Zhaoxin's ZX-F/KX-7000/KH-40000 and beyond

Page 9 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

mikegg

Golden Member
Jan 30, 2010
1,755
411
136
I think a lot of people are underestimating the Chinese's ability to produce a competitive chip in the future.

Sure, they're well behind right now. But it wouldn't surprise me if they're competitive within 10 years.

This trade war has reinforced the idea that China needs to develop their own silicon or risk getting blackballed by the U.S.

And Trump has decided to virtually ban work visa immigration overnight, which means a lot of top talent will be bought out by Chinese companies.

I equate this to Korean car makers who used to make really crap cars and got made fun of. They kept trying and now both Hyndai and Kia are top 10 car makers.
 
Last edited:

DrMrLordX

Lifer
Apr 27, 2000
21,619
10,827
136
I think a lot of people are underestimating the Chinese's ability to produce a competitive chip in the future.

Sure, they're well behind right now. But it wouldn't surprise me if they're competitive within 10 years.

How long is that going to take? VIA/Zhaoxin has been at work for years and haven't gained much ground. In fact they appear to be losing ground. Now they're going to try dGPUs too? At this point, if I were a "pro-China" shop I would be looking at the 8c/8t Kunpeng boards and trying to do things in Linux.
 

mikegg

Golden Member
Jan 30, 2010
1,755
411
136
How long is that going to take? VIA/Zhaoxin has been at work for years and haven't gained much ground. In fact they appear to be losing ground. Now they're going to try dGPUs too? At this point, if I were a "pro-China" shop I would be looking at the 8c/8t Kunpeng boards and trying to do things in Linux.
But it wouldn't surprise me if they're competitive within 10 years.
China's goal is to move up the value chain. If they want to focus on semiconductors, it won't surprise me if they're competitive within 10 years.

For example, they decided 5G was important so they blew the competition out.

As someone who has lived in China, Americans don't understand just how competitive the Chinese people are. There is no communism left. It's pure capitalism. Dog eat dog. The competition inside China is brutal and this inside competition is what is driving much of their competitiveness globally.

In addition, they have something the U.S. has: a huge domestic market. Their companies can grow into multi-billion dollar entities serving only the domestic population and then wreck the competition with their sheer size globally, ala Facebook, Amazon, Apple, Microsoft, etc.
 
Last edited:
  • Like
Reactions: KompuKare

DrMrLordX

Lifer
Apr 27, 2000
21,619
10,827
136
For example, they decided 5G was important so they blew the competition out.

Yeah and spawned 1000+ articles written by jingoistic political hacks in the US about how we're "losing at 5G" even if they have no idea of what it is or to whom it matters.

China threw as many resources as they were prepared to throw at semicon years ago starting with Longsoon/Godson, and really got nowhere with it. They've poured money into VIA, Huawei, and other ventures as well. So call me unconvinced. I'm not saying it's impossible, but in 10 years, we may be at the point of physical limits of semiconductors anyway, which will cause any lead non-Chinese design firms have to slowly evaporate just from being unable to rely on node improvements.
 

Thunder 57

Platinum Member
Aug 19, 2007
2,674
3,796
136
Yeah and spawned 1000+ articles written by jingoistic political hacks in the US about how we're "losing at 5G" even if they have no idea of what it is or to whom it matters.

China threw as many resources as they were prepared to throw at semicon years ago starting with Longsoon/Godson, and really got nowhere with it. They've poured money into VIA, Huawei, and other ventures as well. So call me unconvinced. I'm not saying it's impossible, but in 10 years, we may be at the point of physical limits of semiconductors anyway, which will cause any lead non-Chinese design firms have to slowly evaporate just from being unable to rely on node improvements.

There's a new word for me. And remember, 5G causes COVID-19. /s
 
  • Haha
Reactions: lightmanek

mikegg

Golden Member
Jan 30, 2010
1,755
411
136
Yeah and spawned 1000+ articles written by jingoistic political hacks in the US about how we're "losing at 5G" even if they have no idea of what it is or to whom it matters.
Conspiracy theory much? :confused_old:

Edit: I will stop the conversation here since you've decided to devolve into the conspiracy theories realm.
 
Last edited:

DrMrLordX

Lifer
Apr 27, 2000
21,619
10,827
136
Conspiracy theory much? :confused_old:

No. It's exactly what played out in the newspaper editorial pages. WE'RE LOSING AT 5G OH NOOOOOOO. Practically nobody who wrote those stupid fluff pieces even understood the implications of Chinese firms rolling out 5G infrastructure earlier. There is or was no conspiracy.

Point being, "winning at 5G" didn't cost the Chicoms nearly anywhere near as much as catching up to Intel, AMD, Apple, ARM Holdings, etc. would cost them, and I see no specific evidence that it's going to work out for them either. You can't just throw money at the problem.
 

ThatBuzzkiller

Golden Member
Nov 14, 2014
1,120
260
136
Point being, "winning at 5G" didn't cost the Chicoms nearly anywhere near as much as catching up to Intel, AMD, Apple, ARM Holdings, etc. would cost them, and I see no specific evidence that it's going to work out for them either. You can't just throw money at the problem.

Pretty sure Huawei got tens of billions of dollars worth of subsidies over the years from the state and the creation of their own wireless standard, TD-SCDMA was also a failed costly endeavor too so I don't think you understand just how much aid or patience was needed for Huawei to succeed in wireless network technology. Before China didn't have any choice but to accept the American technology standards like CDMA if they wanted mobile networks so it was an entirely one-sided exchange. It took Huawei years of participation in the 3GPP group to get to the point where they are now and it still took them years to make fairly big contributions like TDD-LTE technology which was dated back early last decade.

Zhaoxin by comparison didn't even exist until well after 2 decades Huawei first opened! If AMD was able to to create competitive x86 CPUs upto this point with assets totaling less than $10B of which I highly doubt that expenses even came anywhere near close to that figure for their architecture design team then it's far from impossible for Zhaoxin to catch up. The only reason why Zhaoxin isn't seeing as much progress as either Huawei, Tsinghua Unigroup, or even SMIC is mostly down to priority ...

I think the Chinese state was originally set on coasting with foreign designed x86 CPUs until 2030 but then plans quickly changed so they now want to see results by the end of 2025 which will coincide with their 14th five-year plan being in motion. The CCP have all the time in the world to oversee their plans coming into fruition as a benefit of been indefinitely in power and they have the patience to obtain the results they want ...
 

DrMrLordX

Lifer
Apr 27, 2000
21,619
10,827
136
Zhaoxin by comparison didn't even exist until well after 2 decades Huawei first opened!

Huawei has likely absorbed more subsidies for their semiconductor research than Zhaoxin by a large margin. There's also the money that the PRC has thrown at SMIC, and other smaller efforts in the past. Part of "catching up" to the durn furriners will involve narrowing the node disadvantage between SMIC and TSMC, assuming, you know, TSMC's ownership doesn't change hands. Zhaoxin is a bit player in comparison.
 

podspi

Golden Member
Jan 11, 2011
1,965
71
91
China's goal is to move up the value chain. If they want to focus on semiconductors, it won't surprise me if they're competitive within 10 years.

For example, they decided 5G was important so they blew the competition out.

As someone who has lived in China, Americans don't understand just how competitive the Chinese people are. There is no communism left. It's pure capitalism. Dog eat dog. The competition inside China is brutal and this inside competition is what is driving much of their competitiveness globally.

In addition, they have something the U.S. has: a huge domestic market. Their companies can grow into multi-billion dollar entities serving only the domestic population and then wreck the competition with their sheer size globally, ala Facebook, Amazon, Apple, Microsoft, etc.

It's interesting because I've heard the same thing in relation to the U.S. and Europe. Not sure if it's something to be proud of...
 

Antey

Member
Jul 4, 2019
105
153
116
i dont know if its already done but i couldnt find it so i measured CNS core area and it is 4,367mm2 on TSMC 16FF

Total área is 195mm2 or 1014*938 pixels or 14,52mm*13,43 (more or less)
1 mm = 69,834 pixels

NCORE is 295px*567px or 4,224mm*8,119mm or 34,29mm2 (more or less, semiaccurate says its 34,4mm2)
CNS core is 179px*119px or 2,563mm*1,704 or 4,367mm2

 

DrMrLordX

Lifer
Apr 27, 2000
21,619
10,827
136
It took 1349s to render the BMW Blender scene on the KX-U6780a. Classroom took 4274s. Oh my goodness. It beat an old AMD 7870k in CBR20 MT though! Lost badly in ST, but still . . . that's something, right?

 

Kosusko

Member
Nov 10, 2019
161
120
116
It took 1349s to render the BMW Blender scene on the KX-U6780a. Classroom took 4274s. Oh my goodness. It beat an old AMD 7870k in CBR20 MT though! Lost badly in ST, but still . . . that's something, right?



Comparison in CPU-Z and Cinebench R20 with AMD 7870 to 3.9 GHz at 4.1 GHz QuadCore. Did I miss something?


AMD 7870k 3.9 GHz at 4.1 GHz QuadCore vs. Zhaoxin KX-U6780A 2.7GHz OctaCore

CPU-Z: 797,6 (55,37%) vs. 1 441 (+80,67%)
CB20: 743 (79,64%) vs. 933 (+25,57%)

 

Asterox

Golden Member
May 15, 2012
1,026
1,775
136
Another german review:

An exotic from China with eight cores: The Zhaoxin KX-U6780A in the test
source:

P.S. anandtech review is still on the way...

Ok, 8 Core Zhaoxin KX-U6780A/917 vs 2/4 Athlon 3000G/913 in Cinebench R20 Multithread.

Blah, blue is going direct in the garbage.We should wait for new KX 7000 series.

2020-07-24_180904.jpg
 
Last edited:

DrMrLordX

Lifer
Apr 27, 2000
21,619
10,827
136
Comparison in CPU-Z and Cinebench R20 with AMD 7870 to 3.9 GHz at 4.1 GHz QuadCore. Did I miss something?

1). A10-7870k is a joke. Anyone automatically loses if they are hitting near the performance of that chip in 2020, much less losing to it in any capacity
2). A10-7870k scores ~220 in CBR20 ST, Zhaoxin U6780A gets a 121 . . . on a core-by-core basis

One core from a KX-U6780a is nearly 50% slower than one module from an old Steamroller APU (when handling one thread) in a fp benchmark. That ain't good folks. Later let's compare this review to the 8c/8t Kunpeng 920 (sorry, running short on time).