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

I meant more along the lines of your impression on how well it overclocks. What you've learned about it's overclocking behaviour, if there's anything interesting to learn from it that is.

 

I ran the card all of 30 minutes so not alot of learning put forth.
It's just like any other card. It's locked down and power throttling.
If it was unlocked some what it would definitely go farther. The 120% power limit on this card is easily hit before even using the 1.075V
2100Mhz is suspect because it wont hold 2100 at a 97 to 100 percent of the time. Clocks drop as low as 1800 to 1900.

 

I'm surprised to hear about power throttling on 3rd party cards because they normally have adequately increased power limits. Zotac (in my experience) are good for some massive power limit ceilings, so perhaps some extreme Zotac versions of these cards would be good overclock performers then.

 

In my experience this is with 100 percent of the cards.

 

I'm talking about Pascal & 3rd party cards, was thinking this pattern would carryover to Turing. Like my Zotac GTX 1070 has got a max of 268W power ceiling, which is a LOT for a GTX 1070, so thinking Zotac Turing might be good based on what you're saying about power limit being an issue.

 

No, your Zotac has a max power of 300W. Provided you had the right bios. And unlocked the 300W

 

dude im jk! i probably couldnt even match that score with a 2080 Ti, no chance hahaha

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

 

Nah, was talking about the people hidden. There are a few that do not show in the top 98%. Literally

 

(Yeah, fine, I don't know about that, but that's not the point I'm making, I'm just saying that the Zotac cards often have a high 'stock' power limit - so based on your observation that overclocking seems hindered by power limit in Turing cards, then my point was that Zotac Turing cards might be good for overclocking then. I was just trying to stimulate some informative conversation on Turing overclocking, but you missed the point by 'picking holes' instead - doesn't matter.)

 

Poking holes...Interesting. The part I left out on purpose was I had a Zotac 1070.
Also, you didn't say Zotac cards you said your Zotac 1070.
I also was not trying to make points. Just gave my opinion.
A 1070 is not in the same class as a 1080 nor a 1080TI So why try using that as an example in the first place?
My whole post had nothing to do with poking holes in your story. I was just stating my opinion.
And had you used a Zotac 1080TI as an example, then you would have known about the 485W Zotac Power limit.

This is all only information and not jabs and upper cuts against you.

 

Hey, I stand by my last post. You need to understand communication better.

 

No, actually I do not. You miss understood and instead of admitting it you took the low road.

I do not have a problem admitting when I'm wrong or made a mistake. That has nothing to do with my communication.

 

It does actually, you chose to pick holes instead of having an informative discussion with regards to Turing overclocking, which is what I was trying to have with you. You either have some kind of communication deficits or instead choose to 'find petty unrelated fault' in posts I wrote. It's a pattern I've noticed before with you, but I tend not to hold grudges against people (although I do remember, not the same), so that's why I figured I'd try to strike up a convo with you again - considering you're experienced in overclocking & had a go with Turing, I wanted to find out what you had learned & was trying to bounce some ideas off you - you were having none of that obviously. I'll let you get back to running benchmarks on loop then.

 

Okay...Do not have me go back and start re posting all your "communication" post where you tried poking holes in every ones stories. And this is a 100 percent pattern in you. Apparently you seem to take us for idiots. Trying to make points where they were never called for or injecting your self in conversations you should have just sat on the side lines and waited to hear the end results. Not sure who you think you are in here, but hey.... You do you.

 

Ok, we've both said our piece.

 

@Robbo99999
Now that that is all out of the way.

This should be useful, but sad....

 

BIOS mods possible to increase power limits - that might need direct hardware chip flashing like it did with Pascal (like people were doing with some laptop Pascal to increase TDP)? I suppose there's always those hard mods, power shunt mods.

 

You are thinking along the right lines, but hoping there is a far easier way. With pascal (1080 TI's) we were able to flash updated bios files from other cards. When doing it this way you keep your card intact with flashing through the normal channels. Although not sure that is a possibility just yet.

 

Yes, simpler options are more welcome. BIOS from other cards - is that from the LN2 Lightening version for instance? (not sure if I'm right about that, vague recollections)

 

Yes, that was one of the files floating around. That brings me back to Zotac 1080 TI with the 485W power limit, but was not stable on most cards. The Strix XOC bios was pretty much the best one, but didn't really work well on the Kingpin 1080 TI.

 

Well let's see in the future, hopefully someone can post here with first hand experience on flashing different vBIOS on these Turing cards - hopefully some success will happen.

 

I watched JayzTwoCents' review of the ROG Strix 2080 Ti during my morning commute to see if he could point to any decisive advantages over a reference PCB. I took note of what he had to say about the power limit of 2080 Tis:

Do those of you who are more well-versed in Turing reviews and/or have hands-on overclocking experience share this sentiment? I'm guessing sample variation and more exotic cooling solutions can lead to a different result, but does what he's saying seem like the rule of thumb for air-cooled overclocking?

 

What he's saying there doesn't make much sense to me, and I'm not sure how he can really know that unless he's tested with Turing cards that don't have a restrictive power limit. Johnksss mentioned in a post earlier that he was seeing power throttling all the way down to the 1800'sMhz area, and I'm sure Turing can see stable core clocks in the 2050+ zone, so power limit does seem to be a problem. Although Guru3d reviews of cards are showing overclocked boost clocks of 2000-2100Mhz ( https://www.guru3d.com/articles_pages/asus_geforce_rtx_2080_ti_rog_strix_preview,27.html), and there's probably not a massive frequency headroom beyond 2100Mhz, so going by just that article it doesn't look like power limit is an issue. I'm not sure the type of tests that Johnksss did on his card, whether it was Furmark or what - I think for a real world gaming test then a loop of Firestrike Extreme Graphics Test 1 is a worst case scenario while still in the realms of plausibility (as opposed to Furmark). Guru3d use that looped test for their Long Duration Stress Test to show clocks & temps ( https://www.guru3d.com/articles_pages/asus_geforce_rtx_2080_ti_rog_strix_preview,8.html), and they specifically say it doesn't bounce off the power limit, it's not overclocked though. There's probably some answers to our questions at some other review sites, but I've not googled them yet.

EDIT: just found this article; yes, power limitation a problem once you free up the cooling limitations:
https://www.gamersnexus.net/guides/3367-rtx-2080-ti-hybrid-results-nvidia-power-limitations
In games they were bouncing off the power limit & not getting any increased performance from their cards with water cooling - it enabled higher clocks (2145Mhz) due to lower temperatures in non power intensive situations, but during practical gaming it was bouncing off the power limit apparently. He doesn't say how low the clocks were throttling from the power limit, so it's hard to guage how much extra performance can be gained by an unrestricted power limit. But from that it looks like 2145Mhz on the core during gaming is not that crazy for Turing if power limit is removed combined with good cooling.

 

https://www.3dmark.com/3dm/28885431?

 

Ha, this is deja vu, but this time with the RTX 2080ti - congrats again on new purchase! What's your max stable overclock, and does it bounce off the power limit when gaming at that overclock? What's the stock TDP of your particular card, and what's the TDP when you've maxed out the TDP slider? And what is the model name of your card? Sorry for loads of questions, figured I may as well include them all in one post!

 

I'm debating going full water block on my RTX 2080 Ti. Might need some help picking out entry level parts. Anyone have any tips or places to start looking?

Max stable clocks seem to be around +175Mhz maybe a bit more. 200Mhz crashes eventually so I'm close to max at least with the current power limit of 120% which is trash. It bounces off the power limit constantly, all day. In games maxed out I am around 2000Mhz under load, a bit above, a bit below. I'll look into the power limits/slider more today.

I got the Asus Dual OC (cheap cheap cheap card) RTX 2080 Ti. If you watch any of the videos though it seems all of these cards seem to perform within small points of each other since they are all power limited. I am already looking into going full block just because I don't find 70-75C under gaming loads with 75% fans acceptable. The cooler is huge but trash without a single heat pipe. It's just a huge piece of aluminum finning. Asus really cheaped out on this card. The box internals was absolutely the cheapest packaging I've ever seen. It didn't even have an anti static bag. A $1200 without a freaking anti static bag. Unbelievable. Cheap plastic shroud. It doesn't even say RTX 2080 Ti anywhere on the card. It simply says "RTX" which leads me to believe they use the same exact cooler on the RTX 2080. In addition the pictures depicted the card as being blacked out and painted heatsinks. Nope. The card has a huge copper plate running along the bottom of the card which IMO is super ugly on such an expensive card. In all honesty this card is probably going back and getting a different RTX that doesn't have misleading and false advertising.

 

I would go with an EVGA GPU if possible. If you can hold out for another month or two, you might be able to get an RTX 2080 Ti with a HydroCopper full block for hardly any more than one with an air cooler. The Ultra series seems to have a pretty robust air cooler.

 

Thanks! I love the EVGA 2080 I had, but I think I may just go Hybrid Water on this card. Like I said they all perform relatively close to each other once the temps are controlled. My biggest gripe was the cooler, but I wanted to try out water anyways this time around.

Edit:

Card is being returned to Newegg. Never had to return a single Item to them in 15 years but there is always a first I guess. They agreed to a full refund with a paid shipping label after they saw the photos of the card vs their product photos on the product listing.

 

http://hwbot.org/submission/3945233_johnksss_3dmark___time_spy_2x_geforce_rtx_2080_21931_marks

 

I find it amazing that there is not a single heat pipe & it's just an aluminium block on a 2080ti, that sounds so crazy it makes me want to ask if you're sure! I wouldn't really expect just an aluminium block on any GPU let alone 2080ti. Good that you returned it then, didn't sound like very good quality. Did you say it was bouncing off the power limiter even when you maxed out the power slider to 120%? (It was a bit ambiguous the way you wrote it).

You might want to re-think whether it's worth going water on your GPU, as Gamers Nexus said it made no difference to gaming performance due to the power limit being the limiting factor (I linked the article a few posts back). Until that power limit issue is sorted (if it ever is), then the only real benefit of water cooling is lower temperatures & perhaps less noise, but performance no difference currently.

 

Just played R6 Siege and I pulled 310W max with +125Mhz. Power and temp sliders maxed out. Card sat around 2000Mhz under load. Starting out at around 2040mhz. Temps seem to be the cards biggest issue. Even 100% can't quite keep them down to stop throttling down a bit. It settles out around 2000Mhz with max fans and my case is COLD.

 

310W just for the GPU or the whole system? I think the best way to know for sure if it's bouncing off the power limit is if you have GPUz open and look at the PerfCap Reason where it would say Pwr to indicate power throttling. Did that happen? It sounds like you're not sure whether or not the small clock drop from 2040Mhz to 2000Mhz was due to temperature or power. It doesn't matter if you're not sure if you didn't have GPUz open when you tested your card, we already know from Gamers Nexus (& JohnKSS) that 2080ti will bounce off the power limit, but I was just curious for another voice confirming as well as under 'normal' gaming scenarios.

So you were at 2000Mhz when gaming. Gamers Nexus managed 2140Mhz with water cooling on their 2080ti, but was bouncing off power limit in gaming resulting in lower frequencies at times. If we say 2000Mhz is a representative 'normal' overclocked gaming frequency for a power limited card, then perhaps a card with an unrestricted power limit and good cooling (and a bit of a silicon lottery win) would maintain 2140Mhz when gaming. How much of a performance difference would we see by water cooling + unrestricted power limit: 2140/2000 = 7% performance increase. And I would guesstimate that a 3rd party card with good aircooling would be able to control 2050Mhz stable without issue with unrestricted power limit, which would be 2140/2050= 4% increase in performance. So, watercooling might only give you 4% extra performance over good quality air cooling on an unrestricted power card. Just guesstimates. Water cooling doesn't seem necessary from a performance perspective, but cards with unrestricted power would be welcome for a slight bump in gaming performance, but even then it only seems like it would be a couple of percent improvement given you were at 2000Mhz already while gaming. (These are just my extrapolations/guesses). (Actual peformance gains are likely to be slightly lower even, because performance doesn't extrapolate linearly with increase in core frequency.)

 

Well it looks like MSI are indeed gonna be releasing a Lightening version of the 2080ti:
https://www.guru3d.com/news-story/msi-is-working-on-a-geforce-rtx-2080-ti-lightning.html

I'm posting this because a while back JohnKSS was saying that one good option to remove a restrictive power limit is to flash a vBIOS from another card that doesn't have such harsh power restrictions, the Lightening version should be one such card. I guess it remains to be seen whether it will actually be possible to flash it's vBIOS to another card, praps we'll see that happening in this forum.

 

After much debate I decided to ultimately keep the Asus Dual. The card in it's stock form is meh, but the card itself is pretty decent and I decided to do a hybrid cooling solution to the card with a Kraken G12. It is compatible by using the AMD brackets oddly enough and now idles at 27C and tops around 50~C under load overclocked. I also managed to get a working version of Nvflash, flashed the Gigabyte OC vBIOS to the card and it's pulling more power now allowing for slightly better OC. I plan to go full water for the hell of it soon as this hybrid solution is working but relies on air/fan cooling of the VRMs and memory modules. Also this card has VRMs at both sides of the card so the front side isn't getting direct airflow blown on it. Slightly worrying but I am going to grab some heatsinks soon.

 

https://valid.x86.fr/lz1a4t | http://hwbot.org/submission/3946214_

 

This Asus card with the Gigabyte vBIOS at 122% is pulling 360W. Wow!

Kraken G12 setup on RTX 2080 Ti. Looks messy AF but it works haha.

https://imgur.com/a/YkybNHS

 

5.2Ghz, you got there!

Is 360W the max power that the card will pull at 122% on that vBIOS - ie it's bouncing off the power limit? If not, what's the theoretical max power it can pull (TDPx1.22)? What clocks you at now while gaming? You were at 2000Mhz before. (New overclock now or same offsets?)

 

360W at 122% slider, yes that is the max, yes that is banging off the power limits still. Gaming CPU core load around 2075Mhz with lower +100Mhz offset. At 2100Mhz it seems to be crashing but I'm still trying to figure out this new power limit as it changes the boost clocks a bit meaning I need to refine my offsets.

 

Yep, I see what you mean about fine tuning the offsets for the new power limit. It certainly sounds like there's a little more performance to be had by a totally unrestricted power limit, but praps not much considering you're crashing at 2100Mhz, yet gaming at a power limited 2075Mhz - this means that a completely unrestricted power limit would only get you a maximum of another 25Mhz (because it's crashing at 2100Mhz). It does sound like that the max power limit of the cards is very close to the limits of the silicon stability, so not much to be gained by an unrestricted power limit. I suppose if you had a GPU that won the silicon lottery, then you could do with the extra power headroom to drive it to whatever it's limit of silicon stability - ie Gamers Nexus got their card to 2140Mhz, they might have won the silicon lottery.

So far your water cooling mod combined with new vBIOS has given you an extra 75Mhz when gaming (2 or 3% more performance maybe), but I guess it's quieter and it's nice to see to the lower temperatures.

 

UNreal!

 

2000-2100mhz seems to be about where almost all of these 2080 Ti GPUs are stopping. Some even lower. There are a few outliers getting higher bur most stop around 2000-2100mhz. I’ll do testing on FPS gains more when I get home next week I’m sure.

 

Just got a whole stack of new vBIOS to tear through as well. I think I'll try the EVGA 130% vBIOS next. I also installed a second 60mm fan on the other side of the GPU so it has active cooling on the front of rear of the card as well since that side has VRMs and memory modules on these news Turing cards. I can breathe a bit easier when testing now knowing they are cooled.

 

https://techreport.com/blog/34136/g...-begin-playing-nice-with-tr-freesync-monitors

FreeSync is working with Nvidia cards with latest driver. Whaaattt??? Hopefully this wasn't an oversight and Nvidia is quietly rolling out support.

 

Ha, just proves how easy it would be for them to support Freesync if they chose to! It sounds like a bug to me, I think they accidentally enabled Freesync support somehow in their latest drivers. I don't think they're gonna roll out Freesync support.

 

I wonder if the power limit of the GALAX RTX 2080 Ti HoF vBIOS is even higher. According to TechPowerup, the card's got an out-of-the-box boost clock of 1815 MHz. Certainly looking forward to hear how you'll do with 130%! Maybe that's already sufficient to not be restricted.

https://www.techpowerup.com/gpu-specs/galax-rtx-2080-ti-hof-limited-edition.b6193

 

Remember, this is at raster workloads only right now - all the added tensor and RT cores (30% of the total die area?) are snoozing.

30% estimate comes from a rough core per mm^2 calculation here: https://www.reddit.com/r/nvidia/comments/9h9t3u/without_the_rt_or_tensorcores_the_754mm2_tu102/

When some games come along that can actually use RT and DLSS, power use will increase and I expect power limit mods will show much larger gains over stock.

 

Yeah, good point, I see what you mean. It's also possible though in some games where overall performance is severely limited by the relatively low ray tracing power of these cards then this will lower the load/usage of the standard GPU core, thereby reducing the watts consumed by that core, so this might offset the activation of the RT & Tensor Cores - see Battlefield V's claims that they're aiming for 1080p 60 fps for a 2080ti card - the standard GPU core is gonna pretty much fast asleep when it only has to push 60fps at 1080p. In games where the ray tracing & standard rendering of the GPU core are balanced, then this is where your idea would come into play though.

 

You've probably noticed a long time ago, but the Gigabyte Gaming OC vBIOS actually has a higher power limit than the updated EVGA XC Ultra vBIOS. It's 122% over a base 300W, not 260W reference like the EVGA, thus 366W max (EVGA max: 338W). Not counting a private 400W EVGA vBIOS that techtubers have access to, the highest you can currently get is 380W from a KFA2/Galax card. It is available in the following archive:

https://www.overclock.net/forum/27644256-post1262.html

Perhaps none of this is news to you, but I thought I'd share what surfaced after a little digging.

 

I'm rocking the Gigabyte vBIOS at 366w right now, but there is another 380w vBIOS out there now from Galax and I'm thinking about flashing it.

I'm a total idiot after I discovered I didn't properly tighten my water cooler screws on this setup. I have dropped 15C after properly tightening them. Just got home and will be doing some extensive testing.

 

That's great. Thanks for sharing that.