News:

Willkommen im Notebookcheck.com Forum! Hier können sie über alle unsere Artikel und allgemein über Notebook relevante Dinge disuktieren. Viel Spass!

Main Menu

Post reply

The message has the following error or errors that must be corrected before continuing:
Warning: this topic has not been posted in for at least 120 days.
Unless you're sure you want to reply, please consider starting a new topic.
Other options
Verification:
Please leave this box empty:

Shortcuts: ALT+S post or ALT+P preview

Topic summary

Posted by S.Yu
 - June 21, 2020, 13:58:00
Quote from: pete on June 18, 2020, 01:07:36
Quote from: Hyperus on June 16, 2020, 04:28:20
Then where would the heat go to?
As I see it here, there is no way ot dissipate heat from that coprocessor

Maybe because there is a fan on each side of the board if you looked..

Quote from: S.Yu on June 17, 2020, 01:06:09
Sounds like a huge drop in efficiency.

Why is that, a full die dedicated to rasterization.
Well, basically what mvl said.
Posted by pete
 - June 18, 2020, 01:07:36
Quote from: Hyperus on June 16, 2020, 04:28:20
Then where would the heat go to?
As I see it here, there is no way ot dissipate heat from that coprocessor

Maybe because there is a fan on each side of the board if you looked..

Quote from: S.Yu on June 17, 2020, 01:06:09
Sounds like a huge drop in efficiency.

Why is that, a full die dedicated to rasterization.
Posted by S.Yu
 - June 17, 2020, 01:06:09
Sounds like a huge drop in efficiency.
Posted by Hyperus
 - June 16, 2020, 04:28:20
Then where would the heat go to?
As I see it here, there is no way ot dissipate heat from that coprocessor
Posted by Kirill
 - June 15, 2020, 15:09:10
Pretty sure this will be only RTX 3090 specific feature.
Previously Nvidia called x90 only dual GPUs.
Posted by mvl
 - June 15, 2020, 14:07:45
Quote from: dasdkas on June 15, 2020, 12:51:52
This is interesting but then why A100 is a monolithic die?

The A100 die does not feature RT cores, so this could be plausible considering this fact and this fact only.

Neverthless, with that being said, considering that a frame on Turing using RT and Tensor cores goes as follows: FLOAT --> RT CORE --> INT+FLOAT --> TENSOR CORE; it seems inefficient, that the RT part of the frame is offloaded, processed and then brought back again to run through the Tensor cores. Furthermore, a lot of gains will be achieved in the denoising part, meaning, using the Tensor cores.

But I'm no expert. So, maybe someone can add something to the discussion. I'm defnitely looking forward to what NVIDIA engineers come up with!
Posted by Rick
 - June 15, 2020, 14:00:04
I wonder how they can implement that on the mobile versions...
Posted by dasdkas
 - June 15, 2020, 12:51:52
This is interesting but then why A100 is a monolithic die?
Posted by Redaktion
 - June 15, 2020, 10:46:46
According to a Careteks video, NVIDIA might be deploying a dual-sided PCB design, with a ray-tracing traversal coprocessor residing opposite the GPU die. This could potentially provide the RTX 3080 and RTX 30980 with a massive boost to ray-tracing capabilities.

https://www.notebookcheck.net/NVIDIA-Ampere-GeForce-RTX-3080-and-3090-could-feature-a-ray-tracing-traversal-coprocessor-better-ray-tracing-performance-than-PlayStation-5-Xbox-Series-X-and-Big-Navi.470737.0.html