• 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.

RTX 5090 Tweakers

Howling

Howling

Howling

Soldato
Joined
14 Sep 2008
Posts
2,797
Location
Peterborough
Hoy fellow nerds,

Basically I just want to use you all as data points to see where the 5090s like to sit with regards to voltage (under/over) and core/mem clocks.
At the minute I've just gone with a very modest +100 to both core and mem and not touched power yet (ASUS TUF).

VUoaQKe.png


2800Mhz core is the lowest I've actually seen from a game (KCD2), but I think that's because the other games I've been playing lately aren't pushing 100% usage so the core clocks are boosting higher (FFVII Rebirth and Marvel Rivals) which tend to sit around 2950MHz but apparently it's been at over 3GHz at some point.
 
Could be a good source of info but I think you need to use a benchmark on loop so everyone is applying the same load to their GPU's. Games are to varied with what parts of the GPU they will be using depending if you have DLSS or Ray tracing enabled etc.
 
Last edited:
Finners said:
Could be a good source of info but I think you need to use a benchmark on loop so everyone is applying the same load to their GPU's. Games are to varied with what parts of the GPU they will be using depending if you have DLSS or Ray tracing enabled etc.
Click to expand...

As a general rule, I'd agree with you that a standardised test would be the logical choice - but, how to word it.... I'm not looking for the 'best clocks in this one load case' - rather the range of different workloads and different clocks/voltages is more beneficial. Your suggestion has merit and perhaps other people would be interested in that kind of comparison, it's just not what I'm after.
 
Iv got mine stable on +195 to core and +300 to mem. I had the mem higher but benchmarks are pretty useless when it comes to stable clocks as they can go fine but then you get a crash in game. I personally found the resident evil 4 remake chainsaw demo to give me the best way to test stability as that would crash to desktop at any hint of pushing the clocks too far where as once I got it stable in that all my other games have been fine.
 
Matson said:
Iv got mine stable on +195 to core and +300 to mem. I had the mem higher but benchmarks are pretty useless when it comes to stable clocks as they can go fine but then you get a crash in game. I personally found the resident evil 4 remake chainsaw demo to give me the best way to test stability as that would crash to desktop at any hint of pushing the clocks too far where as once I got it stable in that all my other games have been fine.
Click to expand...

Any change to power? Out of interest how high is it boosting to (and seeing different under different games?)?
 
No change to power. I was worried about pushing the power as iv seen it spike over 600 watts with the power higher so just left that at 100%. Iv seen it boost highest to around 3050/3150mhz and around there. Depends on the game though like you said. I'm running games at 5k using DSR downsampling so that pretty much puts the gpu load at max unless I cap the framerates. I would say the lowest the boost has been is around 2900 that iv seen. Iv mostly been playing kingdom come deliverance 2 at 5k maxed out and it was around 90 to 120fps but I capped it at 80 which reduced the wattage to around 400-450.

I did try maxing out the power in afterburner and it gave me a good boost but I'm also not comfortable with it being over 600 watts in power constantly and don't think you need those extra few fps in games.
 
Last edited:
TUF 5090 here. I've got it at +230 on core and +2000 on mem. I get around 14.6k on Steel Nomad 3DMark and my clocks average around 2800ish

I've also done an undervolt, with 0.875mv maxed (2.6ghz) and an 85% power target. With this I get around 14k on steel nomad and I draw around 500ish watts instead of 600.

On games I've tested Portal RTX all DLSS off and I average around 2.75ghz and according to reddit post and from my experience it's the best stability checker, I can pass steel nomad at like +320 but can't even boot into that game at those clocks.

Also in the same game, if I enable DLSS then my clocks shoot up to around 3ghz like you. So there's definitely something to do with DLSS vs none that allows the GPU to clock higher.
 
Last edited:
Also I don't think you can put a higher than 100% target? At least I know I can't with my TUF and GPU Tweak 3
 
DizzyZ said:
TUF 5090 here. I've got it at +230 on core and +2000 on mem. I get around 14.6k on Steel Nomad 3DMark and my clocks average around 2800ish

I've also done an undervolt, with 0.875mv maxed (2.6ghz) and an 85% power target. With this I get around 14k on steel nomad and I draw around 500ish watts instead of 600.
Click to expand...

My FE doesn't have this legroom...

I can do +200 on the core and leave memory at stock to get 14.3k on Steel Nomad, but if I push the memory up, the average core speed drops a bit so I end up scoring less. It's as though there is not enough power to to share between the core and the memory to overclock both of them so the card turns down the core clock to up the memory clock.

I've done a mild +100 on the core and set the power limit to 90%. This gives me the same performance as no overclock at all with 100% power. The +100 core literally just makes up for the reduced performance of setting the power limit down. It's possible I could get better result with the curve editor but I've not used that before so I want to wait until more 5090 FE owners have done it so I have a clearer understanding of what to aim for.
 
no_1_dave said:
My FE doesn't have this legroom...

I can do +200 on the core and leave memory at stock to get 14.3k on Steel Nomad, but if I push the memory up, the average core speed drops a bit so I end up scoring less. It's as though there is not enough power to to share between the core and the memory to overclock both of them so the card turns down the core clock to up the memory clock.

I've done a mild +100 on the core and set the power limit to 90%. This gives me the same performance as no overclock at all with 100% power. The +100 core literally just makes up for the reduced performance of setting the power limit down. It's possible I could get better result with the curve editor but I've not used that before so I want to wait until more 5090 FE owners have done it so I have a clearer understanding of what to aim for.
Click to expand...
Interesting, so I don't really have that problem, what I've opted for, is [email protected] and [email protected], and then 30002 on the mem (+2000) with that I get pretty decent results ~14.3k on steel nomad, and it runs max at 500W and is like 5 degrees cooler if it's 100%, with things like DLSS and in-game etc. it's more like 450W, so basically slightly better than my stock score 14.1k.


On the topic of mem overclocks, so what I read is that the 5080 is already at 30GHz stock but the 5090 is downclocked to 28GHz for the same mem so supposedly most 5090s should be able to do 30GHz fine, it might be power related like you suspect, as I just checked my average clocks, at +230/+2000 it's around ~2730 average, on +230/+0 average is around ~2800 on Steel Nomad but funnily the scores are exactly the same ~14.8k in each instance, so no idea what memory is doing, but yeah in your case I'd recommend undervolting and see what happens.
 
You must log in or register to reply here.
Back
Top Bottom