• Competitor rules

    Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.

NVIDIA RTX 50 SERIES - Technical/General Discussion

I got 15.3k in Steel Nomad at my settings of +270 on core and +2000 on memory (Gigabyte Xtreme Waterforce) but definitely power limited in this benchmark versus what I typically see in games. Pulls the full 600W from the wall and the card doesn't boost as high as it does when it's not power limited. Minus Cyberpunk which similarly draws loads of power although boosts about 150 MHz more than Steel Nomad.



Try increasing your memory overclock. Loads of headroom there generally. There's a way to allow you to do +3000 MHz which I haven't tried but +2000 MHz has been rock solid for me.

This is my 15.3 steel nomad with the undervolt starting at 0.935. Just to say that you'd be getting way better results than mine with an undervolt, instead of wasting voltage and power using the traditional method.
On paper, your card is much better than my Vanguard, so I'm sure you can extract more even from 0.935.

On my Founders 5090 I'm struggling to crack 14,000 in Steel Nomad (13,960) with undervolt settings.

My power draw is a good 20% less over stock, but I wonder if there's a limit with what the FE can do as I'm a 1,000 shy of AIB boards it would seem.
I was able to get 14641 on the FE (which I had until last week), following the [email protected] method. Didn't even touch memory for that one. So, I don't know... What's the average clock during the benchmark? And what about the other 3D Mark scores? Speedway, Port Royale, Timespy Extreme.
 
Last edited:
This is my daily curve by the way. Gets to a nice and round 15000 in Steel Nomad (MSI Vanguard).

W94xj2G.png
 
Last edited:
On my Founders 5090 I'm struggling to crack 14,000 in Steel Nomad (13,960) with undervolt settings.

My power draw is a good 20% less over stock, but I wonder if there's a limit with what the FE can do as I'm a 1,000 shy of AIB boards it would seem.

I was able to get 14641 on the FE (which I had until last week), following the [email protected] method. Didn't even touch memory for that one. So, I don't know... What's the average clock during the benchmark? And what about the other 3D Mark scores? Speedway, Port Royale, Timespy Extreme.

I also get around 14600 on my FE with the 2727 / 0.89 method without touching memory
I'm hesitant about pushing the memory because it hits 80c even with the under-volt

Stock I was getting 13800 but setting the power limit to 90% and putting +200 on the core I was able to get 14200
The under-volt is win win really

I'm getting the same power draw as I was with my 4090 FE (I didn't under-volt the 4090) with a lot better performance!!
 
This is my 15.3 steel nomad with the undervolt starting at 0.935. Just to say that you'd be getting way better results than mine with an undervolt, instead of wasting voltage and power using the traditional method.
On paper, your card is much better than my Vanguard, so I'm sure you can extract more even from 0.935.


I was able to get 14641 on the FE (which I had until last week), following the [email protected] method. Didn't even touch memory for that one. So, I don't know... What's the average clock during the benchmark? And what about the other 3D Mark scores? Speedway, Port Royale, Timespy Extreme.
Yeah for sure. I have undervolted but not too much as I haven’t had time so far to play around with mine. So picked a safe ish starting point to get a feel. I’m back home now though after a week away, so gonna be fine tuning.

From memory, mine was running at 0.98v ish in Steel Nomad so should have plenty of headroom to undervolt more to improve it :). Definitely hope there’s room left when you pick a random high number to OC at and it doesn’t fall over immediately :D
 
I played a bit last night. Picked a given frequency and started to walk down the volts till it crashed.

I was aiming for ~2910ghz. Was stable(ish) till .85 any lower and instant driver reset. Not a hard crash. So I think today I’ll move up acurve point and then test through more games/apps.
 
On my Founders 5090 I'm struggling to crack 14,000 in Steel Nomad (13,960) with undervolt settings.

My power draw is a good 20% less over stock, but I wonder if there's a limit with what the FE can do as I'm a 1,000 shy of AIB boards it would seem.
Steel Nomad is a power hog so it probably needs undervolt to get most out of it. I managed 15500 (not stable settings) just trying to use the minimal voltage that still hits 600W. And ofc silicon lottery to what clocks it can run at those volts.
 
Is there a way passed the 1000mhz limit on the curve editor? As it stands that's my limiting point at the .890mv method. Regardless, here's my results on Steel Nomad with my Zotac 5090 Solid. 2752Mhz represents +1000Mhz, I canny go no higher capn'!

UV Curve 1 is .810 - .890mv @ 2752Mhz. The +mem is exactly as it sounds, +1000Mhz on mem.

1V7lIEv.png

0f276Bn.png

bZBMooG.png

Steel Noman scores

Stock = 14177
UV Curve 1 = 14276
UV Curve 1 +1000 mem = 14341

Overall, for a quick mess around, I'm content with a marginal increase in score at a 75W~ saving.

Also, don't forget, for extra power savings, you can set a 2D and 3D profile in MSI Afterburner. For the 2D profile I just lowered everything as low as it'll go, then used my undervolt profile for 3D.
 
Last edited:
Steel Nomad is a power hog so it probably needs undervolt to get most out of it. I managed 15500 (not stable settings) just trying to use the minimal voltage that still hits 600W. And ofc silicon lottery to what clocks it can run at those volts.
Yeah 100%. Using my normal usage curve and settings which hits 3.2ghz in most games and 3.1ghz in cyberpunk, core boosts to like 2950-2990 in Steel Nomad. Can get it to boost higher by undervolting more though.
 
Last edited:
I also get around 14600 on my FE with the 2727 / 0.89 method without touching memory
I'm hesitant about pushing the memory because it hits 80c even with the under-volt

Stock I was getting 13800 but setting the power limit to 90% and putting +200 on the core I was able to get 14200
The under-volt is win win really

I'm getting the same power draw as I was with my 4090 FE (I didn't under-volt the 4090) with a lot better performance!!

I managed to crack 14000 this afternoon (14057), but it still hovers around there. I think I'm just on the max in the silicon lottery for my card.

Tweaked things a little lower on the volts for better power efficiency which is only about 3fps lower.
 
I got around 14450 with my zotac 5090 solid and a undervolt. Oddly I can't seem to do the normal 2827 @ 0.89 as the curve for the zotac seemed different, it was capping me at around 2747 instead of 2827. So my curve is a bit different going up to above 2800 a bit higher mv. Very happy with the thermal/power results though. Assassin's creed shadows was using as much power as my old 3080 did but waaaay more performance. It's also about 100w lower at the same performance as it was at stock.
 
I got around 14450 with my zotac 5090 solid and a undervolt. Oddly I can't seem to do the normal 2827 @ 0.89 as the curve for the zotac seemed different, it was capping me at around 2747 instead of 2827. So my curve is a bit different going up to above 2800 a bit higher mv. Very happy with the thermal/power results though. Assassin's creed shadows was using as much power as my old 3080 did but waaaay more performance. It's also about 100w lower at the same performance as it was at stock.
Could you take a pic of your curve?
 
I got around 14450 with my zotac 5090 solid and a undervolt. Oddly I can't seem to do the normal 2827 @ 0.89 as the curve for the zotac seemed different, it was capping me at around 2747 instead of 2827. So my curve is a bit different going up to above 2800 a bit higher mv. Very happy with the thermal/power results though. Assassin's creed shadows was using as much power as my old 3080 did but waaaay more performance. It's also about 100w lower at the same performance as it was at stock.
It's because you can't increase it beyond 1000mhz. When you drag it up, you'll notice the +mhz increase, anything you try to increase it to beyond 1000mhz will drop back down to 1000mhz. Whether it be 1 or 500Mhz over.
 
It's because you can't increase it beyond 1000mhz. When you drag it up, you'll notice the +mhz increase, anything you try to increase it to beyond 1000mhz will drop back down to 1000mhz. Whether it be 1 or 500Mhz over.
I'm aware of that but +1000 for other cards is 2827, it's not for the zotac solid and I'm not sure why.
 
I'm aware of that but +1000 for other cards is 2827, it's not for the zotac solid and I'm not sure why.
Because the default curve is lower, only Zotac know why they did that. On my Solid, the stock .890mv point on the curve is 1755Mhz. That means I can't go beyond 2755Mhz. If you want to go further, you'll have to flash a different bios.

Regarding my previous attempts, I used a marginally higher point on the curve to get some more headroom. My latest result, still room for improvement though.

uv 4 = .900 @ 2917Mhz +1200Mhz Mem = 14732
 
Last edited:
Because the default curve is lower. On my Solid, the stock .890mv point on the curve is 1755Mhz. That means I can't go beyond 2755Mhz. If you want to go further, you'll have to flash a different bios.
Well yes that's what I said in my original post that it's different. My point was more I'm not sure why zotac made it lower I guess.
 
Well yes that's what I said in my original post that it's different. My point was more I'm not sure why zotac made it lower I guess.
Ah, I see. The pessimist in me says it was so they could sell their OC version at increased price without having to risk GPUs not making the cut. The OC version only has a 15Mhz core boost, that basically puts it near stock if you account for the lower baseline for the non-OC.
 
Ah, I see. The pessimist in me says it was so they could sell their OC version at increased price without having to risk GPUs not making the cut. The OC version only has a 15Mhz core boost, that basically puts it near stock if you account for the lower baseline for the non-OC.
Yeah perhaps. Doesn't matter too much in the end since the difference seems negligible I suppose. Just a bit odd.
 
Back
Top Bottom