[SOLVED] How can I get published performance from 2080TI's - What am I missing?

[soul_collector]

Hey all, I recently built a new computer around the two 2080 TI's I recently purchased. Using Battlefield 1 on all Ultra settings as my benchmark, I'm seeing 100 fps with SLI, and roughly 60 fps without. I've gone through every setting I could and I don't seem to be able to reach the performance numbers I see being published for a similar rig with these cards. Can anyone please help give me some new avenues to follow? My rig is detailed below:

CPU: Intel Core i7 8700k (Clocked at 4.7)
Mobo: MSI Z370 Gaming M5
Memory: DDR4 32 GB 3600MHz Quad (Clocked at 3600)
PSU: Corsair HX1200 (1200W)
Monitor: Predator 1440p 144Hz Refresh Rate
Liquid Cooled with a 360mm Radiator
HHD: OS is installed on the 512GB M.2 SSD Samsung
HHD: Game is installed on WD 2TB BLD Perf 3.5 7200RPM 64MBuffer

Overclocking performed using MSI's XMP and built in Enhanced Turbo Boost for the safest settings possible with current frequencies verified by CPU-Z.

My rig seems to be similar to the one used to get these Battlefield 1 benchmarks shown here: https://www.techpowerup.com/reviews/NVIDIA/GeForce_RTX_2080_Ti_SLI_RTX_2080_SLI_NVLink/4.html

Any advice would be greatly appreciated.

 

[mortemas]

First thing to note is that SLI won't benefit BF1 and is in fact slower than a single 2080 Ti when at 1440. SLI only benefits that game at 4k. Are you sure the game is set to 1440 res?

Are you sure the SLI settings are correct? Hardware set up correctly, too? Coincidentally, 100 FPS is the single-card performance at 4k:

 

[rohitgupta1975]

I have problems too with overclocking a single Zotac AMP! 2080 Ti by more than 50MHz though people have done like 200MHz overclock for the core clock (on the already overclocked cards and founders' edition). Initially I had connected one splitter PCI supply cable from a single 8-pin PSU socket to BOTH the 8-pin sockets on the card. People advised that might be the problem. Now I have two separate cables. Though that did not help me. How have you connected your card to the PSU?

Also, is your BIOS up-to-date? Is your card driver up-to-date?
Does your AA (anti-alias) setting match with whatever the benchmarks have? (TAA?)

 

[soul_collector]

Thank you both for answering, responses inline:
MORTEMAS:

First thing to note is that SLI won't benefit BF1 and is in fact slower than a single 2080 Ti when at 1440. SLI only benefits that game at 4k. Are you sure the game is set to 1440 res?

You're right, in the telemetry posted, SLI does not yield additional FPS in BF1, however, the frame rates in either scenario posted far exceed what I am able to get. Without SLI I am getting around 60, with it, I'm at about 100. I have no idea why in my case, I AM seeing an increase in fps with SLI, that makes no sense. And yes, I have configured the in-game settings to 1440 144 Hz

Are you sure the SLI settings are correct? Hardware set up correctly, too? Coincidentally, 100 FPS is the single-card performance at 4k:
I don't know what other SLI settings to set. I spoke with the people at NVIDIA and they suggested that I make minor tweaks. Here is how I currently have it setup on the NVIDIA Control Panel -> Manage 3D Settings -> Global Settings:

Ambient Occlusion: Off
Anisotropic Filtering: Application Controlled
Antialiasing - FXAA: Off
Antialiasing - Gamma correction - On
Antialiasing - Mode: Application Controlled
Antialiasing - Setting: Application Controlled
Antialiasing - Transparency: Off
CUDA - GPUs: All
DSR - Factors: Off
DSR - Smoothness: Off
Maximum pre-rendered frames: Use the 3D application setting
Monitor Technology: G-SYNC
OpenGL rendering GPU: Auto Select
Power Management Mode: Prefer maximum performance
Preferred refresh rate (XB271HU A): Highest available
SLI rendering mode: NVIDIA recommended
Sharder Cache: On
Texture filtering - Anisotropic sample optimization: Off
Texture filtering - Negative LOD bias: Allow
Texture filtering - Quality: Quality
Texture filtering - Trilinear optimization: On
Threaded optimization: On
Triple Buffering: Off
Vertical sync: Use the 3D application setting

I also have a 4K 60Hz monitor if you think I should test BF1 on that I can and return with the results.

ROHITGUPTA1975
How have you connected your card to the PSU?
The 2080 TI's require 2 8-pin connections each so I simply used two paired PCI-E connectors. This PSU has a ton of these paired connectors so I didn't get any two sets in particular.

Also, is your BIOS up-to-date? Is your card driver up-to-date?
I checked the BIOS date and looks like there is a newer one but only slightly newer. The one I have was released in mid September with the newer one maybe a week after that. Would such a slightly older BIOS make such a difference? I have the latest drivers from NVIDIA (416.34)

Does your AA (anti-alias) setting match with whatever the benchmarks have? (TAA?)
The benchmarks posted state that the games are run in whatever maximum the game allows. I'm assuming this is done in order for tests to be reproducible without adding any additional variables introduced on the gaming side of things.

If I might also add, on my older alienware, I put these two cards in just to see how it'd perform and I could really hear the fans work hard when running BF1. However when I put them in this new computer, they are a quite as a mouse. Not that I'm complaining about that but could that be an indication of the GPU's not being pushed hard enough in my new rig?

 

[rohitgupta1975]

Perhaps you should disable SLI first and investigate why you are getting low FPS with a single card.


What are your temperatures like at load? If you are exceeding the temp. limit, the core clock will get auto throttled and FPS will drop. What is the core clock at load?
What are the fans' speed at load? At high temperatures, they should have higher RPM (unless they have been inadvertently configured to stay at a fixed RPM). You said they are running quieter. Maybe they are not running at the adequate speed to cool down the GPU?

Have you run TimeSpy and seen your 'graphics' scores for a single card? in SLI? Are they comparable with what others have posted online? (If they are OK, then it's BF1 specific issue)

Check if the second card , when used individually (not SLI), gets you better FPS than the first. (Maybe one of the cards is not performing well)

When you said, "'I simply used two paired PCI-E connectors" did you mean that you used TWO SEPARATE cables (not a SINGLE cable that splits into a pair of connectors)?

 

[mortemas]

Hmmm, interesting scenario. 100 fps with sli and 60 without suggests a shift downward and something broadly affecting your performance overall. Something is holding it back. As rohit suggests, you need to check your stats and see if they are normal. Do you have MSI Afterburner and RTSS installed? If not, set up an overlay with gpu temp, clocks, gpu utilization, power, and same for the cpu. Also, run GPUz while gaming and go to the "sensors" tab and see what the reason is for "PerfCap" and that might tell you what's limiting you. The average 2080 ti out of the box should hit around 1900 MHz and temps are usually a bit above 70c or so. GPU utilization should be at 100% for a single card, but I'm not sure about SLI. My i7-8700k runs at about 50-60c during gaming. Not sure what the cpu clock/core usage should be with high fps gaming though, since I'm at 4k/60Hz, although the TechPowerUp benchmarks I referenced above were achieved with an i7-8700k at 4.8MHz so yours should be able to handle it.

One more thought, no frame limits set in-game, right? Also, try without GSync and see if that has any effect.

 

[soul_collector]

Mortemas / Rohitgupta1975
I did everything you mentioned, checked the cables, downloaded MIS Afterburner with RTSS (including OC Scanner with the new Afterburner Beta), GPU-Z and so much more.

I think I know what's going on. As part of the assignment of settings done by GeForce Experience, Nvidia sets Resolution Scale at 200%. Setting that at 100%, which I believe just allows the game to run at the native resolution set, saw a skyrocket in my FPS. On my 4K, I'm hitting 120+ and on my 1440, 150+. Sure, it doesn't look as insanely sharp but I'm guessing that all of these benchmarks I'm seeing are also keeping the Resolution Scale at 100%? Is that right?

 

[soul_collector]

Just to update, I ran Witcher 3 on 4K on Ultra with both 2080TI's in SLI and definitely not seeing the frame rates advertised by benchmarks out there. There does not seem to be any "Resolution Scale" option in the game so that's not what's causing the lag in performance. I'm getting around 80 to 90 FPS which is not a huge deal since that's still great numbers but quite far from the 140+ cited elsewhere.

Looking at GPU-Z, I'm seeing different items listed in 'PerfCap' with one card showing Reliability Voltage and the primary card showing Temperature and Total Power Voltage. I used the OC Scanner setting in MSI Afterburner Beta to try to get the best OC setting without potentially burning the cards out.