又一颗工业明珠？中国的兆芯 KX-7000 CPU 性能是上一代芯片的两倍，但仍比 AMD 和 Intel 的处理器落后了6年
China's Zhaoxin KX-7000 CPU doubles performance of prior-gen chip, still trails 6-year-old AMD and Intel processors
译文简介
Ryzen 5 3600 处理器已有大约 5-6 年的历史，目前仍在 PC 中使用，性能并不比游戏机弱多少。
正文翻译
China's Zhaoxin KX-7000 CPU doubles performance of prior-gen chip, still trails 6-year-old AMD and Intel processors
中国的兆芯 KX-7000 CPU 性能是上一代芯片的两倍，但仍比 AMD 和 Intel 的处理器落后了6年
评论翻译
很赞 ( 19 )
收藏
A roughly 5-6 year old processor is the Ryzen 5 3600, still present in PCs and not that much weaker than what's on a console. So their performance is about second gen Ryzen? Or entry 8th-9th gen Intel equivalent? That's doable for anything office
Ryzen 5 3600 处理器已有大约 5-6 年的历史，目前仍在 PC 中使用，性能并不比游戏机弱多少。所以它们的性能大约是第二代 Ryzen？或者相当于英特尔第 8-9 代入门级处理器？这已经足以应付任何办公室任务了。
Yeah and it’s not like governments replace hardware every year.
是的，而且政府也不是每年都更换硬件。
I have a 3600 in my PC paired with 32gb ram and a 3060ti.
Not only is it good enough for the vast majority of workloads, it's also fine for games. I almost upgraded the CPU to a 5800x3D this chirstmas, but ultimately it's just wasn't worth it. I'll wait for next year's sales on AM5 chips.
China is definitely catching up fast.
我的电脑使用的是 3600，搭配 32GB 内存和 3060ti。
它不仅能满足绝大多数工作负载的需求，在游戏方面也很出色。今年圣诞节，我差点把 CPU 升级到 5800x3D，但最终还是觉得不值。我还是等明年的 AM5 芯片促销吧。
中国绝对在快速迎头赶上。
原创翻译：龙腾网 https://www.ltaaa.cn 转载请注明出处
6 year old entry level performance is an accomplishment but the gap between that and top tier modern chips is a vast gulf in production capability that depends on unfettered access to the world market and friendly bilateral relations with multiple western countries.
达到 6 年前的入门级性能是一项成就，但它与顶级现代芯片之间的差距是生产能力上的巨大鸿沟，而这又取决于能否不受限制地进入世界市场以及与多个西方国家的友好双边关系。
I’m running the Ryzen 5 3600 with a 3060FE and I still game on it with my son regularly. It plays all the current titles I’ve tried without any issues. When it’s not gaming it is searching for black holes for Einstein@Home.
我现在使用的是 Ryzen 5 3600 和 3060FE，我还经常和儿子一起用它玩游戏。它可以玩我尝试过的所有当前的游戏，没有任何问题。不玩游戏时，它就在为 Einstein@Home 寻找黑洞。
（Einstein@Home，是一个由威斯康星大学密尔沃基分校主办，基于BOINC计算平台的分布式计算项目）
I dunno man it's nice progress but if you check out the article, it's slightly behind the rx2500x.
These weren't even top tier cpus at that time, and this is their best effort with many years of corporate espionage assisting them.
It's notable and not something to ignore, but it also shows a very clear and large gap in their capabilities vs U.S.A.
我觉得还不能这么说，这算是不错的进步，但如果你看看文章，它比 rx2500x 还稍逊一筹。
这些处理器在当时甚至都算不上顶级，而这是他们在多年的企业间谍活动协助下做出的最大成效。
它是值得注意的，不容忽视的，但同时也显示出他们与美国在能力上存在着非常明显的巨大差距。
I’d imagine the 3060Ti is doing most of the heavy lifting when it comes to your gaming performance.
说到你的游戏性能，我想 3060Ti 承担了大部分重任。
I got 2 non overclocked EVGA 980 Ti running in SLI and it can still run most games in 1440p with ok ish frx rates. I don't know if that speak volumes about how upgrading computer hardware every year is more just for those people with too much cash or if the 980 Ti was such a beast of a chip.
我有两张未超频的 EVGA 980 Ti 以 SLI 的方式运行，它仍然可以在 1440p 下运行大多数游戏，帧率还可以。我不知道这是否说明了每年升级电脑硬件更多的是为了那些有太多闲钱的人，还是因为 980 Ti 是一款如此强大的显卡。
原创翻译：龙腾网 https://www.ltaaa.cn 转载请注明出处
It was a beast of a chip, the 1660TI and 2060 are comparable in power for certain games despite being much newer. It also has generous headroom to overclock if you wanted to squeeze a bit more life out of them.
Aside from the obvious lack of power compared to modern GPUs, RTX and DLSS missing is a loss.
GPUs in general were also much more affordable back then for many reasons. If that chip was released in the current environment, I think the price/performance would suffer greatly.
Congrats on still being able to enjoy them!
这是一张野兽级的显卡，1660TI 和 2060 虽然要新得多，但980ti在某些游戏中的性能和它们不相上下。如果你想从它们身上榨取更多的生命力，它还拥有宽裕的超频空间。
与现代 GPU 相比，只有光追和DLSS的缺乏是一种损失。
由于种种原因，当时的 GPU 价格也要便宜得多。如果该显卡在当前环境下发布，我认为性价比会大打折扣。
恭喜你还能享受它们！
Considering im still running on a 4th gen i7. This stuff is perfectly usable for most people.
考虑到我仍在使用第 4 代 i7。对大多数人来说，这东西完全能用。
That's good enough for most productivity use cases.
对于大多数生产力使用任务来说，这已经足够了。
Currently.
Until everyone else is getting local LLMs running on their shitty office hardware.
And the productivity apps start being built with the expectation that an NPU will be there.
That’s the problem with being very far behind: the rest of the world builds stuff to match the state of the market.
“目前”够用。
直到每个人都在低劣的办公硬件上运行本地 LLM。
而生产力应用软件的开发者开始以 NPU 为目标开发软件。
这就是过于落后的问题所在：世界上其他国家都在建造与市场状况相匹配的产品。
Maybe but they are making stuff and one could assume they'll be getting better at it as time goes by.
也许吧，但他们正在进步，而且随着时间的推移，我们应当认为他们会做得更好。
Sure, but all their competitors are too, which makes that a moving target.
当然，但他们所有的竞争对手也是如此，这就使得目标不断变化。
Intel wasn't until AMD showed up in the late 2010s. The moment either of them begin to slack China will surpass them.
英特尔一直裹足不前，直到AMD在2010 年代末突然出现。一旦它们中的任何一家开始懈怠，中国就会超越它们。
AMD has been around a lot longer than that. They put out some good options, but had some missteps and Intel pulled some shady shit to keep them down. They were never 6 years behind, but they were a budget chip complete with budget performance. Then Intel got lazy and AMD hit big with Ryzen.
Lisa Su has AMD in a great place and Intel won’t roll over. They aren’t going to get lazy anytime in the near term.
Chinese chip makers could catch up, nothing is stopping them from innovating, but Intel and AMD are not going to get lazy, they are in a serious battle against each other.
AMD 的历史要比这长得多。他们推出了一些不错的选择，但也有一些失误，而英特尔则采取了一些不正当的手段来压制他们。他们从未落后于英特尔 6 年，而且他们的芯片价格低廉，性能不俗。后来，英特尔开始偷懒，AMD 则凭借 Ryzen 一炮而红。
苏丰姿让 AMD 占据了有利位置，英特尔不会就此罢休。他们在短期内不会再挤牙膏。
中国芯片制造商可能会迎头赶上，没有什么能阻止他们的创新，但英特尔和 AMD 不会偷懒，他们正在进行一场严肃的较量。
In technology it is easier to catch up than to break away, so the distance will get smaller and under 3 years the difference will be almost insignificant...
Still a lot to do but nothing insurmountable.
IMO the real challenge for china is building competitive photolitography machines
在技术领域，追赶比超越更容易，因此距离会越来越小，3 年内的差距几乎微不足道...
要做的事情还有很多，但没有什么是不可逾越的。
在我看来，中国面临的真正挑战是制造具有竞争力的照相摄影机。
I work in this sector, they are catching up so fkin fast. I wouldn’t be surprised if they will take over within the next 10 years. China throw everything at this kinda shit, while murica just fight each other
我在这个行业工作，他们追赶得太快了。如果他们能在未来 10 年内取而代之，我一点也不会感到惊讶。中国在这方面无所不用其极，而美国只会在内部互相争斗
Playing catch up is a bit faster than being on the leading edge. All they have to do is look at what you’re doing and reverse/re-engineer that.
追赶总是比领先要快一些。因为他们要做的就是看看你在做什么，然后反向/重新设计。
原创翻译：龙腾网 https://www.ltaaa.cn 转载请注明出处
Their competitors have been doing it longer and have more experience, so they will get better faster.
Building advanced chips is absurdly hard. If it were easy, more than 2 companies would be doing it well.
他们的竞争对手在这一领域做的时间更长，经验更丰富，所以进步得会更快。
制造先进的芯片难上加难。如果这很容易，就不会只有两家公司做得好了。
(Not disagreeing with your overall point, however there are a few things I did want to point out.) If we are just looking at x86-64 then yeah only 2 of those 2 companies only Intel designs and builds them, however if your willing to consider ARM you get several others. Still insanely hard to accomplish. Also keep in mind even Intel had struggles with this if you remember them getting to "10 nm" was an utter shitshow that allowed AMD to catch up. The other thing to keep in mind is there are many other types of high performance processor types for example GPUs (NVIDIA/AMD/Intel dominate) FPGAs (Intel owns Altera AMD now owns xilinx) Memory (Samsung Micron SK Hynix) and many other types of chips are designed by various companies.
(我并不反对你的总体观点，但有几点我想指出）。如果我们只看 x86-64，那么是的，这两家公司中只有英特尔在设计和制造它们，但是如果你愿意考虑 ARM，你就会看到其他几家公司。但要实现这些目标仍然非常困难。此外，如果你还记得英特尔在“10 纳米”上的努力，你就会知道英特尔在这方面也有过挣扎，这完全是一场灾难，让 AMD 实现了赶超。还有一点要记住的是，还有许多其他类型的高性能处理器，例如 GPU（英伟达/AMD/英特尔占主导地位）、FPGA（英特尔拥有 Altera AMD，现在拥有赛灵思）、内存（三星美光 SK 海力士）以及由不同公司设计的许多其他类型的芯片。
While catching up will become more and more difficult as nodes shrink and the design becomes more complex, this is quite the jump! And most importantly, it is powerful enough for a lot of tasks.
Not everything is CFD level CPU-intensive, and plenty of people do their work with things as basic as celeron powered chromebooks, intel macbooks and more than 6 year old systems after all.
虽然随着节点的缩小和设计的复杂化，追赶的难度会越来越大，但这已经是相当大的飞跃了！最重要的是，它足以胜任很多任务。
毕竟，并非所有工作都是 CFD 级别的 CPU 密集型工作，很多人都在使用搭载赛扬处理器的 Chromebook、英特尔 Macbook 和超过 6 年的旧系统进行工作。
Yep, what are their yields/costs is a heck of an interesting question …
是的，但是他们的收益率/成本是一个非常有趣的问题......
原创翻译：龙腾网 https://www.ltaaa.cn 转载请注明出处
I think its really weird that a ton of the comments immediately went political rather than just discussing the tech side of things.
我觉得很奇怪的是，很多评论马上就变得政治化了，而不只是讨论技术方面的问题。
i thought i was in world news for a second
我一瞬间以为自己进的是“世界新闻”板块
Legitimately thought the same which prompted my comment. People need to let go of the tribalism, even if temporarily.
我也是这么想的。人们需要放下部落主义，哪怕是暂时的。
原创翻译：龙腾网 https://www.ltaaa.cn 转载请注明出处
Just your avg “China bad” comment section. They can’t help themselves.
就是你常见的“中国坏”评论区。他们就是控制不住自己。
I mean this is a political discussion first, technology second, you do realize that? The tech itself isn't cutting edge because of political reasons. Without the political context this wouldn't even be news...
我想说这首先是一场政治讨论，其次才是技术，你意识到了吗？因为政治原因，技术本身并不先进。如果没有政治背景，这甚至都不会成为新闻......
Does tech always need to be cutting edge? They certainly didn't make this chip with gaming in mind.
科技总是要走在时代前沿吗？他们在制造这款芯片时肯定没有考虑到游戏。
原创翻译：龙腾网 https://www.ltaaa.cn 转载请注明出处
So if the news isnt solely about cutting edge technological advancements, it cant be a discussion centered around technology. Got it.
/s for obvious clarity.
所以，如果新闻不完全是关于前沿技术进步的，那就不能围绕技术展开讨论。明白了。
加/s以防误伤。
Nobody is talking about production of an unimpressive half-decade-old design in the absence of political considerations. They only care because it's a gauge in the US-China chip war
在没有政治考量的情况下，没人会谈论生产一款已落后了5年的平庸设计。他们之所以关心，是因为它是中美芯片战的一个标尺
By your own additional they can't compete, yet you call it a War. You are just hate mongering like the rest of the goofs in this thread. I'm done talking about this.
根据你自己的补充，他们无法竞争，但你却称之为战争。你和这个话题中的其他笨蛋一样，都是在制造仇恨。我不想再说这个了。
原创翻译：龙腾网 https://www.ltaaa.cn 转载请注明出处
For someone putting together a gaming rig; this would not be sufficient.
But, as most here are saying it is wildly sufficient for the vast majority of workloads.
What I'm finding with the smaller Chinese chips is that they are innovating in other ways. The rockchip series are fantastic for embedded linux.
Even for fairly advanced users this KX-7000 would still be quite good, running things like basic Solidworks, etc. But, not the CFD and related tasks so much.
Another simple reality is that PC sales have generally tanked since the lockdowns ended; the used market is quite flush. Thus, chips aimed at new desktop/laptop buyers aren't really going to be a big thing.
I see two big markets for chips:
Data centers doing ML sort of stuff where they need the latest and greatest for the vast majority of their stuff.
Embedded systems such as robots, and other smart devices. These chips are perfect for that assuming a great price point.
This last is where the latest and greatest aren't that appropriate as you are often looking for the minimum capability required to keep power demands down. This is also a rapidly growing market.
It is misleading to say they are X years behind. There are some obstacles such as patents, etc,(no comment) but it is generally easier to catch up if someone else has already shown you the path. A simple analogy is if I don't know how to tie knots and I flounder around I will probably tie crappy knots and it will take me years or decades to find some pretty snazzy knots, even if I spend an hour or two per day. But if I read a basic book on knot tying, I will "catch up" in short order. One Knot Tying book could literally shrink a multi decade journey down to week or so.
I suspect that for every year forward, these guys will shrink the gap maybe 20% each year. Basically, as they close the gap their catching up will get slower.
One other interesting factor are things like the supporting chips. I suspect they will go vertical and start producing their own. This could end up reducing the whole board cost by a huge amount. It would be a game changer if they dropped a desktop board of some power for under $100.
Again, it is a pile of little things where they don't need to knock it out of the park. 1G ethernet is sufficient, etc.
对于组装游戏机的人来说，这还不够。
但是，正如这里大多数人所说的那样，对于绝大多数工作负载来说，这已经足够了。
我发现中国的小型芯片在其他方面也有创新。对于嵌入式 Linux 而言，Rockchip 系列是非常出色的。
即使对于相当高级的用户来说，KX-7000 也能很好地运行基本的 Solidworks 等。但是，CFD 和相关任务就不太适合了。
另一个简单的现实是，自从封锁结束后，PC 的销售量普遍下滑；二手市场相当火爆。因此，针对新台式机/笔记本电脑买家的芯片并不会大行其道。
我认为芯片有两大市场：
1. 从事机器学习类工作的数据中心，他们的绝大部分工作都需要最新、最好的芯片。
2. 嵌入式系统，如机器人和其他智能设备。如果价格合理，这些芯片就非常适合。
在最后一种情况下，最新和最好的芯片就不那么合适了，因为你往往需要最低限度的功能来降低功耗需求。这也是一个快速增长的市场。
说他们落后 X 年是一种误导。有一些障碍，如专利等，（不评价）但一般来说，如果别人已经为你指明了道路，你就更容易迎头赶上。一个简单的比喻是，如果我不知道如何打绳结，我就会随便乱打，我可能会打出一些蹩脚的绳结，即使我每天花一两个小时，我也要花几年或几十年的时间才能找到一些打漂亮绳结的方法。但如果我读一本基本的绳结打法书，我就能在短时间内“迎头赶上”。一本《绳结打法》就能把我几十年的学习历程缩短到一周左右。
我猜想，每向前迈进一年，这些人每年都会将差距缩小 20%。基本上，当他们缩小差距时，他们的追赶速度也会越来越慢。
还有一个有趣的因素，比如支持芯片。我猜测他们会采用垂直生产方式，开始自己生产。这样一来，整个电路板的成本就会大大降低。如果他们能以低于 100 美元的价格推出一款功能强大的台式机板，那将会改变游戏规则。
同样，这也是一堆小东西，他们不需要把它弄得多出类拔萃。1G 以太网就足够了，等等。
They obviously need to step up their massive IP theft game.
显然，他们需要加强大规模的知识产权盗窃游戏。
Stealing the IP is the easy part, manufacturing is hard.
Q取知识产权容易，制造不容易。
The US built itself on the land and over the bodies of millions and its wealth on the theft and enslavement of literal human beings by the millions and the stolen resources of many countries with innumerable wars, overthrowing of governments and support of dictator (whatever suite their need)...
Who cares about IP theft especially when it's illegally done to economically crush a rival?
美国建立在偷来的土地和数以百万计的尸体上，其财富来源于对成百上千万人的盗窃和奴役，以及通过无数次战争、推翻政府和支持独裁者（只要符合他们的需要）窃取许多国家的资源......
谁会在乎知识产权被盗，尤其是当这种非法行为是为了在经济上击垮对手时？
Well... Patent Act of 1793, the US barred foreign inventors from receiving patents at the same time as granting patents to Americans who had pirated technology from other countries. US thus became, by national policy , the world's premier legal sanctuary for industrial pirates.
呵呵...1793 年的《专利法》规定，美国禁止外国发明家获得专利，同时向从其他国家盗取技术的美国人授予专利。因此，根据国家政策，美国成为世界上最大的工业盗版者的法律庇护所。
Yeah this is classic “kicking away the ladder” behavior. America enriched itself doing the exact shit it’s now accusing China of doing.
是啊，这是典型的 “过河拆桥”行为。美国正是靠着现在自己指责中国的那些破事发家致富的。
Laugh it up. If theres one thing I’m sure of the Chinese are dedicated to catching us up.
你们就笑吧。如果有一件事我可以肯定的话，那就是中国人致力于赶上我们。
Semiconductors are the closest thing we have to wizards and magic. We literally draw lines on glass using light to channel electricity that powers our entire world's e-commerce system and controls all the things that allows our civilization to thrive.
China lacks access to the bleeding edge magic and has to brute force its way through extremely elaborate and inefficient chants. The fact that they got to where x86 was from 6 years ago is impressive and scary that they're picking up pace and improving. Over the next decade, they're going to reach near parity or parity; and no amount of sanctions and tech access denial will make a difference.
They're a super power after all. If you can't match your rivals eventually, you can't be called a super power.
半导体是我们最接近魔法师和魔法的东西。我们用光在玻璃上画线，输送电力，为我们整个世界的电子商务系统提供动力，并控制着所有能让我们的文明繁荣昌盛的事物。
中国无法获得最先进的魔法，只能通过极其复杂和低效的吟唱来强行实现。事实上，他们从 6 年前的 x86 发展到现在的水平，令人印象深刻，可怕的是他们正在加快步伐，不断进步。在接下来的十年里，他们将达到接近或持平的水平；无论怎样的制裁和拒绝技术准入都不会有任何改变。
他们毕竟是超级大国。如果你最终无法与对手匹敌，你就不能被称为超级大国。