News Loongson Begins to Enable CPUs That Could Rival AMD and Intel Offerings

[Admin]

Loongson has started posting its first Linux patches for the upcoming 3A6000-series processors, which promise to rival AMD's Zen 3-based CPUs.

Loongson Begins to Enable CPUs That Could Rival AMD and Intel Offerings : Read more

 

[flashflood101]

What nonsense.
Loongson is years away from producing anything to rival leading CPUs.
Tom's Hardware should be careful what it allows certain contributors to post - credibility, once lost, is difficult to regain.

 

[bit_user]

Loongson is years away from producing anything to rival leading CPUs.
Tom's Hardware should be careful what it allows certain contributors to post

They were careful to point out:

"it should be noted that IPC alone does not necessarily mean that Loongson's 2nd Generation CPUs that rely on its LoongArch microarchitecture will be as fast as AMD's Ryzen 5000-series or Intel's 11th generation Core processors. Clock speed and other aspects of the platform will also play a role."
​

Even as far as the IPC claims, I agree that we need to take the designer's claims with a grain of salt. I'm reserving judgement until someone manages to get a production-grade sample for independent testing.

 

[craigss]

What nonsense.
Loongson is years away from producing anything to rival leading CPUs.
Tom's Hardware should be careful what it allows certain contributors to post - credibility, once lost, is difficult to regain.

I think folks in the West are beyond dumb if they dont expect China to surpass them in CPU and other tech developments in the near future, after all we have laid the foundations for this by building every single electronics component in that part of the world and financing huge budgets the likes of which most western countries can only dream of dipping into for tech research and development
the Godson 1 was a 233 mhz cpu launched in 2002 here we are 20 years later with multiple cores and there own arch thats outstanding progress, where do you think they will be in 5 years time?
Sanctions just force them to develop there way out of the issue and will backfire, we in the west need to take away the budgets by building our own stuff at home!

 

[newtechldtech]

Let them try .. by the time they reach Intel and AMD performance , Quantum computers will be everywhere and will crush them.

 

[bit_user]

I think folks in the West are beyond dumb if they dont expect China to surpass them in CPU and other tech developments in the near future, after all we have laid the foundations for this by building every single electronics component in that part of the world

Why do you assume they will surpass the West? While the rise of semiconductor fabs in Korea and Taiwan have been truly impressive, most of the fundamental hardware & software involved in semiconductor design & manufacturing still comes from western companies. Western CPU designs still remain unsurpassed.

where do you think they will be in 5 years time?

They've published roadmaps, if you really want to know. It's nothing mind-blowing, from what I recall.

Sanctions just force them to develop there way out of the issue and will backfire,

Sanctions are a short-to-medium -term tool. In the long-term, they were on track to have a competitive domestic industry, either way. The sanctions should be used as leverage. It's not clear to me what's the gameplan, but I sure hope there's something besides the sanctions, themselves.

Let them try .. by the time they reach Intel and AMD performance , Quantum computers will be everywhere and will crush them.

China is actually very competitive, in Quantum Computing. It's a lot easier to take the lead in a new field, than to become competitive in an established one.

Also, quantum computers aren't a direct replacement for classical computers, in most cases. The two technologies are complementary, and will exist side-by-side for the foreseeable future.

 

[edzieba]

Let them try .. by the time they reach Intel and AMD performance , Quantum computers will be everywhere and will crush them.

Thinking Quantum Computers will somehow 'crush'/surpass/etc classical computation devices is like expecting FPUs to make integer logic obsolete.

 

[Firestone]

Doesn't the CCP require these chips to come with hardware back doors so the govt can spy on all data going through the CPU

 

[gdmaclew]

The company expects its upcoming LoongArch-based CPUs AMD's Zen 3 in instruction per clock (IPC), which will enable Loongson to challenge leading processor manufacturers.

I have read this sentence backwards and forwards and I can't make any sense of it. Can somone please tell me what he is trying to say?

 

[newtechldtech]

Thinking Quantum Computers will somehow 'crush'/surpass/etc classical computation devices is like expecting FPUs to make integer logic obsolete.

Not exactly ... imagine CPU without FPU today like in old times ?

 

[bit_user]

Doesn't the CCP require these chips to come with hardware back doors so the govt can spy on all data going through the CPU

With SecureBoot, you don't need the CPU to have backdoors. You can just limit which OS images you sign to those with backdoors at the OS level, and the machine can't boot anything else.

 

[Amdlova]

Every single IC manufacturer can challenge AMD, AMD don't have a fab. AMD only exists for make INTEL competitive on market.

 

[bit_user]

The company expects its upcoming LoongArch-based CPUs AMD's Zen 3 in instruction per clock (IPC), which will enable Loongson to challenge leading processor manufacturers.

I have read this sentence backwards and forwards and I can't make any sense of it. Can somone please tell me what he is trying to say?

I think it should read:

The company expects its upcoming LoongArch-based CPUs to match AMD's Zen 3 in instruction per clock (IPC), which will enable Loongson to challenge leading processor manufacturers.
​

IPC is a common performance metric that's often used as a proxy for CPU microarchitecture sophistication. The relation is a little more complicated than that, but it tends to correlate with sophistication.

 

[bit_user]

Every single IC manufacturer can challenge AMD, AMD don't have a fab.

Neither does Nvidia, Qualcomm, Apple, etc.

 

[Johnpombrio]

In 2016, AMD's Zen technology was sold to China to help keep AMD afloat. This is probably the foundation of their Looongson chip development. Yet China still needs to BUY, BUILD, or MAINTAIN lithography machines and FABs capable of manufacturing these highly complex chips. China certainly has not even caught up to the original Zen, let alone surpassed it. I don't expect they ever will.

 

[bit_user]

In 2016, AMD's Zen technology was sold to China to help keep AMD afloat.

It's very important to understand exactly what they sold, and it wasn't high-level IP. Rather, what they sold were placed & routed netlists, that left Zhaoxin just enough flexibility to integrate their government-mandated cryptographic extensions into the CPU. It was the hardware equivalent of shipping someone a DLL (shared library) and a header file excerpt, with only the required callbacks exposed. AMD was very clear that it was not anything that would impact the future competitiveness of the interested parties.

https://en.wikipedia.org/wiki/AMD–Chinese_joint_venture
​

What was probably a far more relevant technology transfer risk was the VIA/Centaur relationship.

https://en.wikipedia.org/wiki/Centaur_Technology
​

This is probably the foundation of their Looongson chip development.

That's a completely different ISA, microarchitecture, and system architecture, from a completely different company. Chinese companies do compete with each other, you know?

 

[sivaseemakurthi]

Anyone can develop CPUs these days.

 

[A Stoner]

I think folks in the West are beyond dumb if they dont expect China to surpass them in CPU and other tech developments in the near future, after all we have laid the foundations for this by building every single electronics component in that part of the world and financing huge budgets the likes of which most western countries can only dream of dipping into for tech research and development
the Godson 1 was a 233 mhz cpu launched in 2002 here we are 20 years later with multiple cores and there own arch thats outstanding progress, where do you think they will be in 5 years time?
Sanctions just force them to develop there way out of the issue and will backfire, we in the west need to take away the budgets by building our own stuff at home!

5 years? In 5 years they will be lucky to have progressed 2 years worth of process. Without our technology in their hands, they will stagnate. The reason they advanced to where they are now is because they had access to our technology, either through theft or manufacturing agreements.

 

[tomscomments]

there are evolving extremely quicky in quantum and photonic chip, they masterd (and bought) MIPS when it was not wort interest during arm era.
There is no doubt they master how to design a soc. We san how kirin chips were balanced and good
For those loongsons, it would be interesting to know the manufacturer. I guess it is SMIC based in a chiplet architecteture. They certainly have other suppliers wich one are classified.
Folowing their progress since some years know, i don't undestant why some are still in denial about their abiity to advance.

Australian think tank says China is leading the world in tec

US comes in second, rest of the world is a distant third in fields from biotech to batteries

www.theregister.com

"Think tank the Australian Strategic Policy Institute (ASPI) has published an update to its Critical Technology Tracker, and asserted that China has taken the lead in research on 37 of 44 critical or emerging technologies."

The best thing to keep the lead is to focus on own innovation and not underestimate them.

 

[tomscomments]

The company expects its upcoming LoongArch-based CPUs AMD's Zen 3 in instruction per clock (IPC), which will enable Loongson to challenge leading processor manufacturers.

I have read this sentence backwards and forwards and I can't make any sense of it. Can somone please tell me what he is trying to say?

Being blocked (officially) from euv lithography, i expect them to go with smic chiplet and chiplet design. (on 28 nm or 14nm) difficult to say.
By the way, they are among the leaders in quantum and in photonic chips. They may be produce something based on photonic in the coming year and won't put more than necessary ressources on EUV (difficult to say)

It's like with combustion engine cars. They simply focused on a different technology instead of playing excessive catch up. That's smart.
China's tech bothleneck is not conception, it is manufacturing. And they seem to put all their efforts to overcome it. When we say they catch up in ten years, they do it in 4.

i have an intel cpu's of 2017 and who runs quite everything. Even 5th generation intel of 2012, can give a good everyday productivity especilly since wind10 or a linux distribution are lightweight.
So it is more than largely sufficient for them those loongsons if there are on par with an 11 th generation intel or even older. The'll just continue improve the ability to fully produce it on their own, while improving code and performance through linux. (you can see the progress made by deepin V23, quite impressive after some virtual box try on)

All in all their man focus is not to make the best cpu of the world, they want to make a performant, independant on hardware and software aspect processor. They are not trying to export those chips. That's for their inner national security. They can produce desktops or laptops from scratch with no foreign technology, i guess. (or in the close futur)
I don't thing they are obcessed with outselling amd or intel (for now) that's not their priority), their top priority is the 1.4 billion market they want to secure. They know they are less experimented, and more relying on foregin tech. They'll play it low profile until they master everything. That's something they accelerated since huawei sanctions.

if there is anything nasty in the 4 coming years, us and china will work together. As they already do in fact.
European industry, despite having critical subcontractors (mainly dutch, britt and german) is however in a worst shape, they lag behing asia and north american in masering the whole thing. Sometimes producing an average product you completely master is more important than producing top products wich you rely on many unreliable subcontractors.

 

[tomscomments]

Why do you assume they will surpass the West? While the rise of semiconductor fabs in Korea and Taiwan have been truly impressive, most of the fundamental hardware & software involved in semiconductor design & manufacturing still comes from western companies. Western CPU designs still remain unsurpassed.


They've published roadmaps, if you really want to know. It's nothing mind-blowing, from what I recall.


Sanctions are a short-to-medium -term tool. In the long-term, they were on track to have a competitive domestic industry, either way. The sanctions should be used as leverage. It's not clear to me what's the gameplan, but I sure hope there's something besides the sanctions, themselves.


China is actually very competitive, in Quantum Computing. It's a lot easier to take the lead in a new field, than to become competitive in an established one.

Also, quantum computers aren't a direct replacement for classical computers, in most cases. The two technologies are complementary, and will exist side-by-side for the foreseeable future.

korea and japan are smaller than china and are part of the west
Of course western design remain unsurpassed but china should not be confused with korea. That's a huge contry, a huge market. Korea built its technology with western tools. Japan's toshiba was attacked by us tech in the 80's
China is different. It doesn't not seek to be a big korea, it does seek to buid independant ecosystem and that more difficult than producing the end product. I don't think they do it to annoy west, they first want to avoid being blocked by sanctions and gain autonomy. On chinese media, (tech media at least), there is no anti american mood, they are just anxious in producing their own tools (there are many forums that can be read but a translator)
let's not forget, a big part of californian big tech executive are from asian descent (indian or chineese ethnicaly). They have high IQ's . It only a matter of time until they become independant (totally i mean)
that doesn't mean us decline, i don't think so. If us don't loose it's dna (and on top of it its great constitution, freedom and granted property), it can still attracts many talents from all the world (just like steve jobs was from a syrian father. I think one of the X86 conceptor is from iran)

I see many evidence china is rising, but usa isn't declining neither. It all depends on inside, not outside.
As a foreigner, usa is still a dream country for me to visit, just as much as i have great respect for american people, culture and their great constitution. I just don't want to hate others because they do well.
So i do believe us will remain a great power. I do believe China too and i believe in the next decades given all the humanity has to face, those two great countries need to work together
China don't hide from saying their took inspiration from many us aspects, first of all, the classical us capitalism found in china's special economic zones. In their mindset, Chineese are merchants not socialists.
they accept bankrupcy
The private sector in china accounts for 70 % of gdp. That's more liberal capitalisme than many european countries

 

[bit_user]

Anyone can develop CPUs these days.

That depends on whether you're talking about using 3rd Party IP and how competitive you want it to be. If you're willing to simply license IP from ARM, then yes, you can "design" a CPU and have it turn out pretty good. Or, if you simply want something modest, you can even design it all yourself. But, there's a pretty big gulf between the top CPU designs and everybody else.

By the way, they are among the leaders in quantum and in photonic chips.

That's apples and oranges vs. classical CPU design.

So it is more than largely sufficient for them those loongsons if there are on par with an 11 th generation intel or even older.

You're confusing server CPUs and workloads with desktop ones.

 

[tomscomments]

That depends on whether you're talking about using 3rd Party IP and how competitive you want it to be. If you're willing to simply license IP from ARM, then yes, you can "design" a CPU and have it turn out pretty good. Or, if you simply want something modest, you can even design it all yourself. But, there's a pretty big gulf between the top CPU designs and everybody else.


That's apples and oranges vs. classical CPU design.


You're confusing server CPUs and workloads with desktop ones.

"

You're confusing server CPUs and workloads with desktop ones."
nope i'm just talking in general. The architecture is the same. They can use it to built a server wich are more critical to them in first place (banking for instance). and they can use the same technologie for laptops. The cores are the same. The rest is different


"

That's apples and oranges vs. classical CPU design."
That was the answer of an earlier post saying they lag in quantum and in manfacturing process. conception and production are two different things, i must have specified it on this page or another one. That's a tech forum here, you expect people to catch the difference between apples and oranges without having to write a book about it, especially on such a complex topic.

 

[bit_user]

"You're confusing server CPUs and workloads with desktop ones."
nope i'm just talking in general. The architecture is the same. They can use it to built a server wich are more critical to them in first place (banking for instance). and they can use the same technologie for laptops. The cores are the same. The rest is different

Your example involved personal computing needs, but these are server processors. If you want to make a case that server processor requirements can be easily met with older-generation technology, that requires different evidence.

My sense is that server workloads are getting more complex and intensive, year-by-year. That should make it difficult and costly to step down from current-gen AMD/Intel server CPUs to processors equivalent to their ~2017 models. You can do it, but suddenly you're going to need a lot more computers and it'll take a lot more energy to run and cool them.

 

[tomscomments]

Your example involved personal computing needs, but these are server processors. If you want to make a case that server processor requirements can be easily met with older-generation technology, that requires different evidence.

My sense is that server workloads are getting more complex and intensive, year-by-year. That should make it difficult and costly to step down from current-gen AMD/Intel server CPUs to processors equivalent to their ~2017 models. You can do it, but suddenly you're going to need a lot more computers and it'll take a lot more energy to run and cool them.

and that's exactly what they'll do if they don't break the euv bottlneckand get more pressure from us
in fact, they are already doing it.