Overclock.net banner
4,981 - 5,000 of 18,210 Posts
I know that at least some 5090s use unsupported controllers right now. I asked Elmor about the Astral and that was the answer.
Yeah, he hasn't started any development on the 50 series yet and doesn't really have any ETA or plans yet. All 5090s use the same NVVDD/MSVDD controller but differ a bit in what FBVDD (memory) controllers they use and how many

With how mediocre 40 series was with voltage, I imagine 50 series will be the same.

Also ignoring the fact that adding and extra connector or waiting for HOF and lightning is really the only smart way to go about it.

I didn't mind pulling 750w+ on the 30 series but ever since these dumb connectors became mandatory, I haven't really been interested in pushing that hard.
 
So yeah, using the new MSI vBIOS my now daily stable OC went from +218 to +300 giving me a nice little performance boost.

old bios:

new bios:
 
So yeah, using the new MSI vBIOS my now daily stable OC went from +218 to +300 giving me a nice little performance boost.

old bios:

new bios:
Why didn't I get a significant clock increase with the new BIOS... 😭

Can you share a GPU-Z screenshot just out of interest?
 
UPDATE: I just checked again today (February 26), there are multiple new models listed with far better pricing

ASUS
$2329 TUF OC​
$2489 Astral​
$2499 Astral OC​
$2629 Astral LC (AIO)​
$2649 Astral OC LC (AIO)​
Note: These prices are correct.

Inno3D
$1999 X3​
$2319 X3 OC​
$2429 X3 iCHILL​
$2569 iCHILL Frostbite (WB)​
Note: The X3 OC and X3 iCHILL are clearly not correct, the Frostbite is possibly correct.

Gigabyte
$2349 WindForce OC​
$2399 Gaming OC​
$2529 AORUS Master​
$2679 AORUS Master ICE​
$2779 AORUS Xtreme Waterforce (WB)​
$2819 AORUS Xtreme Waterforce (AIO)​
Note: All these prices are incorrect, they've been incorrect since launch day.

MSI
$2149 Ventus OC​
$2469 Gaming Trio OC​
$2589 Vanguard SOC (LE)​
$2599 SUPRIM SOC​
$2569 SUPRIM SOC LC​
Note: Some of these are correct, like Ventus and SUPRIM SOC LC, they were updated recently.

Zotac
$2069 Solid​
$2179 Solid OC​
$2279 AMP Extreme Infinity​
Note: These are correct, they were updated in the last few days, and definitely the real prices we should be seeing, X3 OC, WindForce OC, Gaming OC, Gaming Trio OC will all come down to around $2150 and $2250 realistically.
__
This is how I'd like to see the prices personally, based on how good/fancy they are, and using the ASUS and Zotac prices as reference.

Budget cards
$1999 = X3 / PNY
$2050 = Solid / WindForce / Ventus / X3 OC

Mid-tier cards
$2150 = Solid OC / WindForce OC / Ventus OC / X3 iCHILL
$2250 = Gaming OC / Gaming Trio OC / GameRock OC / Phantom OC / AMP Extreme Infinity
$2300 = TUF OC / Vanguard SOC

High-end cards
$2400 = SUPRIM SOC / Master / Master ICE
$2500 = Astral OC / SUPRIM SOC LC / Waterforce LC
$2600 = Astral OC LC

I want a TUF OC, GameRock OC or Phantom OC, they got Dual BIOS and will get the Heatkiller Ultra block (was the best block on 4090), with this water block the price will be around the same as a Astral OC, but significantly faster (depending on your radiator size).
I would put the Tuf in the higher end category. It is just missing the useless fourth fan and the power plug sensor.
 
I suspected this from the get go, and when i received the trio i could immediately compare with the Suprim 4090 i also have and to my surprise it held pretty good maybe only about 2-3C i estimated as advantage for suprim which I’m happy to see now confirmed! Thanks too for the comparison!
So yeah, using the new MSI vBIOS my now daily stable OC went from +218 to +300 giving me a nice little performance boost.

old bios:

new bios:
Delete
 
It's called clock stretching. All Nvidia GPUs deal with it since 30 series. You can track this in hwinfo.

Also, I would take the "avg gpu clock" from those benchmarks with a grain of salt. It does not start or end precisely when the benchmark does nor does it take into account any swings during scene changes.

I have gotten better scores from lower clocks at different voltage points and vise versa. I remember well on a 3090 getting better scores at a lower voltage point with the same clock speed. After messing with voltage using an evc, it was MSVDD that was at fault.

MSVDD was mostly at fault for clock stretching on the 30 series as it would just be set to the same voltage as NVVDD by default. Problem is, MSVDD was very sweet spotty so you could loose effective clock speed at too high or too low a voltage

That's not to say there aren't Nvidia inspectors settings people use (and gatekeep for some weird reason) but that is mostly in the past now.

Are you using the steam version of 3dmark by chance? Because that will nerf your score as well. If you are, you can generate a free key for the standalone download from 3dmark
appreciate the reply. very familiar with clock stretching on the CPU side, was unaware it was an issue on GPUs, and also unaware of what role voltage played in clock stretching, as well as the specific voltage metrics which were at play.

im on desktop version of 3dmark, will play around with voltages - but with the current bug where it gets stuck at ~2500 core clock and won't boost higher, which I'm currently stuck on again :( , it's a huge pain to adjust voltages because every time you do you have a chance of retriggering this core clock cap. Unless others have found a sure-fire way to avoid that bug?
 

WhOa! and they said it couldn't be done! I can't believe someone managed to machine a block that had to hold 2 PCB's at the same time! absolute witchcraft!
 
appreciate the reply. very familiar with clock stretching on the CPU side, was unaware it was an issue on GPUs, and also unaware of what role voltage played in clock stretching, as well as the specific voltage metrics which were at play.

im on desktop version of 3dmark, will play around with voltages - but with the current bug where it gets stuck at ~2500 core clock and won't boost higher, which I'm currently stuck on again :( , it's a huge pain to adjust voltages because every time you do you have a chance of retriggering this core clock cap. Unless others have found a sure-fire way to avoid that bug?
The voltage adjustment I'm talking about is a hard mod using an evc (which doesn't work with 59 series currently)

And it's not a bug. Some chips are just more sensitive to MSVDD voltage than others
 
4,981 - 5,000 of 18,210 Posts