Y
HN Search
Hacker News Search
new
|
comments
|
top
|
jobs
zepmck
searching PlanetScale…
1.
▲
2.
▲
3.
▲
4.
▲
5.
▲
6.
▲
4 ms
·
1.
▲
by
zepmck
3y ago
NVIDIA is showing real performance through using independent MLPerf benchmark.
2.
▲
Operation of IBM Summit system gets extended to 2024
(olcf.ornl.gov)
1 points
by
zepmck
3y ago
|
1 comments
3.
▲
by
zepmck
3y ago
Does this hide any problem with AMD Frontier?
4.
▲
by
zepmck
3y ago
Software developers, especially for certain type of components, e.g. compilers, are NOT easy to find out.
5.
▲
by
zepmck
3y ago
The main reasons of "monopoly" are long-term vision and enormous effort spent in software development. AMD is lacking behind because they have not yet seriously invested on their own software stack.
6.
▲
by
zepmck
4y ago
The next challenge is software stack.
7.
▲
by
zepmck
4y ago
People pretending to do professional work with low budget/quality workstation. If you cannot afford to operate a 4090 according to specs, you cannot blame NVIDIA if things go wrong.
8.
▲
ARM Neoverse V2 Nvidia Grace CPU
4 points
by
zepmck
4y ago
|
0 comments
9.
▲
by
zepmck
4y ago
You blame CUDA because you have never played AMD + ROCm. Furthermore OpenCL is on dead-end street. With standard programming languages natively supporting CUDA, it has no sense to add extra layers.
10.
▲
by
zepmck
4y ago
The most powerful and unfortunately unusable supercomputer of the world. AMD's approach to GPUs is on a failing track since its inception. The only software stack available is super fragile, buggy and barely supported. Rather than buil
11.
▲
by
zepmck
5y ago
If you have GPUs, NVTabular outperforms most of the frameworks out there: https://github.com/NVIDIA/NVTabular
12.
▲
by
zepmck
5y ago
This not completely correct because while a carrier transport anything indescrminately, social networks manipulate and control information exchange, for instance giving more priority to a content wrt to another.
13.
▲
by
zepmck
5y ago
I think this work is quite ridiculous. As far as I could see, they trained the net with a batch size equal to 1.
14.
▲
by
zepmck
5y ago
In terms of PCI lanes efficiency, there is no competition between Intel and AMD. Intel is much ahead wrt AMD. Don't not be impressed about number of lanes available on the board.
15.
▲
by
zepmck
6y ago
When? Where? How can it be miles ahead if the hardware has not been released yet?
16.
▲
by
zepmck
6y ago
One-API is not completely open source. Support for Ponte Vecchio will be not released open source for many reasons.