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

I'll have to ship it soon...

Hence the benching session...

 

OK, let's take it one step further to prove the theory. Add 1709 into the comparison and it becomes ever the more clear that older is better where Windows performance is concerned.

https://www.3dmark.com/compare/fs/19456810/fs/19456840/fs/19463809

 

And if you added in 1803 into this comparison it would come second best

Btw. Tested the Guinea pig patch update for 1903?

 

Nope. What is that patch supposed to accomplish? If it is for "security" then I have no interest in testing it. I don't care about that nonsense. If it is for "performance" then maybe.

 

I haven't installed this OS version. But if the OS Version who is newly sent out... Isn't as expected, maybe they have added in hided fixes equal Dell prefer. Seee... Hided fixes for burning down Area-51m

https://www.deskmodder.de/blog/2019...523-als-ssu-kb4502374-als-dynamisches-update/

 

OK, it seems those updates actually helped improve 1903. I will test Fire Strike and see if the score on that has improved accordingly.

 

Very interesting. Nice find...

 

Well, every little bit helps. It seems to have helped Fire Strike a little bit as well. Not huge, but at least it was a modest improvement rather than degrading. You never know what to expect with these updates. They frequently and more often than not impair performance, so it's good that seems to not be the case. As we can see here, the CPU performance improved just a tad, which is also what Cinebench, wPrime and CPU-Z benchmark suggested.

https://www.3dmark.com/compare/fs/19456810/fs/19464656/fs/19464726

 

Yeah, I expected so. Totally new version out has to be tuned honed by Micro$lope. Remember The Redmond Morons have struggled with put out stable and good released OS this year, so expect some minor changes from them who won't come in the change log. Good see an small improvement. Thanks for your testing bro Fox

 

It is a lot of fun to run Catzilla, but it's an old and buggy benchmark that isn't supported. You need to use these older drivers to make the errors go away because it hasn't been updated in like 3 years to work with newer drivers.

Download | Catzilla Compatible Drivers (Intel RST and Chipset)

You cannot submit the scores to HWBOT on the Catzilla web page. That has been broken for like 2 years or longer. Copy the URL after uploading your results, then post the URL with your screen shot on HWBOT.

Here is a 1080p Catzilla with the 2080 Ti running stock...

https://www.catzilla.com/showresult?lp=1012688

 

OK, well, 1709 gets closer to my Windows 7 CPU performance of any other Windows 10 version. And, 1709 is the oldest Windows 10 version with GeForce driver support for RTX cards, and that includes no ray tracing or DLSS functionality.

Kill-a-Watt reading was 1035W during these tests.

 

Windows 10 Enterprise LTSC

You already know you can't attempt to beat yourself right.

 

If that still doesn't work for you, downgrade to these chipset drivers and then you should be good to run Catzilla.

https://station-drivers.com/downloa...el_inf_10.1.1.38(www.station-drivers.com).exe

 

I knew it the moment I posted it.
That would get your ass off the couch and deliver on that 35K that you let us beg for for weeks...just need to fix that 13K on FSU.

I am literally so happy now. That totally made my day!

Now give us that 20K TS already!

 

https://www.catzilla.com/showresult?lp=1012706 | https://hwbot.org/submission/4166316_

 

Thanks for the files Brother Fox.

 

Here the 35K on the 1080s:

https://www.3dmark.com/fs/19468898

I know, I'll fix the 48K on the GPUs later, if the 780W AC Adapter let's me

 

Glad to have made your day!
Second card is actually garbage(And i really do hate using that as an excuse!) So it makes running dual cards a hit or miss situation.

I know you will!

 

KPx block out now. Ordered two and now I need a bit more compression fittings and quick disconnects.

 

I was just about to post that they were available. In for 2x as well... what size fittings are you using? I'm debating if I want to change my current ones all around or just pick up some more of what I have now... hmmm...

 

35K / 48K Club Membership achieved:

https://www.3dmark.com/fs/19472917

Now it needs some balancing act...

 

Okay, here is one a little bit better.
https://www.3dmark.com/3dm/36520425

 

Yeah, just a little bit better....

 

I love what you do when given a reason to!

So sad that no one is really benching Notebooks these days and we don't have anyone to play with outside our own team.
The bot is totally dead. They even created a sub-section for Max-Q with not a single upload.
I wonder if they would allow us to bench against Desktop cards. That way we would at least get some points.

Yeah, no chance to beat that with 1080s.

Somewhere around 35.5K should be their max region...

 

I was hoping that computex would bring something competent to the DTR market. I do want back into DTR for OC'ing but there's nothing to lure me back in. I thought the newly announced MSI GT76 with its heatpipes aplenty would be a candidate but alas, BGA.

 

Yeah, I was pretty excited by what I saw till they took what could have been a near perfect system and just pigeon dropped us with mouths wide open....

 

Not sure if Svet could remove MSI’s cancer firmware and battery boost. @Falkentyne

 

All Svet does is unlocks all of the BIOS' options. The cancerous firmware/EC remains, unfortunately. Akin to having a 800hp car with a governor that won't let you go above 80mph but getting it chipped to allow you to see all of the cool options you can't truly take advantage of. I admit the GT76 has one of the most interesting cooling setups I've seen in a laptop in a long time.

 

Whether he can or not is unknown, but he will not.

That is correct. He either cannot or will not (I am not sure which) mod away the cripple cancer feces like Brother @Prema does. He may have some kind of agreement with MSI where he doesn't mess with their cancer so their turdbooks can (mal)function as MSI intends for them to.

It's also worth noting that @Prema worked tirelessly trying to eliminate all of the MSI cancer on the 16L13. While there was dramatic improvement over the mess that MSI produced, it never functioned as well as a Clevo in terms of throttle-free overclocking. There was always a bit of gimping that lingered. Might have been a hardware engineering defect.

Yes, that's an impressive looking cooling system. It's unfortunate that it is being wasted on BGA trash.

 

Fake SLI vs Real NVlink....Now what is wrong with this picture..
https://www.3dmark.com/compare/fs/19479320/fs/16495203

https://www.3dmark.com/compare/fs/19479320/fs/16495203/fs/16465974/fs/19069609

 

Looks like the fake SLI totally screws up the CPU performance. Unless the laptop is overheating and malfunctioning, there's no way the desktop 9900K running at 4.9GHz should be stomping on the Clevo's nut sack like that with it running its 9900K at 5.2GHz.

It also looks like 3DMark is doing some manipulations to alter graphics scores based on platform, but that's nothing new. They seem to have been playing dirty pool like that for a few years now. You're not going to get "real" scores with their broken POS 3DMark Time Spy or Fire Strike, unless you're running hardware and an OS version that they want to have real scores. Everything else is artificially broken or manipulated to look the way they want (or get paid) for it to look.

 

Only talking about GPU scores and how its right up there with other NVlink 2080's.(Now there are plenty of benchmarks where it does not work well at all. Like Catzilla) Going to have to get back at it and see about beating his gpu score since i'm pretty close to it now.

 

Probably because the benchmark algorithm just sees "2080 SLI" and gives it a good score instead of capping the FPS to suppress the score. If it saw "2080 Ti SLI" it would probably uncap it a little bit more based on what it thinks the installed hardware is.

 

Not sure. All firestrike and most of the others work as they are suppose to while Night Raid has like a flickering screen or an overlay but runs. Not like too much voltage, but more like mix matched cards trying to work together./ all catzilla - runs like single card. Port Royal has a flashing screen, but you can see it running...sort of. haha.

Edit: Better explanation is every other frame is jacked up.

 

I have the most trouble with Time Spy, VR Mark and Port Royal. I can run 3DMark 11, Vantage, Fire Strike, Sky Diver and play games with the memory cranked to 1400+ but those DX12 and/or RTX POS benchmarks are totally effed up and misbehave unless I drop the memory down to like 1200-1300. They won't even finish test 1 without their dumba$$ "Ooops an error occurred" error nonsense. Sometimes won't even START test 1. Hardware monitoring makes those problems worse. Can't think of any reason they really "need" to monitor my hardware or anyone else's hardware. None of their damned business. They just need to crunch the numbers, do the math, save the results in their database, and shut the hell up. But, they're probably getting paid for the manipulation and policing scores to keep the "unfit" or obsolete products from winning.

 

Timespy seems to sensitive to memory video memory overclocks. VR Mark is up in the air for me. Top being 16275 on Orange And that goes for Port Royal as well. Solid 10.3's but that seems to be about it for now.

Edit: Not so sure about that. There are plenty of old products kicking butt and taking names.

 

I've got no proof, but I'm pretty sure they're using secret cancer and hidden frame rate limiters to keep everyone in check, and so they can control what the rest of the world sees in terms of benchmark scores. Can't have older products and obsolete Windows versions playing vicious rump ranger and ripping heads off the new trash, cause that's bad for business. Just make it crash if it is too good, or just block it so it can't run, then they don't have to explain any weird numbers that don't make sense to stupid people.

Edit: real benchmarks calculate numbers. Fake benchmark manipulate what is allowed to run and what's not. Can't run Time Spy, Night Raid or Port Royal, or Blue Room, with certain versions of Windows or without certain hardware. Because they have pre-decided the outcome. Can't respect that. It's not worthy of any respect.

 

Well, you got me there. Don't know what to say to that.

 

Yep, I saw the same thing with my GTX 1070 - Timespy is most sensitive to VRAM overclocks, and that's what I used to tune my max stable VRAM overlock (Timespy Stress Test), kind of like a worst case scenario.

 

I'm going to completely disassemble my X299 rig tonight, which includes a LM repaste of my 7980XE, in preparation of the KPx blocks. I'm curious if EVGA used KPX thermal paste on the KPx cards.

 

I just view it as being an unstable product that isn't suitable as a tuning tool. Using it to establish your max stable overclock would be kind of unfortunate, like being limited by the lowest common denominator, when that max stable overclock is higher with everything else that isn't nearly as buggy as Time Spy. It's not just a little bit unstable, but actually quite a lot when a benchmark won't even launch correctly with a memory overclock that runs flawlessly on virtually everything else. You may have noticed that I seldom even make an effort to bench with it because it's a disappointing pile of trash benchmark. I don't waste my time with it very often. Almost everything that uses DX12 is an unstable piece of crap, so it's not just Time Spy (although it's definitely one of the worst examples). The Redmond Retards have had 4 years to iron the kinks out of DX12 and make it a relevant and desirable API... must not be important to them, or they're just too stupid to know how to fix it. This is why most of the AAA game titles that make DX12 available also have a way to disable it. It's not even stable enough to use for playing games at stock clocks more than half the time.

 

Yes, i'm inclined to agree.

@Rage Set
The strips covering them says iitek

@Mr. Fox
You're looking at it as benching. You would more than likely need to use only their stress testing tool. Which takes the best over 20 runs.

 

But, the end result would only be relevant for Time Spy, and maybe some other similarly buggy DX12 garbage. If you use Time Spy as the standard, then you'd be leaving performance on the table by thinking your "max stable" overclock is limited to what Time Spy and its bugs can tolerate when the max stable overclock for other things is potentially much higher. To do this effectively, you need to establish a "max stable" overclock independently for each benchmark or game title, because one size doesn't fit all sometimes.

 

Before I go overkill. I normally use dual D5 pumps (on all of my client builds) when I watercool two (or more) GPU's in SLI. Do you think I should continue that thought process or go with a single D5 pump? The rad is going to be the single MO-RA 360 Pro (equals three 360 rads) and the two KPx blocks. @Mr. Fox I know you have more experience with Hybro Copper blocks. How restrictive are they compared to other GPU blocks? I do know you use two XSPC D5 pump/res combos, but have you tested your loop with only one? I'm probably over thinking this.

 

Okay, let me say it this way.... They have 8 different stress test. So that covers the one size does not fit all. I would also not rely on just one benchmark to as a decision maker in what my card is capable of. Since these benchmarks test different aspects of any part. Although I have found that unstable ram and to high vram over clocks tend to crash Time Spy. Buggy or not.

And as you know, i put everything back to stock once i'm finished. I do not need to overclock to surf the net or play a game, but that's just me.

 

Some of us should keep an eye or two on the development of these products. If they are as good as he states, this could be another cooling weapon.

 

I met 8 Pack also at the Las Vegas Show a while back and he had lot interesting things to say about over clocking in general. A real cool Dude.

Other news....

 

Let me know how those pads work out temp wise. I have several packs of Fujipoly Ultra on hand but if they don't make a difference (with your chiller, won't make any difference), I will just use what is provided.

 

I too have Fujipoly on hand, but the first go round i'm just going to see how well there's performs.

 

Wow what a past few weeks it has been..busy busy... Hope all has been well fellas.

The KP Hydro Coppers should be arriving this week. @Rage Set I've found that the stock pads work pretty well with no noticeable difference vs the fujipoly's when running the card really cool according to what I saw from the sensor read outs.

@Johnksss looking nice...all greased up.

@Mr. Fox ..finally most of the fam has gone back home lol.. it has been chaos the past week or so, but fun accommodating everyone. My poor chiller and goodies have been very patient lol ... now time to go to work...