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

http://hwbot.org/submission/3937447_

 

Cool video...

 

Ha, dunno man, the best bit of that vid was Steve getting continually blasted by LN2 vapour and trying to pretend it's not happening! I'm not really into this LN2 stuff, it's just too extreme to even be halfway practical, so it's just a novelty to me.

 

That's why it's awesome. Like top fuel dragsters and funny cars, and rocket-engine dragsters.

Here are some pix. I still think I am going to use the auxiliary pump, but the one I got from Amazon was a piece of trash. It leaked like a sieve and I had to return it for a refund. So, I will just suck it up and pay for a better one. That way I can just unplug the QDCs and bypass the radiator when I want the CPU extra cold and squeeze an extra 100-200MHz out of the CPU. As @Johnksss pointed out, the radiator actually warms the water closer toward ambient temps (although still much cooler than ambient).

Here are some pics. I got a couple of G.SKILL Turbulence II RAM coolers and robbed the fans off of them for the DIMM.2/RAM cooling and I have my Corsair Air Flow RAM cooler laying over the PCH heat sink and blowing air around and under that. That actually made a huge difference the the temperature (based on feel/touch) of the PCH heat sink. It used to get pretty warm to the touch and now it's like room temperature.

​

 

Well I like your setup there, it's extreme for sure, but is still quite practical I think (although might be a bit noisy), so I like your build there, looks good too - LN2 doesn't do it for me though!

Yes, I can see what you've both been saying about the radiator actually warming up the cooling water - the cooling water is lower than ambient, so it's gotta happen.

 

Thanks for the compliment. I think it would look better with the small fans not oriented in a diamond shape and installed parallel to one another, but I need them that way to blow air on both sides of the DIMM.2 cards. I did not realize they were LED. The product information did not say they were and did not have pictures of them lit up. I kind of wish they were not.

It's funny that the air getting blown out of the CPU radiator feels like air conditioning since the water it is warming is so cold.

I think I am going to replace the Kryosnaut with Conductosnaut and see if that works better at moving the heat from the IHS to the water block faster under load. I notice in Cinebench and wPrime 1024M the temps rise fast and then fall, like there is a bit of a lag in the Kryosnaut moving the heat or something. Will be interesting to see if it improves with Conductosnaut, and if so, how much.

But, the good thing is it is getting nowhere close to the dew point for normal benching like the 3DMark stuff. I only need to make in colder for stuff like wPrime and Cinebench. And, it's not loud at all now. I removed the Vardar Furious fans and have only the quiet fans on it now. And, the chiller makes very little sound. It's like one of the small refrigerators you would see in a hotel room. Just a faint hum when it is running.

Nice 3DMark 11 Physics score.

 

Hey, well that's great re the noise, it's actually quite quiet then. Yeah, maybe changing the thermal paste to the liquid metal would help - the temperature difference between the cooling water and the CPU die has probably been increased since you've been using chilled water, so the heat transfer is wanting to happen really fast, so I could imagine the thermal interface material being more critical/limiting in that scenario - you might see greater gains with liquid metal than you otherwise would have done with normal water cooling, that's my hunch based on my heat transfer knowledge.

 

https://www.3dmark.com/sd/5233239 | http://hwbot.org/submission/3937768_

 

Making some progress now that I can keep it cool with chilled water. Amazing what some cool water can do.


Global Rank...

United States League Rank...


But, some of these points are coming from the HP ZBook that I have been doing a little benching with. I took a few first and second place scores with the 4800MQ CPU.

 

http://hwbot.org/submission/3938409_


http://hwbot.org/submission/3938410_


https://www.3dmark.com/3dmv/5712394 | http://hwbot.org/submission/3938422_

 

How many radiators in this loop?

 

Two 360mm radiators and a 1/2 HP water chiller.

 

Do you happen to have a fan blowing the back of the cpu socket?

 

No I do not. It is actually two separate loops. The CPU and GPU are independently cooled, each on it's own loop.

 

http://hwbot.org/submission/3939630_


http://hwbot.org/submission/3939631_

 

5.1Ghz, good good! Don't be blowing up your CPU now! ;-)

 

Not blowing anything up yet.

Got the new pump and with the chiller connected directly to the CPU and no ambient temperature warming from the loop, temps are looking pretty sweet.


Got the CPU IHS surrounded with kneaded eraser just as an extra precaution for condensation. Also replaced the Kryosnaut with Conductosnaut. So far only seeing some minor evidence of frostiness, but watching it like a hawk and checking with a flashlight every few minutes until I get the dew point settings fully figured out.



http://hwbot.org/submission/3940145_

http://hwbot.org/submission/3940140_

@jaybee83 @D2 Ultima

 

Wow, fantastic temperatures, less than 10 degC effectively! How much difference did the metal Conductonaut make vs the traditional Kryonaut paste? Have you decided on a max core voltage you're gonna use at your new sub 10 degC temperatures, and how much more voltage is that in comparison to your previous allowable max?

 

I probably won't exceed 1.500V for brief benching. Maybe I will shoot for 5.2GHz at 1.500V next.

The Conductonaut seems to have brought the temps down about 6-10°C with the full loop. Not much change with the chiller pumping water onto the CPU without the loop... maybe 2-3°C.

I cannot bench the CPU at 5.1GHz with the loop involved because I cannot get it cold enough to be stable at 5.1GHz. At 5.1GHz Cinebench and wPrime 1024M just crashes with errors or closes abruptly if I have the water running through the loop. Due to the temperature reduction, my voltage requirement is about the same at 5.1GHz without the loop as what it needs for 5.0GHz with the loop.

And, man... these quick-disconnect fittings are so sweet. They make changing it back and forth a cinch. I can make the change between using the loop and direct CPU chilling in like 10-15 seconds with no drips or drops.

 

Well the metal Conductonaut made quite a difference, I thought it would be more of a difference than normal given the amount of heat that wants to be transferred now with your chilled water, but didn't expect such a big difference! Yes, 1.5V is traditionally a dangerous voltage for CPUs, with some resources I read stating that degradation can occur over a matter of minutes & hours at that voltage, but perhaps at your low low temperatures then this effect is slowed, but I think you're right to limit it to only short benches. So your chilled water loop might give you an extra 200Mhz in overclock if you end up getting 5.2Ghz stable.

 

id call that pretty nice scaling at 3.95x CB15 points with 4x the cores when comparing that with my best score *lol*

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

 

https://imgur.com/a/hUzv7df

Just waiting on my RTX 2080 Ti FTW to be delivered from EVGA now. Ugh first batch order but damn I doubt with these delays it will ever arrive.

 

https://www.3dmark.com/3dm/28707122?

Still don't have access to higher power limits because EVGA is slow with their Precision X1 beta releasing today to RTX cards.

Edit:

12,100 is near FE 2080 Ti stock graphics score (100 points shy) I'll take it.

 

http://hwbot.org/submission/3940735_


http://hwbot.org/submission/3940736_

 

You can use a beta version that Guru3d used:
https://www.guru3d.com/articles_pages/msi_geforce_rtx_2080_ti_duke_review,28.html

And here's the direct download link copy & pasted from that page above:
https://www.guru3d.com/files-details/msi-afterburner-beta-download.html

 

Newest Beta is working. Version released yesterday was bugged with an LED firmware update issue. They have fixed it today with a new version. All good now!

 

EVGA just blessed me with a 130% power limit vBIOS.

https://www.3dmark.com/3dm/28733672?

Faster than a stock 2080 Ti FE. This was a first run, let me get to work and see what I can come up with.

 

http://hwbot.org/submission/3941528_


http://hwbot.org/submission/3941529_


And, with cache bumped to 32x...

http://hwbot.org/submission/3941546_

 

Fire Strike seems to not really love the RTX series nearly as much, but still a good showing and well top result with RTX 2080 and 8700K.

https://www.3dmark.com/search#/?mod...or&gpuName=NVIDIA GeForce RTX 2080&gpuCount=1

https://www.3dmark.com/fs/16442977

 

Not bad. How overclockable is it? How does it do with Time Spy?

 

https://www.3dmark.com/spy/4471027# | http://hwbot.org/submission/3942052_

 

See post above #3253

 

So, 13% higher Timespy Graphics score than stock Founders Edition RTX 2080: 12444 vs 11002. Stock Founders score taken from here: https://www.guru3d.com/articles_pages/geforce_rtx_2080_founders_review,33.html

That's not a bad increase (13%), because Founders Edition is already overclocked by nearly 100Mhz as part of it's specs on the boost clock when you buy it.

 

That does not answer my question. I want to know how well and how far it overclocks on core and memory, or if it does not. Stock performance is nice, and so is an optimized vBIOS, but not enough of a reason to buy on fave value. For $1,200 I would expect/require some insane overclocking abilities to match the insane price. Wonder how much lube we will need for the Kingpin edition?

 

I’m not sure I’m able to follow you full on this. Isn't the $1200 graphics the 2080Ti ? The tested graphics is the cheaper RTX 2080. He have probably not found max overclock yet, if this was what you meant.

Edit. https://www.3dmark.com/search#/?mode=advanced&url=/proxycon/ajax/search2/cpugpu/spy/P/2283/1208/500000?minScore=0&minScore=10&gpuName=NVIDIA GeForce RTX 2080&gpuCount=1&cpuName=Intel Core i7-8700K Processor

 

Yes, but they should have similar overclocking capacity, just the Ti would do a little better and have more Cuda cores. Does anyone here actually have a 2080 Ti yet?

Sorry, I am traveling and trying to read this stuff on a smartphone, which always sucks massively.

Is 3DMark actually reading correctly for a change? What is the normal max boost clock, or are these the stock boost and memory clocks?

Core clock: 2,130 MHz
Memory bus clock: 8,200 MHz

 

Mr. Fox is not the only one who got confused. He did write Ti at first:

Now all of a sudden having matched the Ti FE is a lot more impressive.

 

Sorry about the confusion bud. This is my EVGA 2080 XC card that I got first. It's not going to be mine permanently and will be going to my brother instead once my FTW RTX 2080 Ti arrives. I got in on the first preorder with those direct from EVGA and was told it should ship around last week or first week of October, but that was pre-delay from Nvidia.

For now I have this RTX 2080 (non-Ti) and its an excellent card. In all honestly I am beginning to have doubts about my need or desire for the $1300 FTW Ti version. I really haven't had as much time for gaming lately, and this card more than powers through any game I have at high/ultra details on 1440p.

As for the overclocks it seems around 2130Mhz on the core are about my limits with current access to voltage or lack there of. I have a 130% power limit vBIOS from EVGA and that certainly helped but still Nvidia has us locked out of voltage for the most part. Temps are excellent so I'm not worried there. The memory clocks are reporting correct. I hit a wall at 8250Mhz. 8300Mhz it crashes.

 

1800 MHz boost clocks and 14 Gbps GDDR6 memory for the factory oc'd RTX 2080FE model. 2130MHz and above 16 GB for memory so this is in 2080 Ti territory.

It's claimed that Turing should keep higher or prolonged boost with overclock vs. Pascal. Aka better overclocking but we have to see when the results ticking in.

Ëdit... One of the better
MSI GeForce RTX 2080 Ti DUKE - Overclocking with OC Scanner



Stock GeForce RTX 2080 Ti Founders Edition


Stock RTX 2080 Founders Edition



Here’s How Much NVIDIA RTX MSRP Will Be Affected By Tariffs Wccftech.com

Example: MSI’s NVIDIA RTX Lineup New Pricing After Trump Tariff
Here is one of the lists we received with post-tariff pricing going live very soon (if not already):

• RTX 2080 Ti Gaming X Trio: $1231 to $1310
• RTX 2080 Ti DUKE 11G OC: $1212 to $1290
• RTX 2080 Ti VENTUS 11G OC: $1203 to $1280
• RTX 2080 Gaming X Trio: $849 to $900
• RTX 2080 DUKE 8G OC: $840 to $890
• RTX 2080 VENTUS 8G OC: $830 to $880

 

http://forum.notebookreview.com/index.php?posts/10798805

 

Those after-tariff prices are too low, unless the seller is absorbing a portion of the 10% tariff, you'd multiply 1.1x price to get new post-tariff cost then add local tax for total cost - not sure if you'd pay tax only on base $ cost or entire base+tariff $. Tax for CA, USA 6.0% + 3.25% +.25% = 9% for Los Angeles, for example - on entire cost or pre-tariff cost, not sure:

• RTX 2080 Ti Gaming X Trio: $1231 to $1310 => $1353 + tax
• RTX 2080 Ti DUKE 11G OC: $1212 to $1290 => $1333
• RTX 2080 Ti VENTUS 11G OC: $1203 to $1280 => $1323
• RTX 2080 Gaming X Trio: $849 to $900 => $934
• RTX 2080 DUKE 8G OC: $840 to $890 => $924
• RTX 2080 VENTUS 8G OC: $830 to $880 => $913

And, after Jan 1 2019, tariff's go up to 25%:

• RTX 2080 Ti Gaming X Trio: $1231 to $1310 => $1539 + tax
• RTX 2080 Ti DUKE 11G OC: $1212 to $1290 => $1515
• RTX 2080 Ti VENTUS 11G OC: $1203 to $1280 => $1504
• RTX 2080 Gaming X Trio: $849 to $900 => $1061
• RTX 2080 DUKE 8G OC: $840 to $890 => $1050
• RTX 2080 VENTUS 8G OC: $830 to $880 => $1100

It's not going to be clear if prices are inflated due to tariff's or if they are jacked up to look like tariff's - once stock is depleted for items already shipped the prices would go up, not immediately - and when a product source changes from China to another country the price should drop, but may not.

European prices may be affected as well, but they are usually already jacked up via VAT and other local tariff's + taxes, so hopefully not.

Nvidia Thread
http://forum.notebookreview.com/threads/nvidia-thread.806608/page-76#post-10798737

 

They will absorb & eventually avoid some of the cost:
https://www.notebookcheck.net/NVIDI...ariffs-will-affect-their-prices.334262.0.html

 

Not at all, reportedly AIB's can't afford to absorb even $1 of the tariff's, each $1 absorbed is $1 off the bottom line, which can't be allowed.

The AIB's complaint for so long is that the AIB partners have a very slim margin, Nvidia allows only slim margin's, and so AIB's can't afford any discounts, so this is a nightmare for AIB's... or is it another golden opportunity so soon after losing the crypto-suckers?

In fact MSI is bumping their prices up higher than what I calculated as the 10% increase for their 2080ti Trio, grabbing a convenient "handling fee" - a nice taste - for themselves:

" For a start, it appears that the tariffs will cause MSI to announce relatively high pricing for this new line. The RTX 2080 VENTUS 8GB OC'd product will be as much as US$880 after the tariffs. This increases to up to $1290 for the RTX 2080 Ti DUKE 11GB OC'd model, with an upper limit of $1390 for the RTX 2080 Ti Gaming X Trio card."

• RTX 2080 Ti Gaming X Trio: $1231 to $1310 => $1353 + tax

So MSI is already adding a nice "bump" to pocket of $1390 - $1353 = $37, which doesn't seem like a lot until you consider the entire "bump" including the tariff is $1390 - $1231 = $159... wow.

And, what kind of "bump" will MSI add when the tariff's go from 10% to 25%? 2.5x $37 = $92, which adds up with the real tariff to a total "take" of $397.

It's gonna be a giant greed fest - with every step along the way adding a bump for themselves, followed by nobody with half a wit buying **** all... why it's not already so is a total mystery to me.

 

Hey, I'm just stating what they say in the notebookcheck article, take it up with them if you don't agree with it. I think what they say in the article about eventual absorption & avoidance of tax costs makes sense. Ha, but RTX launch could definitely do without further price increases due to the tax situation!

 

@Talon - here is a ray tracing benchmark if you want to give it a try...

https://drive.google.com/uc?id=1UweHpbzEvoKpFGtA1kg3X-nyRTxGoWU9&export=download

Futuremark will be releasing their ray tracing benchmark soon.

 

http://www.3dmark.com/spy/4489837

 

cmon john, only 98th percentile? shame on you xD

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

 

Congrats on the RTX 2080.... I think! (Ha, a lot of negative press about this new architecture) What are your impressions on it from an overclocking point of view?

 

Yeah, huh.
https://www.3dmark.com/search#/?mod...=0&gpuName=NVIDIA GeForce RTX 2080&gpuCount=1

As far as I can tell, it's faster than a 1080 TI at some things, but certainly not all. I would not go from a 1080 TI to this card if I were in the market. This card is like maybe a hand full of frames faster in some case's and and that goes for being slower. A 2080 TI is a different story.

 

Will do when I get home Thursday morning.