For all that matters at scale. HPC, hyperscale, large enterprise and now out to the edge.
Joined October 2015
- Tweets2.8K
- Following6
- Followers9K
- Likes1.1K
The Next Platform retweeted
Forget the Top500 supercomputers. We drill down in the new systems for the third time in a year, and that gives better a sense of what is happening in HPC. $NVDA $AMD $INTC
nextplatform.com/2025/06/10/…
The Next Platform retweeted
Optical transceiver power and thermals are eating into the datacenter budget, and now Nvidia can justify the move to co-packaged optics for its switches. $NVDA
nextplatform.com/2025/03/18/…
The Next Platform retweeted
HPE has won the deal to supply LRZ with a pre-exascale machine based on future Nvidia CPUs and GPUs and its next-gen Slingshot interconnect. $NVDA $HPE
nextplatform.com/2024/12/16/…
The Next Platform retweeted
Just because Nvidia, AMD, and Intel are pumping money into Ayar Labs does not mean they are going to use its TeraPHY transport and SuperNova laser for silicon photonics interconnects. But they might.
nextplatform.com/2024/12/11/…
The Next Platform retweeted
The US Commerce Department has finally figured out that HBM, not compute engines, is the linchpin of AI and a lot of HPC. $MU #Samsung #SKHynix
nextplatform.com/2024/12/02/…
The Next Platform retweeted
Dell’s AI server business cooled in Q3 F2025 as some customers await systems based on Nvidia Blackwell GPUs. $DELL $NVDA $AMD
nextplatform.com/2024/11/27/…
The Next Platform retweeted
Guess who has more aggregate FP64 compute in the Top500 supercomputer rankings than Nvidia? Yup, it’s AMD. $NVDA $AMD
nextplatform.com/2024/11/18/…
The Next Platform retweeted
Oct 26, 2021: @NicoleHemsoth/@TheNextPlatform break the story that China apparently has achieved 1 EXAFLOPS on two systems, but won't be submitting to @top500supercomp for political reasons
nitter.cf/NicoleHemsoth/status/1…
#supercomputing #HPC
The #HPC benchmark for Top 500 was run on two Chinese supercomputers in March. One reached 1.3 exaflops (peak-1.05 sustained) as did the other. The results will not be reported publicly but we were able to confirm tests and provide architecture conjecture nextplatform.com/2021/10/26/…
The Next Platform retweeted
On October 24, join moderator Timothy Prickett Morgan from @TheNextPlatform along with industry experts to discover the strategies that will shape the future of AI hardware.
Register for the webinar:
bit.ly/3XQZZS8
#optical #photonics #semiconductor #AI
The Next Platform retweeted
In the six years of the first wave of the GenAI boom, the world will consume $2 trillion in AI systems – and it will change everything. $AMD $NVDA $INTC $ARM
nextplatform.com/2024/10/18/…
The Next Platform retweeted
We need nuclear fusion for power more than we need nuclear fission and fusion for war. And that is why NERSC should have the biggest supercomputers in the United States.
nextplatform.com/2024/10/08/…
The Next Platform retweeted
How much money is being spent this year on AI chips, and how much will be spent in the future? Good questions, and ones with a lot of different answers. $NVDA $AMD $INTC
nextplatform.com/2024/07/15/…
The Next Platform retweeted
AI systems need lots of HBM memory and flash, and that means Micron has a chance to substantially grow revenues and maybe even profits in the coming years. Ah, but will it? $MU $NVDA $AMD $INTC
nextplatform.com/2024/06/27/…
The Next Platform retweeted
"Catching up to and keeping pace with [the competition] in server GPUs is another thing entirely – and AMD has definitely done that and will be keeping pace for years to come." - @TheNextPlatform
AMD beating Intel in server CPUs was inevitable as the latter had foundry follies. Keeping pace with Nvidia in server GPUs is another thing – and AMD is definitely going to do that for years to come. $AMD $NVDA $INTC
nextplatform.com/2024/06/03/…
The Next Platform retweeted
Nvidia has mapped out its GPUs, CPUs, NICs, DPUs, and switches for the next four years. It looks like a two-year cadence for GPUs, CPUs, switches, and NICs with a mid-cycle memory bump for GPUs and a radix bump for switches. $NVDA
nextplatform.com/2024/06/02/…
The Next Platform retweeted
If you want to simulate billions or trillions of atoms, an exascale supercomputer is great. But if you want to see further into the future in a simulation of nearly a million atoms, maybe a waferscale compute engine is better.
nextplatform.com/2024/05/15/…
The Next Platform retweeted
Meta’s Llama 3 model gets good grades on various benchmark tests, but the company is not interested in wrapping up a commercial version and selling support for it. Who is going to make money from this? Odds are, Big Purple – the combination of Big Blue and its Red Hat. $META $IBM
nextplatform.com/2024/04/22/…
