Discussion in 'Videocards - NVIDIA GeForce' started by XenthorX, Sep 24, 2020.
The "One" bios i live
Kingpin 1000w Rebar
Lol, I decided not to go this route, the gains with the latest cards are very limited with a huge increase in power unless cooling can be improved substantially.
With the current high clocks, it takes a huge amount of extra power to get much gain, and that can only happen if temps are kept in check.
Tough life now with maximum pre-overclock cards being sold.
Water cooling has to be the only way a 1KW bios is of use, and even then its gotta be a good setup.
Wow so Thermagic ZF-EX is probably a better paste than Kingpin KPx then LOL, I decided to go with Kingpin KPx on my 3090, Hot spot delta is 12.3C
Using similar paste, I reckon the biggest difference will be caused by luck on the GPU draw and how well the thermal paste is applied.
If you dont manually spread the paste, its hard to know how well it spread.
You can of course check by pulling it apart but that risks being worse on the second attempt and losing as good contact with the thermal pads, unless you start afresh.
Before finding the paste was badly applied I thought my GPU wasnt that good but it seems alright now.
I'm appalled at Gigabytes quality control, their GPU paste application was terrible!
Even worse with it being their Halo product.
Its worth noting, I didnt clean the GPU sink and GPU with alcohol, I wiped them well with good quality toilet paper until no more residue came off lol.
(I had alcohol and lint free cloth ready but something told me not to bother)
Maybe there is mileage in this method?
While wrong thread, I've noticed my EVGA 1080 Ti FTW3 has a hot spot of 92C. It hasn't been repasted, still the factory stock applied pads and paste.
The hot spot value is when compared to normal GPU temp.
As a standalone figure it doesnt tell us much.
Hotspot delta should be 10C (perfect) to 14C (average) when gaming and <16C when mining.
maybe under liquid nitrogen lol
You might not be following this right...
The delta is the difference between hotspot temp and GPU temp.
delta = hotspot temp - GPU temp
In posts 937 and 940:
I showed my delta was 23C. This was bad enough to be worth investigating.
After taking the card apart I found the heatsink paste was badly applied by the card mfr, some of the GPU was bare of paste. This area was running a lot hotter than it should, giving the large hotspot delta.
I reduced the delta to 10C by repasting, see posted images for evidence of the result.
** This is with air cooling.
Liquid nitrogen will likely lower it somewhat being a lot closer to zero Kelvin but isnt relevant to the discussion.
The figures given are for normal cooling methods, 10C delta and above are realisable without extreme cooling.
The take home point from this is, check your hotspot delta.
If its a lot above 11C your heatsink paste likely has a problem.
This can make your GPU hotter, less stable, less clockable and prone to earlier failure.
With a 95mm fan (probably a 10cm fan) running at 7V (silent) blowing on the 1080ti heatsink, memory temp dropped another 4C.
This gives a total drop in memory temp of 10C by adding the 1080ti cooler to the cards rear plate.
(I tried a 140mm fan at full speed but not being able to direct the airflow made it 2C worse.)
Running the 10cm fan at 12V makes it rather loud (much like the 1080ti fan at full blast) but is a good way to see the max capable performance of the 1080ti heatsink in this situ.
After 20mins It dropped temps another 4C, giving this cooler a max 14C benefit.
It also dropped GPU temp 1C.
I think 10C drop silent isnt bad, I'll stick with that.
No doubt the 1080ti cooler can be made to perform better in 2 ways
1) place the cooler over the exact location of the memory.
I didnt do this because there wasnt an easy way to tie wrap it there.
2) drill countersunk holes in the backplate to mount the 1080ti heatsink with screws.
This will make a much better contact and could drop temps a lot more.
I reckon those 2 extra mods could net another 6C+ temp drop.
But I'm not doing them
1000w ReBar vbios.
If you're not familiar with flashing or overclocking, don't use this bios.
Ah someone beat me to it.
That 92C hot spot was during the UK heat spell, as the GPU was hitting 80 - 82C. Under normal conditions it's about 80 - 82C hot spot with the GPU 70 - 75C.
So i installed the kw bios last night im liking it so far think im gonna keep it as my 24/7 bios im running a evga 3090 kingpin hc card temps stay under 50c gaming and 35c mining. One thing id like to ask you guys do you get a heavy loss in cpu score when you use rebar with 3dmark timespy ?
heres my best port royal so far been trying to break 16k
I remember such a thing was happening on Rocket Lake systems, on some games (i do not know about benchmarks) when Rebar first came out, which wasn't happening on Comet Lake systems. But I know absolutely nothing about this or what causes it so I can't help.
i see so far this new kw bios is nice in cyberpunk im using +179 and +1600 and not crashing or stuttering staying around 2160 in the game only uses around 370-400watts in game.
What are you using to measure the wattage? Are you sure that the memory is not self-correcting? I don't think I've ever seen anything above 2100 for some fleeting moments with a 400W BIOS. Can you share your game settings?
hwinfo64 and my card can run up to 1800mhz benchmarking but around 1600-1700 gaming depends on the game my card is a evga 3090 kingpin hc with a mp5works backplate cooler on it also.
heres some bechmarks recently on my card
I can't wait to own one. But it's overpriced now. ;(
Heres a msi ab output after gaming for 2 hrs in cyberpunk i did have to dial back clocks a bit so now it avgs about 2145 most the time going up to 2160 and down to 2130 no lower running +163 +1600. Im running everything maxed 2 settings on psyco at 4k with dlss set to perfmormace suprised game only uses 8gigs at 4k.
I'm at work now and I cannot run 3D Mark, but with my setup and a 390W BIOS with a custom undervolt curve, I'm in the 1900-2000 in most games at 4k120, with +800 VRAM. I need to do a couple of runs to see what I get score wise. I don't think that my card ever dropped in the 1600 range even with the default Inno3D cooler. Seems to me like you might be overvolting a bit perhaps?
This is great actually, how it maintains like that, but I've found that usually the hardest games to keep stable are relatively simpler titles that maintain higher framerates.
Have you manually undervolted parts of the curve? Can you share it?
That's the part that took me longer to set up with my card. I still see sometimes that I am voltage limited for some situations, and not power limited, but I don't want to touch it much, to be honest. I'm getting great performance and the temps are also very good (45-49C core, 55-62C hotspot, 60-72C VRAM, with 26C ambient and the fans are on like 60-70% on the worst concurrent CPU+GPU load).