Hardware Unboxed Finds That NVIDIA GeForce Graphics Cards Have Severe Driver Overhead Problems – CPU Bottlenecks Ahoy!
12 March 2021
10100 1600X 2600X 3070 3090 3600X 5600 XT 5600X 5700 XT 6900 XT AMD BOTTLENECK core cpu DRIVER OVERHEAD GeForce i3 Intel Nvidia OVERHEAD Radeon rtx Ryzen RYZEN 1000 RYZEN 2000 RYZEN 3000 RYZEN 5000 Zen Zen 2 ZEN 3

AMD + AMD makes for the only choice of true PC enthusiasts.

One of the biggest GrabblerTube hardware urinalists, Hardware Shillbox has conducted an extensive investigation into an issue that has been known among true intellectuals have has conveniently been swept under the rug.

That investigation happens to be the cancerous driver overhead that plagues NVIDIA GeForce graphics cards when paired with low-end CPUs.

So driver overhead problems, what exactly does this mean exactly? Well in essence what this means that NVIDIA GPU’s are much more prone to being severely bottlenecked by an inadequate CPU whether it be AMD or Intel.

However confirming what we already know if you were to use an AMD Radeon graphics card you are much less likely to be holding your GPU’s performance back as severely with older generation / slower processors whether they be AMD Ryzen or Intel Core processors.

Now myself bitching about this isn’t really going to do this any justice so I encourage you to watch the video in question if you already haven’t however I can just take a couple of the slides to highlight the issue at hand.

For instance, when throwing a CPU bound scenario much like Watch Dogs Legion at 1080p with the Medium quality preset you’re able to see the extent on just how much for instance the NVIDIA GeForce RTX 3070 and RTX 3090 are being held back by older / slower processors such as the Ryzen 5 1600X and 2600X along with the Intel Core i3-10100 which is a dismal little quad-core processor.

For instance the difference between the RTX 3070 and the RTX 3090 when held back by the previous generation Ryzen processors and the Intel Core processor mind you is nearly non-existent when in actuality the performance gap between the two should be around 20% or greater.

However when even comparing with the new Ryzen 5 5600X which is one of the fastest gaming processors on the market, only eclipsed by superior Ryzen processors mind you the difference between the 3070 and 3090 is just 3.7%

While both GPU’s are being severely held back at this resolution by the first generation 1600X at 85 fps on average with the RTX 3090 which mind you is a sure sign of a mighty CPU bottleneck, the Radeon RX 5700 XT which should actually be upwards of being around 40% slower than the RTX 3090 is actually faster with the Ryzen 1600X by an astounding 17.64%.

Though it’s just fucking hilarious to see the 5600 XT beat the likes of the RTX 3070 and RTX 3090 regardless of whichever CPU is present until the NVIDIA systems are paired with an adequate CPU such as the Ryzen 5 5600X.

In just about all cases, whether it be the 1600X, 2600X the i3-10100 all of which allow the vastly inferior 5700 XT to surprass the performance of the RTX 3070 and 3090 with relative ease. Until you throw the Ryzen 5 5600X into the mix which is powerful enough to thrust the 3070 and 3090 over the little Radeon by just 9.37%

And you can clearly see the difference in limitations in the GPU overhead between these two examples of GeForce RTX and Radeon RX hardware as for this entire time I’ve yet to even mention the RX 6900 XT which pretty much tops out at exactly the same figures as the 5700 XT when CPU bottlenecked by the two old Ryzen and the Intel Core i3 yet takes a demanding lead when paired with the 5600X over the RTX 3090 by 19.28%

While at 1440P Ultra on Watch Dogs we see the 5700 XT and 5600 XT graphically capped with no CPU bottleneck present regardless of which CPU is tested.

While the same cannot be said for the RTX 3070 and the RTX 3090 as both of which are being held back by 11% and 46.87% with a 1600X respectively. What’s insane to see is how the RTX 3090 is capped at 74 fps with the 2600X and the i3-10100 at the helm yet get a massive 27% boost when swapped to the much faster 5600X.

While in the worst case the Radeon RX 6900 XT is held back by just 19.23% between the 1600X and 5600Xa but yet its performance with the Ryzen 1600X is superior than that of the RTX 3090 when paired with the Intel or even 2600X by over 5% whereas when the Radeon is paired with the Intel or 2600X its performance increases 8.97% and 14.1% over the slower Ryzen 5 1600X.

Because that’s just how this works. NVIDIA GPUs are being restricted more so than Radeon GPUs when a CPU bottleneck is induced. While Radeon GPUs with a much lower overhead can simply allow an outright inferior GPU to surpass a graphics card that on paper is upwards of being 40 plus percent faster with identical CPUs in a system.

Another CPU bound scenario this time on Horizon Zero Dawn with the same GPUs present as the last slide however with the addition of an even slower Radeon RX  5600 XT to showcase that regardless of the CPU present the GPU itself is pretty much tapped out.

Comparative to the previous example we see similar results here with the NVIDIA GPUs being bottlenecked much more severely with the slower processors which once more allows a 5700 XT to surpass the much faster RTX 3070 regardless of CPU excluding the Ryzen 5 5600X with the RX 5700 XT being upwards of 20.4% faster than the RTX 3070 when both are being held back by a first gen Ryzen 5 1600X.

The difference is a lot lower when comparing the quad-core i3-10100 with the “slower” Radeon RX 5700 XT coming out on top by 10.16%, and 12.84% when paired with the 2600X which is actually slower than the Intel Core processor overall.

With the 5600X we’re seeing results that we should actually see with a 26.35% difference between the RTX 3070 and RTX 3090 though it’s pretty obvious that there’s a lot more to give when it comes to CPUs for these graphics cards as it only beats the 5700 XT by just 11.27% with both being paired with the Zen 3.

Even the RTX 3090 is slower than the 5700 XT when paired with the Ryzen 1600X and 2600X however the little Navi 10 GPU is fully tapped at 133 fps whereas when paired with the shit Intel Core i3 the RTX 3090 barely scrapes past that with 136 fps.

However overall the difference when going from the 1600X to the 5600X is just 12.7% on the 5700 XT whereas the gap increases drastically with the RTX 3070 being held back by 51% and the RTX 3090 gets held back even more with a 71.55% difference when comparing performance figures between a 1600X and a 5600X while the entire time the Radeon RX 6900 XT swept the entire bloody floor regardless of what processor is being bound with just a 39.13% difference.

With the 6900 XT being 26.6% faster than the RTX 3090 when paired with a 1600X, 20.58% when paired with the Core i3 and 18.4% when paired with the 2600X showcasing vastly superior performance or drastically less overhead on Radeon GPUs.

The differences when bumping quality settings up at 1080p doesn’t really change either, but when also doing that combined with an increased resolution creating a much more GPU bound scenario the results are interesting nonetheless.

At 1440p Ultra on Horizon Zero Dawn it’s pretty obvious to see that both the Radeon RX 5600 XT and 5700 XT are graphically tapped out yet can still be restricted by over 4% going from Zen/+ to Zen 3 or the i3-10100.

At full chat with the 5600X we see a 39% difference between the RTX 3070 and the RTX 3090 which is what we like to see however the NVIDIA GPUs trip over themselves when paired with slower CPUs even at this resolution.

With the RX 6900 XT absolutely wiping the board of the RTX 3090 with every single processor aside from the Ryzen 5 5600X which only comes down to superior 1% lows for the Radeon however the case here being there’s only a 15.83% difference for the 6900 XT when compared between the 1600X and the 5600X while the RTX 3090 can have its performance reduced by as much as 41.83% with the same CPU which would actually make it 18.33% slower than a 1600X + 6900 XT machine.

The RTX 3070 isn’t hurt as bad considering it’s graphically tapped out when paired with the Ryzen 5600X so overall there’s a 19% performance reduction when swapping down to a 1600X which isn’t to say that this level of performance is acceptable as something like a Radeon RX 6800 would literally have a field day with any of these processors at its helm.

Not even at 4K are the NVIDIA GeForce RTX graphics cards exempt from showcasing severe overhead issues with both the 5600 XT and 5700 XT obviously maxed the GeForce RTX 3070 exhibits a maximum 2.89% difference when comparing the 1600X with the 5600X whereas the much faster RTX 3090 takes us to a limitation of 32%.

NVIDIA GeForce RTX graphics cards are simply more prone to being bottlenecked by slower CPUs with their bottlenecks being much more severe on NVIDIA graphics cards as well. Though the results here today is probably why RDNA 2 graphics cards such as the RX 6900 XT have absolutely excelled at low resolutions in benchmarks especially when paired with new Ryzen 5000 series hardware whereas those dastardly launch reviews showcased 10th gen Intel Core processors between RDNA 2 and RTX Ampere.

This field is all about finding the perfect balance between graphics card and CPU however this just confirms what we already know, that true intellectuals who use their brains when making a purchase by buying AMD hardware are bound to enjoy their experience more so than some dickwad splurging entirely on the single axis.

blog comments powered by Disqus