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

@electrosoft

I resolved the power shutdowns with my PC. I use a 1200 watt Seasonic prime platinum PSU and I was running an XOC 2K watt bios on my 2080Ti. And during gameplay under heavy loads, it would literally power off an restart my PC. Even though I was only using like 800-900 at the wall.

So, I also run a XOC bios on my X299 dark lol. I know, I’m all about every drop of performance.

Anyways, I flipped my motherboard bios switch to Bios #1 which is the newest. And I no longer have power shutdowns.

So problem solved! I emailed Seasonic and they made this suggestion to try a newer motherboard bios. Sure enough, it is some sort of issue with running both XOC bios together?


Have you tried updating your motherboards bios? It resolved my issue 100%. And my PC pulls crazy amounts of power.

I also ran the OCCT PSU stress test. This managed like 1,013+ watts of power for over 20 minutes. No issues. I stopped the test.

Unfortunately the standard evga motherboard bios really suck. And my CPU idles 5C hotter running 200Mhz slower. Makes no sense at all. I guess you can’t have it all...

 

Hmmmm, I don't know if that resolved the issue. Seems like it dialed back the performance. What happens if you run the XOC on the motherboard but run a stock vBios on the 2080ti?

As for me, I'm on the newest BIOS for the FTW Z390, but the problem is the lower M.2 slot is fried and the board will rev up and down sometimes like it is trying to turn over along with the CPU fans going bonkers (only fans connected, as I'm running a pseudo open bench) for no reason now. This is across two PSUs (My defective EVGA T2 850w, Buddy's 1300w). So it will be going in for repair/swap as soon as EVGA gets off their keisters and replies to even my primary help ticket to give me a # to issue an RMA for my PSU as I'm going to also have to get the MB RMA'd too.

Basically my desktop is a hot mess atm. I even yanked my 3090 and have a 1060 in there atm till I square away the PSU and MB as it ran like butter on my buddy's rig. It's torture not playing WoW during launch week but I'm not going to run around on a gimpy desktop with a 1060 @ 4k. Sad thing is even when I did load up WoW with the 1060 either the MB or PSU is emitting a very high pitch whine now that is slow nails to the brain annoying.

Everything was running GREAT for a a few weeks then I heard that mild "pop" from my PSU and it's been all downhill since then....

I'll have something sorted out in a few days (I have one more 6700k system to rebuild/tweak) as I have this small pile of "Pimp My Mess" waiting for me tomorrow:



Budget MB that got great reviews for what it is and I went mid line budget on the PSU and used the last of my 10% discount codes for the i9 I got at MSRP since I haven't mastered the art of crapping gold (yet).

 

If I run the XOC X299 Dark motherboard bios, and run my normal Galax 380 watt 2080Ti bios with shunts I don’t have any issues at all. “Never have” The GPU still draws insane power too. Just limited to much lower voltage and the power draw is within reason around 400+ watts.

But I could run my system totally stock and run the Galaxy 2K XOC 2080Ti bios, and I would still get power shut downs. “While using XOC motherboard bios”

I also plugged in the supplemental PCI-E (6) pin power connector for the motherboard at the bottom near the pci-e slots. “Usually only recommended for multi GPU setups” my single 2080Ti does pull power like (2) I guess.

Performance is still good. I broke 17,500 in timespy graphics. And 10,600 in Port Royal. Those are outlandish numbers for a 2080Ti.


Power draw is wild on this card. I’ve exceeded 545 watts in Timespy extreme with just the 2080Ti. I prefer the XOC 2Kw bios due to having 1.125V and I can run 2,175Mhz and it’ll sustain 2,160. Which is the edge I need to be competitive and match modern day GPU performance.

So it’s running great so far. My 2080Ti has both resistors soldered and stacked and my reference 2080Ti has a power ceiling of 2,800 watts. More than enough lol.

 

I'd like to see your system run with a 1600w PSU full throttle and see where that goes. You're entering that kind of territory.

 

It seems like I would use 1,200 watts or even more, but I don’t. If I fully loaded my CPU to 100% load, and my GPU to 100% load then sure I could easily exceed my 1200 watt power supply. But in reality there isn’t a single application that I use that pushes them both simultaneously like that. It’s almost always one or the other that is using some high wattage.

From what I understand you can safely pull 1,300+ watts or maybe even more from these 1,200 watt Seasonics platinums for hours on end without any issues at all. I almost always stay under 800-900 watt range. And I’d have to intentionally try to overload it with something like R20 and furmark running at the same time to exceed the 1,200 mark.

 

Even if you did, you still have to factor efficiency

 

This is the supplement 6 pin power connector that I plugged in. I removed my entire PSU and checked all of the connections too. I am now running both XOC bios without power shutdowns. I dunno if maybe the 6 pin fixed it?? Or just checking all of my connections and moving them around to new places. But I’m solid with the X299 Dark XOC bios 7980XE at 4.8Ghz/ Mesh 3.2Ghz, and the 2080Ti 2KW XOC bios too.


Who knows? Maybe I had a bad connection, crimped wire. Or maybe that 6 pin is really required at a certain power draw?

@Mr. Fox what do you think about this 6 pin? Did you use it? My system is totally solid now.

 

Lapped my ihs and water block. You can see I had a lump on the side of my CPU causing it to be lifted on one side. My core to core temps were so bad.

 

Okay, so I got my Ultra Sonic Cleaner the other day and took it for a test spin. I must say I should have bought this a long time ago!
Before


After
I'll try to take better pictures of the after as it's hard to see, but it's completely clean Vaseline is totally gone.

I Sat it in there for like 25 minutes and wipe it once with a toothbrush and let it sit for a few more minutes. Took it out and dried it and it looks nothing like it did before.

 

Nice, I can see what you mean by the uneven IHS. That surface looks a bit scratched though considering you lapped it, is it worth using a finer grain on it to get it more polished? How are you temps now after the lapping?

 

The picture was in the beginning of the lapping process, I worked with a higher grit for quite a while afterwards to make it a little smoother. But I just did a pretty quick job on it for the most part. Temps are much better though. I knew my liquid metal application was still good underneath the IHS. Because my temps were amazing before I went full custom loop. I have delidded this CPU twice, and I have applied liquid metal twice. So, the LM is solid and good for long term underneath the IHS.

Core temps are 54C-65C now during looped R15 or R20 runs, with that odd ball core hitting 71C maximum.

The temps are quite good. But I am going to lap it some more after seeing such amazing results. I am shocked that lapping the CPU IHS and waterblock cold plate has helped so much. I am just glad that my assumption was right, and my CPU was extremely uneven.

 

I would be afraid to do this on a 2080Ti. Didn’t Nvidia glue/epoxy the memory modules closest to the pci-e slot? I guess if it has the glue it’s ok, but if not some cards were easily affected by the space invader artifacts?

Either way, it looks amazing!! No one will know the punishment you put it through. Hitting like 18K in timespy graphics lol.

Are you selling this 2080Ti?

 

Afraid to do which part? The card is covered in vaseline to protect from condensation or water from loop, but when it's time to put back stock for rma or for sale. One needs to make sure the board is clean. This took way less time and looks cleaner.

If asked I would tell people it was over clocked. Most want a guaranteed card, not a toss up. Plus the card has been power modded and bios flashed.

 

I do not doubt that card for one minute, if I were buying a used 2080Ti, I would want one that was stressed and tested to its fullest. I was only referring to how sensitive the memory modules are on these 2080Ti's. Nvidia put epoxy on them to fix the space invader issue? And, I was thinking maybe the ultra sonic cleaner could dissolve this. I dunno, maybe maybe not. Just a thought.

My 2080Ti has expoxy on the memory modules closest to the PCI-e slot. I had a Gigabyte 2080Ti that didnt have the epoxy. " It had space invader artifacts and major issues."

 

If I had to guess....I would assume they use epoxy to bond the transducers for the Ultrasonic Cleaner. And considering a few overclockers use it to clean all forms of pcb's I'm going with it should be safe, given one knows what they are doing. There are a few ways one could go horribly wrong with using one.

Edit:
So I'm running the 2080 TI and it's getting 135 FPS with the exact same settings as the 3090.

Anyone care to answer how that is possible?

 

I could use a ultra sonic cleaner. My last Clevo MXM GTX1080 was hideous. I used K5 pro for thermal pads, and liquid metal stains, thermal paste stains, and it’s near impossible to clean that stuff up by hand “K5 Pro”. Even with alcohol and small brushes. I suppose the only hope was using a ultrasonic cleaner. So I can see how useful one of these machine can be.

 

hummm, not sure about liquid metal...Let me test the theory a little later today.

 

Well all is well that ends well....

I spent a good chunk of time yesterday with my dremel gutting my Corsair case even further since I never use 2.5" drives anymore and I hadn't touched my DVDRW drive in over a year so I wanted all the extra gunk gone. I also wanted my cables
routed such that removing and installing MBs is literally a 5 minute job which it is now.

I guess you get what you paid for. I went uber cheap with the MSI z490-Pro ($120), and it wouldn't boot with the debug light showing "bad or not detected CPU." Under my jeweler's light pins looked fine so the problem was elsewhere.
Usually it's always the motherboard in these situations though. It was missing clamps and screws too.

Returned it to best buy that didn't have any in stock and the only two MBs they had in stock was an MSI Z490 Gaming and an Asus Strix z490. After the CPU issue with the previous MSI I broke down and went with the Asus (Did I mention I haven't played WoW since launch?).
Figured if it didn't boot with a bad CPU debug light, it must be the CPU but of course it booted right up. Spent the morning getting everything all hunky dory with the system build out before heading out.

Spent the afternoon with my buddy to have access to lots of his hardware and to sort things out:

A> Definitely the Z390 FTW is dunzo on all levels. Still waiting to hear back from EVGA. Starting to think it was the problem all along. From the M.2 slot shot to the fans running wild everywhere for no reason that was the tip of the iceberg.
B> I didn't realize we tested my 850w T2 on his i9-10980XE not a i9-10900k. No wonder it sputtered hard as he was (trying) pulling 1200w+. PSU actually seems potentially ok. I'm sticking with the HX1200 though. It is ridiculously quiet.

It's funny with the 10900k because I had just finished working with a 6700k + gigabyte and even OC'd to 4.6-4.7ghz (after I delidded it), temps were ~49-52 under full CB20 loop load with a single fan that barely moved on the NH-D15. Then I turn around and set AI on this Asus Z490, I'm told I get an extreme OC of ~45% and watch my dual fan Noctua full blast NH-D15 crash and burn under CB20 It finally hit a wall...HARD @ 5.3ghz all cores. I'll end up fine tuning it manually down the road.

I installed the beta BAR enabled BIOS. Now waiting on Nvidia...

 

@tps3443
So trying to game with fans and 2000W vBios is not the way to go as it keeps crashing just like how your card was.
I found that I had to use nvidia-smi to set the max power limit and then lower the voltage alot, for the card to want to run normal as it is now. And it still beats my 3090 on this board.

 

Mine doesn’t shut down anymore. I plugged in the 6 pin power cable at the bottom of my X299 dark below the PCI-e slots. And I’m solid! I run the XOC 2KW bios daily at full 1.125V. I just hate that I can’t save profiles with MSI AB. I literally plugged in the 6 pin, Moved around some wires, changed bios back and forward on my motherboard. And I am 100% solid. I dunno what fixed it to be honest. But it works at full power.

So what do you mean by beats 3090, with the X79 setup lol?

Due to not being able to save profiles I am going to try the kingpin XOC bios it’s locked at 1.125V but has MSI AB profiles working.

 

Here are some scores I achieved on my desktop. I ran these tests the day before my daughter graciously shared her COVID-19 with me. Family.

Overclocked 3950X to 4.2 gHz via the BIOS; used TurboVcore to boost all core to 4.3 gHz. For daily use, I have the 3950X overclocked to 4.1 gHz all core.





Manual fan curve:



Default fan curve



All in all, not record breakers. But the process showed me what I could do with a capable system and tools not too far beyond what I could understand. Looking forward, I would try to get the GPU under a water block. That would allow me to run the card close to the power limit but without the frequency penalty for temperature (about 15 mHz drop in clock every 10 degree rise above 600 --> 1995 mHz vs 1920 mHz

 

Here is my RTX2080Ti. These cards do so well with power, and low temperatures. I am only running my 7980XE at only 4.7Ghz. I am totally solid at 4.8Ghz but it hurts the water temp in my loop. And the extra 100Mhz isn't worth it.

 

Hummm, maybe I should plug in my 4 pin adapter as I don't have that plugged into the motherboard. I was thinking that didn't have anything to do with it since its only 1 card, but maybe your right. I'll give it a go.

My 3090 is getting 95 FPS in Cold War Multiplayer, while my 2080 ti is getting 135 FPS with the exact same settings.

 

Try it. This is the first time I have ever used the bottom feeder 6 pin on this X299 dark. I literally plugged it in out of boredom, and sure enough it fixed my issues. My thinking was that the card is pulling juice for two lol.

 

@GrandesBollas

Be well!

 

@Johnksss

Did you get the XOC 2KW bios game stable?

 

Alright guys. I’ve got my custom loop in place. I really want to run a water chiller now, so
I’m revisiting the subject. How much power do I need though? Is 1/10th HP enough? Do I need 1/4 HP?


I need to work on my core temp deviation some more too. Should I get a new IHS? Or lap Cold plate and IHS some more? My hottest core is 18-21C hotter than my coolest core. This is at 4.6Ghz. Any ideas?? I didn’t lap my entire IHS to bare copper. I only removed the major high spots.

 

Check this out guys. I ran the Metro benchmark, and I set my settings identical to there settings used in a RTX3080 review. "1440P / Ultra / DX12 / Hair works OFF"

My 2080Ti reflects a 37% performance boost over their stock RTX2080Ti that they were testing.

I am still a little shy from matching or beating the RTX3080. But, wow! I am literally right there!



free pic hosting

 

We all have our own preferences. I do the extreme end of water cooling, having two MO-RA3 radiators and a crap ton of fans. @Mr. Fox is radiator cooling for everyday usage and water chilling for world records. Johnksss is at the extreme end of exotic cooling.

I was going to purchase a water chiller, the same one that Mr. Fox has but it is sold out everywhere. For your CPU and mine as well, we need the 1/2HP water chiller for benching but if you're not going to push your CPU like Mr. Fox does, the https://www.performance-pcs.com/wat...tt-cooling-capacity-waterchiller-hc-300a.html will work.

I purchased a copper IHS for my 7980XE and it worked very well for me. I do know some people don't like them. I also don't reseal my IHS with adhesive.

 

I want to run a water chiller for daily usage, I’ll put it under my desk, something that’ll maintain a ambient water temp all the time.

Can the Mora3’s do that? I want my 2080Ti under 40C all the time preferably like 36-38C max temps, or better. I can pull that off with a stock 2080Ti with the normal power budge, and normal voltages. But I want under 39C with high voltage and high wattage.

 

When I had my 2080 TI KPX cards, I had separate MO-RA3 loops, one for my CPU and one for the two KPX cards. I was pulling over 500W for each card. No, it could not keep up with both. They still reached about 60C+ with an ice cold garage in the middle of a Northeast Winter night. MO-RA3 are amazing radiators but their performance is dictated by your ambient temps. What you want is the water chiller.

 

So a water chiller would still be pretty effective with my (2) 360MM ekwb classic radiators? Maybe adding another pump and reservoir with the chiller and it would perform as needed.

 

Yes, by setting a lower power max. This is something Brother @Mr. Fox pointed out a long time ago during his and our many conversations. That, along with lowering the voltage.

A water chiller will do that for you. It should be a minimum of 600W cooling for normal use case scenarios. So 1/3-1/2-3/4-1HP. The bigger the chiller the less time it needs to run to get desired temps. The chillers have small reservoirs along with your pc reservoir. Now this part was pretty important to me...The pump needs to pull the water from the chiller into the pc reservoir and out to the components. That way the components are taking 100% of the coldest water all the time. And have the rads after the components, but before the chiller. Reason being, the rads will heat up the water if you have them after the chiller.

With a chiller you don't really need rads. The bigger your reservoir the longer it takes for the chiller to cycle back on.
Or
You can use the rads for normal stuff then when gaming or benching you could lower the temps on the chiller and allow it to kick on and run.

Edit:
the 2080TI is no longer beating my 3090. It was something funny going on with the drivers.

 

It's the only vBIOS I use on my 2080 Ti FTW3. You have to lock the voltage at the max if you want it to run correctly with the power limit maxed out. Otherwise, it tries to pull more power than the default voltage can cope with, pretty much the same concept as trying to overclock a CPU without enough voltage. And, it won't really work right on a GPU that has not been shunt-modded. Or, you can tone it down so it runs like a stock GPU. There's not much reason to overclock a 2080 Ti for gaming unless you're just wanting to for giggles.

We need to have another one before too long. Going nuts at work right now and not having much time for anything but work at the moment.

 

Did you double check that by unplugging that connector and seeing if you got reboots again?
You said that you did multiple changes at once and you didn't verify that it was plugging in the connector that fixed your problem.
So the easy test is just to unplug it, see if you get the crash/reboot, then plug it back in again.

How many PCIE cards do you have? I was under the impression you only needed that connector if you had more than 2 PCIE cards installed.

 

I have not tried unplugging it again. I have to remove my lower case PSU shroud, then remove my power supply, then I can unplug the bottom 6 pin. It’s in a very tight spot, one of the reasons why I’ve never plugged it in. If you don’t plug it in during the assembly of the PC, you’ll never get it done. My last case would’ve required removing the video card, motherboard, and PSU just to plug it in.

It doesn’t crash anymore, and I tried virtually everything you could imagine.

CPU running stock, memory running stock, GPU running stock. The XOC 2080Ti 2KW bios just caused power shut downs. Now it is very reliable, no need to trouble shoot any further. It certainly wasn’t my PSU, as the same issue occurred with johnksss.

All I’m saying is, (2) different people were getting power shut downs with the same bios on (2) totally different systems. I only have (1) PCI-e card in my system.

 

I mostly overclock for gaming, I like the extra juice. But, I do extreme testing to find 100% stability across multiple games. It is tough.

 

New personal best.

https://valid.x86.fr/hf3tvw

 

@Mr. Fox
I'll be gone for the next 4 days as my birthday is tomorrow, bit i''l catch up with you next weekend.

 

Have a happy birthday, Brother John.

Sounds good. It will need to be Saturday as Sunday I am flying up to WA State to see my two newest grandkids (twins delivered today). That makes nine now, LOL.

 

Tired of messing with this 7980XE's temps. I have clean run out of thermal paste on this thing. I have lapped it (3) times, the IHS is very flat, I also lapped the cold plate on my water block, there must be something going on underneath the IHS. These temps are not as good as I'd hoped. But this is where it sits at 4.8Ghz. I have gotten a little closer to where I want to be though. And this is acceptable for daily usage.

 

Happy birthday @Johnksss

 

Thanks guys!!!!!!!!!!!!!!



Now I just need a water block

 

@Johnksss and for anyone else interested.

Galaxy HOF tuning software works with the 2080Ti Galax XOC 2,000 watt bios.

^ You can adjust power limit down from 100%, change voltage offset, voltage, boost frequency, and memory frequency, and MOST IMPORTANT OF ALL even save profiles “without a BSOD”

 

@Falkentyne

I noticed this today. It says my PCI-e slot power draw is nearly 100 watts. Maybe that’s why people are getting power shutdowns. And maybe that’s why using the supplemental pci-e 6 pin has resolved the issue.

Nothing has burned up so far. So I will continue running it like this.



free image hosting

 

If I had known that you did not know about the Galaxy HOF Extreme Tuner and using the extra power connection for the PCI-e slots I would have told you. I have been using both since the beginning with no issues. Sorry you had to figure both things out through trial and error. But, that is how you remember so that much is good.

 

Yikes. I never knew that a stock 2080 Ti would go past the 75W slot limit.
And a shunt mod simply makes it report lower power draw than it's really running (then you have to multiply it by the shunt reduction modifier, e.g. 5 mOhm stack-->5 mOhm= 2x, 5 mOhm stack-->3 mOhm=2.67x)

I can see why you needed that connector now.
My shunt modded 3090 FE pulling 600W doesn't even reach 100W through PCIE after shunt multiplier (maybe 93W after 1.65x multiplier)...

I went ahead and plugged in the Z490 Maximus 12 extreme "Ez-plug" molex for PCIE, last night, because I don't know why, and it didn't do anything noteworthy. Maybe the PCIE slot voltage is a bit higher now...maybe 0.03v higher.....I don't know...but I sure didn't need it at all...but it can't hurt right?

I only used the 1.61x multiplier for GPU Power (Board Power Draw) in Hwinfo...didn't bother with the other values...but PCIE is probably close to 95W....

 

THAT THING IS A BEAST. It has another 30% on my crazzyy power hungry watercooled 2080Ti.

I get about 17,498 timespy graphics consistently. And a I can actually game stable with the numbers. Its pretty cool!

 

Yeah, and hwbot answered back. The old rules with verification link only for Global Top 20 is now history. Too difficult for UL to make the output interface a bit smaller and less ugly.

Verification link mandatory for 3DMark and VRMark benchmarks

Recently UL has changed the output interface of their renown 3DMark and VR Mark benchmarks.

Users with a 1080P resolution will have hard time to be able to fit everything on the screen as you can see on the below example screenshot:

Hope this can be changed without ongoing bugs...