Why Nvidia’s new GPU performs worse than integrated graphics | Digital Trends


One would think that a GPU that costs over $40,000 will be the best graphics card for gaming, but the truth is much more complex than that. In fact, this Nvidia GPU can’t even keep up with the integrated graphics solutions.

Now, before you get too upset, you should know that I’m referring to Nvidias H100, which houses the GH100 (Grace Hopper) chip. It’s a powerful data center GPU built to handle high performance computing (HPC) tasks and not to power PC gaming. It has no display outputs, and despite its large capabilities, it also has no coolers. That’s because, again, you’d find this GPU in a data center or server, where it would be cooled with powerful external fans.

While it only has 14,592 CUDA cores (which is less than the RTX 4090), it also has an insane amount of VRAM and a huge bus. In total, the GPU features 80GB of HBM2e memory, split into five HBM stacks, each connected to a 1024-bit bus. Unlike Nvidia’s consumer GPUs, this card still has NVLink, which means it can be plugged in to work seamlessly in multi-GPU systems.

The question remains: why exactly is this type of GPU so bad in general use and gaming?

To prove the case, YouTuber Gamerwan received four of these H100 graphics cards to test and decided to pop one into a regular Windows system to check its performance. This was a PCIe 5.0 model and had to be paired with an RTX 4090 due to a lack of display outputs. Gamerwan also 3D printed a custom designed external cooler to keep the GPU running smoothly.

It takes some work to get the system to recognize the H100 as a proper GPU, but once Gamerwan managed to navigate the hurdles, he was also able to turn on ray tracing support. However, as we’ll find out later in testing, there isn’t much support for anything else on a non-data center platform.

In a standard 3DMark Time Spy test, the GPU only managed 2,681 points. For comparison, the average score for the RTX 4090 is 30,353 points. This score places the H100 somewhere between the consumer GTX 1050 and GTX 1060. More importantly, it’s nearly equal to AMD’s Radeon 680M, which is an integrated GPU.

Gaming tests also fared poorly, with the graphics card averaging 8 frames per second (fps).Red Dead Redemption 2. The lack of software support rears its ugly head here, although the H100 can run at a maximum of 350 watts, the system can’t seem to push it beyond 100W, resulting in severely reduced performance.


There are a few different reasons for this poor show of gaming powers. For one thing, while the H100 is an ultra-powerful graphics card on paper, it’s very different architecturally from the AD102 GPU that powers the RTX 4090. It has only 24 raster operating units (ROPs), which is a significant downgrade from to the RTX 4090’s 160 ROP. Additionally, only four of the 112 texture processing clusters (TPCs) can render graphics workloads.

Nvidia’s consumer GPUs get a lot of software-side support to perform well. This includes drivers, but also system support from developers in both games and benchmark programs. There are no drivers that optimize the performance of this GPU for gameplay and the result is, as you can see, extremely poor.

We’ve seen driver power before with the Intel Arc, where the hardware remained the same, but improved driver support provided performance gains that made the Arc an acceptable choice if you’re shopping for a budget GPU. With no Nvidia Game Ready drivers and lack of access to the rest of the Nvidia software stack (including the ever-impressive DLSS 3), the H100 is a $40,000 GPU that has absolutely no business running any kind of game.

Essentially, this is a computing GPU and not a graphics card in the same way most of us know it. It was built for all types of HPC businesses, with a strong focus on AI workloads. Nvidia maintains a strong lead over AMD when it comes to AI, and cards like the H100 play a big part in that.

Editor’s Recommendations

#Nvidias #GPU #performs #worse #integrated #graphics #Digital #Trends

Leave a Comment