Overclock.net banner
7,141 - 7,160 of 18,210 Posts
Which card are you running?: Zotac Solid OC RTX 5090

Frequency at 1.065v at 30 C: 2940
idle clock at stock: 195
Difference between idle clock at stock and FE idle clock at stock (your idle clock -180 MHz): 15
Adjusted frequency at 1.065v (the point on your graph minus the difference in stock clocks): 2925

Is it overclocked?: Yes
If Yes, how much is it overclocked?: +460 Core, +3000 Memory (Steel Nomad), +430 Core, +3000 Memory (Super Position, 460 unstable randomly in it)
What is the power limit set to?: 104%
Did you run Superposition 2 times and screenshotted the second result?: Yes
Did you run Steel Nomad 2 times and screenshotted the second result?: Yes
Tag me: @zhrooms
Show us a picture of your card in your PC:
Electronic device Computer hardware Peripheral Computer Personal computer
Graphic design Advertising
Screenshot Font Software Web page


Nomad Result: I scored 15 413 in Steel Nomad
 
I got some more progress all the sudden on my nomad results, idk why when started messing with just doing flat + mem, and super position and nothing else I was able to consistently and finally break 15200, and also higher voltage/frequency.

The frequency showned and all that still seems very suspect. As mentioned above post. ( I scored 15 413 in Steel Nomad )
As I've scored similar with far lower frequency. As well Zotac card is still kinda bouncing all over place in clock speeds so Im hoping some vbios comes along or option to address this as other cards don't seem to have this issue? Unsure I know Jay mentioned it being problem holding him back.

I was able to push past +460 but didn't see score keep increasing, +460 seemed to be the 'sweet spot' but score variance per run was still a bit of a thing for sure.
With that I was seeing voltages between 995 and 1030 largely.
 
Hi everyone,

I just got my RTX 5090 Astral OC and I've been trying to push its overclock further on air cooling. I'm currently running the GPU in a room consistently maintained at 20ºC with air conditioning.

My current system:
  • 13900KS @ P5.9 / E4.7 - Delidded & wc
  • DDR5 7600 CL36
  • Asus Z790 Apex

Here are my current results:

Plot Diagram Screenshot



I'm trying to determine whether my current results align with what they should be or if I'm doing something wrong. I'm using the latest official NVIDIA drivers (572.83).

Comparing my results with others on 3DMark, I’ve noticed that my card appears to run at higher clocks, yet the final scores don’t reflect that performance boost. I'm not sure what I might be missing or if there’s something I’m overlooking. Also, I'd love to know if it's possible to push past 3,000 MHz on the memory and how to achieve that.

Here is an example with @Alacrity30:

Text Font Number Screenshot Web page


Text Font Number Screenshot Document


Could it be due to the CPU and memory?

The drivers?

I see that the temperature is lower, but it doesn't seem like it could be the cause of everything.


Thanks in advance for your help!
 
Yea, new driver rocks, at least on lighter stuff that doesn't slam the PL too hard. I think that may be why SN doesn't seem to improve for people as it's the most power limited of PR/SW/SN.

Over 400pts higher and my temps aren't quite as good. BTW, if people are wondering, I tried bumping my CPU to 61/45/53 and my score didn't improve much at all, so not CPU limited at all as far as I can tell (on a modern CPU of course).


CPU clocks increased, (just within run to run variance, but I guess a little better)

 
Yea, new driver rocks, at least on lighter stuff that doesn't slam the PL too hard. I think that may be why SN doesn't seem to improve for people as it's the most power limited of PR/SW/SN.

Over 400pts higher and my temps aren't quite as good. BTW, if people are wondering, I tried bumping my CPU to 61/45/53 and my score didn't improve much at all, so not CPU limited at all as far as I can tell (on a modern CPU of course).


CPU clocks increased, (just within run to run variance, but I guess a little better)

Ι guess the 14900 pushing more than amd cpus the 5090 at pr.bench. or i need to make a fresh os(format) cause i have it for about 2 years
Btw nice score!!! 👍
Here is mine for example.

 
Ι guess the 14900 pushing more than amd cpus the 5090 at pr.bench. or i need to make a fresh os(format) cause i have it for about 2 years
Btw nice score!!! 👍
Here is mine for example.

Very likely the OS config, etc... and not the CPU. OS can have a big impact sometimes. I always use a completely clean install for this stuff. I actually have 4 different OS installs on my machine.
 
Yea, new driver rocks, at least on lighter stuff that doesn't slam the PL too hard. I think that may be why SN doesn't seem to improve for people as it's the most power limited of PR/SW/SN.

Over 400pts higher and my temps aren't quite as good. BTW, if people are wondering, I tried bumping my CPU to 61/45/53 and my score didn't improve much at all, so not CPU limited at all as far as I can tell (on a modern CPU of course).


CPU clocks increased, (just within run to run variance, but I guess a little better)

In the top 10, those 14 points could move you up some places in the future lol
 
Did you get a 5090 at msrp? Thought you skip this gen? Guess the temptation too much? lol
Sorta. I had a friend get one for me but I did give him a finders fee as I thought that was only fair for his trouble. But between that, Asus price gouging, and the Orange Man's tariffs, it was stupidly expensive.
 
Yea, new driver rocks, at least on lighter stuff that doesn't slam the PL too hard. I think that may be why SN doesn't seem to improve for people as it's the most power limited of PR/SW/SN.

Over 400pts higher and my temps aren't quite as good. BTW, if people are wondering, I tried bumping my CPU to 61/45/53 and my score didn't improve much at all, so not CPU limited at all as far as I can tell (on a modern CPU of course).


CPU clocks increased, (just within run to run variance, but I guess a little better)

Nice

Pushing the 14900 definitely doesn't play any role here.

Tried all combinations even up to 63/49/54 ht on. Gave nothing compared to ht off.
The latest PR or Steel runs were done while using the daily ht off profile.
 
  • Rep+
Reactions: yzonker
Holy cpu speed, what temp is fluid at?
The 16/17° water used daily.
It's 3dmark so apart from cpu temp ramping really dirty at the loading screen it doesn't put any significant load during the run. Of course it's the only thing the chip would run at those frequencies.

Will work on the curve approach after a clean update to 572.83.
 
  • Rep+
Reactions: Gamerman
The 16/17° water used daily.
It's 3dmark so apart from cpu temp ramping really dirty at the loading screen it doesn't put any significant load during the run. Of course it's the only thing the chip would run at those frequencies.

Will work on the curve approach after a clean update to 572.83.
For the GPU only benchmarks I set a low PL1/PL2 (like 200-250w) to keep friggin System Info from crashing as it spikes the CPU one or more times.

Even TS/TSE I'll set PL1/PL2 just high enough to not throttle the CPU test.
 
RTX5000 is voltage unlocked ? It wasn't on 4000/3000/2000 etc ...

I'm able to push 1.1V+ with Astral OC ? Normal ? :)

It makes no sense without the XOC bios, even at 1.1V, which will be PWR limited but ... ?

Electronic device Technology Electronics Computer hardware Hardware Programmer
 
RTX5000 is voltage unlocked ? It wasn't on 4000/3000/2000 etc ...

I'm able to push 1.1V+ with Astral OC ? Normal ? :)

It makes no sense without the XOC bios, even at 1.1V, which will be PWR limited but ... ?

View attachment 2703585
It varies from card to card. The real question is, do the cards with higher voltage limits actually perform better, or are they poor bins and hence a higher factory voltage? Requiring more voltage/power for the same clocks.
 
It varies from card to card. The real question is, do the cards with higher voltage limits actually perform better, or are they poor bins and hence a higher factory voltage? Requiring more voltage/power for the same clocks.
I would assume largely means have a more efficient silicon that potentially can also boost higher under a strict power limit.
Where like I'm largely stuck around 1000-1040 seems like, and unremarkable scores very power limited since can't get voltage much further.

Idk I'm pushing over +460 in steel nomad now which seems high but i feel like cards like that astral are keeping way higher clocks. in load.
 
I would assume largely means have a more efficient silicon that potentially can also boost higher under a strict power limit.
Where like I'm largely stuck around 1000-1040 seems like, and unremarkable scores very power limited since can't get voltage much further.

Idk I'm pushing over +460 in steel nomad now which seems high but i feel like cards like that astral are keeping way higher clocks. in load.
Yea, I'm not sure. My Astral throttles hard in 4k native (i.e. Steel Nomad). It not only drops voltage due to the PL, it also drops clocks a lot relative to the VF curve like others have noted. Max voltage for my card is 1075mv and only seen with no load and just locking the VF curve to 1100mv.

This is why when I tested CP2077 on the new driver, I didn't see any performance gain I think. It was not only dropping voltage to the 975-1000mv range, it also reduced clocks significantly relative to the VF curve in that voltage range. I'd need to do a better comparison to be sure, but it didn't appear to be running at a higher clockspeed on the new driver despite using a much higher offset.

But on lighter stuff like the Monster Hunter bench, the card runs at a much higher clockspeed on the new driver (100mhz+).
 
7,141 - 7,160 of 18,210 Posts