The important new information that I got from this is that AMD Zen 6 supports FRED (Flexible Return and Event Delivery), which was introduced first in Intel Panther Lake, at the beginning of this year.
FRED is a great improvement over the extremely inefficient mechanisms for handling interrupts, exceptions and system calls of x86-64, which had been constrained by inheriting most of their features from Intel 8086 and 80286, from more than a half of century ago.
FRED should decrease the overhead caused by the operating system for x86-64 CPUs.
I am not sure if I am more excited for the Zen6c 256 Core 512vCPU or the 88 Core Nvidia Vera. I have been looking forward to this for so long unfortunately we are in an AI Boom which means capacity for these will be limited while demand are skyrocketing.
In terms of per system / server that is easily an 50%+ increase in performance with Core count and IPC from last generation Zen 5. And there are plenty of server still on Zen 4 and Zen 3.
On x86 side IPC continues to lack behind ARM. Vera is showing great results without a much increase in die space, and it is still on 3nm. I only wish we could have 256 Core Vera.
I hope we get an update from Netflix team on streaming 1.6TB/s per box with these new chips.
Take also into account that the SPEC benchmark does not show the significantly greater performance that can be obtained on Zen 6 with programs optimized for AVX-512. With such programs, the advantage over Vera would be much greater.
Vera seems to have outstanding performance only for the narrow applications for which it is intended, AI training, inference and agentic work.
It is not clear how much of the great performance demonstrated by Vera in those applications is due to the hardware and how much is due to well tuned and matched to it NVIDIA software.
It is very likely that the Vera Olympus cores have a poor performance per die area, which is why they have put only this relatively small number of cores per socket.
The Olympus cores seem to be a greatly enhanced version of the Arm Cortex-X925 cores used in the previous NVIDIA products. Cortex-X925 already had excellent IPC, but very poor performance per die area in comparison with cores like Cortex-X4/Neoverse V3 (which is used in the best Arm-based server CPUs, e.g. Graviton5 and Cobalt 200).
Thus Cortex-X925 was unsuitable for a throughput-oriented CPU, and NVIDIA Olympus seems to be the same kind of core, i.e. a core that is useful in a server only if the server contains other components, i.e. GPUs, which do all the high throughput work, and the CPU only orchestrates the work of those components, without having to do much work itself.
Jeez, I wish I could get AMD to give me one of these to try out xc [1] on it with some RX9070XTs installed - those benchmarks would be a thing of beauty :) The problem with high-end hardware is that (especially these days with the RAM prices) it's just so damn expensive.
The important new information that I got from this is that AMD Zen 6 supports FRED (Flexible Return and Event Delivery), which was introduced first in Intel Panther Lake, at the beginning of this year.
FRED is a great improvement over the extremely inefficient mechanisms for handling interrupts, exceptions and system calls of x86-64, which had been constrained by inheriting most of their features from Intel 8086 and 80286, from more than a half of century ago.
FRED should decrease the overhead caused by the operating system for x86-64 CPUs.
I am not sure if I am more excited for the Zen6c 256 Core 512vCPU or the 88 Core Nvidia Vera. I have been looking forward to this for so long unfortunately we are in an AI Boom which means capacity for these will be limited while demand are skyrocketing.
In terms of per system / server that is easily an 50%+ increase in performance with Core count and IPC from last generation Zen 5. And there are plenty of server still on Zen 4 and Zen 3.
On x86 side IPC continues to lack behind ARM. Vera is showing great results without a much increase in die space, and it is still on 3nm. I only wish we could have 256 Core Vera.
I hope we get an update from Netflix team on streaming 1.6TB/s per box with these new chips.
Even in single-threaded performance, where Vera is much better than in multi-threaded performance, it is inferior to Zen 6.
See e.g.:
https://chipsandcheese.com/p/amds-zen-6-family-feeding-the-b...
Take also into account that the SPEC benchmark does not show the significantly greater performance that can be obtained on Zen 6 with programs optimized for AVX-512. With such programs, the advantage over Vera would be much greater.
Vera seems to have outstanding performance only for the narrow applications for which it is intended, AI training, inference and agentic work.
It is not clear how much of the great performance demonstrated by Vera in those applications is due to the hardware and how much is due to well tuned and matched to it NVIDIA software.
It is very likely that the Vera Olympus cores have a poor performance per die area, which is why they have put only this relatively small number of cores per socket.
The Olympus cores seem to be a greatly enhanced version of the Arm Cortex-X925 cores used in the previous NVIDIA products. Cortex-X925 already had excellent IPC, but very poor performance per die area in comparison with cores like Cortex-X4/Neoverse V3 (which is used in the best Arm-based server CPUs, e.g. Graviton5 and Cobalt 200).
Thus Cortex-X925 was unsuitable for a throughput-oriented CPU, and NVIDIA Olympus seems to be the same kind of core, i.e. a core that is useful in a server only if the server contains other components, i.e. GPUs, which do all the high throughput work, and the CPU only orchestrates the work of those components, without having to do much work itself.
https://www.storagereview.com/review/nvidia-vera-cpu-review
Where real applications were tested on both Vera and Zen 6.
Jeez, I wish I could get AMD to give me one of these to try out xc [1] on it with some RX9070XTs installed - those benchmarks would be a thing of beauty :) The problem with high-end hardware is that (especially these days with the RAM prices) it's just so damn expensive.
1: https://news.ycombinator.com/item?id=50151160