too high on some components, it will error out at boot with an "Over-Voltage Protection" error and I have to press F1 to enter the BIOS and lower whatever value is set too high, or confirm my intent to ignore the warning.

 

You willing to go higher than your 1.52V then for benches, I'm guessing when you get better cooling you'd go higher than 1.52V? 1.6V your limit for short benches for example with good chilled water cooling? I understand what you mean by saying you'll explore the absolute max, and then tone it down for everyday use, rather than continuing to bench it & just reproducing the same scores.

 

Yeah, I am willing to go up to like 1.600V if nothing bad starts to surface, but only in very small increments to observe the behavior that accompanies the changes along the way. I'm not going to just make a big jump in voltage and cross my fingers. Measure twice, cut once.

It's kind of weird how voltage changes are needed in stages. By that, I mean you might not have to change voltage to increase the multiplier by one or two sometimes, then the next bump requires a TON more voltage for that 100Mhz. It does not necessarily seem to scale in direct proportion to the multiplier. It does to some extent, just not evenly across the board.

The nice thing about being able to use excessive voltage is the fact that I can if I want to. Laptops have a lower limit, and when your core ratio exhausts the stunted amount of voltage the system will support, you're basically done benching. That is one of the reasons the P870DM3 was sold. I hit the wall on the CPU voltage and there was no point in benching it any more. I was competing against myself in a battle that could not be won once it was maxed out.

 

When I read that LN2 GPU overclocking article that Papusan linked the other day - it was quite amazing to see how much difference the really cold temperatures -100 degC could make in terms of a lot less voltage needed for any given clock. Are you expecting with a chilled water system to see the same effect, albeit the CPU will only be around 10-20 degC (guessing) rather than -100 degC, but are you expecting to be able to run a greater Mhz at the same given voltages in those chilled water conditions?

 

So, you have to look at the temp of the loop and components and find the dew point with the relative humidity at the ambient temp. Below that point, condensation becomes an issue. With being in the desert, you have a lower relative humidity, which will allow for a lower water temp before condensation forms. If trying to solve the problem of condensation, break out the eraser, the liquid electrical tape (backside of MB, and potentially graphics card, although using a conformal coating can do the same, if not at more expense), some foam pipe jackets for condensate on the hoses, and some blue shop towels when you drop it to guaranteed condensation causing temps and you are ready to roll!

I remember having to deal with that on the Skylake build. Had it in the high 1.5V range for some runs. Fun!

Also, we saw voltage shelves like that for awhile now (including Haswell). So not too new. And sorry I didn't have one to go against you. Even now, you have an Intel 6-core and I have an AMD 16-core, so not directly competing. Meanwhile, if Johnksss got a high core count chip, he'd have it under phase change or cascade, if not LN2, so not in the same class there either. But still loving the work being done! Want to lend me some graphics cards to do some better scores for the 1950X? LOL!

 

Are you talking about running just water with inhibitors and biocides, or are you talking about methanol mixtures, etc. If using something to do chilled water sub-zero, you then have to look at the additive and the effects on different materials in the build, such as hosing, effects with corrosion, etc. It starts getting a lot more complicated really fast! Considering my build is pulling so much in a house with older wiring, I can't even dare here. Maybe when I buy a house, I'll do like a friend and run 2x20amp lines with the appropriate thickness on wires in the walls to really do it up right. But time and future expenses.

 

Any time you lower the temperatures in a meaningful way requires less voltage. Even a 20°C drop allows you to do more. If you can control the temps, the need for voltage does not fluctuate dramatically. So, yes it would be possible to overclock higher and colder temperatures (and more stable with controlled temperatures) even without sub-zero temperatures. Just not as far as with sub-zero. If I could get the core max around 35 -40°C benching at 5.5GHz would likely be not very difficult. And, with chilled water I suspect that would be doable.

As inconvenient as using a desktop case it, I could also use that to my advantage turn it face down to minimize the risks of condensation. If and when any condensation formed it would drip onto the side panel window instead of running down onto the mobo. I have never experienced any condensation with any of my AC benching. I checked for it just to be safe, and everything stayed as dry as a popcorn fart at all times. It was never possible to get it cold enough to reach the dew point using AC in the desert climates I have lived in for the past 30+ years because of the very low humidity.

 

Now this is where I have to intervene.. You don't need to go out and get better hardware. You just need to go get more hardware. Get versed in what old cards get the most points. New hardware is just too hard to keep up with, unless you have deep pockets. And then those points do not hold over time unless you are benching 2 to 4 cards in all categories.

 

Ha, yeah, with threadripper & 1800x those CPU Watts are gonna grow large (can't remember which you have!)! Yep, just asking Mr Fox if he's expecting a Mhz increase for any given voltage point by using chilled water. I suspect that if you go subzero chilled water then you can expect even more.

Yes, would be interesting to see when/if you get round to it.

Yeah, I can see someone who's into overclocking challenges & benching wanting to experiment with old hardware too, makes sense. A cost-effective way to keep the overclocking urges indulged!