|
A Joust to the finish.
The CPU market these days is an interesting place, we have three price segments and three markets. Low, Mid, High (duals and quads) and Budget, Mainstream, Professional. The tactics that AMD and Intel have been using have proven to be quite interesting.
Not too long ago AMD introduced the Phenom II line of quad core processors which undercut Intel’s Core2Quad line in price and brought similar performance. However, Intel at that point still remained the performance king with its high clocking E8000 dual core line, the formidable Q9550 quad, and untouchable Core i7. The Phenom II line quickly sprouted off in multi-core directions, bringing the Phenom II X2 and X3 into the fray as well as the X4 955 and X4 965. It seemed like, for the first time in a long time, AMD was back on top, at least in the mid-range.
The Core2 line still technically provides better overall performance but fails in two important areas for its mid-range segment: price and gaming. The Phenom II architecture is a heavy hitter when it comes to gaming, giving nearly on par clock-for-clock performance against the Intel flagship; the i7. So when AMD branched its Phenom II line into dual and triple core forms they had effectively stolen the segment from Intel. This, in addition to backwards compatibility and promise of long term socket support, gave AMD a large edge and command of the mid-range. Good times for AMD.
The good times, however, would not last. A few months later Intel introduced socket 1156 which would power their 32nm Clarkdales in the future, as well as the newly introduced Core i5 line. The Core i5 line was a crippling blow to AMD’s domination. The cheapest i5-750 was essentially an i7 without Hyper-Threading and triple channel memory. These losses are nearly nonexistent and with Intel pricing the i5-750 at around $200 it chopped off AMDs mid range quad domination.
AMD struck back at the low-end with the Athlon II X2 and X4 line. The Athlon II X2 250 brought a 3GHz dual core to the low end segment which performed around 8% worse than the Phenom II X2 550 at more than 20% less cost. The Athlon II X4 620 marked the first time a brand new quad core would sell for under $100 while keeping up, and in some cases, besting Intels $160+ quads. With the Athlon II X2 250 dominating the low-end, the Phenom II X2 550 BE dominating the high-end duals, the Athlon II X4 620 dominating the low-end quads, and the Phenom II X3 720 BE being the best bang for buck processor, AMD seemed to still be in good shape. In addition to this, 2010 would see the introduction of AMD 45nm hexa cores and 32nm octo cores.
As good a year 2010 seems to be for AMD, it will be even better for Intel. Early next year, Intel will be introducing its answer to AMD’s remaining mid-range domination: the 32nm Clarkdale. Early reports of Clarkdale told only half the story. Intel would release a 32nm dual core to combat AMD’s Phenom II quads? Surely, a losing proposition. As the release of Clarkdale grew nearer, revelating rumors began to surface. Clarkdale will have Hyper-Threading. Clarkdale will have Turbo.
Currently, Clarkdale will be separated into two lines: the i5 which will be high end ($170-$280), the i3 which will be mid range ($120-$140), and the Pentium which will be low end ($90). The i5 and i3 lines will include HT and Turbo. To make things interesting (and quite frankly, practical) let's assume the Clarkdale architecture will match the Phenom II clock-for-clock and core-for-core in gaming performance and surpass it in everything else. With the slowest mid range Clarkdale clocking in at 2.93GHz, what this would mean is that AMD would lose the dual core market which is, in essence, a majority of the gaming market. That’s not even taking into account Turbo mode.
But what about those gamers who also multitask heavily? Intel answers that in the form of Hyper-Threading which turns Clarkdale into a virtual quad core. This practically destroys most of the incentive for the average gamer to go with Phenom II. A higher clocking dual core is going to perform better than a lower clocking quad core while consuming less power and producing less heat, and unless you are doing something that requires 100% of one core, Hyper-Threading will provide adequate real world multitasking.
This really only leaves AMD with consumers looking for incremental upgrades for their AMD motherboards and the low-mid range quad core market for people who actually need 4 physical cores; a very small segment (practically nonexistent) as most people who require 4 physical cores will move up market.
What's AMD's next move?
|
Osaka27118 Posts
As someone who does not follow the trends of hardware much, this was a really interesting read. I'm interested in any debate that develops in this thread.
|
I never knew there was such competition between AMD and Intel. If this thread develops into a real debate, I'm sure to learn lots.
Edit: So manifesto basically said it better.
|
I have to say that I agree with the i5 cutting deeply into AMD territory. I was actually going to build a Phenom II x4 machine until I saw the i5 :|
|
Welcome to TL! Had you included some links in there I'd say you were a spam bot or marketer lol.
While there are enough techies around here, you might be on the wrong forum for this kind of stuff (seeing how we also have people deleting their rundll files ).
I guess we'll just have to wait and see? :p Nothing touches the i7 series at the moment, but I'm sure AMD is brewing up their own new line of CPU's.
For people wanting to do more research.. check these benchmarks! In some areas the i7 series are more than twice (!) as fast compared to the best AMD competitor. http://www.tomshardware.com/charts/2009-desktop-cpu-charts/benchmarks,60.html
Personally I prefer AMD for desktops & Intel for notebooks. However, if I were to buy a new desktop right now and had a budget to buy a monster comp, I'd definitely go for an i7 tho.
|
|
Nice write-up... its really hard keeping track of new processors.... i feel that the product cycle is really speeding up recently
|
I'm actually thinking about buying some AMD stock, though my main worry is because they've been operating at a deficit the past year. It would be interesting to see what people's thoughts are on the company and its products.
|
I believe AMD has a 6 core proc coming out in 2010
|
It's like a weapons race.
|
thx for the writeup
|
This is AMD's desktop roadmap. Highly relevant.
|
rofl I bet they are all going to use AM3 socket
Except the 2011 cores
|
Somewhat on topic rant:
Anybody who buys AMD/Intel based on the company is retarded. There's a very simple way to buy hardware:
1. Decide on your price range 2. Go to sites like anandtech, techreport, etc and look at reviews of whatever hardware you want (ie CPU). Look for charts/graphics which give a bar graph of the products' performance in that range. There will probably be 10+ cpus in whatever your price range is, and their performance all nicely graphed out. 3. Buy the best performing part in your price range. Alternatively you might find that you can cut off 25% of what you pay for only 5% less performance with a certain part. Etc. The bottom line is you have the performance, you have the price, and you make the call.
This is so obvious but it frustrates me to no end when I see people who post here in hardware / build a computer thread with advice like "Buy intel, they're better" or "Don't get ATI, nvidia is a lot better." Just understand the performance, then look at prices, and go from there.
edit: This wasn't a criticism or response to the OP really but rather to unaware people who might conclude from the OP that you should buy a certain company rather than buying based on facts of the hardware you are paying for
|
yeah, but sometimes there is reasoning behind it. As much as I hate to say it, AMD/ATi drivers sucked whereas nVidia drivers are generally much better. Intel has been dominating the market with it's C2 and i7 lines whereas the Phenom II came out at a horrible time
|
<3 my corei 7, I run crysis full settings/screen and I have never seen CPU usage go above 33%.
|
To me it's amazing how transistor feature size keeps shrinking so fast these days. Since chip size is no longer increasing, the decreasing feature size is the only thing keeping Moore's Law going. Practically every silicon chip on the market, including processors (and except maybe flash memory), can contain more transistors than people really know how to use. Right now everybody's cheating with processors, saying, "Oh, let's just put 2 4 8? cores on there. Since we don't know how to make a better processor with all our available transistors, let's just put the same old crap in duplicate all right next to each other to use up the space and hope people will be happy...oh, and all the huge excess space even after all that, let's make into cache." I mean, sure, extra cache helps for some stuff. And extra cores help for some things.
But I feel while Intel and AMD keep pushing hardware innovation, software to actually fully utilize the hardware is lagging way behind. It's not every task that can be parallelized so easily. People don't really know how to write programs to run in parallel (yes, lots of stuff does it, but it's far from mature).
What I want to know is what happens when transistors of reasonable cost can no longer be made smaller. There are physical limits--you can't make layers less than an atom thick. This really isn't too far on the horizon, so it's something to consider soon.
|
On October 09 2009 13:23 Myrmidon wrote: To me it's amazing how transistor feature size keeps shrinking so fast these days. Since chip size is no longer increasing, the decreasing feature size is the only thing keeping Moore's Law going. Practically every silicon chip on the market, including processors (and except maybe flash memory), can contain more transistors than people really know how to use. Right now everybody's cheating with processors, saying, "Oh, let's just put 2 4 8? cores on there. Since we don't know how to make a better processor with all our available transistors, let's just put the same old crap in duplicate all right next to each other to use up the space and hope people will be happy...oh, and all the huge excess space even after all that, let's make into cache." I mean, sure, extra cache helps for some stuff. And extra cores help for some things.
But I feel while Intel and AMD keep pushing hardware innovation, software to actually fully utilize the hardware is lagging way behind. It's not every task that can be parallelized so easily. People don't really know how to write programs to run in parallel (yes, lots of stuff does it, but it's far from mature).
What I want to know is what happens when transistors of reasonable cost can no longer be made smaller. There are physical limits--you can't make layers less than an atom thick. This really isn't too far on the horizon, so it's something to consider soon. There's a couple proposed routes, the one that I can remember the best is quantum computing. http://en.wikipedia.org/wiki/Quantum_computing I also believe there's another proposed processor that uses lasers somehow.. I'm fuzzy on this one. Edit: This is probably it: http://en.wikipedia.org/wiki/Optical_computer
There are tech articles that come up on this subject every couple of months.
|
I used to keep up with this stuff, but I don't anymore. It was an interesting read.
The AMD/Intel rivalry is a very interesting rivalry. They're both incredibly well managed and well run companies, a perfect example of how competition is good for the consumer.
|
On October 09 2009 13:16 cz wrote: Somewhat on topic rant:
Anybody who buys AMD/Intel based on the company is retarded. There's a very simple way to buy hardware:
1. Decide on your price range 2. Go to sites like anandtech, techreport, etc and look at reviews of whatever hardware you want (ie CPU). Look for charts/graphics which give a bar graph of the products' performance in that range. There will probably be 10+ cpus in whatever your price range is, and their performance all nicely graphed out. 3. Buy the best performing part in your price range. Alternatively you might find that you can cut off 25% of what you pay for only 5% less performance with a certain part. Etc. The bottom line is you have the performance, you have the price, and you make the call.
This is so obvious but it frustrates me to no end when I see people who post here in hardware / build a computer thread with advice like "Buy intel, they're better" or "Don't get ATI, nvidia is a lot better." Just understand the performance, then look at prices, and go from there.
edit: This wasn't a criticism or response to the OP really but rather to unaware people who might conclude from the OP that you should buy a certain company rather than buying based on facts of the hardware you are paying for Not necessarily true, I would be willing to pay a bit extra to a company I trust and that I know the shit works. For example, I have had a terrible experience with amd in the past(Support issues/driver), so I am less likely to purchase their products even if one of their products is a bit cheaper/better.
|
|
|
|