Intel or AMD, some guidelines **updated 21-11-03**

As some of you might know, there have been quite some discussions about Intel vs. AMD CPU’s. Often, users asked for help because they couldn’t make a choice. After posting a thread and getting various answers, they still did not know what to choose. One of the reasons why, is the continious flaming between AMD and Intel fans.
That’s why I made this sticky topic: to help you choose the CPU that fits you the best.

Introduction:

For the time being, Intel is reaching way higher clockspeeds than AMD is doing. This scares some people, although this is not really a bad thing. Intel chose to go for high clockspeeds and a lower efficiency. AMD chose for lower clockspeeds, but higher efficiency. In technical terms, the IPC (number of instructions per clockcycle) of an Intel CPU is way lower than the IPC of an AMD CPU. Because of the higher clockspeeds, this isn’t a real problem.

There are various ways to rate the speed of a CPU. The best known way to rate a CPU’s speed, is to mention the clockspeed it’s running at. For the time being, Intel does this for all CPUs, while AMD only does this for their cheapest CPU’s.
AMD uses three kinds of ratings nowadays: the AMD Athlon 64 FX, the AMD Opteron and the AMD Athlon XP each got a rating system of their own. The ratings of the Athlon 64 FX and the Opteron CPUs are somewhat similar, as they are just numbers with a certain meaning, while the rating of the Athlon XP really points at performance that would be achieved with another CPU running at that clockspeed. How these ratings works, will be explained in the ratings paragraph.

Besides the capacity of the CPU, there are also other factors responsible for the overall speed of the system. Those will be mentioned later as well.

Type of CPU:

AMD:
Since some time, AMD’s latest series of CPUs, the AMD64 series has entered the market. This line of CPUs is based on the x86-32 (32bit, also known as Intel Architecture 32, or short IA32) instruction set, but extended with the AMD64 instruction set.
You might think dat you don’t need 64bit, as this is normally only used for server systems. Well, you’re wrong :). There are quite some programs (especially those who have to do some heavy computational work) that can make use of this 64bit registers (this will increase the performance of the system) and the nice gimmicks that come with the technology.
There are some different types of the AMD64 CPU. The first version to be released was the AMD Opteron CPU. Although it was intentionally only meant for server systems, due to it’s high performance (especially in SMP setup), it also entered the professional workstation market. Now, all the models of the AMD64 series are becoming available. We have the AMD Athlon 64, the AMD Athlon FX and of course the Opteron.

  1. AMD Athlon 64. This is the cheapest version of the 64 bit CPUs. As you already might have thought, it’s also the CPU with the lowest performance. Nevertheless, it’s quite a fast CPU. The main difference between this CPU and the two other Athlons is, that the Athlon 64 doesn’t support dual-channel memory. This CPU can be put on socket 754 mainboards.

  2. AMD Athlon 64 FX. This CPU is the faster one, meant for both work(fun) stations, and little server systems. This version is in fact identical to the Opteron, but doesn’t have a pin for ECC memory. The socket for this CPU is socket 939, but it can also be put on socket 940 mainboards.

  3. AMD Opteron. This is the most expensive CPU of the three. Although it’s quite expensive, it performs equal to the Athlon 64 FX. The only difference is that it needs a socket 940 mainboard and it can’t work (yet) with unregistered memory (that explains the 1 pin difference between the Opteron and the FX).

Of course, the 32 bit AMD CPUs are still available and they still can satisfy the customers that need quite some power. In the 32 bit CPUs, there are three ways you can go:

  1. Fast AMD Athlon, based on the Barton core. This Athlon XP CPU comes with a 333 or 400Mhz FSB (2x166/200) and 512KB cache. It’s available at ratings between 2500+ and 3200+. This is quite a good performer (and the ones at lower stock speeds are excellent overclockers) and comes for a fair price.

  2. Slower AMD Athlon, based on the Thoroughbred B core. This Athlon CPU comes with a 266 or 333Mhz FSB (2x133/166) and 256KB cache. It’s available at ratings between 1700+ and 2700+ (<2200+ is getting obsolete). Due to it’s smaller cache, possibly lower FSB and lower clock speeds, the performance isn’t as good as the Barton’s performance, but it still very satisfying. A nice thing to know is, that some of the slowest of these Tbred-B CPU’s (1700
    +, 1800+, 2100+) can be overclocked more than 100%. So if you are a real overclocker, you know what to get :).

  3. Budget AMD Duron CPU, based on the Applebred core. This CPU is clocked at lower speeds (yes, for this CPU’s, the clockspeed is mentioned, instead of a rating!) and has only 256KB of cache memory. These CPUs come at speeds like 1600Mhz and have a 266Mhz FSB. Their performance isn’t all that good, but their price is that low, that this leaves nothing to complain about.
    Intentionally, this CPUs were only meant for the Asian market, but as this CPU can be a very nice and cheap upgrade for quite some people, it’s available in the rest of the world as well (not as widely spread as the Athlon XP CPUs, but ok…).

Intel:
On the other hand, there’s of course Intel. The brand that everybody knows and sells very well (mostly because it’s a known brand). Intel offers quite some CPUs as well.

At first, we have the high performance models:

  1. For the most power-hungry people, there’s the Intel Pentium 4 Extreme Edition. This special edition of the P4 CPU comes with a load (3MB) of cache memory and is equipped with Hyperthreading (see optimizations) technology. For the time being, this very expensive CPU is only available at 3.2 GHz, but most probably, the 3.4 GHz version will enter the market very soon. This CPU is pointed towards gamers, but everybody that needs a lot of CPU power (let’s say for video/photo editing or so) can make this enormous cache useful.

  2. If you don’t want to spend an enormous amount of money, but still get a good CPU, you can go for a P4 CPU with the 800Mhz FSB and Hyperthreading technology. This CPU gives still a very good performance (especially when it’s used for photo or video editing) but is significantly cheaper. At first, only the P4 at 3.06 GHz was equipped with the Hyperthreading technology, but as many people wanted to have HT, but didn’t want to spend $700 on a CPU, Intel started releasing slower P4 CPU’s with Hyperthreading. Nowadays, you can get CPU’s from 2.4 GHz to 3.06 GHz.
    For people who got a somewhat older mainboard, there’s the P4 with the 533Mhz FSB with Hyperthreading technology. It can’t compete with the faster 800Mhz FSB models, but it may be worth it, if you still are running a somewhat slower CPU.

  3. Last, but not least, there’s the Xeon CPU. This CPU was only intended for light server systems, but it’s quite suited for workstation needs as well. There are two different versions of this CPU: the Xeon and the Xeon MP. The difference between the two is that the MP is equipped with Hyperthreading technology. The standard Xeon is only available with a 400Mhz FSB, while the Xeon MP is available at 400Mhz and 533Mhz. As these CPUs are quite expensive, motherboards are rare (for normal customers that is) and almost no CD Freaker would buy such a CPU (except for Domi of course;)), I’ll leave this one further out of the picture.

If you want to go Intel, but don’t want have to spend an enormous amount of money, there are two options to consider.

  1. At first, there’s (once again) the Pentium 4 CPU. This time I am talking about the slower ones. These CPUs come with a 400 or 533Mhz FSB (some even only have 256Kb cache, instead of 512Kb) and don’t have the Hyperthreading technology. Nowadays, models from 2 GHz to 2.8 GHz are available (the slower ones are quite hard to get btw).

  2. Intel’s ultimate budget solution is the Celeron CPU. Although the Celeron reaches speeds up to 2.8 GHz, it’s 400Mhz FSB and tiny cache (128Kb) make it a not-so-good performing CPU. Although the performance is low and the price is relatively high, this is a very popular CPU. Why? It doesn’t generate too much heat, doesn’t cost that much and so is ideal for all kinds of office applications.

Ratings
As I pointed out earlier, AMD uses three different performance rating systems.

The simplest rating is that of the Athlon 64 FX series. The first CPU to be released was model “51”. The number 51 doesn’t mean a thing, it’s just a number used to point at a certain CPU. The next FX CPU to be released will have “52” as rating number… etc…etc… you get the picture :wink:

The rating of the Opteron is somewhat more complex. It consists out of three digits. Just like the FX, the last two digits of the rating number mean a certain speed. The first digit stands for the number of CPU’s that can be used at max on a single mainboard. For example, the first Opteron CPU, the 140, was a 1.4Ghz model, designed for single CPU use. It’s bigger brother, the 848 is designed for 8-way systems, running at 2,2 GHz.

Last, but not least, there’s the rating system of the Athlon XP. Many people think that an AMD Athlon 2800+ is running at 2800Mhz, but that isn’t very true. No, it’s running at a much lower clockspeed…
Often people say that an Athlon XP CPU with PR-rating of 2200+ is as fast as a P4 at 2,2Ghz. This is not true, the PR-rating is the speed of the CPU as it would be if an AMD Athlon Thunderbird-C would run at that clockspeed!

Optimizations
Both AMD and Intel CPU’s have some optimizations implemented. These optimizations can fasten some processes.

On the moment, the AMD Athlons (and Opteron as well) have:
-MMX
-SSE
-3DNow!
-Extended 3DNow!

Intel has:
-MMX
-SSE
-SSE2
-Hyperthreading

As you can see, the CPUs differ quite a lot from each other.

The 3DNow! instructionsets of the AMD CPU’s are somewhat like the SSE instructions of Intel. MMX was implemented later into the AMD CPUs.
SSE2 is a technology that soon will be adopted by AMD in Q1-2004 and is created by Intel. It works very well, but has only been adapted by some developers. If you work with software like Adobe Photoshop and Premiere, an Intel CPU has quite a nice advantage here.

Another neat feature is the hyperthreading technology. Theoretically it can fasten programs that are capable of SMT (using more than 1 CPU at the same time) a lot. Although this is a nice technology, not all programs know how to handle it. Some programs are even slowed down up to 30%!
If there would be new versions of many programs that support and use the HT-technology, this would really be an incredible performance gain (up to 30%)…!
If you are running an OS that can’t profit from HT, it’s better to disable it in the BIOS, as you’ll loose performance by keeping it enabled. Windows XP is one of the few OS’ that is capable of using HT. Another one is Redhat Linux 9.

Cooling
AMD cpu’s are known for their enormous energy consumption and production of heat. The modern AMD CPUs are still consuming a little more energy and producing slightly more heat than the modern Intel CPUs, but the difference is getting smaller and smaller. The next big CPU release from Intel, the Prescott CPU, will eat over 100W. It’s the first time that a desktop CPU goes over 100W. The former number 1, was the AMD Athlon Thunderbird-C at 1,4Ghz, consuming 80W.

Intel CPUs do have a better heat protection technology, but as most modern AMD mainboards have such a built-on protection (works pretty well) this does not have to be a problem anymore.
The new (64 bit) CPUs don’t produce as much heat as the older Athlons did and they are equipped with a heatspreader, just like Intel’s CPUs. This is better for the conduction of the heat, as well for protecting the CPU by damage done by a wrong installation of the heatsink.

With some decent cooling the heat an AMD CPU produces is no problem at all. Just make sure you pick a cooler that doesn’t make too much noise, if you like silence;).

Be carefull: overheating might damage a CPU.
AMD CPUs do not have any active overheating protection and can fail easily, especially without a heatsink mounted. Although modern AMD mainboard do have some nice protection technologies implemented, they cannot save a CPU without a heatsink. Saving the CPU that is overheating due to things like a failing CPU fan should not be a problem.
Intel’s P4 has a nice overheating protection. They can even run without a heatsink on them (although the speed is loweved by serveral hundred megahertzes!)… no damage there!

Chipsets

Intel:
The most popular chipset for the Intel CPU, was the Intel i854-series. Since some time, Intel released two new chipset series: i865 and i875. The main difference between the i865 and i875 chipsets, is that the i875 supports PAT (performance acceleration technology). With this technique, the memory access can be handled faster. Some mainboard manufacturers found out that it was possible to enable PAT on i865 chipsets as well, making it useless to buy the more expensive i875 chipset. As Intel didn’t really like this, they made a 2.0 version of the i865 version of the chipset, making it impossible to enable PAT. So if you want a cheaper chipset with PAT, make sure you get a “classic” i865-based mainboard.
Another feature that was added to the new chipsets, is the 800Mhz FSB and Prescott support. Although it is nice the chipset can handle a Prescott CPU, I really doubt if any recent mainboard could work with a Prescott on it, as the boards are designed to deliver up to 75W for the CPU (as I said, Prescott needs over 100W).

Another chipsets for the Intel platform are build by SiS. They have many budget models (like the 650 and 655 series) that are mostly used for laptops or cheap barebone systems. In the near future, SiS will release a high performing chipset for the Intel platform. If that ever happens, I’ll update this sticky :slight_smile:

AMD:
Some time ago, the KT133 and KT266 series from VIA were quite popular. Since the KT333 series, VIA lost their touch in making fast chipsets. Their latest chipsets, the KT400a and KT600 are doing great, except for the mediocre performance.
Luckily, there was NVidia with the NForce1 and 2 series. Although the NF1 series weren’t really popular, the NForce2 series has gained a vast majority in the AMD chipset market. Nowadays, the NForce2 chipset is the most popular chipset for AMD CPUs. There are quite some versions of the NForce2 chipsets. Besides all the features NF2 chipset has (LAN, 6.1 audio, GF4MX), there are different versions. The newer series, NForce2 400, are certified to run at a 200Mhz FSB (effective: 400Mhz), while the older NForce2 chipset isn’t.

For the 64 bit CPU’s, there are a few chipsets available. For now, there’s the AMD chipset (that’s mainly being used for Opteron based server systems), the NVidia NForce3 (pro) and the VIA KT800 series. Although they perform about the same, the VIA seems the best way to go, as it has more features than the NVidia chipset has.

Memory
Nowadays there are 3 kinds of memory: SDR (133Mhz), DDR (266/333/400Mhz +) and RDRAM (400/533Mhz).

AMD CPUs can be used with SDR and DDR memory. Since the Athlon XP (the only model that’s being sold right now), DDR memory is a must, if you want the full performance. As you can see, it comes in different speeds. Depending on the CPU you’ve got you need a specific kind of memory. Since the new Athlon XP was released (based upon the Barton core) there are 3 different busspeeds: 266, 333 and 400Mhz. For the 266Mhz FSB Athlon XP, 266Mhz memory (PC2100) is the bottom line, faster memory will give you a small leap of performance. . The 333Mhz FSB CPUs from AMD need 333Mhz memory (PC2700) to give full performance, 400Mhz FSB CPUs need 400Mhz (PC3200) memory.

NB. If yoou want to make use of a NForce2 chipset with integrated graphics, you can’t use any memory faster than PC2700!

Intel CPUs can be used with SDR, DDR and RDRAM(RIMM, Rambus) memory. If you want real performance, RDRAM is the way to go. Intel recently decided to let RDRAM go. If you want to make use of RDRAM on your P4, you should look out for a somewhat older Intel board, or a board based on another chipset (e.g. from SiS) with RMRAM support. In the near future, RDRAM might becoming important once again, as RAMBUS’ latest technology, Yellowstone memory, is extremely fast. Since RDRAM is really expensive, most people nowadays buy DDR memory. As the P4 CPUs have FSB speeds 400 , 533 and 800Mhz, you should take the fastes memory you can find (PC3200 is fine) to make use of the full potential of the CPU.

I won’t tell anything about the CPUs combined with SDR memory, since that’s really outdated and hardly used anymore.

A nice technique that is becoming popular (once again) is dual memory. By installing 2 equal DDR modules you can (if your chipset can handle this) bundle them as a memory-array (comparable with RAID-0) and have double speeds (theoretically). This’ll give both CPUs a nice performance gain…!

If you’re up to buying a new system right now, you could consider DDR400 (PC3200) memory, especially if it’s memory with lower latencies (the bigger modules you choose, the slower modules get).

If you’re into overclocking, you might want to go for extremely fast memory, as PC3500 and PC3700. Although these memories aren’t JEDEC certified, you can use this for cranking up the CPU FSB and run the memory in 1:1 with the CPU FSB.

Usage

Caution: I didn’t update this section for some time. Why? As both Intel and AMD both are releasing new CPU’s, its still hard to tell who’s the best for what, as it needs some time for the hardware (mainboard and chipset) and software producers to adept to the new technologies.

The former situation:

Desktop applications:
Both AMD and Intel CPUs are perfect for these applications. In many applications the AMD is the faster choice (compared by price, not by clock speed), Intel is the quiet choice. AMD can be quiet as well, but it takes a little effort.
For the most desktop applications the modern, faster CPUs are overkill. That’s not bad, but it’s a waste of capacity (and so money).

Gaming:
In most situations AMD wins the contest here. Comparing processors that cost about the same, AMD clearly wins in 90% of the cases.

Audio, video and picture editing:
Here Intel wins the contest. The main reason is the implementation of SSE2 that is used well by Adobe etc.


Well I hope I could help you out a little with this thread. If you got any questions left (there are loads of them that are not covered by this topic), feel free to ask it on the forum!

If there is anything in this story that isn’t correct or whatever, please contact me by PM!

One last tip: CPU cores are fragile. The Intel CPUs are protected by a heatspreader that is almost impossible to be damaged. The AMD CPUs don’t have this protection and can be damaged easily, if the heatsink is installed wrong! Only do this yourself, if you know how to do this. Please do not forget the thermal grease!

Dee-ehn