1. PC had SSD for years, we don't need to improve anything!
Shattered by RTX IO: specific hw/sw solution to solve the IO problem.
2. PC have beefy CPUs to decompress anything!
Shattered by RTXIO: otherwise 24 CPU cores are needed at full speed just to decompress 7->14GB/sec.
3. PC have huge RAM pools, all assets are loaded into RAM. No need for fast SSD streaming.
Shattered by miniscule VRAM upgrades and RTX IO, NV seems to think that using SSD as a VRAM extension is viable.
4. RDNA2 RT cores are a bolt-on solution. Turing RT is superior!
Shattered by the mention that Compute/Graphics pipelines cannot run in parallel with RT on Turing and Ampere fixes it. RDNA2 had it all along.
Anybody has other examples?
Shattered by RTX IO: specific hw/sw solution to solve the IO problem.
2. PC have beefy CPUs to decompress anything!
Shattered by RTXIO: otherwise 24 CPU cores are needed at full speed just to decompress 7->14GB/sec.
3. PC have huge RAM pools, all assets are loaded into RAM. No need for fast SSD streaming.
Shattered by miniscule VRAM upgrades and RTX IO, NV seems to think that using SSD as a VRAM extension is viable.
4. RDNA2 RT cores are a bolt-on solution. Turing RT is superior!
Shattered by the mention that Compute/Graphics pipelines cannot run in parallel with RT on Turing and Ampere fixes it. RDNA2 had it all along.
Anybody has other examples?
Last edited: