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

Thats very high for us also. Usually a storm moving in when it gets like that.

 

Yes, your weather is similar to Kennewick, where I used to live. (It is just a little bit colder where you are, but not usually by much.) Humidity is normally low and seldom any rain. I really miss that area of the country. Nice weather most of the time. I don't like rain or snow, and I rarely had to put up with the experience of either one in the Tri-Cities.

There is another person in this thread that is very familiar with that area.

I miss having winter. My favorite season, and what they call winter here isn't. It's more like a second Fall season. Winter down here is lame, and all of the snow-birds converging on the Valley of the Sun to escape their cold weather at home just makes it more lame. The best thing about Phoenix is that it hardly ever rains. One of the not-so-nice things (other than the horrible summer heat) is that when it does rain, it rains way too much (monsoon season). Going from no rain to excessive volume, suddenly and all at one time, isn't nice.

 

What are you planning to build there - will share some pics of my small builds



My “main” rig

9900k @5 GHz with an alphacool eisbaer 240
Asrick z390 phantom gaming
2x8GB GSkill 4133 CL19
RTX titan
Corsair 750W sff
512GB Samsung 950PRO nvme
2TB WD Blue s-ata
Case is Ghost S1 Pangea + large tophat (for AIO) and I have short Custom cables



And the crew



Velka 3 - 6700k with RTX 2070 mini 4.1 L (max) volume
Velka 5 - 7700k with Titan X (Pascal) 5.7 L volume
Nouvolo Steck with addin - 8700K with Titan Xp - 11.2 L volume
Ghost S1 Ash with large tophat - 9900k with (temp) RTX 2060 - 11.04 L volume

None of my 1080TI / 2080TI are fitting in those cases (if I want to keep using AIO for CPU)

 

Yeah, you shouldn't complain. 81% average for December. Today only 100% And average for the full year 76%

 

http://i.imgur.com/CP1uhRil.jpg

https://thermalbench.com/2016/05/23/alphacool-eiszapfen-quick-disconnect-fittings/3/

Check that out. Notice how the pressure shoots up with the increase in flow rate.

http://thermalbench.com/2016/05/22/koolance-quick-disconnect-fittings/3/

So if you are truly shooting for 3GPM, you should consider thicker tubing and the QD4. The QD4H would take an extraordinary flow rate to need.

Hope that shows why I used the 1.5GPM flow rate to compare the fittings.

 

FYI - the hyperlink is broken. I can copy and paste the text and that works.

http://i.imgur.com/CP1uhRil.jpg

 

Thank you so much !!!! I will definitely do a custom loop (or better say loops) with chiller in the equation.

What scares me a bit is:

1. the power draw ...

System + chiller ... seems I need some powerful lines not to melt down the cables (I have no exact clue how the power distributor is set as I am just renting )

2. Heating produced ...

I already have challenges in the hot NC summer in my “office” with the computers that I am using daily.

If I add another separate portable AC unit in the picture point nr1 will move from a little bit to very scary.

 

You're welcome.

If you normally have a window open in your office, place the chiller on a table that is window height and point the back of it out the window and you should be no worse off than you have been without the chiller.

My wife was freaking out that the electric utility bill was going to be crazy high with me running the chiller and benching. She called our power company for information and checked the power consumption on the days I was at home benching with the chiller running against the days I went out of town on business and the difference in the total kW of electricity used was negligible. She apologized to me for making a big stink about it.

The chiller is basically like a small refrigerator that you would see in a hotel room or use in a wet bar at home. So, it's not too crazy.

 

Yeah the problem is my benchmark rig is nowhere near a window ... (probably 6-8 feet away)

Anyway .. I am very excited on what future will bring now the SFF itch is gone ... I need a new enthusiast project before the new platforms are out and we start everything all over again

 

That is nothing that an extra 15 to 20 feet of flexible tubing won't cure. I'm looking forward to seeing the results from your adventures.

 

Very nice! As for my situation with this m-itx...

"Uncle uncle can we play xbox?"

"No more xbox kids... give me a minute and I'll put together a big boy console for you guys to play on.."

@Mr. Fox no time for that sissy stuff right? LOL ... consoles are literally set to "below low," graphic settings... compare Red Dead Redemption 2 on console vs PC and it's miles apart...

The result after grabbing some spare parts lol... nothing fancy, just slap together and play...

It's hilarious how small the m-itx board is inside of the case.



Big Boy Console:

8700K Delidded OC'ed
2080 Super OC'ed
360 AIO
xbox Elite Controllers

Great stuff..thanks for the links...

 

I almost had to look twice, as I thought that was a Prometeia Mach 2 GT

 

Ok, got up early and started benching with the fan in the window....brr CPU Idle temp 22c, load temp (CB) 75c. Getting close to @iunlock Wprime record, @4600mhz and did some fine CB runs



https://hwbot.org/submission/4314870_

 

Nice job! There is just no substitute for cold.

 

Awesome Bro! Crack into the .700's ... you can do it!

 

Ironically I just saw a really nice review on the Velka 3 and then checked back at your post to see if that was one of the cases that you had lol..

Wow talk about small form factor eh... 4 liters ... that's like a gallon of milk lol (3.78l) ...

I'm really considering that case to make an actual console killer console...

I'd be curious to get your thoughts on it..pro's / con's etc... are you on discord by any chance?

How about them specs eh? Not bad....

 

What is your score like without Benchmate? I really don't like using it. It's just way too bloated (consumes too much memory) and lowers benchmark scores. I can see why most overclockers are not very accepting of it for that reason. Is is required for Ryzen because of their RTC bug not being resolved?

I ran a few using Benchmate last night and my scores are definitely higher without it. It literally DOUBLES my memory resources utilization while it is running. I go from about 5-6% to 10-12% memory utilization and that lowers benchmark scores. It chews up more memory than AdGuard and AIDA64 combined.



All Cinebench scores are lower across the board with Benchmate running in the background.

 

For some reason it doesnt affect Wprime, but does all of the CB's. All my scores are higher without BM running. Im using it, because I fear that it will become standard practice at the bot. Especially with timer bugs, and cheating.
https://hwbot.org/submission/4314927_

 

I haven't tested it with wPrime. I hope they don't make it required at HWBOT. I am not aware of much cheating going on, or that there are any cheat for things like Cinebench and wPrime. But, between W8+ and pre-Skylake Intel/AMD RTC timer bugs, I can see it being required for some systems. Enabling HPET on W8+ also hinders performance.

On another note, I don't know why my GPUPI scores are just so horrible. I've never been able to figure that out. Probably some kind of driver thing or user error on my part, but they have never been as good as others running the same hardware and the same or even lower clock speeds. It bugs me that I don't know why.

 

For cpu benching? Have you tried messing with the batch size, and reduction size? I believe you can submit with any combination.

I need to install the AMD driver, and run it. Tried sitting here on my laptop...haha

 

I don't like using BM either, but due to there being a separate category for it, plus with the CPU being AMD I'm using it just because. It'd be much better without having to use it though. Standalone is better for sure and a lot lighter.

I've notice that too in that with wPrime it seems to run pretty much 1:1.

 

I have that issue on superpi, which makes no sense because I do really well with CPU and memory optimizations.

Couple tips for GPUPI for CPU:
1) try out multiple openCL drivers! That is number one, first and foremost, and not always is the most recent driver the best. You can even try AMD openCL drivers, but generally Intel's do better (and the ones for AMD can be hard to find, I use one off github).

2) go to overclock.at which makes it to get the command line structure. Run the automated command line to find the optimal batch, etc., to run in the window version, whether doing 100M or 1B

3) make sure you are raising priority (does not need to be real time, cannot remember which benches are helped or hurt with standard high priority versus normal versus real-time ATM).

Those are the easiest ways to improve one's GPUPI for CPU. Similar can be done for GPU runs as well, but openCL drivers often are in the graphics driver itself and finding the right one can take time.

What have you tried so far?

 

Awesome, thanks for the tips. I don't even think I have OpenCL drivers installed now that you mention it. If I do, it's whatever NVIDIA bundled, which probably isn't ideal. I will have to check, and if I do try other versions.

Do the OpenCL drivers affect the Cuda results as well?

Yeah, if you get points for running the same benchmark a second time (separate category) using Benchmate... might as well. I saw they are giving hardware points for it. (I normally won't run benchmarks where no points are given unless it has some sort of entertainment value to it.) That's exactly why I did it last night. Points are points, and we shouldn't really care where they come from, LOL.

 

You can run either cuda or openCL on the GPU. I cannot remember if Nvidia bundles openCL or not (AMD used to with their graphics driver, but I cannot remember if they still do). I do know that using the Intel openCL did well in the past. I haven't done just GPU in so long that I cannot remember which I used when I was going back and forth with the 980m years ago.

 

I can vaguely remember using Intel OpenCL drivers helped a lot for me a few years ago, so you're probably right. I think Brother @Johnksss suggested that and the score improved a lot. I totally forgot about that because it was quite a while ago.

 

AMD has discontinued development of OpenCL, and the driver on github is downloading @20k. Should have it in about 2 hours. SMH. Ive just been doing some research on it, and did see that it used to be bundled in the old catalyst software. As far as I can see, the only driver available is the one on github, as the AMD page for it is 404.

 

Yes indeed. On hwbot all the submissions for BM show as being beta so I hope that it never gets out of beta lol... I'm all for verified validations, but to corner us into using bloat isn't really progressive. There has to be another light way to go about it for AMD benching etc...

It's funny how some youtuber's are pushing BM and there's a funny used car salesman flare to how it's being promoted...odd...

But at the end of the day... points are indeed points... it's just nice that the unadulterated benching is still intact.

 

Don't speak until you understand why. During work on benchmate, they discovered new ways to cheat on Intel CPUs as well, not just ryzen rtc skew.

And to answer Mr. Fox, CB was one of the easiest to cheat on by replacing the image with a lower detailed one to be processed.

Go into the forum for HWBot, specifically the thread on Benchmate. Even dufus (0.o) contributed. And if everyone has the same handicap, it doesn't matter if the scores are lower.

What benchmate does is monitor the clock skew, but also validates a lot of other things. It also will replace having to potentially use CPU-Z windows in the future.

So this isn't as simple as it looks.

 

Interesting. I never noticed that. Maybe it is primarily people that are using different specs with lower performance capacity that are pulling those shenanigans. I say that only because most of the scores higher than mine are systems on LN2 and their clock speeds are higher. So, nothing above me looks out of whack or incredible. The higher scores are commensurate with the cooling and clocks (not a massive difference that would cast any doubt on the scores).

I've never really understood the rationale for cheating at stuff like this because there is no actual accomplishment, and if the scores are inflated everyone that sees them is going to question the score and the integrity of the person submitting it.

 

Very true. I don't see a reason to cheat, but people do. Dufus found a way to make the system clock check look correct with the scores making no sense in reality. An Intel timer skew was found on 4th gen Intel CPUs on windows 7. And some things showed up as odd with combinations of certain hardware, like certain motherboards.

Basically, this quest for creating a program to validate has gone down a rabbit hole that has exposed numerous previously publicly unknown ways one can cheat.

Hell, you know my scores and that I don't cheat, but even my scores on my 1950X are considered outliers as few get that high with that hardware (there were a couple others that could achieve similar, so it wasn't an aberration, but).

Now, this was birthed and is being designed by people that made GPUPI. They started creating a software where you could load all the benchmarks in to be ran in sequence while things were monitored, thereby being able to quickly spot if a configuration was on while gathering data for reviews (non-public program). That grew into benchmate, which can help to validate scores, take the screenshot, and upload the scores for you, thereby automating the process a little.

With that said, there are drawbacks, like performance impact. But if the sandbags are equal and everyone has to carry them, then it is an equal playing field, which is the most important thing in competition.

Also, it is why we can't compare scores with and without benchmate, unfortunately.

So it is trying to create a better playing field that the scores are more accurate, less ability to cheat, and equal sandbags.

 

Yeah, most of that makes perfect sense.

I don't like the concept of using software to defeat performance advantages created by using specific hardware (like a particular model of mobo). I mean, people pay more to get more from better hardware with firmware that has been tweaked and optimized for increasing performance and software that removes that advantage rubs me the wrong way. I look at benching as a form of racing and (cheating aside - never cool) using special or user-modified hardware has always been acceptable. Know what I mean? An enthusiast that pay $500+ for a high end mobo should not be held back and crippled to make his system behave exactly the same as some gamer-kid that pays $150 for a crappy low-budget mobo. Same for those that use firmware mods and tweaks like GPU shunt mods. It's no longer competitive if they go that far. Not saying that is what is taking place, only saying I hope that is not one of the outcomes.

 

You misunderstand. What I meant is not specialized firmware being leveled, I meant like clock skews happening on certain boards, etc.

It isn't what you said here that was the issue I referenced. It was that invalid scores would happen with one board but not another.

It is fine to use custom firmware or optimized hardware so long as it isn't winning because it throws invalid scores.

Does that make more sense? In other words, it checks better to make sure the scores are valid is all. Sometimes bad firmware can cause invalid results or the interaction between the firmware and MB.

They have no intention to purposely cripple anything. The program has some overhead, sure. But if that overhead is applied equally, the lower end boards can be dragged down as well.

Also, that is a misnomer that the more expensive boards perform better. Anymore, they just have more features and sometimes better memory overclocking, sometimes not. Look at how crappy Asus has become, where you pay a premium for broken firmware with more switches versus an EVGA Dark, which is truly a nice platform. Then you have Gigabyte coming back with decent boards even in the midrange, which I surely didn't see coming. Etc. Another example is the MSI Tomahawk B450 board, which was a budget delight but could, short of exotic OCing, hang with the X470 line to midway or slightly above in that stack.

So it is more making sure that if something is off, like mb mfr firmware causing errors or skewing something, the score is invalidated.

But, overhead is overhead, and there is a chance that quad cores are hit harder by having fewer resources free, but who cares about quads now?

 

I'll speak because of the fact that I do understand why.

I think you've failed to read my previous posts as to how I am all in favor of verifying and validating. If you go back some pages you'll see where I specifically state that.

I never said that it was simple, so again you are making an assumption there friend.

Yes indeed. It'd be very obvious for a cheater to get called out given the nature of the hardware drag race we all enjoy. Cheaters are clowns...rather cowards for that matter.

This emphasizes what I've said about recognizing the different classes and respecting the accomplishments within those classes.

I'm all for proper measures for validation.

I've always wondered about the exploits within benching software and if one with the knowledge to manipulate it could make one little tweak in an exploit to alter their scores? - I don't think it exists as much as we think, but with software anything is possible, unless the code itself is encrypted and inaccessible. Even then never say never.

 

Working on 47x... It's hitting the limitations on water, but I'll try to get it bench stable.

 

New wPrime v1.55 PR.

3970x

 

Windows 8.1? Wprime likes anything but windows 10 ...lol

 

Yea W10 is so clunky lol...

 

Unless something has recently changed, there is absolutely no driver support for Turing video cards with Windows 8.X and even INF mods do not work. The only options are Windows 7 and 10 if you are using Turing GPUs. Windows 7 driver support for Turing will likely be removed in the next couple of months. Sad days. But, at least we can use old drivers with Windows 7 if we want to. I am not aware of a solution for Windows 8.X.

CPU performance is crippled in Windows 8.X just as much as it is under Windows 10. It had nothing to offer in terms of performance. UWP bloat aside, I think Windows 8.X is worse than 10 in most ways. There was absolutely nothing that I could find good or likable about Windows 8.X including the more hideous UI.

 

For CPU type benches W8 seems to work better for me so far, but it could very well be limited to just those benches. I had only resorted to testing out different OS's other than W10 because I was having issues trying to get W7 installed on the board with the 3970x.

The older drivers work great, in fact I rarely use the latest Nvidia drivers; like M$ windows virus kb/updates it's often worse.

I hope to get W7 working on the bench... that'll be fun.

 

Did you find a 2080 Ti GeForce driver that works with Windows 8.X somewhere?

 

Brother @Rage Set did you post this video from your channel here in the thread anywhere?
I don't remember this one. Nice setup, bro. Love those radiators.

 

I think I might have found something that might work... will test it out tomorrow or maybe even tonight after NYE. I'll let you know.

 

@ajc9988 - that's what it was... openCL drivers were not installed. Thanks for mentioning it.

https://hwbot.org/submission/4316725_

https://hwbot.org/submission/4316728_

https://hwbot.org/submission/4316758_

https://hwbot.org/submission/4316764_

 

That was a video I posted to see how well a video can grow organically without adding in a custom thumbnail and etc. I love those radiators too, I can even run the computer without fans with those on light productivity tasks.

 

The videos do grow organically. Tags are important as well, so that it shows in search results. I saw it off to the right side on one of my videos and I'm like... wait, I know that guy, LOL.

I'm going to have to get one of those radiators pretty soon. I was looking at them not long ago at performance-pcs, but they were out of stock at the time. Just one would be better than my dual 360 setup, but I might just add one along with those and leave it on the floor like you did. I might even be able to mount both of the 360s on brackets attached to one of those.

 

Blessings my friends. May 2020 be filled with great and wonderful things for all of you. Can't believe 2019 is gone already.

 

Something everyone can relate to, including turdbook jockeys.

What's Your Current Favorite Mouse?

 

Happy New Year fam. I hope the new year brings lots of overclocking and broken records!

 

Happy New Year 2020 for all NBR Family.
http://forum.notebookreview.com/threads/happy-new-year-2020.831412/

 

Happy New Year all! I've been a part of this forum since Summer 2013 when I first modified my Alienware laptop with a non standard GTX 670MX, which was the point I became a fan of all things modified! Enjoy 2020, may we have some exciting hardware inbound!