Wednesday, May 17th 2023
Enablement Continues for Chinese Loongson 3A6000 CPUs Poised to Compete with Intel Willow Cove and AMD Zen 3
Chinese company Loongson, specializing in creating processors for usage in mainland China, has been steadily working on enabling its next-generation Loongson 3A6000 CPUs. Aiming to provide the performance level of Intel Willow Cove and AMD Zen 3, these new CPUs will use Loongson's custom LoongArch Instruction Set Architecture (ISA) with a new set of 64-bit superscalar LA664 cores. Today, thanks to the report from Phoronix, we find out that Loongson has submitted some Linux patches that enable the upcoming 3A6000 CPUs to work with Linux-based operating systems at launch. Interestingly, as the new CPU generation gets closer to launch, more Linux kernel patches begin to surface.
Today's kernel patches focus on supporting the hardware page table walker (PTW). As PTW can handle all fast paths of TLBI/TLBL/TLBS/TLBM exceptions by hardware, software only needs to handle slow paths such as page faults. Additionally, in the past, LoongArch utilized "dbar 0" as a complete barrier for all operations. However, this full completion barrier severely impacted performance. As a result, Loongson-3A6000 and subsequent processors have introduced various alternative hints. Loongson plans to ship samples to select customers in the first half of 2023, so we could see more information surfacing soon.
Source:
Phoronix
Today's kernel patches focus on supporting the hardware page table walker (PTW). As PTW can handle all fast paths of TLBI/TLBL/TLBS/TLBM exceptions by hardware, software only needs to handle slow paths such as page faults. Additionally, in the past, LoongArch utilized "dbar 0" as a complete barrier for all operations. However, this full completion barrier severely impacted performance. As a result, Loongson-3A6000 and subsequent processors have introduced various alternative hints. Loongson plans to ship samples to select customers in the first half of 2023, so we could see more information surfacing soon.
23 Comments on Enablement Continues for Chinese Loongson 3A6000 CPUs Poised to Compete with Intel Willow Cove and AMD Zen 3
As you can see:
Loongson’s 3A5000: China’s Best Shot? – Chips and Cheese
So they need to massively increase IPC and Clock Speed. I don't doubt that they can eventually achieve it, I just think it's just too much to do in a single generation.
They are simply meant to be an independent cpu for China, and they are that, at least. They don't care terribly if they are a gen or two behind. Obviously they'd love to beat us, but that's not the primary goal here.
CCP defense contractorChinese Tech Firm?Oh, I'm so very surprised... /s
Still, one usually gets flamed for bringing up such things, most places.
It's like the years of cheap consumer goods somehow 'bought off' consumers to shill and defend Chinese interests...
In this case though, the product isn't even equal.
I really wish I could play with some of these CPUs its a shame they are so hard to get a hold of, not this one specifically but in general. I know they have like a zen 1 or 2 performant in x86 land IIRC. Would be neat to play with it.
Id really really like to play with the Altra stuff: www.ipi.wiki/products/com-hpc-ampere-altra one day. I need to get rid of things in my office before wife will let me buy new stuff.
Previewing China’s Loongson 3A5000 with Performance Counters – Chips and Cheese
Ah yes, Zhaoxin thats the one I was remembering.
Don't most "benchmarkable" programs require support for the specific ISA?
www.amazon.com/QNAP-TVS-675-8G-US-High-Speed-KX-U6580-Connectivity/dp/B09KMMP9SD
As an example, the latency of some of the SIMD instructions of Loongson are below, comparatively Zen 1 has 3-cycle latency for FADD and FMUL while the same for Multiply-Add, while also clocking higher. I believe that AMD has improved that in Zen 2 and 3.
Loongson will have to improve that while greatly increase clocks and other parts of the architecture(likely also including stuff in the SoC as it seemed to have issues with handling the bandwidth of DDR4).