RTX 3090 [H] Owner's Official 3DMark Time Spy Leaderboard

Made some improvements.

https://www.3dmark.com/spy/18451316

Memory up to 20500, core clock set to +105. Still using stock fan curve, this is basically an every day setting. 19,891 overall score and 20,848 graphics score.
I imagine you still have a bit of improvement left on the card. My 3090 XC3 on air can hit 20800-20900 pretty easily and can just barely crack 21K. My card is still on the limited 366W BIOS so your FE with 400W should be able to get over 21K without too much issue.
 
Yeah, I'm just taking it slowly so I can ensure that everything is stable.

I knew these cards were power hungry but wow, it easily pulls 400W, I've seen up to 410W in Afterburner's power monitor and ~775W system usage in games on my UPS.
 
Yeah, I'm just taking it slowly so I can ensure that everything is stable.

I knew these cards were power hungry but wow, it easily pulls 400W, I've seen up to 410W in Afterburner's power monitor and ~775W system usage in games on my UPS.
They may be power hungry and cost and arm and a leg, but I have to add that I for one am very pleased with what these 3090 cards deliver... and they also OC nicely. This is especially the case when it comes to 4K gaming performance. It sure is refreshing to not have to turn down anything to make a game playable or maintain a solid 60+FPS, even with all the eye candy set to "yes please, may I have another?!". :D
 
They may be power hungry and cost and arm and a leg, but I have to add that I for one am very pleased with what these 3090 cards deliver... and they also OC nicely. This is especially the case when it comes to 4K gaming performance. It sure is refreshing to not have to turn down anything to make a game playable or maintain a solid 60+FPS, even with all the eye candy set to "yes please, may I have another?!". :D
Well, with the exception of CP2077 where I have to use DLSS Balanced with RT on Psycho... lol. Worth it though for global illumination.

Everything else tho at 4K maxed? Card is a beast!!!

Not a "massive" leap from my 2080Ti, but very much an improvement with no compromise in settings. RT is clearly where this card is made to reign. :)
 
Last edited:
Finally received my EK waterblock and just finished build.

GPU Score: 21,305
EVGA RTX 3090 FTW3, watercooled
Driver: v461.40

https://www.3dmark.com/3dm/58569653?

I think my question is why is my cpu score so low vs others of same processor.
I noticed this. I am on the right.
1613801487941.png

Do you have hyperthreading/SMT off?
 
Is anyone overclock fine in timespy bench but in actual gaming is not? My overclock that is solid in bench marks, fails in cyberpunk. Funny thing is, it's the sole game that the overclock is not stable in.
 
Is anyone overclock fine in timespy bench but in actual gaming is not? My overclock that is solid in bench marks, fails in cyberpunk. Funny thing is, it's the sole game that the overclock is not stable in.
Yup, which also means it is technically NOT stable. I had the same thing, benchmarks worked perfectly, even some games, but CP2077 did not. However; after gaming a bunch on other games, BFV ended up crashing too after about 2+ hours in a single session, so I guess CP2077 is a great way to find a unstable GPU overclock quickly! ;)
 
Yup, which also means it is technically NOT stable. I had the same thing, benchmarks worked perfectly, even some games, but CP2077 did not. However; after gaming a bunch on other games, BFV ended up crashing too after about 2+ hours in a single session, so I guess CP2077 is a great way to find a unstable GPU overclock quickly! ;)
Yep. I had some pretty high clocks that worked fine in the benchmarks and even in most of my games with no issue. Even the Bright Memory benchmark would work fine running for extended periods. But then I would put on RDR2, which is a game that already has questionable stability, and it became even more temperamental with the higher clocks. Some times it works fine for a while or through a whole session, other times it would crap out in 10 minutes. While it may not be the clock speeds, it did seem that bringing the clocks down reduced the number of fits the game might have. It kind of reminded me of the Batman Arkham games that were very picky about overclocks. On my older machine, I would have a solid overclock that would work day in and day out without issue in both day to day activities and in most of my games. But when playing one of the Arkham games, I would have to lower my clock speeds because the game would randomly crash if left overclocked. Yet I had more taxing games, like GTA V, that had no problems. Go figure.
 
Yup, which also means it is technically NOT stable. I had the same thing, benchmarks worked perfectly, even some games, but CP2077 did not. However; after gaming a bunch on other games, BFV ended up crashing too after about 2+ hours in a single session, so I guess CP2077 is a great way to find a unstable GPU overclock quickly! ;)
Everything has a limit. Its just a matter of how what your running stresses the GPU. I can clear port royal just fine and not pass time spy. Just like how the OC will be fine in CP2077 and not in COD Warzone.

COD Warzone seems to be the most unforgiving.
 
Yeah, I mean CPU, Memory and GPU stress programs are great for seeing if you are "pseudo-stable", I use them all the time when overclocking. However; I have found long term, gaming on "AAA" games can be the best stress test of them all.... lol. Unless the game you are playing is utter garbage, if you crash, it's typically one of your overclocks. Once I got my system 100% stable for gaming, I have never seen a game crash on me, not once.
 
Yeah, I mean CPU, Memory and GPU stress programs are great for seeing if you are "pseudo-stable", I use them all the time when overclocking. However; I have found long term, gaming on "AAA" games can be the best stress test of them all.... lol. Unless the game you are playing is utter garbage, if you crash, it's typically one of your overclocks. Once I got my system 100% stable for gaming, I have never seen a game crash on me, not once.
Yep, and it can be misleading too. Early on when I started playing Cyberpunk it crashed 2 or 3 times in the exact same spot. Seemed obvious it was a game bug, but I tried resetting my card to stock and it went right through without an issue.

As far as games unforgiving to OCing, I'd say in this order of the ones I've been playing with my 3090,

FS2020
RDR2
Cyberpunk

The kicker is that any of them may run for an hour or more before crashing with a slightly unstable OC.
 
Yep, and it can be misleading too. Early on when I started playing Cyberpunk it crashed 2 or 3 times in the exact same spot. Seemed obvious it was a game bug, but I tried resetting my card to stock and it went right through without an issue.

As far as games unforgiving to OCing, I'd say in this order of the ones I've been playing with my 3090,

FS2020
RDR2
Cyberpunk

The kicker is that any of them may run for an hour or more before crashing with a slightly unstable OC.
Haha. For me Cyberpunk is the only one. FS 2020 works flawlessly. Goes to show how each PC is different than the next even with identical hardware. Like previous user said, Cyberpunk is my new benchmark tool. I'm serious. I'm going to be using it to test the overclock on my other pc.
 
Tweaked a few things. Best so far and broke the overall 20k mark. Not sure there is anything left in the tank for the gpu. +1000 mem, +175 core. Some people look to have won the silicon lottery for the chips.

https://www.3dmark.com/3dm/59124135?

GPU 21516
Under water.
Huh, is something wrong with your 5950x? Your clocks are WAY higher than my 3950x but I'm scoring much higher in the CPU test. I'm not bragging here but rather just concerned for your setup. https://www.3dmark.com/compare/spy/18797456/spy/18731905#
 
Sorry GM, my wiener is exactly 99.9910785975555% shorter than yours. Nothing I can do about it. :D

Seriously though, I'm thinking silicon lottery has a lot to do with just how high the numbers go once you get to these levels with "normal" cooling... and like I mentioned previously, I'm not really interested in shunt mods or anything beyond water cooling. Your GPU simply holds up slightly better than mine, and I'm OK with that.
Plus I have you and most everyone else on the list here by a bit when it comes to overall Time Spy score! :p
 
Last edited:
Sorry GM, my wiener is exactly 99.9910785975555% shorter than yours. Nothing I can do about it. :D

Seriously though, I'm thinking silicon lottery has a lot to do with just how high the numbers go once you get to these levels with "normal" cooling... and like I mentioned previously, I'm not really interested in shunt mods or anything beyond water cooling. Your GPU simply holds up slightly better than mine, and I'm OK with that.
Plus I have you and everyone else on the list here by a bit when it comes to overall Time Spy score! :p
But its only two points!
 
Huh, is something wrong with your 5950x? Your clocks are WAY higher than my 3950x but I'm scoring much higher in the CPU test. I'm not bragging here but rather just concerned for your setup. https://www.3dmark.com/compare/spy/18797456/spy/18731905#

When people overide the boost by xxx(0-200) mhz, if it goes too far, it seems to hurt performance instead of increasing it.

EG I was getting close to the same score as he was for CPU with my 5900x boosted at +175, when I Dropped it down to +100 instead , my CPU score shot up to 15175 at a lower peak mhz.
 
Last edited:
They may be power hungry and cost and arm and a leg, but I have to add that I for one am very pleased with what these 3090 cards deliver... and they also OC nicely. This is especially the case when it comes to 4K gaming performance. It sure is refreshing to not have to turn down anything to make a game playable or maintain a solid 60+FPS, even with all the eye candy set to "yes please, may I have another?!". :D
Yeah they really hit their full stride at 4k class resolutions, still wish the 1440p was a little stronger. Don't get me wrong, I get 144fps mins in all the games I play at 1440p.

https://www.3dmark.com/3dm/59157834?

GPU score 21430 fans cranked; normally scores 20k-20500.

I have my LOD to mip level -3 for dlss gaming mode. Score doesn't change with it.
Nice work!
Sorry GM, my wiener is exactly 99.9910785975555% shorter than yours. Nothing I can do about it. :D
buy a second 3090!
Haha. For me Cyberpunk is the only one. FS 2020 works flawlessly. Goes to show how each PC is different than the next even with identical hardware. Like previous user said, Cyberpunk is my new benchmark tool. I'm serious. I'm going to be using it to test the overclock on my other pc.
Cyberpunk seems to randomly crash for me completely unrelated to any hardware settings. I may see a slight uptick with all the ray tracing cranked. Don't have this issue in metro or any of the other more demanding ray tracing titles.

Added another radiator to the front of my case and bumped my core offset up 30 mhz more.

22183

https://www.3dmark.com/spy/18510546
I haven't put mine under water yet, but it's so tempting!
 
So close yet so far. ..

https://www.3dmark.com/spy/19068413

22352

I think I could top 22.4k if I hadn't lost the VRAM silicon lottery. Other than lowering temps, not much left. My Corsair block is just so so for deltaT. A better block might get me 5C or so. A repaste might help a bit too.
 
Back
Top