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

View attachment 295737


Board power limit Target: 370.0 W Limit: 390.0 W Adj. Range: -73%, +5%


safe?
It's safe, but you probably don't want to run it with a higher power limit unless you have free electricity. It's fun for benchmarking but no real reason to for gaming. Try to edit your post instead of double posting so the thread doesn't get too huge.
 
New high score: 18,788

https://www.3dmark.com/3dm/52621081

All I did for this was crank the fans to 100% and increase the Power Limit to 114% and the Temp Limit to 90C. I didn't change any of my other case fan profiles nor touched my settings on my CPU. But looking at my average temp in the benchmark (56C), it looks like the fans do a pretty damn good job at keeping things cool (albeit at the expense of noise). With no forced overclock, my average clock speed jumped up from 1,812MHz on my previous best score to 1,906MHz on this one.

I can only imagine what this card is capable of if watercooled. I wonder how much radiator it would need... I currently have a 240mmx54mm radiator in my case cooling my Ryzen 9 3900XT. I'm looking at putting a second 240mm radiator in there, but only 30mm thick, to tie it into the system. My original plan was to fit two 240mmx54mm radiators into the case and have it cool my CPU plus the video card (whenever the water blocks became available). But a slight measuring error with my CPU water block made me find out too late into the build that the second radiator is too thick and won't fit, but a 30mm radiator should fit fine. My error in measurement was the fittings. Most water blocks I saw the fittings were left to right, but my water block the fittings have to be up and down. As such, the upper fitting hits where the fan on the radiator would go. But I wonder if 480mm of total radiator surface area would be enough for the CPU and GPU with mild overclocks.
 
Last edited:
New high score: 18,788

https://www.3dmark.com/3dm/52621081

All I did for this was crank the fans to 100% and increase the Power Limit to 114% and the Temp Limit to 90C. I didn't change any of my other case fan profiles nor touched my settings on my CPU. But looking at my average temp in the benchmark (56C), it looks like the fans do a pretty damn good job at keeping things cool (albeit at the expense of noise). With no forced overclock, my average clock speed jumped up from 1,812MHz on my previous best score to 1,906MHz on this one.

I can only imagine what this card is capable of if watercooled. I wonder how much radiator it would need... I currently have a 240mmx54mm radiator in my case cooling my Ryzen 9 3900XT. I'm looking at putting a second 240mm radiator in there, but only 30mm thick, to tie it into the system. My original plan was to fit two 240mmx54mm radiators into the case and have it cool my CPU plus the video card (whenever the water blocks became available). But a slight measuring error with my CPU water block made me find out too late into the build that the second radiator is too thick and won't fit, but a 30mm radiator should fit fine. My error in measurement was the fittings. Most water blocks I saw the fittings were left to right, but my water block the fittings have to be up and down. As such, the upper fitting hits where the fan on the radiator would go. But I wonder if 480mm of total radiator surface area would be enough for the CPU and GPU with mild overclocks.
I think I'm pretty much at my limit in terms of CPU and GPU. I was pretty much foot to the floor on both in terms of OC.
 
  • Like
Reactions: erek
like this
I managed to squeeze out a higher score with some mild overclocking done added to my previous settings (+50 GPU and +200 Mem vs stock with both at 114% power limit and 90C temp limit), but I will wait for someone to post a newer score before adding another attempt at the scoreboard.

This time I tested with both the fans at Auto and at 100% (like before). The difference was about 100 points less with fans on Auto vs fans at 100%, but higher temps (mid-upper 60s C vs mid-upper 50s C), 21 MHz lower average GPU clock, and a lot quieter. Still no curve adjustments nor any added voltage.
 
I managed to squeeze out a higher score with some mild overclocking done added to my previous settings (+50 GPU and +200 Mem vs stock with both at 114% power limit and 90C temp limit), but I will wait for someone to post a newer score before adding another attempt at the scoreboard.

This time I tested with both the fans at Auto and at 100% (like before). The difference was about 100 points less with fans on Auto vs fans at 100%, but higher temps (mid-upper 60s C vs mid-upper 50s C), 21 MHz lower average GPU clock, and a lot quieter. Still no curve adjustments nor any added voltage.
that is good to know, about the fans. I was wondering how much that would affect things... the fans use power and therefore contribute to the power limit, but apparently still net positive. Those things really scream at 100% don't they :D
 
that is good to know, about the fans. I was wondering how much that would affect things... the fans use power and therefore contribute to the power limit, but apparently still net positive. Those things really scream at 100% don't they :D
They do but it isn't as annoying as the blower on the 1080 FE at full speed. And on top of that, they seem to do a really good job at keeping temps in check along with the massive heatsink. To manage to keep the GPU in the mid 50s C when under load and overclocked is saying something. Even keeping it in the 60s under the same situation but with fans at auto is impressive.
 
18,809
https://www.3dmark.com/3dm/52733778

A little OC

oc11072020.PNG
 
Sooo crazy, near 400w on air.
These gpus are gonna bring the death curtain for multi rail psu. All hail single rail!
 
Sooo crazy, near 400w on air.
These gpus are gonna bring the death curtain for multi rail psu. All hail single rail!
My 10900X draws 425 W at 5 GHz - more than the RTX 3090 FE's 400 W power limit.
+750 on memory? Damn.
I recalculated and found that +748 brings the memory on the RTX 3090 to exactly 21.5 Gb/s, which is Micron's rated speed for these GDDR6X modules. I previously thought it was +750. I noted performance drops both below and above +~750.
 
Last edited:
19,098
https://www.3dmark.com/spy/15120662

Non-OC'd 9900KS keeping up with the big boys. Gotta get in the mix before the people with Zen3 roll in!!! :D

100% Core Voltage, 114% Power Limit, 90 Celcius Temp Limit, +135 MHz Core Clock, +700 MHz Memory Clock, 100% Fan
 
Last edited:
I’d be careful with those high memory clocks, they could be hurting your score. So I’d be curious if lower clocks yield any improvements at all.

Remember, these cards have ECC memory and the way the new DDR6X works is it will more than likely run at the higher speeds without crashing, but all the internal error corrections and pull backs it will do can actually hurt your performance. It just doesn’t crash.
 
Yes, started hurting the score after +700.
I was thinking about this after I posted:

Are you actually adding +700 or is it really only +87.5 (+88), which 8 fold is +700 (704)? That would make more sense.

Just like my +200 is really only a +25 bump on the actual clock speed. Yeah, that is probably more like it now that I am thinking out loud. I forgot that the memory clock is 8x the actual clock.

Anyways,here is my latest:
19,143
https://www.3dmark.com/3dm/52772266

Core Voltage: 100%
Power Limit: 114%
Temp Limit: 90C
Core Clock: +100
Memory Clock: + 500
Fans at 100%

Looks like my average GPU temp was 54C, average GPU clock was 1,958MHz (with 2,115 MHz peak).

I might be able to push it more, haven't tried yet. Figured seeing the others results and clocks on here I could meet somewhere in the middle.
 
Nice. I’m pretty pegged at 19,098. I’ll have to start OCing the CPU for anymore gains.

+700 in Afterburner - which it says is MHz but it could be calculated differently. +500 gave me lower performance.
 
Last edited:
On GDDR6X, MT/s is 16x the memory clock. Afterburner incorrectly shows 8x, but the overclock is also calculated at 8x, so it's okay. Afterburner also reports MT/s calculated with 8x, but incorrectly labels it MHz. Multiply the Afterburner values by 2 to get the effective MT/s.
 
Mine is a stock 9900KS. No OC.
Some variable probably causing low clock speed. 5950x has 3.4ghz base clock. I wouldn't be surprised if a 5900x is faster than the 5950x in some games stock. But more samples will tell the full story.
 
10900k @ 4.7 ghz
3090 Founder's Edition @ 530W max TDP, +150 core, +500 memory.

https://www.3dmark.com/3dm/52920622?

Can't post on leaderboards because I'm using the hotfix 456.98. Newer drivers are buggy as hell in Warzone and Nvidia admitted finding a bug with crashing if you started the game with a browser running.

timespy.jpg
 
Nice score - get that validated!!! Perhaps after the next driver release.
 
https://www.3dmark.com/3dm/53267587

19,624 with my new 10850k @ 5.0GHz!
+114 Power, +85 Core +400 mem on the GPU
32GB 3200MHz RAM

I have 4x8GB sticks of Corsair Vengeance Pro 4000 coming in tomorrow. I am interested to see what performace boost the faster DDR4 will bring!
 
Your extra two cores kill me

Yeah, I wasn't even going to show up to this thread using my 8700k, hahaha. I bet the Zen 3 people will smoke us both using this damn benchmark. :D. I need the Empire to strike back next year with Rocket Lake!
 
https://www.3dmark.com/3dm/53267587

19,624 with my new 10850k @ 5.0GHz!
+114 Power, +85 Core +400 mem on the GPU
32GB 3200MHz RAM

I have 4x8GB sticks of Corsair Vengeance Pro 4000 coming in tomorrow. I am interested to see what performace boost the faster DDR4 will bring!
Duke, so this can be put into the proper category, what kind of cooling are you using to keep the GPU so chill?

mare you the one that mentioned having an AC unit ducted to your case? Or is this water cooling?
 
Back
Top