From the point of view of support for SSE extensions, Intel and AMD go, albeit in different, but sometimes overlapping paths. For example, the Phenom and Barcelona processors offer support for SSE4a instructions, which are inferior in instruction set to SSE4.1 supported by Intel Penryn generation processors. Nehalem processors will offer support for SSE4.2, and AMD is going to implement SSE5 support in its Bulldozer generation processors, which will be released no earlier than 2010. Intel is ready to respond to SSE5 with a set of AVX instructions, which will be supported by Sandy Bridge generation processors.
By the way, in this context, we can also mention the SSSE3 extensions, which are supported by 65 nm Intel processors of the Conroe generation. AMD ignored them, but the situation could change if you believe the document published recently on the official AMD website. In the new revision of the CPUID specifications, AMD admits that its processors may receive support for SSSE3 (not to be confused with SSE3):
We would venture to assume that SSSE3 support will appear in AMD processors of the Shanghai generation. Along the way, AMD announces support for SSE5, which will be implemented no earlier than 2010, according to preliminary estimates.
Finally, AMD is ready to take on the SSE4.1 extensions used by Intel 45nm processors:
In theory, this will increase the speed of AMD processors with video and 3D graphics. Obviously, support for SSE4.1 will also appear in 45 nm processors of the Shanghai generation. As always, the practical benefits of using all of these extensions will depend on software optimization. In this sense, the software industry lags somewhat behind processor manufacturers.
Admittedly, this document is a programmer's guide and does not directly guarantee that the listed extensions will appear in AMD processors. Nevertheless, the company's interest in them suggests that some new instruction sets will be added to the next generations of AMD processors.
A few months ago, AMD introduced a new architecture that will be used in new processors starting in 2011. The new architecture is named Bulldozer and is completely different from the current AMD64 architecture that AMD has been using since 2003.
The Bulldozer architecture will inherit some of the technology solutions introduced with the AMD64 architecture, such as an integrated memory and bus controller HyperTransport for communication between the processor and the chipset.
Bulldozer Is an architecture codename, not a specific processor name. As is usually the case, the first release of processors will be focused on the server market, then a release for the market for expensive high-performance computers, then for the mid-price segment, and at the end for the budget-level market.
Although AMD did not disclose the specifications of the new processors, they noted that the first desktop processors will be based on the new socket. AM3 + which will be compatible with existing AM3 socket. However, Socket AM3 + will not be compatible with Socket AM3 motherboards.
The Bulldozer architecture will feature Intel's Turbo Boost-like technology, which automatically overclocks the processor.
Before talking about the internal architecture of Bulldozer, let's look at the set of instructions supported by the new architecture.
The Bulldozer architecture, in addition to being compatible with the x86 instruction standard, will support the following additional instruction sets:
- SSE4.1 and SSE4.2
- AVX (Advanced Vector Extensions) with two additional instructions XOP and FMA4
- AES (Advanced Encryption Standard) - advanced encryption standard
- LWP (Light Weight Profiling)
SSE4.1 and SSE4.2
Finally, AMD processors will support the SSE4 instruction set. AMD processors do not currently support this instruction set, which improves performance in multimedia applications (such as image and video processing applications). Currently, AMD processors support their own instruction set called SSE4a, which is not the same as SSE4.
AVX (Advanced Vector Extensions)
At one time, AMD offered to use the new SSE5 instruction set. That is why Intel decided to create its own implementation of what was called SSE5 and called this instruction - AVX (Advanced Vector Extensions). AMD decided to add this instruction set for the Bulldozer architecture.
AVX instructions will also be supported by Intel's new Sandy Bridge processors.
The AVX instruction set adds 12 new instructions and increases the size of the XMM registers from 128 bits to 256 bits.
In the Bulldozer architecture, AMD decided to use some of the instructions that were proposed for SSE5. Thus, the use of AVX in the Bulldozer architecture is more complete than that of Intel. These additional instructions are called XOP and FMA4. AMD also noted that AVX has a subset of FMAC (Fused Multiply Accumulate) commands, but in fact, it is part of the XOP instruction set.
AES (Advanced Encryption Standard)
This instruction set is already used in new Intel processors based on the "Westmere" architecture (except for Core i3), and consists of six new instructions related to encryption. Intel calls this instruction set AES-NI.
LWP (Light Weight Profiling)
LWP instructions can improve the performance of multi-threaded software running on multi-core processors. The LWP includes six new instructions.
Often modern software or games require the processor to have SSE 4.1 - 4.2 instructions. If there are none, the required application cannot be launched, some error crashes, or nothing happens.
FarCry 5 swears at the lack of SSE 4.2
At the same time, the processor power may be quite enough for a more or less comfortable game (for example, some Xeon processors for the 775 socket are still capable of producing tolerable FPS in new products), and the requirement for instructions is sometimes necessary not even for the game itself, but for work copy protection. For example, Denuvo's protection did not allow owners of older processors to play Assassin's Creed Origins, although the game itself did not require the latest instructions.
Other popular games or their components also require SSE 4.1 or 4.2: No Man Sky, Far Cry 5, Dishonored 2, Mafia 3 and others.
Nevertheless, there is a solution, although not 100% guaranteeing success. To run the desired application, you can use the emulator sde external, which can be downloaded from the link (choose the version for windows) or at the bottom of this article.
How to use the SSE 4.1-4.2 emulator
- Download the archive with sde external and unpack it so that sde.exe is in the folder with the desired game or program
- Create a shortcut for sde.exe. Then we open the properties of the shortcut and add the object in the parameter - the required file.exe... For example: D: \ Games \ No Man "s Sky \ Binaries \ sde.exe" - NMS.exe. There must be a space after the last quotation mark, otherwise the system will not allow saving the shortcut.
- Also, in the properties of the shortcut on the "Compatibility" tab, you should put the option "run as administrator".
- Save the shortcut and run it. A black window appears, you can close it. After a while, the required application should start.
SSE(Streaming SIMD Extensions) – set of instructions and SIMD extension processor (Single Instruction, Multiple Data, one instruction - lots of data) first introduced in processors Pentium III (Katmai), in response to instructions issued a year earlier 3DNow! from AMD... Initially, the set was called KNI(Katmai New Instructions), but later acquired its current name. The technology was first introduced into mass production in 1999 year, with the release PIII (Katmai).
Instructions are applied when you need to perform the same actions on different data. Thanks to SSE a very large performance gain is obtained due to the parallelization of computational data streams (4 calculations per clock cycle).
Unlike MMX and 3DNow! Instructions, SSE can simultaneously use its instructions together with coprocessor instructions, since the registers SSE and coprocessors are not shared. This is another indisputable advantage, however, as well as the ability to work with non-integers ( MMX could only work with integers).
Usage SSE only possible if the operating system supports these instructions.
SSEconsists of:
SIMD the extension adds eight (or 16 if the processor 64 -bit) 128 -bit registers XMM(XMM 0-XMM 7) divided by four 32 -bit (single precision) floating point register. Also, the line-up includes 32 -bit (in x 86-64 – 64 -bit) control and monitoring register, which is also needed to check the status SSE instructions.
Since each register has 128 -bit, can store everything 4 from 32 -bit floating point numbers (1 — bit sign, 8-bit exhibitors, 23 -x bit mantissa). In fact, in one clock cycle, any data that fits into 128 bits (letters, integers, bytes, double precision, etc.) and it all depends on program optimization for a specific task.
SSE
SSE can perform 70 new instructions that work on these 128 bits, MMX and even ordinary 32 -bit registers if SIMD idle idle.
SSE 2
SSE 2 instructions, appeared with the release of the first Pentium 4 and significantly improved SSE first generation. Includes the ability to use operations on scalar and packed data types. There are instructions for streaming integer processing MMX data in 128 -bit XMM registers SSE... New set of instructions greatly simplified the work programmers and even a saying was born - “ SSE should have been born like SSE 2».
SSE 3
Also had the name PNI (Prescott New Instructions). Added additional 13 instructions that expand mathematical capabilities DSP and flow management instructions. More like cosmetic improvement than the practical one that was the previous update.
SSSE 3
Additional update for SSE 3 which includes 16 new instructions... Lets multiply 16 - these bit numbers with a fixed point and correct rounding, enter words into instructions and rearrange bytes in them.
Added 54 instructions but with reservations. V SSE 4.1(appeared in Intel Penryn) contains a truncated set of 47 instructions and in SSE 4.2- full, from 54 (+7 added in architecture Nehalem).
Instructions only work with 128 -bit XMM 0-XMM 15... Added POPCNT and additional integer instructions & etc. SSE 4 brought almost as many innovations as the first SSE... The main direction in the development was the expansion of multimedia capabilities - decoding games, accelerating physical and vector calculations for games, accelerating the processing of photos, music, etc.
Hello everyone. Let's talk today about how to find out which SSE instructions the processor supports. But what is SSE you know? I don’t know, and it’s not that I don’t know, I can’t even understand what it is. Well, that is, I understand that this is a processor instruction that is needed to optimize its operation, that is, so that with the same frequency of percentages with this instruction it can process more commands. But this is so, roughly speaking so to speak ...
About SSE I don’t know at all where in life it is needed, maybe for games? I know what Hyper-threading is (though it's not a processor instruction, it's a technology), what VT-x, VT-d is, what EM64T is, I don't know what SSE is! Well, these are the pies guys
In short, guys, I’ll tell you right away that there is a small scrappy with this case, I mean that with standard Windows tools such a thing as SSE cannot be found out whether it is or not. Here you need to download a special program. But don't worry, this super duper program is free, it weighs very little, the computer does not load at all, but at the same time it is MEGA USEFUL and its name is CPU-Z (by the way, you can download it here: cpuid.com/softwares/cpu-z.html , this is the official site).
So guys, downloaded CPU-Z, installed it and then run it. And right away you will find out how many of these SSEs I have:
Not one or two, but as many as six, dear fellows
By the way, as you can see there is still a lot of useful information, see? If you urgently need to know something about your process, then you quickly launch CPU-Z and oops, everything you need is at your fingertips! I say that the CPU-Z program is one good thing! Don't believe me? Well, no problem, I'll prove it to you right now. Look, do you know when this or that memory bar was released? Well, that is, the date of its release at the plant, so to speak. Or is it not interesting to you? Well, some people are very interested, for example, I am very interested! And now the CPU-Z program can show such information! So guys, look, you launched CPU-Z, go to the SPD tab, select the slot with the bar (on the left), that is, the connector where it is installed and look at the information on the selected bar. I have one bar for 8 gigs in the fourth slot and this is what the CPU-Z program showed:
Here you can see that my plank was released in the 30th week of 2014. It is also written that my manufacturer is Hyundai Electronics, well, this is the name of the Hynix bar
Well, in short, CPU-Z is super, if you need to quickly look at the most important information about the hardware of a computer or laptop, it will show all this without jokes! In short, I recommend guys!
And yet, I forgot to write something about SSE. You cannot enable or disable SSE. For this instruction is either there or not. For example, Hyper-threading can be enabled / disabled, but SSE is not!
That's all guys, I hope that everything here was clear to you, and if something is wrong, then I apologize. Was this info useful to you, honestly? I hope with all my heart that yes! Good luck in life, so that you are healthy and not sick, good luck
09.12.2016
