By using this site, you agree to our Privacy Policy and our Terms of Use. Close
Pemalite on 30 July 2020
Edit Thread
Captain_Yuri said:
AMD Radeon Instinct MI100 ‘CDNA GPU’ Alleged Performance Numbers Show Its Faster Than NVIDIA’s A100 in FP32 Compute, Impressive Perf/Value

https://wccftech.com/amd-radeon-instinct-mi100-cdna-gpu-alleged-performance-benchmarks-leak-faster-than-nvidia-ampere-a100/

CDNA is basically just an enhanced GCN part geared even more towards compute.
GCN was always amazing at compute operations.

Captain_Yuri said:

Looks to be the Crysis of this gen jeez. Brings 2080Ti + 3900X to it's knees at 4k. Overall looks to be Intel + Nvidia favoured. This is still in beta so there could be performance improvements on release. They will apparently add even more visual enhancements such as ray tracing and this is probably the first implementation of the "cloud" that actually makes sense. It is very scalable on both GPU and CPU though.

Steam page is also up:

https://store.steampowered.com/app/1250410/Microsoft_Flight_Simulator/

It made me moist, lets put it that way.

Literally "Petabytes" worth of texture and mesh data exists and is loaded in via the cloud... Compression can only take you so far when dealing with such expansive amounts of data. *looks at console SSDs*

JEMC said:

Also, that MI100 article... well, that will be the first card with using the new CDNA architecture from AMD, which will be different from the RDNA for the gaming parts. Therefore, we shouldn't take any conclusions about Navi based on it.

Also, and from Videocardz article:

The website further claims that the MI100 will feature 120 Compute Units. Assuming that the CDNA architecture features 64 Processors per Cluster, it would mean that the accelerator has 7680 cores in total (if each Compute Unit had 64 cores). We are intentionally not calling them Stream Processors, because we are unsure if this is the exact name for this architecture.

What does not make sense, however, is the 42 TFLOPs claim on the slide. This would put make MI100 more than twice as fast as NVIDIA Ampere A100 (19.5 TFLOPs). To achieve 42TF, it would require either: 7680 cores running at 2.75 GHz or 15360 cores running at 1350 MHz. The latter would suggest that each CU contains 128 cores, not 64.

Or it has 240 Compute units @64 stream processors each.

Or it could be none of the above.

192 CU's @ 1.72Ghz could get you there as well... That would essentially be 3x Fury or Vega 64.

We need to remember that CDNA is ditching allot of the rasterization baggage and dialing home the compute, that's a ton of free extra transistors that can be dedicated to the task of more shader cores.

This will be a big win for Cryptocurrency though.

JEMC said:
vivster said:

That's not too terrible.

Though we shouldn't forget that newer GPUs will always have much higher improvements on the most extreme scenarios. In moderate loads it's probably not much higher than 30%.

And we should also remember that the 2080Ti was actually CPU limited at anything lower than 4K by any processor except the 9900K, and the newer and more powerful cards could fare the same problem, probably even worse.

Zen 3 is coming this year as well.






--::{PC Gaming Master Race}::--