Jump to content

Dual Core Double Delicious

Recommended Posts

As usual Steve, you write a great article.


The one thing you failed to mention as a huge factor about the rise in dual core processors is.


You can't buy a computer today without one :P


I personally have a q6600 quad core in my machine. Someone wanted me to compare it to their e6850 in a test that only uses one core.


I didn't do it, as obviously nothing but the higher click speed comes into play in such a test and isn't the reason I chose the quad core.


That e6850 was mine, and I sold it to him, and kept the slower clocked q6600 because I do a lot of multi-task computing, and also build applications and the build process is just incredibly fast because I can use all four cores. :tup: For me the more cores I have the better, as they actually do increase my speed and in some instances more then double my speeds.


I do agree with you that many people don't understand whats in play when it comes to multiple cores and what kind of gains they should expect.


I also think that "most tests" don't actually test the capabilities of multiple cores, so they don't give accurate results, and don't reflect the way computers are used on a daily basis.

Link to post
Share on other sites

Of course we read that Jacee. What self respecting, aspiring computer geek, would have missed that. LOL.


I don't understand it but for some reason a $650.00 motherboard and $2000.00 worth of processors doesn't sound like that much. Besides I don't have to buy the case, powersupply,or Optical drives. LOL :overclocking:


Keep in mind I am still wearing the same tennis shoes as I was in 2001. Now that I think about it, the same laces too.


Bruce, you're exactly right. The choices for single core are very limited now. :tup:

Link to post
Share on other sites

Well im not to sure with pricing anymore, but i bet that 3.7ghz processor is a pretty penny. Im running on a 939 socket dual core opteron 165, stock clock at 1.8, i have it clocked to 2.7 and if i had better ram i could run 3ghz without volting. this chip is $100 and ive ran the overdrive says im top 30% in the world and my processor is in the top 1%. but i would say when it comes to money u cant beat my setup, under $500 for what im running.

Link to post
Share on other sites

Personally, I use multiple computers to do multiple things. If I'm rendering a video, I use a different computer to play games while the video is rendering on a different machine. Same thing with downloading. If there is something I need to download, such as a Linux ISO, I use my file server to download it. My hardware isn't always the "latest/greatest", but I do upgrade various components maybe every 9 months or so. That is why I have multiple machines. They are all running dual-core processor by now, all from the AMD variety.


I'm just now seeing a benefit for dual cores in gaming, at least with the benchmarks various games come with. Does it perform any different than my friends who are still on single core? Barring video card differences, no, not really. As for the other things the other machines do, I don't really notice very much of a difference with dual core since I typically set it going and forget about it until it's done.

Link to post
Share on other sites

Original Article-

"Is it twice as fast? No, not in the way most perceive fast, not as it relates to our daily computing. All else being equal a 2200MHz, dual core processor is no faster than a 2200MHz single core processor. Bear in mind I said "all things being equal." There are other things that come into play, cache, memory speed, front side bus, and yes there are faster and slower architectures. Does your new dual core 2200MHz rig surf, instant message, and open programs as fast as my friends old 3.73GHz EE (extreme edition) single core? Nope, it's not even close. Does the processor burn a DVD while rendering a photo or downloading a program, as fast? Yes, it beats it do death"


All things being equal would mean you were using a dual core with the same architecture as the old 3.73GHz EE (or as I liked to call it the over priced space heater). Many of the new Dual cores are new designs and even at a lower clock speed and in single core versions, more efficient than older faster clock speed cpus. Clock speed is not an accurate mesure of perfomance any more. Anyone who has compared AMD and Intel products can tell you that. If you do one thing at a time, and you are comparing 2 cpus with the same core design, on a dual core and the other a single core with a much higher clock speed, then yes the faster single core wins. The reality is that most operating systems are not only doing one thing.


Compare a Core2Duo or Quad version at a lower clock speed even on a single item task and it is faster, part of it is the cache, part of it is the more efficient design. More over the newer cpus with lower clock speeds run cooler and use less energy than that extreme edition pig.


The fact is there is no all things being equal between the old P3, P4, and the new Core# cpus.

Link to post
Share on other sites
  • 3 weeks later...

Depending on what you are doing, you can get from 40% increase to almost 100% when you have some fairly memory independent process going on in one processor, and that is over a single processor of the same speed. On any given task, with CPU's from the same base technology and without multi-threading, the single processor will work slightly faster than one of those of equal clock speed in some multi-core processor. That's because of the overhead of making sure that reads and writes from multiple processors are properly sequenced, amongst other things. Now, assigning some of the overhead tasks to another processor can theoretically improve performance, but the overhead of so doing largely negates the gains. And, some processes are sequentialized, no matter how distributed, causing extended wait states for something happening on another processor or in an unprocessed thread, or waiting on an external process, be it a user at the keyboard or a piece of data from an HD. Even nominally independent tasks can get into resource conflicts at various levels. Most of the gains in Intel multi-core processors are to be found in their greater cache, where the cache is not partitioned, and is available fully, or almost fully to any one of the processors. In the current Intel Quads, you really have two dual core processors bus tied on a single chip, so the cache is partitioned between the two dual core processors.


Without some really detailed information and some very careful characterization, it is impossible to know just where and how all gains and losses happen. Most such characterization these days is done at the application level, yielding anything but generally applicable information. That is why there is such a wide range of relative performance statistics from application to application on the same processor, much less from processor to processor, even those in the same family. Plus such detailed information is jealously guarded by all parties. Then on top of all that, there is process variation, which can affect different parts of a processor in different manners. It may be within spec, but that does not mean that it cannot have significant effects in particular situations. Try reading the processor errata at Intel or AMD, and you will get a faint idea of just how much goes wrong that must be handled at the BIOS or the OS level.


Then, there are those BIOS's and OS's. I would say that current code in those has enormous performance implications, and the way in which they operate changes from motherboard to motherboard, and so on and so forth. Really, the only place that even a baseline can be established is in the reference boxes provided by the manufacturers. From there, we aggregate the thousands and tens of thousands of tests done on various boxes with various software, and we start to get a feeling for the range of performance and the strengths and weaknesses of each processor. Even though Intel kicks butt on AMD these days, the AMD processors have a huge advantage in RAM bandwidth and interprocessor communication, which is not taken advantage of, and that can only be due to weaknesses in BIOS and OS and application code.


Intel's huge advantage is the fact that they are always one process step ahead of everybody else, allowing them to put more cache on the chip. Also, their cache access has always been wider than AMD's, allowing them to take advantage of the economies of block movement from RAM. Also, because of the limitations of their total RAM bandwidth, they have come up with more sophisticated pre-fetch algorithms. They also have room for all that logic on their chips because of their process advantage. What is most interesting is that Intel's most recent families of processors have been based on AMD's ideas of how to design processors. Intel was headed into a dead end with their whole P4 architecture, which almost cost them a collapse. Only their financial muscle and monopoly actions in the market have allowed them to take advantage of everybody else's intellectual property, while suing people who get even close to their own. It is an uneven playing field. AMD and others do not have the process or research dollars to cover the territory that Intel does, and they must give up what advances that they do make to Intel so that they can compete at all in the monopoly market. In return, Intel only gives them the right to use the basic outlines of Intel instruction set design.


To shift to another arena, the point is proved by Intel's ability to stop IEEE 1394 at 400 Mhz with a piece of junk like USB 1.1 at 12 Mhz, with only the promise of USB 2.0, which finally arrived at 480 Mhz. In addition, USB is a client-server architecture, making it much less flexible and powerful than the peer to peer architecture of 1394, or Firewire, as it is known. Now, 1394b 800 has almost zero penetration, though it is more than twice as fast as USB 2.0. Actually, that is a technical lie. USB 2.0 runs at 480 Mbits, while 1394b 800 runs at a bit below 800 Mb, making it not quite twice, but like 1394 400, it will transfer more data per Mbit of rated capacity than will USB. That is the power of monopoly.

Link to post
Share on other sites

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now
  • Create New...