Your GeForce card is not as fast as you think it is. [Driver Overhead Test] HWUB

Never played this game, but having a crocodile as a pet while you are on a killing spree is such a cool thing.

 

You proved him correct... unless you meant to.

 

Clearly not, thanks for validating the title.

 

None of this matters jack if the game has DLSS/FG.

Why is this even a discussion? Things are really getting slow over at AMDunboxed.

 

Anyone stupid enough to use 4090 or 7900XTX at 1080p for single player games should not be using a PC at all

Such people should buy consoles

Also AMD Unboxed Steve would never dare to post anything against AMD, like this one saying 7900XTX is a furnace when playing competitive games at competitive settings LOL

 

funny, the other Steve just released this recently as well

 

I love the insecurity guys, please feed more.

 

Only in some workloads, in games its about 30.7tflops

 

Very cool to see Intel moving forward.
Looks like genuine efforts.

Eventually they will leave AMD treading water in the GPU market.

AMD needs some new blood.
They cant rely on new titles leaving out DLSS/XeSS forever, and shouldnt.

 

Says who? What if the user wants more frames, shouldn't he have the final say if there is more juice to squeeze?

 

Hm...thread title is also wrong, it should be "Your Geforce is faster than you think it is", since 4090 is a whopping 39% faster than 7900XTX at 4K

 

They had to do 360p resolution in order to reveal the CPU bottleneck. And no, "users IRL don't play in that resolution, according to Steam data blablabla" isn't an excuse to use 1080p when you are trying to reveal scientific data.
Just like how 7900xtx was faster than 4090 in Star Wars Jedi Survivor 1.0 due to immense CPU bottleneck, we'd be able to see a relevant data.
I'd wager Nvidia would choke the CPU before AMD would.

 

Thanks for proving me right. I have a 4090 and a 7800x3d and I play FC6 at max settings 4K and even without RT it still doesn't maintain 120fps and in that video the fps drops into the 80's in parts.

 

A shame HWU didn't use this tool to show driver overhead graphs, the video was disappointing considering Intel contributed to this open source tool to determine actual driver overhead.

 

I think this explanation makes the most sense, as the performance varies too much from game to game and even from different locations within a game to be explained by *driver overhead."

Games with good CPU optimization don't have these problems as they easily supply the GPU with enough data to achieve a high utilization rate.

 

CPU overhead with NV GPUs on midrange/ lowerend CPUs is brutal. Especially with stuff like 5600x/12400f/13400f. Anything higher and including 3060ti will incur massive bottlenecks in certain regions of a wide variety of games. Usually the first couple sections and tutorial sections are super GPU bound and super CPU light (Hogwarts Legacy, Jedi Survivor and more. Probably some cheap trick/tactic to lure people into thinking they won't massive performance issues). But further into the game, these midrange/lowend CPUs fail colossally in most recent titles.

This is my problem with these tests. They do not cover actual end user experiences. Yes, you can still find issues here and there with a review sweep but real problems occur at actual gameplay, in further sections.

Take Cyberpunk's Jig Jig street. Even a 5800x drops to 40 or so frames there with ray tracing enabled. And NVIDIA's driver overhead is full on assault in that specific case. But no one focuses on such instances.

 

Radeon and Nvidia has the same CPU-limited performance in BG3

 

isn't it because cdpr haven't totally fixed the amd smt issue on the game and you had to enable it through some mods?

 

I don't know what the appropriate or most correct term is for the behavior -- maybe calling it driver overhead is not really what it is (there seems to be a lot of debate over this point) -- but it's apparent in a good number of titles, when CPU bound, Nvidia is doing worse than AMD. Is it all games? No. Is it a lot of DX12 and Vulkan games? Seems to be the case going off the tests I've seen at least. A 15-20% difference in CPU bound performance is no joke. DX11 of course Nvidia GPUs still do better than AMD when CPU bound -- usually this AMD lead when CPU bound seems to happen more commonly in low level API games from what I've seen.

 

Yeah, it's not really consistent for all games it would seem. Some games AMD pulls way ahead in CPU bound performance (15-20% even), in other games they're about even, and in especially DX11 games Nvidia blows AMD out of the water when CPU bound though that's become much less relevant these days.