What game do you think represents the new Crysis?
What game do you think represents the new Crysis?
People in the community are still debating what Crysis represents today. Some think it’s more about being unoptimized, while others highlight how its graphics feel ahead of their time. Previously, I favored Cyberpunk for its newer PT mode, though it still struggles on older hardware like a 4090. The launch bugs were also a concern. Now, Alan Wake 2 seems to offer better visuals, but overall performance is lacking, especially with GPUs lacking mesh shader support.
Crysis was born when everyone assumed CPUs would reach 10GHZ in just a few years. It was the CPU expectations that made it so difficult to run. Not entirely the graphics side. There shouldn't be any games with such high ambitions. Sure, they look impressive now, but they don't completely strain the CPU. Most are designed for consoles lacking a 6ghz processor.
The issue with Alan Wake 2 is that it requires just over 7GB of VRAM at the lowest resolution. However, the VRAM consumption changes from 10GB to 12GB when moving from 1080p maximum to 4K maximum. It seems unlikely that the game would require more than a 20% rise in VRAM given that it is handling four times as many pixels at 4K.
Many recent releases appear to suffer from significant optimization problems. It's unclear if any title truly struggles to run smoothly on top-tier hardware like Crysis did when it launched. When Crysis was released, my old 8800GTS with 640MB of RAM could manage it just a bit, around 40 FPS at 1024x768? That performance was only slightly behind the leading 8000 series model that was released shortly beforehand. I rarely played it because I worried my GPU would overheat, often reaching temperatures near 100°C while trying to run the game. Remember, back then 1080p was considered similar to today's 4K resolution. In essence, a game that struggles to maintain 60 FPS at 4K with a 4090 at high settings is what you're likely seeking.
I've heard that the GTX 260 was the initial single GPU capable of handling Crysis effectively. It seems Jedi Survivor's release at launch could have been suitable for the new game, since even though the graphics weren't groundbreaking (still decent), the title was having issues on a 4090 and adjusting settings didn't help.
Reviews still indicate the GTX 260 has issues at 1900x1200 with maximum settings, only achieving about 38 FPS without anti-aliasing. At 1280x1024 it seemed manageable for most users. Crysal was special because it was hard to run on computers when they were first released, yet it remained out of reach as resolutions rose. Most findings point to games using ray tracing, and those without RT or DLSS often hit CPU limits. However, if you accept RT in titles like "New Crysis," "Portal RTX," "Cyberpunk 2077" with RT, and "Metro Exodus" with RT, they generally perform well on a GTX 4090.
The only games I play with RT are Minecraft RTX and Jedi Survivor. I support using RT here because modern GPUs have improved significantly in handling traditional rasterization. Portal RTX achieves 26 fps at 4k native on maximum settings. I doubt my 6800 would handle Portal RTX at 1080p natively well. Edit: according to the techpowerup benchmark, a 6900xt—22% faster than my 6800—gets about 5 fps at 1080p max, which is concerning. Even the 7900xtx manages roughly 10 fps. I know this game is sponsored by Nvidia, so it shouldn't perform well on AMD cards, but I'm surprised a $1000 GPU can't reach 30 fps at 1080p max.
The 8800GTS, as noted by techpowerup, performs roughly 20% better than the 8800GTX. It also lags behind the GTX 260 by just 16%.
It's often difficult to define what performance should look like. You can't simply state that something "runs well" or is "poorly optimized" without clear performance goals. These goals can also be open to interpretation, especially if your targets are very high, such as 4K at 240FPS with maximum detail. Every game might then appear suboptimal or not fully optimized.
For instance, when adding a new physics model to an in-house flight simulator, we needed to demonstrate that the software was functioning as intended. This required thorough testing to verify our expectations. Writing requirements around such changes is essential, since improvements—even by a tiny margin—can be misleading if not measured properly.
The goal was to ensure the new physics model met specific tolerance levels compared to real-world data, based on standards like those from the FAA. However, focusing too much on technical specs can overlook practical issues.
Sometimes tweaking settings can drastically affect performance, regardless of algorithmic efficiency. For example, a calculation that seems fast under normal conditions might slow down significantly when increased to maximum settings. This highlights the need for careful consideration of data volume and its impact on speed.