Loongson 3A6000 CPU Reportedly Matches AMD Zen 4 and Intel Raptor Lake IPC

[AleksandarK]

China's homegrown Loongson 3A6000 CPU shows promise but still needs to catch up AMD and Intel's latest offerings in real-world performance. According to benchmarks by Chinese tech reviewer Geekerwan, the 3A6000 has instructions per clock (IPC) on par with AMD's Zen 4 architecture and Intel's Raptor Lake. Using the SPEC CPU 2017 processor benchmark, Geekerwan has clocked all the CPUs at 2.5 GHs to compare the raw benchmark results to Zen 4 and Intel's Raptor Lake (Raptor Cove) processors. As a result, the Loongson 3A6000 seemingly matches the latest designs by AMD and Intel in integer results, with integer IPC measured at 4.8, while Zen 4 and Raptor Cove have 5.0 and 4.9, respectively. The floating point performance is still lagging behind a lot, though. This demonstrates that Loongson's CPU design can catching up to global leaders, but still needs further development, especially for floating point arithmetic.

However, the 3A6000 is held back by low clock speeds and limited core counts. With a maximum boost speed of just 2.5 GHz across four CPU cores, the 3A6000 cannot compete with flagship chips like AMD's 16-core Ryzen 9 7950X running at 5.7 GHz. While the 3A6000's IPC is impressive, its raw computing power is a fraction of that of leading x86 CPUs. Loongson must improve manufacturing process technology to increase clock speeds, core counts, and cache size. The 3A6000's strengths highlight Loongson's ambitions: an in-house LoongArch ISA design fabricated on 12 nm achieves competitive IPC to state-of-the-art x86 chips built on more advanced TSMC 5 nm and Intel 7 nm nodes. This shows the potential behind Loongson's engineering. Reports suggest that next-generation Loongson 3A7000 CPUs will use SMIC 7 nm, allowing higher clocks and more cores to better harness the architecture's potential. So, we expect the next generation to set a bar for China's homegrown CPU performance.



View at TechPowerUp Main Site | Source

 

[Vayra86]

'Father, what is this strange chip?'
'On this CPU, tasks still take a bit loong son'

I'll see myself out now

 

[Crackong]

A 4c8t Max 2.5GHz CPU with a TDP of 80W in 2024
Products from the major manufacturers in this CPU power is about 5W at MAX

Good luck selling this thing outside of CN government contracts driven by patriotism.

 

[Haile Selassie]

A 4c8t Max 2.5GHz CPU with a TDP of 80W in 2024
Products from the major manufacturers in this CPU power is about 5W at MAX

Good luck selling this thing outside of CN government contracts driven by patriotism.

View attachment 334340

But that's the whole point of this CPU. It's a domestically built CPU that will be used as drop-in replacement for critical infrastructure, most likely for internet facing devices.
Security through obscurity.

 

[Owen1982]

Well... matching IPC is one thing but when the base i3 from 4+years ago still performs better using less power whilst using a slightly larger process then it's not really anything to shout about is it???

 

[qcmadness]

Frequency target is part of the architecture preferences. See Zen 4 Vs Zen 4c

 

[lilhasselhoffer]

But that's the whole point of this CPU. It's a domestically built CPU that will be used as drop-in replacement for critical infrastructure, most likely for internet facing devices.
Security through obscurity.

So...is this about security at all? It seems more like the Chinese government stating to the west "we can be fine without you." It's then turned around, and in order to show they're fine they had to make their own relatively obscure benchmark, disregard floating point calculations, and forget that they're running at half the effective clock speed and with so much poorer power efficiency that it's a joke.

Yeah, I can make a car just as good as yours. It's going to weight five times as much, consumed fuel will be something akin to a boat (gallons per mile, instead of miles per gallon), but the interior will be just as comfortable and luxurious as a middle-tier Honda. What, that's not really viable? The car we've designed isn't really a socketable replacement and thus would also require its own redesigned roadway with lanes 30% wider. Yeah, but all of these are small details....look at that interior.



I see this collapsing. It's exactly the same proposition as changing instruction sets...and decades later after RISC we still run x86. Sometimes in demonstrating that "you don't need" a thing you highlight why it's necessary. I'd suggest that this CPU does exactly that...and the Chinese market isn't deep enough to support an entirely different and significantly less efficient chip that will likely never be able to be produced in enough volume to ever actually support a separate market. Funny that...it's almost like the core issue is that to succeed you might need more than one "equal footing" claim with a mile of qualifiers and an insufficient manufacturing volume to create your own independent market that will have to compete with the incumbent. It's almost like why ARM is so big because it created its parallel market focusing on an emerging market (mobile) and extreme niche (power efficiency) to carve out something x86 couldn't competed with...and this new processor is, by their definition, gunning for direct competition with established and easily better hardware.

Maybe if these things are dirt cheap, and rapidly iterate, they've got a chance. I...just don't see it. What I see is another claim from a company that wants the world but will flame out when they realize what it costs to genuinely compete. Please note that a decade ago the news stories were about how China was going to have their own home grown processors and OS for governmental entities. It's 2024...and the story now is that soon they'll compete with western tech. In both cases the evidence was basically a technical demonstration without any additional plans beyond "we'll throw money at it." A story that always ends in failure because all that thrown money lines pockets until the project is eventually buried as a "success" that looks more like a failure. Talk about "with Chinese characteristics" being the kiss of death.

 

[Denver]

Well, I don't give a damn about ARM chips, and I don't care about this either... but I'd like to see Risc-V take on ARM in smartphones.

 

[aktpu]

They have the budget to iterate on this for a long time unlike western companies, you might not care now but maybe in 5-10 years time

 

[TheLostSwede]

Well, I don't give a damn about ARM chips, and I don't care about this either... but I'd like to see Risc-V take on ARM in smartphones.

This is a xinese MIPS variant though.

 

[kondamin]

How expensive are they to buy and to make?

if you cut out advertisements with an add blocker you can be very productive with 15 year old hardware

 

[Icon Charlie]

This is a xinese MIPS variant though.

Unfortunately, I can not trust anything tech coming out of the country. The Tofu Dreg Construction concept is fully engrained into the culture.

 

[R0H1T]

'Father, what is this strange chip?'
'On this CPU, tasks still take a bit loong son'

I'll see myself out now

Spoiler

 

[LFaWolf]

I think for starter they will put this cpu for office work computers. Along with their variant of Linux os and open source office suite, now they can create an environment that doesn’t depend on the West.

 

[Baum]

just the bad "West" who made all the open source software in the first place :-9

 

[gruffi]

Pointless. First, IPC might not scale linearly with clock speed. And second, SPEC is synthetic. Real world apps are a different story. Not to mention the power consumption. 80W TDP for just 4 cores at such low clock speed is horrible, even on 12nm. Looks to me like a desperate brute force design. And not like something that will be competitive anytime soon.

 

[Crackong]

even on 12nm.

True.
Major manufacturers had 2.5GHz CPUs even at 130nm process era.
This 3A6000 must he limited by other aspects (like FSB) so the clock speed maxed at 2.5GHz without breaking other things.

 

[Lew Zealand]

80W TDP for just 4 cores at such low clock speed is horrible, even on 12nm.

This here. The 4c4t i5-4690 in my Dell Optiplex uses max 50W under all-core load at 3.7 GHz and that's a 22nm CPU. There's more than a bit of catching up needed when it comes to efficiency.

 

[DavidC1]

This here. The 4c4t i5-4690 in my Dell Optiplex uses max 50W under all-core load at 3.7 GHz and that's a 22nm CPU. There's more than a bit of catching up needed when it comes to efficiency.

i5 4690 is a 89W part, not 50W. Just because modern CPUs have advanced power management and in lot of cases are under TDP doesn't mean the TDP part is wrong.

Further, Zen 4 level PPC means it's more than 50% faster per clock than the Haswell in that i5, thus a 2.5GHz 3A6000 is equal or even slightly better than your 3.7GHz i5.

Competition-wise, even being just 1 year behind is serious deficiency, but in the big picture, reaching Golden Cove/Zen 4 levels are respectable.

Pointless. First, IPC might not scale linearly with clock speed. And second, SPEC is synthetic. Real world apps are a different story. Not to mention the power consumption. 80W TDP for just 4 cores at such low clock speed is horrible, even on 12nm. Looks to me like a desperate brute force design. And not like something that will be competitive anytime soon.

SPEC is a complicated enough benchmark that it reflects modern usage cases pretty well, especially the 1T Int part.

Also, more than brute force is needed for high PPC, otherwise everyone would have done it already. 2.5GHz quad core Zen 4 is more than enough for replacing for vast majority of use cases when the inevitable full sanctions against China is implemented after WWIII starts, and having a fast CPU ends up being the LAST thing anyone cares about.

 

[Lew Zealand]

i5 4690 is a 89W part, not 50W. Just because modern CPUs have advanced power management and in lot of cases are under TDP doesn't mean the TDP part is wrong.

No. I'm talking about actual power use, not TDP. Reported power for this Chinese CPU is actual power use.

The i5 4690 has a listed TDP of 84W which I have tested it extensively and all core loads put it at about 50W max unless you're running a specific power virus and then it only gets to about 60W. That is its actual power draw. Intel's current philosophy of draw as much power as your PSU can handle at all costs was not always the case.

Even the higher spec i7-4790 (also 84W TDP) doesn't exceed 65W (75W for power virus) in all work scenarios and I have 4 copies of that CPU which perform similarly.

Further, Zen 4 level PPC means it's more than 50% faster per clock than the Haswell in that i5, thus a 2.5GHz 3A6000 is equal or even slightly better than your 3.7GHz i5.

Looking at CPU-Z single core (easiest thing to quickly find), Zen 4's IPC is less than 20% higher than Haswell. Per Clock. Modern performance is as much about increased clock speed as same-speed instruction issue. Sure, other benchmarks will put the difference higher than CPU-Z but that's on you.

And IPC wasn't the point. The point was power efficiency which Haswell has all over this new CPU, on a fatter node at higher clocks. It's new and they will improve it but right now it's very inefficient. Apple has shown that you can focus efficiency instead of clock speed and achieve good performance and this processor hasn't done that yet. But I'm sure they will continue to improve it, we'll see how it goes.

 

[Frick]

So...is this about security at all? It seems more like the Chinese government stating to the west "we can be fine without you." It's then turned around, and in order to show they're fine they had to make their own relatively obscure benchmark, disregard floating point calculations, and forget that they're running at half the effective clock speed and with so much poorer power efficiency that it's a joke.

Yeah, I can make a car just as good as yours. It's going to weight five times as much, consumed fuel will be something akin to a boat (gallons per mile, instead of miles per gallon), but the interior will be just as comfortable and luxurious as a middle-tier Honda. What, that's not really viable? The car we've designed isn't really a socketable replacement and thus would also require its own redesigned roadway with lanes 30% wider. Yeah, but all of these are small details....look at that interior.



I see this collapsing. It's exactly the same proposition as changing instruction sets...and decades later after RISC we still run x86. Sometimes in demonstrating that "you don't need" a thing you highlight why it's necessary. I'd suggest that this CPU does exactly that...and the Chinese market isn't deep enough to support an entirely different and significantly less efficient chip that will likely never be able to be produced in enough volume to ever actually support a separate market. Funny that...it's almost like the core issue is that to succeed you might need more than one "equal footing" claim with a mile of qualifiers and an insufficient manufacturing volume to create your own independent market that will have to compete with the incumbent. It's almost like why ARM is so big because it created its parallel market focusing on an emerging market (mobile) and extreme niche (power efficiency) to carve out something x86 couldn't competed with...and this new processor is, by their definition, gunning for direct competition with established and easily better hardware.

Maybe if these things are dirt cheap, and rapidly iterate, they've got a chance. I...just don't see it. What I see is another claim from a company that wants the world but will flame out when they realize what it costs to genuinely compete. Please note that a decade ago the news stories were about how China was going to have their own home grown processors and OS for governmental entities. It's 2024...and the story now is that soon they'll compete with western tech. In both cases the evidence was basically a technical demonstration without any additional plans beyond "we'll throw money at it." A story that always ends in failure because all that thrown money lines pockets until the project is eventually buried as a "success" that looks more like a failure. Talk about "with Chinese characteristics" being the kiss of death.

Do they want the world though? According to wikipedia "The processors are mainly used in Chinese computers; in 2021 it was reported that Loongson supplies CPUs for most desktop computers procured by the Chinese government, and 80% of the Chinese government's servers." That's a lot of chips right there.

 

[马嘉伟]

So...is this about security at all? It seems more like the Chinese government stating to the west "we can be fine without you." It's then turned around, and in order to show they're fine they had to make their own relatively obscure benchmark, disregard floating point calculations, and forget that they're running at half the effective clock speed and with so much poorer power efficiency that it's a joke.

Yeah, I can make a car just as good as yours. It's going to weight five times as much, consumed fuel will be something akin to a boat (gallons per mile, instead of miles per gallon), but the interior will be just as comfortable and luxurious as a middle-tier Honda. What, that's not really viable? The car we've designed isn't really a socketable replacement and thus would also require its own redesigned roadway with lanes 30% wider. Yeah, but all of these are small details....look at that interior.



I see this collapsing. It's exactly the same proposition as changing instruction sets...and decades later after RISC we still run x86. Sometimes in demonstrating that "you don't need" a thing you highlight why it's necessary. I'd suggest that this CPU does exactly that...and the Chinese market isn't deep enough to support an entirely different and significantly less efficient chip that will likely never be able to be produced in enough volume to ever actually support a separate market. Funny that...it's almost like the core issue is that to succeed you might need more than one "equal footing" claim with a mile of qualifiers and an insufficient manufacturing volume to create your own independent market that will have to compete with the incumbent. It's almost like why ARM is so big because it created its parallel market focusing on an emerging market (mobile) and extreme niche (power efficiency) to carve out something x86 couldn't competed with...and this new processor is, by their definition, gunning for direct competition with established and easily better hardware.

Maybe if these things are dirt cheap, and rapidly iterate, they've got a chance. I...just don't see it. What I see is another claim from a company that wants the world but will flame out when they realize what it costs to genuinely compete. Please note that a decade ago the news stories were about how China was going to have their own home grown processors and OS for governmental entities. It's 2024...and the story now is that soon they'll compete with western tech. In both cases the evidence was basically a technical demonstration without any additional plans beyond "we'll throw money at it." A story that always ends in failure because all that thrown money lines pockets until the project is eventually buried as a "success" that looks more like a failure. Talk about "with Chinese characteristics" being the kiss of death.

Well......China does not intend to use it to replace Intel or AMD. Loongson used the MIPS to make 3A4000, and start with the 3A5000, they use self-develop LoongArch ISA, not X86 or ARM. This CPU is mainly used in the military and government, it isn't a "to c" product. For this reason, it will continue to roll out the next generation of products steadily, no need to worry. Reason is China doesn't trust Windows and American cpus, just like US doesn't trust Huawei. That is why Loongson designed this.
If you want to try a CPU other than I A, you can expect the Zhaoxin KX7000. It uses X86 ISA, 8 "世纪大道" cores, can run Windows even crysis. It's more like a "to c" product. It is manufactured by 7nm from SMIC, single core matches 3770K, and multi core matches 10400 (maybe).