Computer Memory: A Beginners Guide – Everything You Wanted To Know But Were To Lazy To “Google” There always seem to be a lot of questions about memory so I decided to throw together this guide for beginners. I also included a few links at the end for additional explanations for the curious or technically inclined. Not a "Beginner"? Scroll down the post to: Computer Memory: The Next Step – Adjusting Memory Timings and How Your Memory Effects CPU Overclocking Types of RAM: DDR, DDR2, GDDR3, etc. These are the most common types of RAM you will see listed, or talked about, with DDR being the most common of the three. DDR – This memory type is the current standard when it comes to computer memory. DDR stands for Double Data Rate and gets this name as it transfers data at double the speed of the previous memory standard. DDR has pretty muched reached it's limit though at 400MHz, although some will go faster. DDR2 - Similar in design to DDR and theoretically transfers data twice as fast as DDR memory, up tp 800MHz, and maybe more, hence the “2” in the name. DDR2 technology is improving and Intel has adopted DDR2 memory as its standard and AMD will be introducing it into their lineup in 2006. Once AMD is completely switched over, say good-bye to DDR! GDDR3 – The G stands for graphics and GDDR3 memory is pretty much the standard in video cards. And the DDR3 part should be self-explanatory by now. GDDR4 graphics cards are rumoured to coming out near the end of 2006. Also, as far as the memory interface on your motherboard is concerned, DDR and DDR2 are completely different seeing as they have different pin counts. DDR = 184 pins vs DDR2 = 240 pins Some memory says ECC. What is that? ECC stands for Error Correction Code and is exactly that. It’s memory that has the ability to correct most of the errors that your memory may make. Yes, your memory is imperfect! However, today’s memory makes errors so infrequently that for the average joe like you and me, ECC memory is a waste of money, as it is more expensive. Also, if anything, it will be slower due to the error checking cycle. ECC memory is mostly utilized in servers or for the likes of CAD operations where they are pumping through huge amounts of data and accuracy is extremely important. Buffered and Registered Memory In unbuffered memory, the chipset controller on your motherboard deals directly with the memory. There is nothing between the chipset and the memory as they communicate. Buffered modules contain a buffer to help the chipset cope with the large electrical load required when the system has a lot of memory. Registered modules are unbuffered modules that contain a register that delays all information transferred to the module by one clock cycle and is usually done on modules with a lot of memory to help ensure that the data is properly handled. Like ECC memory, buffered and registered memory is typically used only in servers and other mission-critical systems where it is extremely important that the data is properly handled, and is therefore not for the average user. Memory Speeds You will see RAM speed listed in two ways, examples of each are PC3200 and DDR400. In actuality, PC3200 and DDR400 is RAM running at 400MHz, so they are the same speed. This speed is how fast the memory and the CPU send data back and forth and PC3200 is the current standard. Listed below are some common conversions: PC2100 = 266MHz PC2700 = 333MHz PC3200 = 400MHz PC4000 = 500MHz Also, throw in a “2”, and we’re talking about DDR2 RAM: PC2-3200 A simple way to remember is to take the PC number and divide by 8. Ex: PC2700 / 8 = 337.5 (333MHz) It’s not always exact, but very close. Which Speed Do I Need? Well, that depends on your motherboard. Motherboard manufacturers will list the speeds of RAM that the motherboard will support something like this: DDR 400/333/266 or PC3200/PC2700/PC2100. Obviously to maximize the potential of your system, you want to put in the fastest RAM your motherboard will support. You can put in faster RAM than your motherboard supports but it will only run at the fastest motherboard supported speed. Ex: If your motherboard supports PC3200/PC2700/PC2100 and you put in PC4000, then your PC4000 RAM will only run at PC3200 speeds. Can I Mix and Match RAM? Yes*, but it will only run at the slowest rated speed in your system. For the example, if you have a motherboard that has a top rated speed of PC2700 then with: 1) One stick of PC3200 your memory runs at PC2700 2) One stick of PC2100 your memory runs at PC2100 3) One stick of PC3200 and one stick of PC2100, both sticks will run at PC2100 *You can not mix and match DDR with DDR2. You motherboard will support one or the other, not both. How Much RAM Do I Need? The current standard is now 1GB and should work very well for the most gamers; but for the not-so-rich 512MB is the absolute minimum for any gaming system (but not recommended). However, newer games like BF2 perform better with more RAM and 2GB of RAM is going to be the standard in the not-to-distant future. Dual Channel RAM – The Myth This is currently the most widely misunderstood aspect of RAM. Let me make this perfectly clear: THERE IS NO SUCH THING AS DUAL CHANNEL RAM!!! Dual channel is a mode your memory can run in and is controlled by the memory controller, which is located on your motherboard or CPU, depending upon which system you operate. Dual channel effectively doubles the amount of data your memory and CPU can transfer. IE: It makes a one-lane highway into a two-lane highway. Read the article below to see how dual channel mode affects your computer. But I bought Dual Channel RAM? Yes, all this means is that the RAM has been tested to ensure that it will run in dual channel mode if your system supports it. In order for RAM to be able to run in dual channel mode, you need a pair of identical sticks of RAM set-up properly on your motherboard. One 512MB PC3200 and one 512MB PC2700 = no dual channel possibility One 512MB PC3200 and one 256MB PC3200 = no dual channel possibility One 512MB PC3200 and one 512MB PC3200 = dual channel possibility Two 512MB PC3200 and two 256MB PC2700 = dual channel possibility (and remember, all sticks would run at PC2700 speed in this set-up) Three 512MB PC3200 = no dual channel possibility I say “possibility” because occasionally you can have two identical sticks of RAM that will not work in dual channel mode, due to minor differences in the manufacturing process. These sticks are not damaged; it’s just that they are not compatible to operate in dual channel mode. That is why manufacturers sell “Dual Channel Memory”; these sticks have been tested to ensure that they will run in dual channel mode. Also, motherboards generally come with four slots to put your memory in, and some come with only three. As you now know, you need a pair of identical sticks of RAM in order for the dual channel mode to work. You also need to put them in the correct slots so consult your motherboard manual to ensure you have the sticks set-up in the correct slots. Some want them in Slot 1 & 3, some in Slot 1 & 2, etc. What About Those Other Numbers? - The Timings You often see a set of numbers like 3-3-3-6 attached to the memory description. These numbers represent the timings that regulate the speed of data transfer within the memory module itself. The faster the memory can process the data internally, the faster it can get it back to the CPU. I won’t go into much detail here as these numbers are generally for more advanced users and overclockers. However, for the beginner, the first number is the one you want to pay attention to. It is the CAS (Column Address Strobe) or latency number and is the primary controller in the module. You want this number to be as low as possible, with 2, 2.5, and 3 being the choices. Of course CAS 2 RAM will be more expensive than CAS 3 RAM, since lower is faster. I think that’s it! I hope that this proves to be useful and that I didn’t make too many mistakes. (I’m sure you’ll let me know if I did!) Also, for those interested in overclocking and how your RAM can affect it, read the first article below. Enjoy! How memory timings (CAS, etc) can effect FPS Memory Timings Explained DDR vs DDR2 Dual Channel vs Single Channel Note: Dual Channel Kicks Butt In HL2!!!! Making Dual Channel Work Thanks Danny M Computer Memory: The Next Step – Adjusting Memory Timings and How Your Memory Effects CPU Overclocking First, let’s get the old disclaimer out of the way: Overclocking any of your computer components may void the warranty of the components, as well as damaging your computer and/or the components. OVERCLOCKING -> DO SO AT YOUR OWN RISK! Memory Timings: 2.5-3-3-7 (as an example) So what do these numbers mean? Well, this article (listed above also) explains it well enough, but it still confuses me so I won’t go there. In practical terms though the lower (tighter) the numbers, the faster your computer will be. Conversely, higher (looser) numbers will result in a comparatively slower computer. Any stick of RAM will come with pre-set timings, the SPD (Serial Presence Detect) numbers, which your computer will identify and implement. However, most computer systems have the ability to change these numbers. The quest of course is to get these numbers as low as possible and still have your computer work, and work well. Yes, if you set them too low, the memory may start making errors or worse yet, your computer may not even start at all. Changing The Timings (or, One Way Of Overclocking Your Memory) To do this you need to go into your computer’s BIOS and if you are reading this I will assume you know how to do that. In the BIOS there may be a segment that allows you to change the memory timings. I say “may” since some BIOS’s will not have this feature at all, while other will “hide” it and have you press “Ctrl” + “F1” (for example) to enable this feature. You will need to consult your motherboard manual to see whether it is supported or not, and to see how to access it. Once you have found the appropriate segment, changing the numbers is simple. Just type in, or select, a new number. But…. Which Timings Should I Choose? This is the fun part and is simply done by trial an error. Say your SPD numbers are 2.5-3-3-7, you could try 2.5-3-3-6, or 2-3-3-7, or 2.5-3-2-7, etc, etc. Just change one number, save your BIOS, and then reboot your system. Now one of two things will happen. One, your system will boot up normally. If this happens you can skip down to “Testing Your New Timings.” Two, your system will NOT boot up, which is usually accompanied by a lot of beeping! If this happens, don’t freak out, just turn off your computer. But now you have a problem, don’t you? How can you change those timings back to ones that will work if your computer won’t start? Well, your motherboard has something called the CMOS, which you can use to reset your BIOS back to the original settings. You will need to consult your motherboard manual to see how this is done but it is very simple and usually just involves removing a battery for a few minutes or resetting a jumper. Once this is done your BIOS will be back to the original settings. Now that your computer will start, you can do one of two things. Either you can increase the voltage going to your memory, or try with some different numbers. Notice that I said “different” and not “higher” or “looser”. I have experienced, and also read, where some timing combinations would just not work with a certain set-up. For example, 2.5-3-2-6 would not work, but the tighter timings of 2.5-2-2-6 did work. Keep this in mind also when you are Testing Your New Timings (see below). As for voltage, your BIOS may also give you the option to increase the voltage going to your memory. Increasing the voltage by one or more steps may stabilize the new timings, or it may not. As I said before, trial and error! However, increasing the voltage has the side effect of producing more heat, which is where Heat Spreaders come into play. Heat spreaders are metal (aluminum or copper usually) fins that fit over the RAM to help dissipate the heat. There is debate as to whether enough heat is produced to make heat spreaders necessary but two things are certain, they can’t hurt and they look way better than a plain stick of RAM! Testing Your New Timings Now that your computer has new timings and has successfully rebooted, you need to test your system and make sure it is stable. This is done by running programs that stress your memory and check to see whether it makes any errors. Two tests you want to run are MemTest86 and Prime95. Download these programs from their websites and follow the instructions. Now I’ve read articles saying you need to run these tests for 30 minutes and other articles have said 24 hours. Myself, I let them run overnight or when I’m at work. How much time you run them for is up to you but one thing you can’t control is the test of time. Even if your new timings pass these two tests your computer may lock up five minutes into your next session of CS:S, or it might lock up in a week. If your new timings do fail, then it’s back to picking new timings. Obviously it’s a good idea to keep track of the combinations you have tried so as not to duplicate your work. The bottom line: Finding stable timings for your memory takes time and patience. It can take days, weeks, or even months, depending on the free time you dedicate to the task. Enjoy yourself and have fun doing it. So How Much Faster Is My Computer With These New Timings? You can run benchmark programs such as SuperPi, 3DMark05 or 3DMark06, and Aquamark3 to test your system performance. See the Sticky by T Rush to read all about benchmarks. How Memory Effects CPU Overclocking (and How Overclocking Your CPU May Result in Overclocking Your RAM) As I mentioned back near the very beginning of this post, the most common speed for RAM currently is 400MHz, and this is the speed with which it communicates with the CPU. This communication takes place via the FSB (Front Side Bus) and the most efficient communication takes place when the memory and FSB are at the same speed. Hence, the most common FSB speed currently is 400MHz. However, one of the most common ways to increase (overclock) your CPU is to increase the speed of the FSB. The problem with this is that it also overclocks your memory by running it faster than it was meant to, and can result in instability. If your system gets unstable with a higher FSB you either need to adjust the FSB speed or you need to start playing with the memory timings again (loosen them most likely) to find the tightest timings your memory can run with at this increased FSB speed. If your memory just will not work at the desired FSB, you can overcome this problem by using a divider. This is another option your motherboard may have and simply, it allows you to run the FSB and your memory at different speeds. This allows your memory to operate within its limits, but at the price of inefficient communication with the CPU. For CPU overclockers though, the resulting increase in CPU speed is more than likely to compensate for the inefficient CPU/memory communication. Actually, there is one another way….buy better RAM! (Unless you already have the best of course.) Like most things in life, you generally get what you pay for. If you have some El-Cheapo brand memory, chances are it won’t go much faster than it’s rated speeds of, for example, 400MHz and 3-4-4-8. But if you have some More Expensive brand memory, odds are it will be able to run faster than it’s rated speeds, and therefore be able to match the FSB speed and maintain the 1:1 ratio that CPU overclocker’s desire. Or at the very least, minimize the divider, or difference the FSB and memory run at. So there you have it (I think). I hope all this makes sense and that once again, the mistakes are few and far between. Have fun tweaking! **Clockspeed does not determine performance.** * Benchmarks Look here to see how P4s stack up to Athlon64s. Also, read these: Contemporary CPUs and New Games: No Way to Delusions! Games against CPUs. Part II. * General Info The Athlon64 and Pentium4 are both x86-64bit CPUs. They both are 32bit CPUs in a 32bit OS and are both 64bit CPUs in a compatible 64bit OS. A true 64bit CPU (Itanium, PowerPC, etc) require emulators to run 32bit x86 code and take a HUGE performance hit because of this. Anyways, Moral of the story: A new Pentium4 will NOT spank any Athlon64 (both x86-64 CPUs). The Athlon64 is a more effecient CPU. It can do more clock-per-clock than P4. Since the Athlon64 only has like a 13 stage pipeline, a prediction error is no problem. The Pentium4 was designed strictly for gigahertz. It is not effecient, produces lots of heat, and is overall poorly designed. The Pentium4 Prescott has something like a 34stage pipeline. This long pipeline allows the P4 to be clocked high allowing for good performance under the right conditions. If there is a prediction error, the whole pipeline has to be flushed and it takes a huge performance hit. The Athlon64 is a superior design, period. The Netburst arch. of the Pentium4 was designed by the marketing department of Intel and not by the chip architects. The Pentium M is a much better design. The Athlon64 and Pentium4 are both very good general purpose CPUs. Your average person can not go wrong by picking either of them. However, there are some areas that the P4 performs stronger in and there are some areas that the A64 performs stonger in. * Hyper-threading and Dual Core Hyper Threading is a way of exploiting the Netbursts long pipeline to allow it to do two threads at once. HT makes the CPU appear as two logical CPUs. Dual Core is when there are actually two physical cores or CPUs on the same die. Dual core is better than HT. Both AMD and Intel offer dual core products. Hyper-Threading Technology Quote: Hyper-Threading Technology enables multi-threaded software applications to execute threads in parallel. This level of threading technology has never been seen before in a general-purpose microprocessor. Internet, e-Business, and enterprise software applications continue to put higher demands on processors. To improve performance in the past, threading was enabled in the software by splitting instructions into multiple streams so that multiple processors could act upon them. Today with Hyper-Threading Technology, processor-level threading can be utilized which offers more efficient use of processor resources for greater parallelism and improved performance on today's multi-threaded software. Hyper-Threading Technology provides thread-level-parallelism (TLP) on each processor resulting in increased utilization of processor execution resources. As a result, resource utilization yields higher processing throughput. Hyper-Threading Technology is a form of simultaneous multi-threading technology (SMT) where multiple threads of software applications can be run simultaneously on one processor. This is achieved by duplicating the architectural state on each processor, while sharing one set of processor execution resources. Hyper-Threading Technology also delivers faster response times for multi-tasking workload environments. By allowing the processor to use on-die resources that would otherwise have been idle, Hyper-Threading Technology provides a performance boost on multi-threading and multi-tasking operations for the Intel NetBurst® microarchitecture. This technology is largely invisible to the platform. In fact, many applications are already multi-threaded and will automatically benefit from this technology. However, multi-threaded applications take full advantage of the increased performance that Hyper-Threading Technology has to offer, allowing users will see immediate performance gains when multitasking. Today's multi-processing aware software is also compatible with Hyper-Threading Technology enabled platforms, but further performance gains can be realized by specifically tuning software for Hyper-Threading Technology. This technology complements traditional multi-processing by providing additional headroom for future software optimizations and business growth. More Info: Excellent article, HT considered harmful, here, here * AMD's HT - HyperTransport Technology HyperTransport Technology Quote: * HyperTransport technology is a high-speed, low latency, point-to-point link designed to increase the communication speed between integrated circuits in computers, servers, embedded systems, and networking and telecommunications equipment up to 48 times faster than some existing technologies. * HyperTransport technology helps reduce the number of buses in a system, which can reduce system bottlenecks and enable today's faster microprocessors to use system memory more efficiently in high-end multiprocessor systems. * HyperTransport technology is designed to: o Provide significantly more bandwidth than current technologies o Use low-latency responses and low pin counts o Maintain compatibility with legacy PC buses while being extensible to new SNA (Systems Network Architecture) buses. o Appear transparent to operating systems and offer little impact on peripheral drivers. * HyperTransport technology was invented at AMD with contributions from industry partners and is managed and licensed by the HyperTransport Technology Consortium, a Texas non-profit corporation. The full specification and more information about HyperTransport technology can be found at HyperTransport.org. HyperTransport is a licensed trademark of the HyperTransport Technology Consortium. More Info: Here, Here, * The Truth Here is the real story on the P4: Quote: We are proud to offer you the most detailed article on NetBurst architecture and its peculiarities. We applied our own approaches and methodology to studying the NetBurst architecture and we are ready to share some facts and details, no one has ever discussed before. This is the first article of the trilogy: more exciting stuff is coming within the next couple of days with absolutely exclusive material about what actually slows down Pentium 4 processors. You will not believe your own eyes when you see this! So, stay tuned and don’t miss the three exciting parts of X-bit’s indepth NetBurst Architecture investigation! Prescott: The Last of the Mohicans? (Pentium 4: from Willamette to Prescott) Prescott: The Last of the Mohicans? (Pentium 4: from Willamette to Prescott). Part II What You Should Know About the Past, Present, and Future Of Computers “A device that shuffles numbers around from place to place, reading, writing, erasing, and rewriting different numbers in different locations according to a set of inputs, a fixed set of rules for processing those inputs, and the prior history of all the inputs that the computer has seen since it was last reset, until a predefined set of criteria are met which cause the device to halt.” Wow, that sounds really boring; however, this is the definition of the computer according to Jon Stokes in his article “Understanding the Microprocessor.” One of the most important inventions of all time, the computer is a device that people just can’t live without this day and age. Despite the vast importance of computers in our lives, your average person know very little about computers and needs to be educate a bit in the matter. The computer has changed our past, evolved to the necessity that it is today, and will stick around to aid us in the future. Computers are comprised of several main pieces of hardware: the microprocessor (CPU), the motherboard (mobo) and chipset, memory, hard drive, video card and any add-in cards. To keep things simple, let’s just focus on the microprocessor. The CPU, or processor, is the brains of the computer. “In terms of computer power, the CPU is the most important element of a computer system” (Webopedia). All modern day processors are merely extension of the x86 architecture invented by Intel in 1978. In the lovely year of 1978, Intel introduced a 4.77-MHz 8086 microprocessor – the first x86 processor (Beard). Intel worked hard developing new variations of their first x86 in hope of hitting it big, which is exactly what happened when Intel created the Pentium processor. The successor of the 486, deemed the Pentium processor, improve slightly on the performance of its predecessor and kept x86 compatibility (Stokes, Pentium). The Pentium processor was introduced on March 22, 1993, at 60 and 66MHz, manufactured on a 0.8-micron (or 800nm) process, containing 3.1 million transistors and 16kb of L1 cache. This processor, by all means, “wasn’t exactly a blockbuster,” especially by today’s standards and was only popular among DOS users for writing word and spreadsheet documents. It was not until the introduction of the P6 architecture that Intel’s processors really caught on (Stokes, Pentium). On November 1, 1995, Intel introduced the P6 architecture in the form of the Pentium Pro. The Pentium Pro performed significantly better than the Pentium and proved to be extremely scalable (Stokes, Pentium). The P6 architecture began as a .60 micron, 150mhz processor and died as a .25 micron, 1400mhz Pentium III processor with almost twice as many transistors (Stokes, Pentium). As all good things come to an end, so did this architecture when it fell off the maps because it could not scale any faster than 1.4ghz – this was not enough to catch up to Intel’s rival AMD (Stokes, Pentium). Today’s processors are just a bit more complex than those based on the P6 architecture. There are only two main competitors in the mainstream CPU industry today, Intel and AMD, and each has its own boxer so send into the ring that is the consumer market. In Intel’s corner, we have the Pentium 4 “Prescott” processor based upon the Netburst architecture; and in AMD’s corner we have the Athlon 64 based on the K8 architecture. When Intel could not get any more life out of their P6 based processors, they created the Netburst architecture designed for sheer megahertz madness. The Netburst architecture’s greatest difference from the P6 is in the amount of stages in the pipeline. “Pipelining,” as Shimpi of AnAnd puts it, “is to a CPU as the assembly line is to a car plant. A CPU’s pipeline is not a physical pipe that data goes into and appears at the end of, instead it is a collection of ‘things to do’ in order to execute instructions. Every instruction must go through the same steps, and we call these steps stages. The more pipeline stages you have, the less work is done per clock and thus the higher you’re able to clock your CPU; this is the reason the 20-stage Xeon is currently at speeds of 3Ghz, compared to the 12-stage [Athlon 64] which is debuting at 1.8Ghz” (Shimpi, Opteron/K8). Even though the P6 had a short, 12-stage pipeline, it is no match for the multi gigahertz P4s of today, despite their lengthy 31-stage pipeline (Stokes, Pentium). Now lets compare the original Pentium Pro to today’s Pentium 4 Prescott. The Pentium Pro, at the peak of its career, ran at 200mhz, a front-side bus (FSB) of 66mhz, was manufactured on a .35micron process, consisted of 5.5million transistors, had a total of 16k L1 cache and 512k L2 cache (Stokes, Pentium). The Pentium 4 Prescott runs at 3600mhz, a quad pumped bus of 800mhz (4x200), is manufactured on the worlds first .09micron process, consists of 125 million transistors, has a total of 28kb L1 cache and 1mb of L2 cache, and features MMX, SSE, SSE2, SSE3, and Hyper-threading (Schmid). Intel’s biggest and only surviving competitor today is Advanced Micro Devices (AMD). Although AMD’s processors used to be quite similar to Intel’s, with the introduction of the K8 architecture AMD finally has something of their own. AMD’s CPUs have a unique instruction set known as AMD64 or x86-64. What this is is a 64bit extension of the x86 platform, allowing users to run both 32- and 64-bit applications at the same time (FastSite). So how does this benefit us today? The most immediate benefit is that it allows the computer to address more than 4gb of RAM; more importantly once programs are written to use AMD64, we will see some huge increases in performance. Just to give people an idea of how much better 64-bits is than 32-bits: “64-bit computers can process numbers that are 4.3 billion times as large as those processed by their 32-bit counterparts” (Stokes, Introduction). That’s quite a difference! Another new technology that AMD has implemented in its newest processors is HyperTransport. HyperTransport is “a high-speed ‘point-to-point’ date transfer bus developed by AMD.” HyperTransport “is used to connect the processor and the chipset, different parts of the chipset developed by AMD for the [Athlon 64], and different processors in multiprocessor systems by means of additional HyperTransport controllers built into the processor” (FastSite). The HyperTransport bus replaces the conventional bus, like that of Intel’s processors, and gives the Athlon 64 a 2000mhz (2x1000) front-side bus. In order to lower memory latencies, AMD moved the memory controller off of the chipset and onto the CPU itself (Shimpi, Opteron). This allows even a Single Channel (less bandwidth) DDR Athlon 64 setup to perform on par with a comparable Pentium 4 system with Dual Channel (more bandwidth) DDR. And probably the greatest difference between the Pentium 4 and the Athlon 64 is the pipeline. Because of its 12-stage pipeline (as well as other factors), a 1.8ghz Athlon 64 can perform better than or equal to a Pentium 4 clocked at 3ghz (Shimpi, Opteron). Although AMD has not yet moved on to a 90nm (.09micron) process yet, they are able to squeeze the same 1MB of L2 cache onto its .13 micron as the P4 (L2 caches take up a majority of the space on a processor). The future of the computer world, although filled with many hurdles, looks very promising. As the processor dies shrink, megahertz goes up and new technologies are created to boost the x86 platform and extend its life just a little longer. However, the x86 platform and CMOS processors are coming to an end. Companies are investing great sums of money into the fields of quantum computer, molecular electronics, and nanotechnology. With these new technologies, computers will be able to do things we never imagined. Computers have become integrated into our everyday lives. Almost every appliance people use today somehow relies on computers to work. Computers have become more than just a personal tool. These incredibly machines allow communication across the whole world, provide entertainment through games, movies, or music, and provides a great medium to conduct business. Without the computer, mankind would struggle to survive. It is because of this that people need to be educated about their investments in computers. *As you all will notice, this was written quite a long time ago therefore some things have changed. Works Cited "CPU." Webopedia. 15 Jan. 2003. <http://www.pcwebopedia.com/TERM/C/CPU.html> (13 July 2004) Beard, Jason. "Computer Chronicles: From Stone to Silicon." 1998. <http://library.thinkquest.org/22522/about.html> (6 July 2004) FastSite. "A Glance at the Future: AMD Hammer Processors and x86-64 Technology." 14 Spet. 2002 <http://www.xbitlabs.com/articles/cpu...r-preview.html> (6 July 2004) Gavrichenkov, Ilya. "Intel Prescott: One More Willamette-like Slow Processor or a Worthy Piece?" 1 Feb. 2004. <http://www.xbitlabs.com/articles/cpu.../prescott.html> (7 July 2004) Schmid, Patrick. "Intel's New Weapon: Pentium 4 Prescott." 1 Feb. 2004. <http://www.tomshardware.com/cpu/20040201/index.html> (7 July 2004) Shimpi, Anand Lal. "Intel's Hyper-Threading Technology: Free Performance?" 14 Jan. 2002. <http://www.anandtech.com/cppuchipset...oc.aspx?i=1576> (6 July 2004) Shimpi, Anand Lal. "Intel's Pentium 4 E: Prescott Arrives with Luggage." 1 Feb. 2004. <http://www.anandtech.com/cpuchipsets...oc.aspx?i=1956> (6 July 2004) Shimpi, Anand Lal. "Part 1: Intro to Opteron/K8 Architecture." AMD Opteron Coverage. 23 April 2003. <http://www.anandtech.com/cpuchipsets...oc.aspx?i=1815> (6 July 2004) Stokes, Jon. "An Introduction to 64-bit Computing and x86-64." <http://arstechnica.com/cpu/03q1/x86-64/x86-64-1.html> (6 July 2004) Stokes, Jon. "Part I: Basic Computing Concepts." Understanding the Microprocessor. <http://arstechnica.com/paedia/c/cpu/part-1/cpu1-1.html> (7 July 2004) Stokes, Jon. "Part I: From the Pentium to the P6." The Pentium: An Architectural History of the World's Most Famous Desktop Processor. <http://arstechnica.com/cpu/004/penti...ntium-1-1.html> (12 July 2004) Stokes, Jon. "Understanding Moore's Law." <http://arstechnica.com/paedia/m/moore/moore-1.html> (7 July 2004) What is a Power Supply Unit (PSU) in a personal computer, what does it do, and how do you know if it is a good or bad PSU? This thread will try to shed some light on this often mystifying topic. First off, a PSU is the heart of your system. It is what pumps the electrons for the CPU, GPU, sound card, fans, hard drives, CD/DVD drives, LEDs, etc. While it has no brains, per se, it is the life blood of your system. It simply converts alternating current (AC) power to direct current power (DC) for your computer. Many people often take the PSU for granted, because it is just there. Or, they spend thousands of dollars on their new rig and buy the cheapest PSU on the web site. Both of these are bad things! Do not buy a cheap PSU if you are spending hard earned money on a computer!. If you are trying to keep your system running at its peak efficiency or trying to make sure that those components will all run with your system or those strange slow downs in apps/games aren’t caused by your PSU, then read on. Ratings and Specifications All manufacturer’s show their Wattage rating for a PSU. Is this the only thing you need to look at? NO! Wattage is only one component of the capabilities of a PSU. Keep in mind that there are both input Wattage ratings and output Wattage ratings (the output wattage rating is what is important for powering your components). The output Wattage rating will also have an ambient (operating room) temperature associated with it. If the ambient temperature goes above this, the rating actually drops by a derating factor based on the quality of the electrical components in the PSU. So where are these ratings shown? If you're looking on the web, hit the manufacturer's web site or if you can't find that, the ratings are always shown on a label that is affixed to the PSU somewhere (side, top, bottom, back?). Just pull your case side panel off and start looking. Label <--- Look at the row that says DC Output Wattage = Voltage x Current (Amperes) There are several different voltages that a PSU must deliver to the motherboard and other components. These voltages are: +5V +12V -12V +3.3V +5VSB Please note that the -5V voltage is no longer needed with the ATX12V specs (emphasis went to the +12V rail). Some PSUs may still have a wire on the main 20/24 pin connector and some may not, but it's not needed with motherboards that support the ATX12V spec. Another rating that should be considered is Power Factor Correction (PFC). This is an indication of how efficiently the PSU can control the phase angle of voltage versus current when converting the power from AC to DC. The PFC is generally provided in a % number. The higher, the better, with 100% being the best. In very simple terms, it helps to reduce the amount of energy needed to “get” the power to the PSU from the electrical ‘grid”. I won’t go into detail here, but all I can say is the higher the PFC, the greener. What about noise? All PSUs need to be cooled to maintain a decent operating temperature. How is this done? With fans! Most good PSUs will have a noise rating, expressed in decibels. The killer PSU link posted below has a noise rating of 32 - 40dB(A). This represents the noise that is generated when the fan is running slow (low temps) up to max fan speed (high temps). The lower the dB(A) number the better if you're concerned about noise. What about protection? Good PSUs will have built-in protection against electrical hickups. These include over-voltage protection and over-current protection. It is important to look at this when evaluating your PSU. This is what ensures your PSU can protect itself and your computer components. Now, onto the good stuff! Besides the above, what else should I look for in a PSU? Currently (no pun intended) the +12V rating is king! PSUs now have single and multiple +12V outputs (rails). If it’s a single rail, it will show the rating as +12V. If it’s a multiple rail, it will show it as +12V1, +12V2, etc. What is king is the current (amp) rating for that +12V rail. The higher the ampere rating the better! 18 Amps or above used to rule, 30+ Amps or above is now desired with the newer video cards and CPUs on the market. Here is an example of what you can expect from one of the highest quality PSUs on the market for voltages, currents on each rail and what is actually being fed power by each rail (rock my socks!). Turbo-Cool 1KW-Quad SLI Technical Specifications +5V @ 30A +12V1 @ 16A (CPU) +12V2 @ 16A (Drives) +12V3 @ 36A (Graphics) +12V1,V2,V3 = 66A (70A peak) -12V @ 0.8A +3.3V @ 30A +5VSB @ 3.5A continuous power = 1000W peak power = 1100W But where do I find my PSU's ratings? Once again, the ratings can typically be found on the manufacturer's web site. If you don't know who manufactured your PSU, then you need to open your case and look for the label affixed to the PSU. If you can't see the label, then it may be on the top or back side and will require you to remove the PSU (4 screws on the back hold it in). Holes <---Bottom left of this drawing So are wattage and +12V currents all that matter? NO! There are a couple of other items that need to be considered. Regulation, Ripple and connectors (this will be discussed below). Regulation is how well the PSU can keep the voltages at or near the proper voltage level. Back to that killer PSU above, it has regulation specs of 2% (+3.3V, +5V, +12V) and 5% (-12V). Ripple is how clean the DC voltage is. In theory, it should be a straight line when viewed with an oscilloscope. In practice there is some, for lack of a better word, waviness. Once again, that killer PSU has a 1% (p-p) rating for ripple. The (p-p) or peak-peak part means from the top and bottom of the wave. Here is the more technical explanation: Ripple The lower the numbers for Regulation and Ripple, the better. So the PSU I'm looking at shows continuous power and peak power? What does that mean? Continuous power is what the PSU can deliver continuously at it's rated temperature. Peak power is what the PSU can deliver for short periods of time, for example, when a hard drive first spins up or you open the CD/DVD tray. Remember the key words here too, Rated Temperature. That killer PSU above has a rated temperature of 50°C! That's 122°F! It is also rated 1000W continuous and 1100W peak! So how much power (Watts) do I need? This is always a tough question! It really depends on all of the components in your computer and their individual power needs. Here are a couple of sites that will allow you to get a feel for what it may take for Watts, but remember, the +12V amp rating is king and these sites don't take that into consideration. Calculator What is this thing about form factor? ATX or micro-ATX? This is basically the physical size of the PSU in height and width to fit into a standard PSU bay. That killer PSU above has an ATX form factor, but look at it's depth, 20" minimum! Here is some technical info: ATX What about these PSUs that say they are modular? Your most common PSUs are what I will refer to as hard-wired. What that means is the power cables are soldered to the PSU's main circuit board. Modular PSUs allow you to plug the power cables into the PSU as you add components to your system. The nice part about this is keeping the system's wire management clean since you only have the cables plugged that are needed. The downside to modular PSUs can be the additional connector that is needed to plug in the cable. Each connector has a certain resistance that will drop the voltage, but that depends on how well the connectors are made. BUT I HAVE A DELL AND I CAN'T UPGRADE MY PSU! Oh yes you can: Dell This is still a stinky in the works. I also want to link to some PSU calculators for basic wattage info and will keep a Steam-powered ranking for PSUs that forum members have tried and either given a thumbs UP or DOWN to. Input, output, throughput, feedback. This is trying to put some new super stinkies in here. I took on the PSU one, if you can't tell. It is a work in progress with your help. Connectors, connectors and more connectors! "Ok, so whatever dude! I got a good PSU like you have talked about, but like there are all these wires and smack coming out of it! Whats even crazier is on the ends of all these wires are all these different little freaky plastic 'do-hickeys'! They look like 'plugs' maybe, so now I want to use this crazy thing I just paid $100 bucks for! Man I just can't make sense of it all! It's like some kinda sick Walt Disney spaghetti!" Well, here's what all those crazy colors of spaghetti are used for (note the color and the description). Wires Now, let me try to clear up your questions on how a PSU connects to the motherboard. Here is a photo of all the wires coming out of the that nice PSU you bought http://i1.tinypic.com/ncxi61.jpg I have [marked] the main ATX power connector 20 pin and the main ATX power connector +4 pin. Here is a photo of just the main ATX 20+4 pin PSU connector. http://i1.tinypic.com/ncxcf5.jpg Notice how they are apart so the "+4 pin" would not be used on a older AGP motherboard for 20 pin use. Here is another photo of that same 20+4 pin connector together for 24 pin use. http://i1.tinypic.com/ncxdab.jpg They fit together quite easily. Here is a photo of that nice new motherboard. http://i1.tinypic.com/ncxfg8.jpg I have [marked] where the 20+4 pin connector (fit together for 24 pin use) plugs into the motherboard. Also, all PSUs have a 12 volt secondary 4 pin power connector. http://i1.tinypic.com/ncxjcz.jpg Notice I [marked] a different 4-pin connector this time which plugs in to the motherboard [marked] here http://i1.tinypic.com/ncxlqo.jpg Now before you ask, yes you will be able to tell the "12 volt secondary 4 pin" apart from the "main ATX +4 pin" connector when you get the PSU (it will be quite easy...as the "+4 pin" connector wires will be 'zip tied' to the wires of the "20 pin" connector). As long as I'm at it, here are the "6 pin PCI-E" and "SATA hard drive" power connectors on that PSU. http://i1.tinypic.com/ncxpb7.jpg You will need to plug the 6 pin connector directly into the PCI-E video card. The thinner SATA power connector is used on most SATA hard drives, some drives still have MOLEX power connectors on them, and some even have both. If they do have both use only one or the other, don't plug both a MOLEX and a SATA power connector into the same drive. "MOLEX? What the heck is that? Is that something you use to eliminate moles?" LOL, No!. It's this. Molex If you have any other questions (even if you think they are silly) don't be afraid to ask The Video Card Tutorial Introduction HELLO! and WELCOME! I have made this thread for all the people who are confused on what video cards are and what they do. I will be covering everything from slots, to video card brands and overclocking ect. so please... DO LEARN SOMETHING =). Also! if you have any questions or problems please send me a PM and i'll see what i can do! Table Of Contents: What does a video card do exactly? 1a. How to Install your Video Card 1b. What Video Card Do i have?! The AGP slot The PCI slot The PCI-E slot nVidia & ATI Workstation cards Integrated Video Cards Directx Drivers Overclocking Flashing Unlocking Pipelines Pipelines and Clock Speeds SLI Conclusion 1. Video Card?!? What does a video card do exactly?? Well yeah it puts an image on your monitor well... yes and then some. Your video card renders everything you see from a tiny pixel to a giant high res image of HL2. It receives a rendering signal from the CPU which is sent to the video card's core itself. After receiving instruction it renders whatever image it's been asked to do. Higher your Core clock speed, the faster it renders the image. The higher memory clock speed, the faster it is able to recall it. 1a. Installing a Video Card Installing a Video card isnt the hardest thing to do, probably a quick 5 minute process, heres a quick run down on how to do it. Remove your current drivers (read how to remove drivers under the drivers section). Open up your case Take out your old card Pop in your new one Lock your case back up Boot up Install new drivers from the video card brand website (ATI/Nvidia) the drivers on the disk provided are usually out of date. Reboot after drivers are installed Enjoy your new and higher fps 1b. What Video Card do i have?1 SIMPLE! follow these steps! Step 1 Step 2 Step 3 Step 4 Step 5 2. The AGP Slot The AGP slot runs in three versions: 2x, 4x and 8x the common ones today run 4x/8x. The AGP or Accelerated Graphics Port does exactly what it stands for -- it allows faster communication between the GPU and the CPU and creating a bigger bandwidth. Note that the difference between these versions are not only bandwidth but also the amount of power going to them. Here are some pictures of the different AGP slots: 2x 4x (4x information by Intel, given to me by Deathknight_NSA - THANK YOU) 8x These slots are designed for certain cards and THOSE CARDS ONLY. If you manage to fit a 4x AGP video card into a 2x AGP slot your computer will not boot up, and you might have problems with your video card. Furthermore, just about every game or program today only uses the bandwidth up to about 4x. 3. PCI Slot Ah yes the old PCI slot. This slot is a predecessor of the old ISA slot which is a giant brown looking thing. Anyways PCI stands for Peripheral Component Interconnect This bus slot is used for most accesories that a computer can accommodate such as sound cards, networking cards, video cards and RAID controllers. Really just about anything. Until the AGP slot came out PCI was the main graphics bus... why don't we use it now? Well in order for a GPU to operate in a PCI slot it has to make an extra stop at the northbridge before it can communicate with the CPU and this creates an abridgement in bandwidth, thus slowing the GPU down SO the AGP slot came along, and sped things up a bit. The white ones The PCI-E Slot Peripheral Component Interconnect Express is the most recent addition to the motherboard buses. It looks like a backwards PCI slot and also runs much faster than the AGP slot, so next generation cards will be obviously put into the PCI-E bus format. Unlike the 8x speed of the AGP slot PCI-E will run at 16x. THAT'S GREAT! it runs twice as fast!!!! Well yes and no. The AGP 8x refers to its bus clock speed, PCI-E 16x means it has 16 individual lanes each supporting 250MB/s per direction adding up for a total of 4GB/s. As for the AGP 8x, its max bandwidth is about 2.1 GB/s; HOWEVER, nothing in today's video card structure and operation uses more than AGP 4x which is about 1 GHz. ( Thank you again Laughingman11). Also here are some links that give good deal of info on the PCI-E slot: Here provided to me by T-Rush! nVidia & ATI These companies are the most popular GPU makers. They create the highest rated equipment and ship it to millions daily. Every year or so these companies release the cutting edge technology and display stunning and amazing graphics, of course, each company trying to out do the other. nVidia MX cards: Direct X 7 Geforce4 400 Geforce4 420 Geforce4 440 Geforce4 440 SE Titanium (TI): Direct X 7-8.1 Geforce4 4200 Geforce4 4400 Geforce4 4600 Geforce4 4800 Geforce FX: Direct X 9.0b *** READ BOTTOM* 5200 5500 5600 5600 XT 5600 Ultra 5700 LE 5700 5700 Ultra 5800 5800 Ultra 5900 XT 5950 5950 Ultra Geforce 6 Series : Direct X 9.0c and Shader Model 3 6200 6200 Turbo Cache 6600 6600 GT 6610 XL 6800 LE 6800 6800 GT 6800 Ultra Geforce 7 Series Direct X 9.0c and Shader Model 3 7800 GS 7800 GT 7800 GTX 7800 GTX 512 (*NOTE: i know that there are many more cards from nvidia that i have not listed, mainly because they're not nearly as popular and so i havent put them in the list, though i believe i have most of them). ****MAXIMUM PC Jan.05 pg50 "A GeForce FX (5000's) card will default to DX8 mode, but it's possible to force the game (HL2) to DX9 mode - just be prepared for a big performance hit....Note that the only "DX9 - capable" hardware that defaults to DX8 mode are NV30-based based GeForce FX cards; NV40-based videocards, such as the GeForce 6800 (and 6600GT), perform admirably in DX9 mode."**** - Thank You T-Rush! ATI The Radeon Family: Direct X 7 7000 7500 Direct X 8.1 8500 LE 8500 9000 9000 Pro 9200 9200 SE 9250 Direct X 9.0c 9500 9550 Pro 9600 SE 9600 9600 Pro 9600 XT 9700 XT 9700 9700 Pro 9800 SE 9800 AIW 9800 9800 Pro 9800 XT (this list was edited/remade by Danny M. THANKS MAN =) The Radeon X Family Direct X 9.0b x300 SE x300 x300 Hyper Memory x600 Pro x600 XT x700 Pro x700 XT x800 GT x800 SE x800 x800 GTO x800 GTO2 x800 Pro x800 XL x700 LE x700 X800XT AIW x800 XT x800 XT Platinum Edition X850 Pro x850 XT x850 XT Platinum Edition The Radeon 1kx Family Direct X 9.0c x1300 Pro x1600 XT x1800 XL x1800xl AIW x1800 XT x1900 AIW x1900 XT x1900 XTX Click here for discontinued ATI cards * important note about ATIs X family cards.. everything from a x700 Pro and down are all PCI-E ONLY, HOWEVER the x800 are avaliable in PCI-E format as well as AGP. As for the regular radeon family they are a mix of PCI and AGP, however most can be found in 4x to 8x AGP format. Work Station cards Where to begin with workstation cards... WELL they are primary for graphics and graphics only and by graphics I mean cinema efffects NOT GAMING. Workstation cards include Quadro FX from nvidia, and FireGL from ATI. I ONLY recomend these GPUs to those who do more editting with 500$ Pixar Toy Story effects programs than gaming. I have seen some workstation cards with 512 mb of video memory -- it ran HL2 @ about 10 frames per second, usless in my opinion. Integrated Video Cards What is an integrated controller? Well it's simple really.. it's a GPU core imbedded onto your motherboard. Sounds good right? WELL only if you have 0 expansion slots. Truly the integrated graphics solution is not the way to go; infact it's probably the slowest form of a GPU and most likely the fastest card you can find that is integrated is going to be a MX440, or some random Intel Extreme, which I might add are not the greatest. To install a new graphics card over an integrated one: FIRST insert your new sexy card into your machine.. THEN boot up with your monitor plugged into your onboard graphics then goto: Start Settings Control Panel System Device Manager Display Adaptors Right click on your integrated card Disable it Reboot OK now you must switch your monitor plug to your NEW graphics card..let the computer boot up.. and there you go.. your screen should look like crap and you will need to install drivers for your new card. Direct X DirectX is Microsoft's program that allows the latest video games to be played and the greatest sounds to be heard. Because this tutorial is on video cards I have decided to tell you which cards are compatible with the latest and greatest. Yes that's right not ALL video cards are able to render and produce the image that the latest versions of directX are capable of. Infact nVidia's past generation cards the Geforce 5 FX line are NOT able to do this. Unlike ATI whose cards are able to do this, such as the 9600 Pro. Please refer to my list of video cards for which card(s) are able to render the different versions of DirectX. The latest version of DirectX can be found here after you download this, let the program run; it will detect what is needed for an upgrade. After the upgrade is complete you will be asked to restart. Little Note on Directx on the Source engine here are the commands to FORCE the engine to render the following versions 6.0,7.0,8.0,8.1,9.0 type these in the console: mat_dxlevel 60 mat_dxlevel 70 mat_dxlevel 80 mat_dxlevel 81 mat_dxlevel 90 mat_dxlevel 95 Drivers Drivers are crucial for video cards to work their magic. Drivers basically tell the card what to do and how to operate ect.. to find the drivers visit Here for Nvidia and Here For ATI. Now there are also another set of drivers called Omega, these basically are modifeid versions of the offical Nvidia and ATI released drivers that specialize in either image quality or performance, your choice. Note that you will be using these at your OWN risk =) they can be found Here Another thing to keep in mind is that if you are switching from one video card maker to another I.E. going from a 5200 FX to a 9800 pro you will suffer the consequences of NOT CLEANING YOUR DRIVERS =) SO go here for driver cleaner and follow the instructions in removing your old drivers. ( Thanks Danny =) also keep in mind, this is for ATI users ATI's Catalyst drivers come in two bundles: 1) Catalyst Control Center - this requires the Microsoft .NET framework be installed (available from Windows Update) or 2) the old style "control panel" version - this is faster, but doesn't offer the same level of control as the CCC. (thanks grommit) ATI Catalyst Tweak Guide Thank you K9-Cop Overclocking Overclocking is where you turn up the Core and Memory clock speeds on the video card for exta performance, Now this is very dangerous and you can permanently damage your GPU by doing this. When you overclock it is HIGHLY suggested to have after market cooling HSF such as a Arctic Nv Silencer. Flashing Flashing means, upgrading or changing the BIOS on the GPU chip itself. It can be a dangerous procedure but as long as you have backups I believe it is recoverable. What you get from flashing is basically a new video card... many people flashed their 9800 pro to 9800 XT speeds. When they changed the BIOS the 9800 Pro thought it was a 9800 XT and so it ran at much faster speeds. I am not an expert at flashing for I have NEVER done it before. Unlocking Pipelines Unlocking pipelines is like opening up a dam... it lets much more stuff through.. and thus allows faster speeds. As far I know you can unlock the pixel pipelines on Nvidia 6800s and ATI x800 Pros. In order to do this you must use Riva Tuner 2.0 -- NOTE that this will NOT ALWAYS work, as I found for myself when I tried to unlock my pipelines.. when I did I received massive artifacts in game. If this occurs when you do it, please reset the pipelines for it may cause damage. Here is a link to a great tutorial on how to unlock pipelines -- check it out! This was given to me by gr8fu1 THANK YOU MAN! =0 Here Pipelines and Clock Speeds Getting a tad bit more indepth with a video card i am going to give a little explaination on the markets 2nd most advertised statistics (aside from amount of RAM) - clock speeds. When talking about clock speeds in GPUs the 2 primary speeds are Core and Memory. The core clock is how fast the GPU is able to render an imagine (there are some other factors than just Core clock, this will be explained later) and Memory clock is really quite simple from a broad standpoint; memory clocks determine how fast the GPU is able to retrieve and or write to the memory banks, high memory clock - higher bandwidith - higher rate of data flow - better fps! Now back to the GPU core, they're a couple factors which either limit or allow the core to render images. A common term amongst these factors is Pipelines. Pipelines are the 'tunnels' or 'pathways' that allow the GPU to communicate through. Its like someone yelling through a microphone - bigger the microphone the more sound is going to be made. An example of how the GPU and the pipelines would work can be shown in Peak fill rates: which is determined by the gpu core clock speed X (multiply) the pipelines. This is really a rather breif explanation if anyone has something to add please reply to this thread or send me a PM. It may also bea tad bit sloppy its 2:41 A.m. SLI What is SLI exactly? Well it stands for Scalable Link interface. It includes a Nforce4 motherboard with 2 PCI-E slots. The way it works is that you have a little bridge connector that connects the 2 video cards for multi-GPU rendering, which in the end gives a massive increase in frames per/second; HOWEVER costs quite a bit. SLI can be done with I believe Nvidia cards (AND NVIDIA CARDS ONLY, ATI CARDS CANNOT USE SLI)... 6600 GT and up. SLI looks like This. This is sort of a return for SLI. It was originally started by 3DFX with Voodoo 2s and could run quake 2 ON MEDIUM GRAPHICS!! Yes medium. Note that when using SLI, it cuts your PCI-E speed in half, so 2 cards in SLI mode will run at 8x PCI-E -- this does NOT MEAN ONE CARD RUNNING AT 16x IS BETTER. Remember 16x is PCI-E means 16 different lanes for bandwidth, so you now have 2 cards each with 8 lanes of bandwidth. Conclusion WEll I hope that this tutorial has give some insight and/or help towards your decision on your next video card or maybe it just helped your general knowledge. Please once again message me if they're ANY mistakes in this I will gladly fix them. ALSO message me if you have any questions I will happily answer them. AND OF COURSE thank you to all the people that have helped me, critized me and informed me. This wouldnt have been stickied without you =). All Credits to: Hardware Dept. SF
