汉英对照:
Chinese-English Translation:
Teraflop即Teraflops（简称TFLOP）用于测量CPU的浮点单元FPU计算性能。含义为每秒执行的浮点指令个数， 浮点运算次数/秒，是衡量计算机性能的一个重要单位。
Teraflop (tflops) is used to measure the performance of floating-point unit FPU of CPU. It means the number of floating-point instructions executed per second, and the number of floating-point operations per second. It is an important unit to measure the performance of a computer.
Teraflop 即 Teraflops（简称 TFLOP）用于测量 CPU 的浮点单元（FPU）计算性能。teraflop 等于 1000 gigaflops，或者 1 万亿 FLOPS，这意味着每秒 1 万亿浮点运算。
Teraflop (tflops) is used to measure the CPU’s floating point unit (FPU) computing performance. Teraflop equals 1000 gigaflops, or 1 trillion flops, which means 1 trillion floating-point operations per second.
含义为每秒执行的浮点指令个数， 浮点运算次数/秒。所谓 teraflop，是衡量计算机性能的一个重要单位，可表示为： 1.每秒 1 万亿次浮点运算 2.每秒 10 的 12 次幂的浮点运算 3.每秒 2 的 40 次幂的浮点运算。
It means the number of floating-point instructions executed per second, and the number of floating-point operations per second. Teraflop is an important unit to measure the performance of a computer. It can be expressed as: 1. 1 trillion floating-point operations per second; 2. Floating-point operations of 12 powers of 10 per second; 3. Floating-point operations of 40 powers of 2 per second.
什么是 TFLOP?
What is tflop?
与测量处理器时钟速度的千兆赫(GHz)不同，TFLOP 是对计算机性能的直接数学测量。
Terahertz (TFP) is a direct measurement of the performance of the processor’s lOGHz clock.
具体来说，teraflop 指的是处理器每秒计算 1 万亿浮点运算的能力。例如，说某个东西有“6 TFLOPS”，就意味着它的处理器设置平均每秒能处理 6 万亿次浮点运算。
Specifically, teraflop refers to the processor’s ability to compute 1 trillion floating-point operations per second. For example, to say something has “6 tflops” means that its processor setting can handle an average of 6 trillion floating-point operations per second.
微软对其 Xbox 系列 X 定制处理器的评级为 12 TFLOPs，这意味着主机每秒可以执行 12 万亿次浮点运算。相比之下，苹果 16 英寸 MacBook Pro 内置的 AMD Radeon Pro GPU 的运算能力最高可达每秒 4 万亿次，而重新设计的 Mac Pro 的运算能力最高可达每秒 56 万亿次。
Microsoft rated its Xbox family x custom processor at 12 tflops, which means that the host can perform 12 trillion floating-point operations per second. In contrast, Apple’s 16 inch MacBook Pro has a GPU of up to 4 trillion operations per second, while the redesigned Mac Pro has up to 56 trillion operations per second.
TFLOPs 对游戏有影响吗?
Does tflops affect the game?
微软最近透露了其 Xbox 系列 X 的细节，声称其图形处理器的性能可以达到 12 万亿次浮点运算。这是 Xbox One X 6 万亿次浮点运算的两倍!该公司称这是“处理和图形领域的一次真正的跨越”。处理器速度并不是游戏性能的全部(例如，看看 PlayStation 5 在存储创新方面的表现)。尽管如此，它仍然是决定游戏玩得如何以及在任何给定时间内可以进行多少图形和动作计算的核心因素。
Microsoft recently revealed the details of its Xbox family x, claiming that its graphics processor can achieve 12 trillion floating-point operations. Twice as many as 6 trillion floating-point operations on the Xbox one X! The company calls it “a real leap forward in processing and graphics.”. Processor speed is not all about game performance (for example, look at Playstation 5’s performance in storage innovation). Nevertheless, it is still a central factor in determining how well the game is played and how much graphics and motion calculations can be performed at any given time.
所有增加的能量将使更高的帧率和分辨率支持，以及硬件加速光线追踪。这一性能将进一步增强微软实现的自定义算法的可变率阴影(VRS)，它根据焦点的位置以不同的细节呈现一个场景。这有助于在最需要的地方最大化性能，使游戏看起来很棒而不需要使用系统的所有资源。
All the added energy will enable higher frame rate and resolution support, as well as hardware accelerated ray tracing. This performance will further enhance the variable rate shadows (VRS) of the custom algorithm implemented by Microsoft, which presents a scene in different details according to the location of the focus. This helps maximize performance where it’s most needed, making the game look great without having to use all the resources of the system.
什么是浮点运算?
What is floating point operation?
浮点运算是衡量计算机计算能力的一种常用方法。事实上，一旦我们开始使用 FLOPs，它很快就成为讨论计算机能力的国际通用标准。
Floating point operation is a common method to measure the computing power of a computer. In fact, once we started using flops, it soon became the international standard for discussing computer capabilities.
浮点数，或“实数”，是所有数字的集合，包括整数、有小数点的数字、无理数(如 pi)等等。从计算的角度来看，浮点计算是任何使用浮点数(特别是小数)的有限计算。这比观察定点计算(只使用整个整数)要有用得多，因为计算机所做的工作通常涉及有限浮点数及其所有现实世界的复杂性。
Floating point numbers, or “real numbers,” are the set of all numbers, including integers, numbers with decimal points, irrational numbers (such as PI), and so on. From a computational point of view, floating-point computing is any finite computation that uses floating-point numbers, especially decimals. This is much more useful than looking at fixed-point calculations (using only whole integers), because the work done by computers usually involves finite floating-point numbers and all the complexity of the real world.
FLOPS 衡量的是一个处理器在一秒钟内能解决多少个涉及浮点数的方程。在各种设备需要的失败中有很多差异。例如，传统计算器的所有操作可能只需要大约 10 次失败。所以，当我们开始谈论兆次浮点运算(一百万个浮点运算)、千兆次浮点运算(十亿)和万亿次浮点运算(一万亿)时，你就可以开始了解我们所说的运算能力了。
Flops measures how many floating-point equations a processor can solve in a second. There are many differences in the failures required by various devices. For example, all operations on a traditional calculator may fail only about 10 times. So when we start talking about trillion floating-point operations (one million floating-point operations), Giga floating-point operations (one billion), and one trillion floating-point operations (one trillion), you can begin to understand what we mean by computing power.
制造商经常把失败作为计算机的规格来说明它们的速度有多快。然而，如果你有一个定制的机器，并且真的想炫耀它的万亿次浮点运算。
Manufacturers often use failure as the specification of computers to show how fast they are. However, if you have a custom machine and really want to show off its trillions of floating-point operations.
所以，更多的 TFLOPs 就意味着更快的设备和更好的图形?
So, more tflops means faster devices and better graphics?
虽然这一假设在某些情况下是正确的，但看到具有较高万亿次浮点运算的 gpu 表现出较低性能的情况并不少见。虽然这看起来很奇怪，但这种现象与我们看到的瓦特数非常相似。你的最终表现取决于多种因素。
Although this assumption is true in some cases, it is not uncommon to see GPUs with higher trillions of floating-point operations show lower performance. Strange as it may seem, this phenomenon is very similar to the number of Watts we see. Your final performance depends on many factors.
以聚光灯为例，除了耗电量，你的表现还取决于亮度和结构等因素。与具有相同功率的聚光灯相比，这些额外的因素可能会使聚光灯看起来不同。万亿次浮点运算遵循同样的规则——你需要考虑核心速度、帧缓冲和处理器。
Take the spotlight, for example. In addition to power consumption, your performance depends on factors such as brightness and structure. These additional factors can make a spotlight look different than a spotlight with the same power. Trillions of floating-point operations follow the same rules – you need to think about core speed, frame buffers, and processors.
但是，作为一个指导原则，更多的 TFLOPS 应该意味着更快的设备和更好的图形。这是一个令人印象深刻的增长迹象。就在几年前，消费类设备甚至还不能达到 TFLOPs 的水平，而现在，我们可以毫不犹豫地谈论设备有 6 到 11 次 TFLOPs。在超级计算机的世界里，这更令人印象深刻。
However, as a guideline, more tflops should mean faster devices and better graphics. This is an impressive sign of growth. Just a few years ago, consumer devices didn’t even reach tflops. Now, we can talk about devices six to eleven times. In the world of supercomputers, this is even more impressive.
研究人员正在讨论超过 100 千万亿次浮点运算的超级计算机，而 1 千万亿次浮点运算就是 1 千万亿次浮点运算。目前，性能记录由 ibm 在橡树岭国家实验室建造的一台超级计算机保持，该计算机每秒运算能力为 200 千兆次。
Researchers are talking about supercomputers with more than 10 billion floating-point operations, and 10 billion floating-point operations are 100 billion floating-point operations. At present, the performance record is maintained by a supercomputer built by IBM at Oak Ridge National Laboratory, which has a computing power of 200 gigabits per second.
