Chinese CPU Maker, Phytium Unveils Next-Gen Architecture, 3 GHz Chip on Par With 4 GHz AMD Zen 3

[Nan Yang]

Chinese CPU Maker, Phytium Unveils Next-Gen Architecture, 3 GHz Chip on Par With 4 GHz AMD Zen 3​

Muhammad Zuhair•Oct 18, 2023 08:00 AM EDT151

Image Source: Phytium

Chinese chipmaker, Phytium, has unveiled its next generation of high-performing CPUs, competing with the likes of AMD and ARM.

Phytium Showcases Next-Gen CPU Architecture Performance For China's Server Markets, Strongly Competes Against AMD Zen 3 Architecture​

While you may be hearing about Phytium for the first time, in terms of market presence, they have been here for almost a decade, however confined to China only. The company started out by integrating the outdated SPARC instruction set and later on shifted its focus towards the ARM camp since this enabled it to reach high-performance targets and a more "well-developed" ecosystem.

Progressing with time, Phytium made noticeable strides in generational improvements, however, the company looks to make a decisive entry in the upcoming years. By now, Phytium has focused on three different industry domains, ranging from server applications to desktop and embedded products. The company has recently defined its future developments at the 2023 Integrated Circuit Application Innovation Forum, where it aims to compete with market offerings.


Image Source: CNBeta

Based on the performance data officially disclosed by the company, it is revealed that they are prepping for their next-gen CPU release, called the "FTC870". The processor shows a whopping 20% generational improvement over its predecessor the "FTC860". Moreover, the FTC870 has also been benchmarked at the SPEC CPU2017 application, which is widely utilized for server CPUs. Official data reveals that the CPU is able to hold up against the ARM Neoverse N1, and even gives tough competition to the Neoverse N2 in some scenarios.

Most importantly, the chip competes favorably against AMD's Zen 3 CPUs. The FTC870 with a clock speed of 3 GHz is able to get within the reach of the AMD EPYC 7443 which runs at higher clocks of 4 GHz. The CPU also surpasses the performance of the ARM Neoverse N2 in the floating point tests while getting really close in the Integer benchmarks.


Image Source: CNBeta

Now what Phytium has managed to obtain here, is indeed a breakthrough for Chinese server markets, since they have been "cut off" from cutting-edge components for a long time now. The harsh US sanctions have promoted the growth of "homegrown" equipment, and Phytium's FTC870 is a prime example of it. While this is official whether the disclosed figures are actually true, oftentimes, it is a completely different story when Chinese CPUs are benchmarked by individual reviewers.

Phytium's accomplishment would prove to be vital for Chinese server markets, especially since they are expected to grow tremendously in the upcoming years. It will be interesting to see how Phytium's architecture is received by the Chinese firms, since by the looks of it, they do pack great performance.

News Source: CNBETA

 

[epebble]

For some reason Loongson CPU never became successful in the market. I was curious to get a board if it became available, just to check out different architecture. But somehow it never came out of China, at least in quantity. Hope this succeeds.

 

[Menthol]

Seeing how USA is blockading China chip development.

Phytium achievement is quite amazing!

Using old technology but able to closing the gap against the world best.

 

[gambit]

For some reason Loongson CPU never became successful in the market. I was curious to get a board if it became available, just to check out different architecture. But somehow it never came out of China, at least in quantity. Hope this succeeds.

In semicon industry, there are three types of customers:

- Secured​

- Committed​

- Contracted​

A secured customer is like a "gentleman's handshake".

A committed customer is like a promise to buy.

A contracted customer is the best -- signed. And usually with a long term agreement (LTA).

A customer can specify to buy only from specific fab(s), like Apple does. The longer a customer re-sign an LTA, the more open to accepting products from other fabs.

So from this perspective, if I go to Alphabet and offer to sell my products for their Youtube data centers, most likely Alphabet would decline my proposal. Alphabet or Apple already have contracts with multiple suppliers of wide variety of semicon products, from CPU to storage memory to controllers and whatever else. Not only that, they already planned their markets for several yrs into the future. They cannot afford to change those contracts. It could take up to two yrs for them to qualify my products and in that two yrs, they will not suspend their current contracts but continue to renew them, after all, they are customers as well as being suppliers. So why should they accept my proposal to test my products?

Someday someone will take a chance on China's CPU.

 

[epebble]

So from this perspective, if I go to Alphabet and offer to sell my products for their Youtube data centers, most likely Alphabet would decline my proposal. Alphabet or Apple already have contracts with multiple suppliers of wide variety of semicon products, from CPU to storage memory to controllers and whatever else. Not only that, they already planned their markets for several yrs into the future. They cannot afford to change those contracts. It could take up to two yrs for them to qualify my products and in that two yrs, they will not suspend their current contracts but continue to renew them, after all, they are customers as well as being suppliers. So why should they accept my proposal to test my products?

Someday someone will take a chance on China's CPU.

But technology does not give a huge advantage to incumbents. See how quickly AMD + TSMC humbled Intel, and ARM is humbling Intel too without even fab. If the China chip gives, say, a 30% advantage in price/performance and more importantly energy efficiency, they will be in data centers in no time at all. Data center business is like a commodity; small cost differentials affect profitability a lot. If not in U.S., surely, they can sell a lot of data center boards elsewhere.

 

[Menthol]

But technology does not give a huge advantage to incumbents. See how quickly AMD + TSMC humbled Intel, and ARM is humbling Intel too without even fab. If the China chip gives, say, a 30% advantage in price/performance and more importantly energy efficiency, they will be in data centers in no time at all. Data center business is like a commodity; small cost differentials affect profitability a lot. If not in U.S., surely, they can sell a lot of data center boards elsewhere.

USA is blockading China from obtaining high performance chips.

And this is a good opportunity for Phytium to grab the market share in China.

EU fears for losing car market share in China to the local brands.

But USA is giving it away for free.

 

[gambit]

But technology does not give a huge advantage to incumbents. See how quickly AMD + TSMC humbled Intel, and ARM is humbling Intel too without even fab. If the China chip gives, say, a 30% advantage in price/performance and more importantly energy efficiency, they will be in data centers in no time at all. Data center business is like a commodity; small cost differentials affect profitability a lot. If not in U.S., surely, they can sell a lot of data center boards elsewhere.

AMD is already well known alternative to Intel and TSMC already had plenty of customers as a contract manufacturer. And sorry to say, there is no such thing as 'in no time'. It will take at least two yrs to engineer qualify a new product.

ASUS or Gigabyte, on their own, can divert some of their resources to test the Chinese CPU, but if not, then it will take a major semicon consumer like Alphabet or Apple to ask the electronics manufacturers, like Asus or Gigabyte, to test the Chinese CPU. Price cuts are worthless of the new product does not perform like the current.

 

[Menthol]

AMD is already well known alternative to Intel and TSMC already had plenty of customers as a contract manufacturer. And sorry to say, there is no such thing as 'in no time'. It will take at least two yrs to engineer qualify a new product.

ASUS or Gigabyte, on their own, can divert some of their resources to test the Chinese CPU, but if not, then it will take a major semicon consumer like Alphabet or Apple to ask the electronics manufacturers, like Asus or Gigabyte, to test the Chinese CPU. Price cuts are worthless of the new product does not perform like the current.

Go AMD! Go!

I am a huge fan and supporter of AMD!

Because I'm poor man, living in third world country, my only hope for cheaper PC is AMD...

Despite my everyday thin laptop is using Intel. Lol

 

[TNT]

The main hurdle for china is getting hands on latest EUV lithography machines. The design is no more an issue.

 

[Ali_Baba]

Chinese CPU Maker, Phytium Unveils Next-Gen Architecture, 3 GHz Chip on Par With 4 GHz AMD Zen 3​

Muhammad Zuhair•Oct 18, 2023 08:00 AM EDT151

Image Source: Phytium

Chinese chipmaker, Phytium, has unveiled its next generation of high-performing CPUs, competing with the likes of AMD and ARM.

Phytium Showcases Next-Gen CPU Architecture Performance For China's Server Markets, Strongly Competes Against AMD Zen 3 Architecture​

While you may be hearing about Phytium for the first time, in terms of market presence, they have been here for almost a decade, however confined to China only. The company started out by integrating the outdated SPARC instruction set and later on shifted its focus towards the ARM camp since this enabled it to reach high-performance targets and a more "well-developed" ecosystem.

Progressing with time, Phytium made noticeable strides in generational improvements, however, the company looks to make a decisive entry in the upcoming years. By now, Phytium has focused on three different industry domains, ranging from server applications to desktop and embedded products. The company has recently defined its future developments at the 2023 Integrated Circuit Application Innovation Forum, where it aims to compete with market offerings.


Image Source: CNBeta

Based on the performance data officially disclosed by the company, it is revealed that they are prepping for their next-gen CPU release, called the "FTC870". The processor shows a whopping 20% generational improvement over its predecessor the "FTC860". Moreover, the FTC870 has also been benchmarked at the SPEC CPU2017 application, which is widely utilized for server CPUs. Official data reveals that the CPU is able to hold up against the ARM Neoverse N1, and even gives tough competition to the Neoverse N2 in some scenarios.

Most importantly, the chip competes favorably against AMD's Zen 3 CPUs. The FTC870 with a clock speed of 3 GHz is able to get within the reach of the AMD EPYC 7443 which runs at higher clocks of 4 GHz. The CPU also surpasses the performance of the ARM Neoverse N2 in the floating point tests while getting really close in the Integer benchmarks.


Image Source: CNBeta

Now what Phytium has managed to obtain here, is indeed a breakthrough for Chinese server markets, since they have been "cut off" from cutting-edge components for a long time now. The harsh US sanctions have promoted the growth of "homegrown" equipment, and Phytium's FTC870 is a prime example of it. While this is official whether the disclosed figures are actually true, oftentimes, it is a completely different story when Chinese CPUs are benchmarked by individual reviewers.

Phytium's accomplishment would prove to be vital for Chinese server markets, especially since they are expected to grow tremendously in the upcoming years. It will be interesting to see how Phytium's architecture is received by the Chinese firms, since by the looks of it, they do pack great performance.

News Source: CNBETA

interesting and an excellent performance - but China will only ever be in control of its destiny which it has its own instruction set with control of its own intellectual property. Given the size of the chinese market, and its manufacturing base, it could develop its own instruction set for devices that do not require a "western operating system".

ARM could well be owned by the Americans soon(!) and they could withdraw licenses granted to China, unless USA goverment pressure.

 

[gambit]

Go AMD! Go!

I am a huge fan and supporter of AMD!

Because I'm poor man, living in third world country, my only hope for cheaper PC is AMD...

Despite my everyday thin laptop is using Intel. Lol

AMD is the near perfect example of what I said in post 4.

But first, am not a fan of Intel, but neither do I hate them. I worked with Intel before, from Production up to the Fab Director level. Whatever problems there are with Intel, they came from Corporate, not Engineering/Technical.

The Engineering/Technical level is what I said in post 4 applied. AMD products may, and does have, some advantages, but overall, the main reason why Intel dominates is because their shid works and works well. Does not have to be perfect, just well or slightly above well. The consumer is the end and no one wants to jeopardize their profits at this point. For the consumers, it is not their job to know how any product, a CPU or even a toaster, works. We just want the thing to work as claimed. Once a product is entrenched into the capitalist ecosystem, this is where any challenger will have issues even if the alternative is just as good or even better. YOU have to convince me that your product works as I use the current product(s) from your competition to make money, let alone profits. The mistrust is high. The doubts are evident. After all, my livelihood is at stake here, and in the case of NASA, literally lives are on the line, so why should they take a chance on a new product.

Take the Intel-AMD issue and apply to what the Chinese semicon industry is trying to do. The Chinese semicon industry simply cannot flood the market with low(er) prices products and expect the industry leaders to adopt. The reason why China became the world's semicon manufacturing center is low labor cost, nothing more. The product designs, manufacturing methods, and standards were developed outside of China and introduced into China, so of course, the trust and confidence levels will be high to start. Literally, if a new die shrink is available, the new product would have been R/D, experimented, and refined for at least two yrs before it is allowed to be mass produced by any Chinese fab, then it will take another two yrs of clients (plural) qualifications before products from that Chinese fab is crossed off the restriction list.

But now, the semicon industry in particular and the end consumers in general, are looking at a new semicon product from China that was designed essentially with no Western inputs. People will not like that. Call it biased or even racism, if you want, and I say 'you' generically, but the reality is that since the Industrial Revolution, global technologies came from the West, bluntly speaking, not from Asia. Post WW II, JPN became a notional member of the West while China stagnated, at best. The end result is that products from JPN are usually quickly accepted and elevated to high consumer standards.

China's J-20 and C929 are illustrative. Who are the intended end consumers of the J-20? One -- the PLA. Who are the intended end consumers of the C929? The rest of the world. Which product MUST have the most Western influences before acceptance? Everything about the J-20 can be 'state secrets', but everything about the C929 must be transparent. The scope of the end consumer(s) determines the acceptance.