mirror of
https://github.com/gomson/TimothyLottes.github.io.git
synced 2026-08-04 22:58:49 +00:00
41 lines
1.3 KiB
HTML
41 lines
1.3 KiB
HTML
<html><head><link rel="stylesheet" href="style.css"></head><body><div class="page">
|
|
<h1>20150624 - AMD Fury X (aka Fiji) is a Beast of a GPU Compute Platform</h1>
|
|
<br>
|
|
|
|
|
|
|
|
<b>Compute Raw Specs</b>
|
|
<br>
|
|
Raw specs from wikipedia adjusted to per millisecond: comparing what the two vendors built around 600 mm^2 on 28 nm,<br>
|
|
<br>
|
|
<tt>
|
|
AMD FURY X: 8.6 GFlop/ms, 0.5 GB/ms, 0.27 GTex/ms<br>
|
|
NV TITAN X: 6.1 GFlop/ms, 0.3 GB/ms, 0.19 GTex/ms</tt><br>
|
|
<br>
|
|
Or the same numbers in operations per pixel at 1920x1080 at 60 Hz,<br>
|
|
<br>
|
|
<tt>
|
|
AMD FURY X: 69 KFlop/pix, 4.0 KB/pix, 2.2 KTex/pix<br>
|
|
NV TITAN X: 49 KFlop/pix, 2.4 KB/pix, 1.5 KTex/pix</tt><br>
|
|
<br>
|
|
<i>Think about what is possible with 69 thousand flops per pixel per frame.</i><br>
|
|
<br>
|
|
<b>HBM</b>
|
|
<br>
|
|
HBM definitely represents the future of bandwidth scaling for GPUs:
|
|
a change which brings the memory clocks down and bus width up (512 bytes wide on Fury X vs 48 bytes wide on Titan X).
|
|
This will have side effects on ideal algorithm design: ideal access granularity gets larger.
|
|
Things like random access global atomics and random access 16-byte vector load/store operations
|
|
become much less interesting (bad idea before, worse idea now).
|
|
Working in LDS with shared atomics, staying in cache, etc, becomes more rewarding.
|
|
<br>
|
|
<br>
|
|
|
|
|
|
|
|
|
|
</div></body></html>
|
|
|
|
|
|
|