I'm actually impressed by the opengl performance. I would have expected a bigger gap.

Here's the Gigabyte RTX 3090 Gaming OC running at stock on Ubuntu 20.04.1 with an Intel Core [email protected] all core and DDR4@3600Mhz with tuned sub memory timings. I'll get to overclocking the 3090 in the next day or two...

Quoting: The_Aquabatdo you guys think that nvidia will release a TI on this generation?

They might do another "super".

Quoting: The_Aquabatnice score! congrats!:smile:
I did the maths and it's 31% faster than your previous 2080TI.

The RTX 2080Ti benchmarks I've posted here are using an overclocked 2080Ti. Comparing RTX 2080Ti at stock (59.47 FPS) and the RTX 3090 at stock (83.59 FPS) the uplift is 40% :smile:

I had a tinker last night and overclocked the RTX 3090 to gain 5% over the stock 3090 performance :woot: I also tuned Ubuntu 20.04.1 which delivered an additional +40 points on top of the overclocks alone, and game-mode was enabled.

Intel Core [email protected] all core and DDR4@3600Mhz CL16 with tuned sub memory timings

• +210 Mhz GPU Clock Offset (1935 Mhz - 1950 Mhz)
• +200 Mhz Memory Transfer Rate Offset (9950 Mhz or 19900 Mhz @ double data rate)
• +20W Power Limit (390W target)
• 100% Fan duty (both GPU fans)

NVIDIA Settings tweaks

• Force Composition Pipeline - Disabled
• OpenGL Image Settings - High Performance
• PowerMizer Preferred Mode - Prefer Maximum Performance

Last edited by Wimpress on 27 Sep 2020 at 2:50 pm UTC

isnt the CPU the bottleneck for 1080 on a 3090?

Quoting: mylkaisnt the CPU the bottleneck for 1080 on a 3090?

Not for Superposition 1080p Extreme. It's a GPU killer.

you might get better results OC'ing the memory more and not clocking the GPU that much, will have more headroom for the powerlimits and memory OC seems to help Superposition a bit more :)

Nice beast btw! Just insane power usage.

Changed the thermal paste on my 2080 Ti today. Not 3090 levels, but a reasonable uplift nonetheless.

Quoting: Xpanderyou might get better results OC'ing the memory more and not clocking the GPU that much, will have more headroom for the powerlimits and memory OC seems to help Superposition a bit more :)

So, I gave it go:

• +180 Mhz GPU offset
• +950 Mhz memory offset
• +20W power limit
• 80% fan duty

not a huge difference it seems. maybe ampere has so beastly memory that it doesn't change much anyway.
my 1080Ti gets a bit more with memory OC... iirc with +150 core and +300 mem i only got around 5450 and with +80 core and +900 mem i got like 5600 something. Anyway its a fine balancing act, to fit into the powerlimits with all the OCs, so it wont start throttling.

Last edited by samurro on 23 Dec 2020 at 2:10 pm UTC

Here is an update on my little vega 56, still going strong ! :smile::heart:


Last edited by GaGaGaeland on 27 Dec 2020 at 8:40 pm UTC

My current score with R5 3600 CPU is:


Which is not that much better than my previous CPU (R3 1200).


Clearly a GPU-bottlenecked benchmark as during the test it showed 100% GPU utilization almost all the time.

I have a question, is my low cpu mhz indication for some misconfiguration? I see other R5 2600 go for much higher mhz on this bench.

Quoting: samurroI have a question, is my low cpu mhz indication for some misconfiguration? I see other R5 2600 go for much higher mhz on this bench.

i don't think this number there changes much. it might have just captured the clocks when your CPU was idling. you can ofc try to force your CPU to performance mode by using cpufreq, cpupower or whatever have you on your distro. or maybe even feral gamemode. I doubt it makes much difference in this bench anyway though, as at the default settings its just taxing the GPU the most anyway.

Quoting: samurroI have a question, is my low cpu mhz indication for some misconfiguration? I see other R5 2600 go for much higher mhz on this bench.

It seems to just report what the kernel says it is, which is generally the CPU at base clock, not the maximum boost clock.

Okay, appreciate the replies.

Updated the system with new mobo/CPU/memory and installed new Mint 20.1. There is no significant differences as the GPU is the same, similar to Alm888's post.

Old system:


New system:


links are a bit broken but can be clicked at least.
anyway yeah.. it shows that CPU doesnt matter in this bench.

Quoting: Xpanderlinks are a bit broken but can be clicked at least.
anyway yeah.. it shows that CPU doesnt matter in this bench.

CPU, RAM or HDD/SSD I must add.
My version of GTX 1660Ti seems to be sh*ttier… Maybe because it is (I have the cheapest one from ASUS with only a single fan) and it goes to throttling pretty fast. In fact, both times I measured my second run because I wanted the card to pre-heat first. :)