Pc Hardware in a Nutshell Douban

Select Your Region

The browser version you are using is not recommended for this site.
Please consider upgrading to the latest version of your browser past clicking one of the following links.

• Safari
• Chrome
• Edge
• Firefox

Retentiveness Performance in a Nutshell

1. Performance Comeback Opportunities with NUMA Hardware
2. NUMA Hardware Target Audience
3. Modern Memory Subsystem Benefits for Database Codes, Linear Algebra Codes, Large Information, and Enterprise Storage
4. Retentivity Performance in a Nutshell
5. Information Persistence in a Nutshell
6. Hardware and Software Approach for Using NUMA Systems

In that location is little conceptual difference between storing data in a reckoner and storing things in a home.

Important Characteristics of Movable Objects

• Space: A kitchen shelf holds less than a basement.
• Placement: Some things are on mitt, while others are deeply buried on a back shelf in the distant corner of the basement. Cooking implements are stored in the kitchen, near where they are used; the backyard mower is stored in the garden shed, because it is used near there. Christmas tree lights spend most of the year on the shelf in the basement but are moved to the living room merely in time, just to render to the basement a few weeks later.
• Latency: The dishwasher takes just equally long to launder ane plate every bit it does a full load, so we fill information technology before turning it on unless we need something in it correct now.
• Bandwidth: We carry laundry in a handbasket because we can bear more during i trip.
• Density: Balloons can be stored and transported inflated or deflated, and – unless they are filled with helium which is not available at the destination – nosotros unremarkably buy, store, and move packages of deflated balloons, inflating them near the play location.
• Removal vs. storage: Sometimes you lot need to determine whether to throw something abroad and make a new 1 when needed, or shop it in the attic where information technology takes up space and hinders access to other things.

Whether you are getting the guest room ready or running the calculator application predicting tomorrow's weather, yous are trying to become something done earlier you need it done, at as low a cost every bit possible, given the circumstances.

You lot try to minimize the space required to shop things, the fourth dimension moving things, the idle time waiting for things to go far, and the difficulties involved in using things in one case they arrive.

Important Characteristics of Computer Data and Storage

Space and Placement

A typical modern computer can shop a few hundred bytes of information in registers within each core of the processor. The information in the registers may be explicitly moved to or from more distant devices, or may be created by the core and lost when overwritten. The movement starts with reading or writing a register from/to an L1 (Level one) cache. Each cadre commonly has a individual L1 cache that can contain tens of thousands of bytes (32 KB is a typical size).

Other private and shared caches are usually located on the path betwixt the L1 cache and main retentivity (although non-temporal loads and stores can featherbed them). These caches range in size from the private 256-KB caches and many-MB shared caches in processors, to the many-GB caches stored in Multi-channel DRAM (MCDRAM) and High-Bandwidth Retention (HBM) memories or in dual inline-memory modules (DIMMs).

The MCDRAM and HBM memories of Intel® Xeon Phi™ processors can exist used as caches for more distant DIMMs, and these caches contain on the order of 16 to 60 GB of information.

The principal memory of a personal computer or server tends to be in the 4-GB to 1500-GB range.

Futurity 3D XPoint™ DIMMs may make it practical for chief memory to hold terabytes – 6 TB (6000 GB) is predicted. 3D XPoint DIMMs volition probably have a slower bandwidth than double information rate (DDR) DIMMs, perhaps with their contents cached in MCDRAM, HBM retention to recoup for this. Such DDR DIMM caches could be about 10% of the capacity of the main memory, so these caches can exist 600 GB in size – a far cry from the 4-KB master memory on the machines from the early 1970s.

The fundamental deviation between caches and chief retentivity is:

• Caches await somewhere else for data they practise non hold.
• Main memory is the end of the line – if asked for data it does not have, it does not expect elsewhere for it.

Data can survive in registers, caches, and master memory while the figurer has power.

Across the master retentivity are devices where information can survive when a reckoner loses power. These devices can usually hold more data than principal retention. Hard drives, solid-state drives (SSDs), and removable media such as DVDs are all examples of such devices for persisting information.

If these devices were faster than main memory, y'all would use them as main memory. You practise not, because the processor cannot access the data on them as quickly as it can from main memory – it is hindered past latency and bandwidth.

3D XPoint technology, and any DIMMs built using information technology, may continue contents across power failures, so you tin employ them as both main memory and as a persistent information shop. 3D XPoint technology may be used in Optane SSDs.

Latency, Bandwidth, and Density

Latency is how long information technology takes from when you commencement a request for data until the data arrives.

It is hard to measure latency in many situations because both the compiler and the hardware reorder many operations, including requests to fetch data. They as well reorder instructions to practice other things while waiting for data to arrive. They may even predict what the fetched data is going to be and deed on that prediction, so the arrival of the data simply makes the results of this prediction visible. Of course, if the prediction is wrong, all that piece of work must be redone. Because of this, latency matters virtually when the compiler and hardware cannot find useful work to do while waiting.

Bandwidth is the rate at which the data arrives, however long that is after it is requested. The usual example involves sending a ship conveying 1 million DVDs across the sea every twenty-four hour period. The latency might be six days, just the bandwidth is tens of GB per second.

As for density, many data transfers occur in 4 bytes, 8 bytes, or even more bytes betwixt the core or memory devices. But not all this data may be used at the destination.

Size, Latency, and Bandwidth of Memory Subsystem Components

Bold you have a large processor (most 16 cores), the following summarizes, for 2016, approximate data totals nowadays in and moving through the system.

Retentiveness	Size	Latency	Bandwidth
L1 cache	32 KB	1 nanosecond	1 TB/second
L2 cache	256 KB	4 nanoseconds	1 TB/second Sometimes shared by two cores
L3 cache	8 MB or more	10x slower than L2	>400 GB/2d
MCDRAM		2x slower than L3	400 GB/second
Main memory on DDR DIMMs	4 GB-1 TB	Similar to MCDRAM	100 GB/second
Main retentiveness on Cornelis* Omni-Path Fabric	Express only past cost	Depends on distance	Depends on distance and hardware
I/O devices on retentivity bus	6 TB	100x-1000x slower than retentiveness	25 GB/second
I/O devices on PCIe bus	Express merely by toll	From less than milliseconds to minutes	GB-TB/hr Depends on altitude and hardware

Summary

The previous article, Modern Retentivity Subsystem Benefits for Database Codes, Linear Algebra Codes, Large Data, and Enterprise Storage, aligned new memory subsystem hardware technologies with the needs of applications. This article provides a deeper agreement of hardware capabilities when used for usual variables and heap allocated information of an awarding – information that did non exist before the application started and evaporates when it ends. The next article, Data Persistence in a Nutshell, introduces using non-volatile retentivity to supplant the utilize of files for keeping data from i execution of an application to the next.

About the Author

Bevin Brett is a Chief Engineer at Intel Corporation, working on tools to help programmers and system users improve application performance. He was born and raised in New Zealand, where he earned a B.Sc (Hons) in Mathematics earlier moving to Australia and and then New Hampshire, pursuing beginning an instruction then a career in software engineering.

Product and Performance Information

Share this post