Benchmark Drops/Dips

[CARLS_CAR_LOL]

Not sure where else to post this…

Do these benchmark results look abnormal to anyone? My concern are the random dips to 40-90fps. Also, sometimes I don’t get a “full render” in the benchmark.

I’ve run frame view and checked logs on a few multiplayer games and I don’t see any frame drops, but I do get strange hiccups where it feels like I lost frames.

DDR4 @ 3500MHz (stable)
No OC on CPU other than turning on XMP for the RAM.
NVCP Settings: Gsync on, Vsync on, triple buffer on, maximum power, ultra low latency on

If you need any further PC specs, let me know.

ASYNC on

ASYNC off

[GhostofDelta2]

Why are you benching with vsync on?

@AC_Wireless ar e you still having performance issues with your 3090?

[CARLS_CAR_LOL]

does it matter if it’s on or off? in any case, the results don’t change if i turn it on/off.

i read @ blurbusters that having vsync ON with gsync is the best way to go. whether or not it is on in-game doesn’t matter because NVCP settings override it.

[GhostofDelta2]

@CARLS_CAR_LOL

does it matter if it’s on or off? in any case, the results don’t change if i turn it on/off.

Your first pic shows a cap of 144 and sync on, which means it wouldn’t render any frames above 144. Which means any performance that would hit above that target is lowered. to that cap, which in turn impacts on your average FPS (but prevents screen tearing).

Your second pic has it (effectively) off (as you have set to unlimited FPS at the point where it should stop) and has higher average FPS and higher lowest 5% margins.

@CARLS_CAR_LOL

i read @ blurbusters that having vsync ON with gsync is the best way to go.

This depends greatly on what you are wanting for each game and what performance you can hit. There isn’t a “best” setting. Gsync and Vsync on will prevent tearing but will add input latency. Gsync on but vsync off (where you are hitting above your monitors refresh rate) won’t have any added input latency (and can minimally decrease it) but then you risk screen tearing.

Triple buffer Vsync adds even more input latency (though it should be fairly minimal). Perhaps turning this down to a single frame would help your issue? It may not but I would say this is worth a shot.

@CARLS_CAR_LOL

whether or not it is on in-game doesn’t matter because NVCP settings override it.

As long as you have it set to do this, yes it will (barring any stupid bugs).

Are you using a SSD or a Hard drive? My friend with a 3090 and similar performance issues to you has a Samsung 970 evo plus and has these issues so I don’t think drive speed is an issue here but figure it is worth asking, and if you both have the same SSD and same GPU there may be an issue here.

[GhostofDelta2]

Oh sorry forgot to mention, can you try with Vsync “OFF” everywhere (in game and Nvidia control panel) and see if you get the same dips.

[CARLS_CAR_LOL]

Thanks for your responses man, I’ll try messing around with the settings a bit more. I’ll drop a line and @ you when I get done.

I have Samsung 970 Pro 512GB, by the way.

[GhostofDelta2]

Nice SSD mate!

And please let me know how it goes

[CARLS_CAR_LOL]

@GhostofDelta2 Thanks man

I turned off Vsync in game and in NVCP. I also turned off triple buffer in NVCP, and turned low latency to On (1 frame). I tested with Vsync OFF/ON in-game. I got nearly identical results. Notice that results look nearly identical to my previous settings as well. However, with my previous settings, I saw less of the “raining” effect on the graph and more consistency.

As for input latency, I don’t really notice anything horrible. I’d rather have a smooth experience over screen tearing.

My question is, is it abnormal to see the graph looking it’s “raining”? That’s my main concern here.

[GhostofDelta2]

You got slightly better results but its within margin of error, and it would be a similar story with the input latency, so I would set things however you prefer.

I will link a few benchmarks for gears 5 for comparison

I don’t think there is a lot to worry about on your results, the dips do sound annoying though, but do you get them in campaign?

[CARLS_CAR_LOL]

Thanks for tagging along on this ride with me.

Those graphs look a lot cleaner and more consistent than mine. I know I’m getting phenomenal results, however in comparison, my plot looks far more erratic than any of those which leads me to believe there may be an issue. Am I wrong to assume this?

Edit: Haven’t played campaign since putting in the 3090. Actually, haven’t touched campaign since I beat it back after it released.

[GhostofDelta2]

It does look a bit odd, but if it isn’t impacting on ingame performance I wouldn’t worry too much

Edit: If it does worry you perhaps try a clean install of the drivers if you haven’t already.

[III_EnVii_III]

Your GPU is all over the place,

Similar to my 3090,

I believe some cards are having issues whereas others aren’t, the drivers don’t help either.

The RTX 3000 on the whole don’t seem to work smoothly with Gears.

@IDONTuseMODz has a near flawless benchmark.

Did you do a TimeSpy benchmark?

Also,

VSYNC should be on in the NCP, not in game.

You need to limit in game FPS 3 below your max RR.

Also,

Don’t set a minimum FPS.

A working 3090 should get around 180-195 FPS at 1080p Ultra.

A not working so well will get around 130-150, with lots of dips.

[IDONTuseMODz]

Thanks for the tag, @III_EnVii_III, I don’t really have anything to add here other than the “rain” effect you’re seeing is definitely abnormal, at least compared to my benches.

These are all the pertinent parameters I can remember from when I made those benchmarks:

• 9900KS locked to 5GHz, RAM @ 3600 16-18-18-38, all C-States disabled
• Mild GPU overclock of +50 Core, +450 Memory
• Power Management Mode set to Maximum Performance in NVCP
• All background processes closed, Game Bar/background recording disabled, etc
• Reduced Buffering set to ON and Async Compute set to OFF in game

Bear in mind, these were very much “best case scenario” type benchmarks but even without most of these tweaks my graphs don’t “rain” like yours do. They remain pretty tight, just lower average and 5% minimums. I’d maybe try some of the above changes in addition to making sure your card isn’t thermal throttling and maybe a fresh install of drivers.

[III_EnVii_III]

Yeah,

That rain effect as you put it, is showing an issue somewhere, it shouldn’t be like that.

My Ti was similar to @GhostofDelta2 - I nice tight consistent line.

Perhaps optimised settings and a clean install is needed.

Something definitely isn’t right.

What’s odd is my CPU frames dropped a lot too, without any changes, after I installed the 3090.

Will see what happens when I get the 6800XT / 6900XT.

[CARLS_CAR_LOL]

@III_EnVii_III @IDONTuseMODz

Ok, so I think it may be my RAM. I switched to my CL 14 kit (16 GB) and I get some amazing results. The 4K bench was a little lower than I was expecting, but I’ll be doing a small OC on it when I put it under water.

I went and turned off Gsync as well and just let the GPU go wild. I also tested w/Gsync ON and Vsync ON (in NVCP).

As you can see, much more stable graph. The second one is with Gsync on and limited to 141 FPS.

Perhaps my CL16 32GB memory kit has some issues, might need some tweaking. MEMTEST86 shows it as having no errors and it’s a kit that is compatible with my motherboard. I understand CL14 is faster, but I didn’t think the difference would be this vast… The results are night and day.