Question Upgraded 4090 System - Worse Performance

[Jmunk]

Sort of feel like I'm going a bit crazy her. Ever since upgrading to my FE 4090, I never really felt like I was getting the performance that I think I should have/everyone else talked about. At the time, I had a 9900k, so I thought maybe it was CPU related. Upgraded to a 7700x, still felt like I was underperforming. Today, I upgraded to a 7800x3d, and somehow, all my benchmarks are WORSE than on my 7700x. What in the world is going on here? Temperatures never went above 67C on 7800x3d during userbenchmark's test. Not to mention, in CoD MW2's in-game benchmark, I only went up 2 average frames, and it actually said I was GPU bottlenecked. Is that even possible?? Also, before some of you say "what's the difference between 312vs314 frames, this is the benchmark, not the actual in game frames you'll see in Warzone. There are plenty of locations in WZ2 where I'll drop below 140 frames.

Not to mention, I think it's more about the system not performing nearly as expected, more so than actual numbers. It feels like something is just off or something is potentially wrong, but would be interested in your guys' perspective.

7700x benchmark:

7800x3d benchmark:

CoD Benchmark (GPU Bottleneck):

Voltages/Temps during benchmarks:

 

[Crazyy8]

What about GPU temps? FE cards don't have as good coolers as board partners. Is Vsync or frame cap on?
edit didn't see the temps stuff

 

[Jmunk]

What about GPU temps? FE cards don't have as good coolers as board partners. Is Vsync or frame cap on?
edit didn't see the temps stuff

Just posted the screenshot of max GPU temps, nothing went above 60C.

Vsync off, no frame cap.

 

[Crazyy8]

Just posted the screenshot of max GPU temps, nothing went above 60C.

Vsync off, no frame cap.

Defective card? are GPU power cables fully plugged in? is PSU high enough wattage? What about drivers? your pic shows AMD radeon graphics, did u have radeon card prev and haven't uninstalled drivers?

 

[Jmunk]

Defective card? are GPU power cables fully plugged in? is PSU high enough wattage? What about drivers? your pic shows AMD radeon graphics, did u have radeon card prev and haven't uninstalled drivers?

I mean, how could I even tell it if it was defective? It seems to be odd that it runs everything "well" just what feels to me not, "4090 well", if that makes sense. PSU is the 1000rmx.

Nope, previous GPU was a 3080. I'm assuming that's just the onboard graphics?

 

[Crazyy8]

You could tell if it was defective if it performed noticeably worse than reviews and other samples. Other option is that you got screwed in the silicon lottery.

 

[Jmunk]

You could tell if it was defective if it performed noticeably worse than reviews and other samples. Other option is that you got screwed in the silicon lottery.

That's the thing, in the userbenchmark photos it says the 4090 is performing below potential but still smack in the average range. So like, what exactly does that tell us? How is it even possible that going from a 7700x to a 7800x3d reduces performance, and shows 98% GPU bottleneck, but it's still within "normal" performance range.

 

[Crazyy8]

That's the thing, in the userbenchmark photos it says the 4090 is performing below potential but still smack in the average range. So like, what exactly does that tell us? How is it even possible that going from a 7700x to a 7800x3d reduces performance, and shows 98% GPU bottleneck, but it's still within "normal" performance range.

Don't trust userbenchmark's results. They are biased towards Nvidia and Intel. Try cinebench and other benchmarking software that don't have bias.

 

[pete_101]

That's the thing, in the userbenchmark photos it says the 4090 is performing below potential but still smack in the average range. So like, what exactly does that tell us? How is it even possible that going from a 7700x to a 7800x3d reduces performance, and shows 98% GPU bottleneck, but it's still within "normal" performance range.

The 7700x is always going to beat the 7800X3D in CPU benchmark tests where 3D memory isn't as important as clock speeds. In gaming, the 7800X3D usually wins, but again sometimes clock speed is the winner, it all depends on the game.

 

[sycoreaper]

Your card only drew 357w.

How is the card being powered? There should at least be spikes of ~475w under benchmark load.

Edit: I was thinking of the GPU Rail Power.

 

[Jmunk]

Your card only drew 357w.

How is the card being powered? There should at least be spikes of ~475w under benchmark load.

Edit: I was thinking of the GPU Rail Power.

4090 is being powered with 4 seperate cables straight from the PSU (rmx1000), into the adapter.

Sorry, unfamiliar with rail power - so you're saying the rail power seems low?

I also noticed it seemingly topping out at around 350w, and saw the 4090 usually hits 450w under load. Could definitely be contributing to less than optimal performance but I'm not sure what the cause of that would be.

 

[sycoreaper]

4090 is being powered with 4 seperate cables straight from the PSU (rmx1000), into the adapter.

Sorry, unfamiliar with rail power - so you're saying the rail power seems low?

I also noticed it seemingly topping out at around 350w, and saw the 4090 usually hits 450w under load. Could definitely be contributing to less than optimal performance but I'm not sure what the cause of that would be.

I was backing away from my statement. GPU Rail Power is a separate value. Could you try everything again using HWInfo64? It shows more power readings.

Also, I was discussing in another thread if this mattered or not but I'm going to ask anyway:
Are you using a end to end 12VHPWR cable or are you using the 12VHPWR to PCIe adapter? If using the adapter, do you have dedicated PCIe plugs connected? In other words, PCIe cables usually have the main cable and then a looped "pigtail". Make sure you're Not using a pigtail on the adapter.

 

[kerberos_20]

wattage depends on game/benchmark used, but even if theres some power limit going on, it has minimal impact on gpu performance

FE card rarely touches 450watts



if OP is GPU bottlenecked, then replacing CPU obviously doesnt improve GPU performance

There are plenty of locations in WZ2 where I'll drop below 140 frames.

if its in multiplayer locations, that could be from CPU, but we can just guess as you didnt check it

 

[Jmunk]

I was backing away from my statement. GPU Rail Power is a separate value. Could you try everything again using HWInfo64? It shows more power readings.

Also, I was discussing in another thread if this mattered or not but I'm going to ask anyway:
Are you using a end to end 12VHPWR cable or are you using the 12VHPWR to PCIe adapter? If using the adapter, do you have dedicated PCIe plugs connected? In other words, PCIe cables usually have the main cable and then a looped "pigtail". Make sure you're Not using a pigtail on the adapter.

So I'm obviously using the 4-port adapter that came with the 4090, and I'm using 4 separate PCIe plugs plugged into that adapter, straight into the PSU. I'm not daisy chaining anything.

Hopefully I answered your question.

 

[Jmunk]

wattage depends on game/benchmark used, but even if theres some power limit going on, it has minimal impact on gpu performance

FE card rarely touches 450watts



if OP is GPU bottlenecked, then replacing CPU obviously doesnt improve GPU performance


if its in multiplayer locations, that could be from CPU, but we can just guess as you didnt check it

Is it honestly even possible that a 7800x3d is being bottlenecked by a 4090, though?

 

[kerberos_20]

Is it honestly even possible that a 7800x3d is being bottlenecked by a 4090, though?

with uncapped framerate? yes thats normal, either GPU or CPU bottleneck
nothing like balanced where both CPU/GPU are maxed at same time exists

 

[Jmunk]

with uncapped framerate? yes thats normal, either GPU or CPU bottleneck
nothing like balanced where both CPU/GPU are maxed at same time exists

I understand there will always be a bottleneck, I'm saying I just can't believe that my 7800x3d, if everything is normal, is being bottlenecked by a 4090.

 

[sycoreaper]

@kerberos_20 @Jmunk

Teach me for not looking closer at the original screenshot. I swore I saw a different resolution.

I agree @kerberos_20 , CPU bottleneck is very possible.

 

[Unolocogringo]

Some games and benchmarks will perform a little worse with the 3d models.
They have a lower max boost clock.
Some games GHZ is king and others cache is king. Some dosent make too much difference. [well optimized ]
In your benchmark the video card was the slowest part 98% of the time and the CPU 2% of the time.
Can you run it with no upscaling enabled?

Userbenchmark gives you an idea of system performance compared to same parts.
If you know how to read their graphs and ignore everything else on the page.
Their conclusions are EDIT EDIT EDIT.
Pre edited for sensitive viewers.

 

[drivinfast247]

Don't trust userbenchmark's results. They are biased towards Nvidia and Intel. Try cinebench and other benchmarking software that don't have bias.

Bias aside, his CPU isn't working properly.

 

[drivinfast247]

Sort of feel like I'm going a bit crazy her. Ever since upgrading to my FE 4090, I never really felt like I was getting the performance that I think I should have/everyone else talked about. At the time, I had a 9900k, so I thought maybe it was CPU related. Upgraded to a 7700x, still felt like I was underperforming. Today, I upgraded to a 7800x3d, and somehow, all my benchmarks are WORSE than on my 7700x. What in the world is going on here? Temperatures never went above 67C on 7800x3d during userbenchmark's test. Not to mention, in CoD MW2's in-game benchmark, I only went up 2 average frames, and it actually said I was GPU bottlenecked. Is that even possible?? Also, before some of you say "what's the difference between 312vs314 frames, this is the benchmark, not the actual in game frames you'll see in Warzone. There are plenty of locations in WZ2 where I'll drop below 140 frames.

Not to mention, I think it's more about the system not performing nearly as expected, more so than actual numbers. It feels like something is just off or something is potentially wrong, but would be interested in your guys' perspective.

7700x benchmark:

7800x3d benchmark:

CoD Benchmark (GPU Bottleneck):

Voltages/Temps during benchmarks:

What CPU cooler? BIOS up-to-date?

 

[Crazyy8]

Bias aside, his CPU isn't working properly.

Yes, and using proper benchmarking software can help diagnose the issue, userbenchmark does not like AMD, thus AMD results might be wrong on userbenchmark.

 

[DSzymborski]

Yes, and using proper benchmarking software can help diagnose the issue, userbenchmark does not like AMD, thus AMD results might be wrong on userbenchmark.

Any brand-related bias, even if we accept for the sake of argument exists, is irrelevant here as the OP is comparing an AMD CPU to an AMD CPU, so the relative measures of Intel and AMD CPUs don't really matter. Bias and accuracy are two *very* different concepts; biased data can be highly accurate, while unbiased data can be very inaccurate.

 

[Jmunk]

What CPU cooler? BIOS up-to-date?

iCUE H150i ELITE CAPELLIX​

BIOS is actually out of date. Looks like there's 6+ versions I'm behind on. Funny enough, one of the notes for the update says "Performance optimized for Ryzen 7000X3D series processors".

Maybe worth flashing?

 

[Jmunk]

Some games and benchmarks will perform a little worse with the 3d models.
They have a lower max boost clock.
Some games GHZ is king and others cache is king. Some dosent make too much difference. [well optimized ]
In your benchmark the video card was the slowest part 98% of the time and the CPU 2% of the time.
Can you run it with no upscaling enabled?

Userbenchmark gives you an idea of system performance compared to same parts.
If you know how to read their graphs and ignore everything else on the page.
Their conclusions are EDIT EDIT EDIT.
Pre edited for sensitive viewers.

Sure thing, here's the same benchmark with no upscaling.