Intel 45nm Processors Waiting to Clobber AMD's Barcelona?

DKC writes "Tech ARP's anonymous source claims that Intel is merely waiting for AMD to release their Barcelona processors before they clobber them with their 45nm die-shrinked processors. In fact, Intel is already producing these 45nm processors at one of their fabs in Arizona. AMD and Intel are in for a long and tough battle ahead. Should be an interesting one though."

Comments

Nice

[Anonymous Coward]

I wish other markets were like this. They compete: I win

Re:Nice

[TheRaven64]

Years ago, the latest and greatest 386,486, Pentium X, [insert latest CPU], the introduction price of a new CPU was about the same cost as today. Have you factored inflation into this? Counting inflation, I think the new ones are a lot cheaper, unlike graphics cards (compare the VooDoo 2 launch price with a new ATi or nVidia GPU). More importantly, though, compare the cost of the cheapest CPU that performs acceptably for most peoples' daily needs. This has been dropping considerably for a long time.

Re:Nice

[TheRaven64]

Really, the lowest-end CPUs that either Intel or AMD are builing today are wildly overpowered for most people's needs Intel and AMD aren't the only people making CPUs. PowerPC, SuperH and ARM chips aimed at embedded systems are powerful enough for most peoples use, and can be had for around $10 in bulk. Of course, they don't run Windows (well, they run Wince...)

Re:Nice

[Kjella]

I wish other markets were like this. They compete: I win

Until it degenerates to one of two:
They compete. You win. One wins. You lose.
They compete. You win. They go "wtf, let's both make a killing". You lose.

Granted the last decade has been great, but face it... in that time Intel has made two terrible strategic blunders, the Pentium 4 and the Itanium. AMD did great innovation with the Athlon and 64-bit processors, and yet AMD has barely passed 20% in market share, has lost the performance crown and is a full generation behind on process technology. It's like watching a chess game where one side is way up in material but has only been moving pieces without striking. There's only so long you can live on miracles from your side and blunders from the other. Sure their prices are competitive but it's because they have to, their margins are down the toilet. If the savings they have to make hit the R&D department, their present is saved but the future is screwed. Not to make it a complete doomcasting but I fear for the future of AMD...

Re:So who will win?

[Anonymous Coward]

Egads man, did you read the article? This isn't about horse racing, this is about computer processors! This is serious business!

Well, there is more than one truth

[imsabbel]

Intel cannot switch their production completely to those parts in a few month. they have huge amounts of 65nm cpus in production, plus they dont have to fab capacity to replace that production at 45nm.

Also, seeing that they already are > 3/4 of the (x86) cpu market, and AMD will only ramp up slowly, Intel would most of all hurt the sales of their own established product lines.

Re:Well, there is more than one truth

[RingDev]

Yes, but by switching to 45nm fabrication they are increasing the yield of their production facility, so they can produce more products for the same amount of raw material. Switching to 45nm chips is in Intel's best interests long term. Short term, selling down 65nm stock and spinning up production of 45nm tooling is in their best interest.

That said, I want AMD to come out with some kick ass chips. If it weren't for AMD forcing innovation down Intel's throat we would still be stuck with that crap they called the Pentium 4. If AMD continues to lag behind in performance and sales, it will only lead to slower development tracks from Intel.

-Rick

Re:Well, there is more than one truth

[sjames]

Yes, but by switching to 45nm fabrication they are increasing the yield of their production facility, so they can produce more products for the same amount of raw material. Switching to 45nm chips is in Intel's best interests long term.

Actually, yield is the percentage of manufactured chips that are functional. When you shrink the die, you tend to get poorer yields and have to use more expensive wafers. Essentially, small imperfections (and all chips have them) that don't matter to 90nm parts will render 45nm parts completely non-functional or sub-standard. Of the chips that DO work, more will end up in the slower bins.

In spite of that, switching to 45nm IS in Intel's best long term interests. As they gain experiance and make process improvements, yields will go up and costs will come down. However, they probably won't benefit in the short term.

Re:Well, there is more than one truth

[dave420]

We don't know for sure AMD's pressure on Intel is driving their innovation. After all, the Core/Core 2 Duo chips were essentially developed from an accidental boost in the performance of their mobile chips. It's a pretty big assumption to pin Intel's progress on AMD.

Re:Well, there is more than one truth

[dave420]

We can sit here all day and just guess what's causing Intel to do what it does, or if we actually wanted to understand the CPU market we could study the facts of the situation. Ass-delving for reasons doesn't help anyone.

This is why I wish AMD was still as competitive.

[etymxris]

When Intel doesn't have to even compete with the latest offerings, business logic rules and technical improvements play second fiddle. Here we have "Why should we release this chip now? The old chips are cheaper to produce and since AMD can't even compete with our current lineup we can keep selling them at the same price, ensuring more profit for us."

So 45nm is not innovating?

[ion]

So 45nm is not innovating? If it was so easy to do, then we would have been there a long time ago. And AMD would have 45nm as well.

I think slashdot choose the wrong Subject, it makes it sound like intel is doing it to be evil. It's much more possible that they are waiting with the release to make more money. Some might think making money is evil, but i dont. I like making money.

If intel has the fastest and lowest power consumption now, and AMD is not a threat, so why release a faster CPU, intel can still make lots of money selling the old. When AMD releases their new CPU, intel has a response ready, meaning intel will make more money.

Intel is in the world to make money, not particular to ruin AMD, how ever if making ruining AMD makes more money, intel will most likely try.

Re:So 45nm is not innovating?

[linuxrocks123]

> Some might think making money is evil, but i dont. I like making money.

Whether making money is evil depends on how you make it. In particular, anticompetitive behavior is not a legal or moral way to make money.

Re:So 45nm is not innovating?

[Xiph1980]

Don't confuse a technological advancement with innovation.

45nm process has nothing to do with innovation. It's just the same technology, the same process, on a different scale.


Innovation is seeing a ball rolling, and making a bearing out of it. the 45nm process opposed to the 60nm process is seeing a 30cm diameter ball, and making a 40cm diameter ball.

Re:So 45nm is not innovating?

[Gospodin]

Yeah, try telling that to the IRS.

Re:So 45nm is not innovating?

[crgrace]

of the most impressive technological feats our society accomplishes on a regular basis. The45nm process has nothing to do with innovation. It's just the same technology, the same process, on a different scale.

What you declare is simply not true.

45nm is the result of a huge amount of innovation, just as 65nm was compared to 90nm. There are a lot of technological hurdles to overcome as the length of transistors are scaled. For example, improved high-k dielectrics are required to increase the channel capacitance and reduce leakage. Improved isolation between devices is required. Tighter tolerances for lithography are needed. Better control of ion implant doses are required. More stable silicides are needed to reduce interconnect resistance. Better drain structures are needed to deal with the increased electric field density in the transistor channels. Improved thermally conductive materials need to be developed because the heat density is increasing. I could go on and on and on. Scaling transistors is onere is a huge financial incentive to do so, and tens of thousands of engineers worldwide are attacking the problems from many angles.

What most people don't understand about device scaling is that it isn't a single problem to be solved. It is a huge number of equally challenging problems spanning multiple engineering disciplines.

Re:So 45nm is not innovating?

[servognome]

Innovation is seeing a ball rolling, and making a bearing out of it. the 45nm process opposed to the 60nm process is seeing a 30cm diameter ball, and making a 40cm diameter ball.

Except making the 30cm ball requires radical advancements in materials and processing. The end product may not be innovative, but the steps to make it are.

Re:So 45nm is not innovating?

[Evanisincontrol]

45nm is the result of a huge amount of innovation, just as 65nm was compared to 90nm. There are a lot of technological hurdles to overcome as the length of transistors are scaled. For example, improved high-k dielectrics are required to increase the channel capacitance and reduce leakage. Improved isolation between devices is required. Tighter tolerances for lithography are needed. Better control of ion implant doses are required. More stable silicides are needed to reduce interconnect resistance. Better drain structures are needed to deal with the increased electric field density in the transistor channels. Improved thermally conductive materials need to be developed because the heat density is increasing. I could go on and on and on. Scaling transistors is onere is a huge financial incentive to do so, and tens of thousands of engineers worldwide are attacking the problems from many angles.

To make a 40cm ball instead of a 30cm ball, you need a bigger cast. You'll probably need more laborers, too. You'll need more materials to make the ball because it's significantly larger, and you'll probably need stronger tools to bring in the material. In fact, you may need to start using a new material altogether, because the old material might not be capable of holding a spherical shape when the diameter is increased 33%.

I just made the same argument for different sized metal bearings as you made for different nanometer threading. It wasn't even hard to do, either, because all I had to do was think about the scale. The GP said it best: "It's just the same technology, the same process, on a different scale." Sure, making 45nm chips is complicated, but complicated does not equal innovative.

Innovation would require taking a route that is completely unlike anything we're seeing today. I can't give an example, because if I could, I'd have a million dollar idea and I sure as hell wouldn't be posting it on Slashdot. The point is, though, that just because moving to 45nm is hard, it doesn't make it innovative.

Re:So 45nm is not innovating?

[networkBoy]

How is Intel being anticompetitive?
I highly doubt they would sit on a process till AMD is out with a new product. The road to marketshare is not to wait for your competitor, it is to get your product out as far ahead as possible. Given the options I believe Intel is likely still working out some non-trivial (i.e. no microcode workaround) issues in the 45nM process before releasing.
Which sounds more plausible?
* Intel sits on a new process, risking sanctions, not making money (actually losing money given the cost of running a fab), just to beat AMD at their launch.
* Intel has some bugs to work out, and does not want to relive a PPro style recall for _any_ reason as that would be disaster.

-nB

Re:So 45nm is not innovating?

[quanticle]

While your argument holds at normal scales, at the nano-scale, even small differences in size can make large differences in the chemical and electrical properties of a substance. Therefore, it is possible that the change in size from 65 to 45 nm could create significant technical challenges that require real innovation to overcome.

Re:So 45nm is not innovating?

[Ryan Amos]

The idea itself is not innovative. The devil, as always, is in the details.

It actually does require innovation; old things have to be done in completely different ways.

What you're saying is that if someone created a space ship that could travel at light speed, that would not be innovative. We already have space ships that go slower than light speed, so it's trivial to scale it up. That's obviously not the case.

Ideas, by themselves, are worthless. The real innovation is how to actually do it. That, combined with the ability to do it, is what makes a technology company money. Nanoscale chip fabrication does in fact require real innovation. Materials at this scale have different properties than they do at a larger scale. If it didn't require innovation, we would have been making 5nm chips for years now.

Re:So 45nm is not innovating?

[HiThere]

I don't count scaling as innovative...but I wouldn't rush to say that Intel's new product with unknown specs isn't innovative. It won't be the scaling, though, it'll be what they do with it.

If they've come up with an efficient way to run multiple cores at full speed (or close to it), then they've been quite innovative. If they've come up with a computer that can run a 2 GHz equivalent and sit on your arm, then it's innovative. If they've come up with a way to build chips 3-dimensionally (rather than 2 1/2), then it's innovative.

There's lots of ways that the product could be innovative. They could have a new microcode language. Nobody would notice, but that, in and of itself, could be quite innovative.

N.B.: Scaling, while not, in and of itself, innovative, frequently enables innovation. Consider the digital wrist watch. (Note also that innovations aren't always improvements.)

OTOH, innovation often isn't noticed as such. Consider those phone speaker/mikes that people are now frequently wearing in their ears. That's innovative, even though it wasn't much of a technical advance (if any).

Literally innovation means the making of something that's new...but different people have different standards for what's needed to count as new. For me, scaling doesn't suffice. It enables, but it doesn't suffice.

Re:So 45nm is not innovating?

[hypergreatthing]

I didn't mean to suggest that shrinking the die from 65nm to 45 is a trivial task. What i am saying is that they're taking an old core (c2d) and shrinking it. Yes, there will be some slight alterations for sse4 performance and whatnot, but it's the same old core.

AMD is not taking the shrinking route (lets face it, they were having lots of problems shrinking 90 to 65, hense the delays) but rather creating a whole new core that's a native quad core cpu.

It's not that hard to see AMD applying a 45nm shrinking process to the k10. Maybe they'll do that by the time that nephilim (whatever intel's new architecture is called) is supposed to come out (end of 2008).

Re:So 45nm is not innovating?

[indil]

In particular, anticompetitive behavior is not a legal or moral way to make money.

Yep, except Intel isn't being anticompetitive by creating superior processors or withholding them for the right time. Many businesses, like movie studios and game publishers, delay releasing their content to maximize their profits. It's not illegal to conspire to put your competitors out of business.

Re:So 45nm is not innovating?

[Technician]

Given the options I believe Intel is likely still working out some non-trivial (i.e. no microcode workaround) issues in the 45nM process before releasing.
Which sounds more plausible?

Much more plausable is they are getting the pipeline filled while the manufactures are finishing evaluating the engineerinng samples (the chips marked Intel Confidential) and building a product. In a new product launch, having a shortage of product is bad. Manufacturing has little surge capacity built-in. It looks like a normal product roll-out to me.They are either aiming for the back to school launch or the fall Christmas shopping season. This is less about hitting AMD and more about beating the January market downturn. Just because AMD is trying to hit the same fall release schedule is not an accurate indicater Intel is doing this to hit AMD. Intel would release this fall if possible regardless of whether AMD was there or not. Check their release cycles. The only times they miss a cycle is when they have problems. They aim for back to school or the Christmas buying season. Early spring launches are rare and are usually covered by press coverage of missed launch dates.

Re:So 45nm is not innovating?

[networkBoy]

Do you know how much it costs to produce a single wafer? 200mm for recipe #xxxx (not telling either) $1070/wafer (assuming lots of 25 wafers), rush job on I helped AMD track down an issue (no I am not going to say which issue) Hint? Etch, doping, poly?

(I think that they had a yield of ~10% Pffft.
Worst product ever:
Codename was SH (or SH II, don't quite remember). A step silicon had a severe issue, yield of 2%. B step solved the issue and projected yield to 95+%. Then the bottom fell out of the dot com market and the projected demand fell so badly that the decision was made to sort and dice the A step wafers and sell the good 2%. That was projected to meet all demand for the lifespan of the product.
Ouch.

-nB

where are the Barcelona benchmarks?

[Dr Kool, PhD]

Why is AMD holding back Barcelona? We're less than a month from launch and there are still no benchmarks. Intel allowed its 45nm chips to be benchmarked and they aren't coming out until November... why is AMD holding back?

IMO this does not look good for AMD.

Re:Whatever...

[Gr8Apes]

AMD has been crushing Intel for many years (on the order of at least 15) I'd say about 4, but hey, I like AMD.

and now that Intel slipped SLIGHTLY in the lead...I personally am a HUGE fan of AMD, and feel that their 64 bit technology is FAR superior to Intel's... Their new technology clearly leads, by 10-20%. However, it is their new technology running against 2004 AMD tech, which should be quite interesting when Barcelona finally ships. As for 64 bittedness, how is AMD's superior to Intel's? I'll admit that AMD's overall CPU design is superior, but the 64 bit extensions?

Plus the fact that memory and core bandwidth is so limited in Intel really makes me wonder how much longer Intel will go before going on-die with their memory controller. Intel will go to an on-board proprietary memory controller/architecture at the end of 2008/beginning of 2009, according to their roadmap.

Re:Whatever...

[Pojut]

I consider myself to be an AMD fanatic (haven't owned an intel-based system since my P166 with MMX) However, there is no way that you can deny both synthetic benchmarks and real-world gaming numbers: Intel's shit is vastly superior in performance.

I'm not saying their design is better or not better, I'm not saying they are doing things smarter or dumber, but the PERFORMANCE of their CPU's (at least in the desktop market...I don't really know anything about the server market) more or less decimates what AMD has to offer.

Why wait?

[ArcherB]

If this article is true, it proves my theory that Intel sits on technology, milking every last dollar from the consumer before releasing something better. This is why I don't buy Intel.

Yes, I know, it's good business and makes the stockholders happy. But as a geek, I'm not into the business side of it. I am into the technology and performance aspect. What if AMD never releases Barcelona? Does Intel never release these new 45nm monsters (or only release them in the quantities already produced, at extremely inflated prices)?

It reminds me of the days of the AMD K6. Intel was "stuck" at 266 Mhz. Reaching beyond that was "impossible". Then, suddenly AMD released a K6 at 300Mhz. Within a week, Intel released the 300 and 333Mhz Pentiums (P-IIs I think). That kind of pissed me off. How much sooner could Intel have released the 300? How much further could they have gone? How many people were forced to pay top dollar while Intel sat back and quietly raked in the cash, knowing that they were selling an inferior product marketed as "the best we can do", when, quite frankly, it wasn't.

This is the action of a monopoly, plain and simple.

Re:Why wait?

[jd]

Nothing unusual for Intel. Transmeta's work on efficiency was bettered by Intel suspiciously quickly and easily. More than a few developments have "appeared" shortly after the competition bettered them in something. There are only two exceptions that I know of. The first was maths co-processing, in which Intel lagged the competition on both price AND performance until they eliminated the entire niche by producing the 486DX. The second was the 32/64 processor architecture. In both cases, it took Intel many years to do anything.

Based on those examples, I would say that genuine progress by Intel is slow, and that any sudden shifts are really the result of having already produced the technology and holding it back.

Re:Why wait?

[rhartness]

Be careful what you wish for. There is a slim chance that Intel could be holding onto this technology because they don't want to be 'anti-competitive'. Let's assume Intel could hypothetically release chips that are twice as fast as anything that is out there right now. What would happen? It could kill AMD if Intel can keep up that technological growth at a much superior rate than AMD. Anti-trust lawsuits would follow.

But, before you call the anti-trust lawyers a bunch of SOB's stifeling technological growth, consider this. If Intel did run AMD out of existance. Intel would no longer have a reason to sink as much money in R&D. They could slack off with only moderate growth and nobody could do anything about it.

I dare say Intel understands very well and they are going to do all that they can to remain #1 in the industry while at the same time avoiding all possible litigation that could be brought against them by the competition.

Re:Why wait?

[CaffeineAddict2001]

or in otherwords:
But mommy!! I want it *NOW*

AMD is in precarious condition

[DrDitto]

I really hope AMD survives because then we are effectively down to a single commodity processor company. But AMD is struggling to survive. I don't care what the fanboys say, just look at their financial numbers. Third quarter in a row with massive losses. Intel opened the door a bit when they faltered with their Pentium4/Itanium strategy. But the door is swinging back shut. Nobody can keep pace with Intel on process technology...they will be ahead of the curve for the forseeable future. AMD is on such a tight-rope that they cannot afford a single mistake or major delay. Since acquiring ATI, nVidia has nearly all of the laptop chipset sales. You wonder if AMD overpaid for ATI. The "wow" factor that came with Opteron is not there with Barcelona. I'm skeptical...

Re:AMD is in precarious condition

[Knara]

Actually if you really look close, cash flow for 3rd quarter was positive. They're not making money hand over fist (actually net profit is down, though down less than 2nd quarter), but they're not going anywhere any time soon (their assets exceed their liabilities by ~5billion dollars, and their cash/cash equivalents + short term investments are around 1.5billion dollars as of end of third quarter). Obviously they won't replace Intel anytime soon, but they're not in dire straits, either.

Re:Where are the new motherboards

[Celandine]

Barcelona is supposed to drop into your existing socket F motherboard with a BIOS update.

What we really need to look forward to...

[xgr3gx]

Is getting away from the almost 30 year old x86 architecture.
Don't get me wrong, I love x86, it has been great, and has adapted amazingly into the most powerful computing the world has ever seen.
But, since most software is tied to x86, we are holding ourselves back from hardware advancements. x86 is loaded with archaic instruction sets for compatibility with Windows code that is based on 16bit DOS code.
I'm not laying out flame bait, this is what I read in an article about the future of processors, and moving to specialized co processors.
It was a cool article, I wish I remember what it was.

Re:What we really need to look forward to...

[durdur]

Well, there were a lot of new architectures for a while that did exactly that. Intel had their own .. it was called Itanium.

While Itanium has a niche market, and SPARC and others are still viable, continued bumps in performance on the x86 stack has caused it to continue to be very competitive for many applications. And compatibility is a wonderful thing. It gets more important, not less, as the number of existing x86 systems continues to grow.

Re:What we really need to look forward to...

[the_humeister]

Ha! x86 compatibility is almost nothing compared with the almost 40 years of legacy code that new IBM mainframes have to put up with.

Re:What we really need to look forward to...

[frieko]

Somebody always makes this comment whenever a CPU story shows up on Slashdot. But it's just not true. As painful as x86 code looks to an engineer, it doesn't really affect processor speed. By the time the code hits the instruction window, it's been mutated into RISC microcode, complete with the huge register bank, ortohogonality, everything. x86 has basically turned into a 'compression algorithm' for the actual machine code.

I think a better optimization would be to replace English with Interlingua. And I think it's about as likely to happen as ditching x86.

Re:This is why I wish AMD was still as competitive

[hobbesmaster]

So, is it better to release something while you're getting low yields and have it show up almost nowhere (the case for the first few months of the core 2 release) or to wait until you can actually have a good number on the shelves, and keep pumping them out?

Re:This is why I wish AMD was still as competitive

[shawnce]

Intel is only starting to pump out 45nm parts (ramping up production lines). They cannot fulfill the needs of their first tear customers yet, so they won't officially release them until they can. I however wouldn't be to surprised if Apple, who has lower unit volumes, picks up the 45nm parts ahead of the big guys as part of an off the price list deal (like Apple currently has with the 3.0 GHz quad core Xeons). Intel so far appears to be ahead of what they originally predicted timeframe wise for Penryn / 45nm.

Also given that Intel is investing heavily in 45nm fabs and they need to recoup those costs by using those fabs. Using a smaller process means they can product more units per die which drives per unit costs down (ignoring capital investment in the plant retrofit). So they aren't just sitting back on profits... when better profits are ahead of them when they start to leverage their investment in 45nm process / fabs.

Re:Whatever...

[ArcherB]

AMD wasn't even remotely competitive until 99 when the Athlon came out

AMD released the K6 in April of 1997. After that, they released the K6-2 and even the K6-3 right before the Athlon came out. The K6 competed very well against the PII. The K6-2 and K6-3 competed will against the P-III until the Athlon was released. Well, you know the story from there.

I try to buy AMD exclusively and this article is a fine example why. I won't buy from a company who is holding back their best from me in order to milk every last hard earned dime they can from me. I'm sure AMD would do the same if they could, but they are not, so I buy from them.

Third Player Will Steal the Gold

[MOBE2001]

AMD and Intel are in for a long and tough battle ahead. Should be an interesting one though.

While these two Goliaths are locking horns and fighting over soon-to-be-obsolete technology, a third player will sneak behind them and steal the pot of gold. Let's face it. CPU architecture is due for a radical change. The computer world is going parallel and the old algorithm/thread paradigm is showing its age. There's a sweet scent of revolution in the air. Who will be the leader of the next revolution? Sun, IBM, Tilera? We'll see.

Re:Third Player Will Steal the Gold

[jrwr00]

I see IBM taking the lead on this one, every looked at what processors the gaming consoles are using? IBM PowerPCs

CELL Processors FTW

Re:$$$ money

[Bazar]

Its only short term, that lower prices are good.

Long term, if AMD doesn't make a profit, and eventually liquidates, Intel will be the only remaining manufacturer of x86 CPUs (At least the only one able to meet demand, at cost effective prices)

They'll have an effective monopoly, which means without a doubt, Intel will raise their prices... Its not like a competitor can spring in to compete. The capital required, both in plant, and research, to enter such a manufacturing market is mammoth, how many billion have AMD invested in their own manufacturing plants, let alone research?

If AMD dies, the only thing that could keep Intel in check dies, and with it, fair prices.
The same can be said with nvidia, since ati's fate is tied to amd.

Welcome to capitalism

[nobodyman]

AMD would do exactly the same thing if the situation were reversed. In fact, they did just that back in the Pentium 4 days. This underscores why competition is such a good thing.

Re:A little dissappointing

[Bert the Turtle]

ATI still have not released their drivers as open, and the closed ones are complete ass.

The open ones are severely limited, no surprise given the lack of help from ATI.

Intel, on the other hand, has excellent open source graphics drivers.

Right, AMD is not competitive.

[Inoshiro]

A friend of mine and myself both upgraded our desktop PCs. They chose an Intel Core 2 Duo because "Intel wins in all the benchmarks." I bought AMD instead.

Their system is based around a E6600 ($270 at the time), mine is based around an X2 3600+ 65nm ($75 at the time). Their system has 2gb of RAM, mine has 4gb of RAM. My motherboard (with nVidia chipset) was $80 cheaper than their P5B Deluxe. Overall my system was $400 cheaper -- with double the RAM. I go into my Asus M2N-SLI Deluxe BIOS and change the clock rate of my CPU from 1.9Ghz to 2.4Ghz with no ill effects and get the same # of 3D Marks as them because I have the same kind of video card (8600 GTS PCI-E). They're happy because they bought "performance" (as sold to them via Intel marketing), and I'm happy because I bought the same performance (as proved by benchmarks) for a lot less.

What's the lesson? For my workstation use in Linux compiling and rendering and working with large images, 4gb of RAM that run at the same speed as L2 cache (thanks to AMD's integrated memory controller) beats the piss out of that Intel setup (which has much lower memory bw and also half the RAM). For gaming use, I get the same # of 3D Marks and similar performance because an Intel 2.4Ghz CPU and an AMD 2.4Ghz CPU happen to be within a few % of each other on the same video card (which is the true bottleneck; don't lie to yourself and say it's that CPU that's 14-18x faster than RAM).

I got the same performance for $400, but with more RAM. My CPU was $190 cheaper. My motherboard was also cheaper. In a lot of ways, it reminds me of all those people who rave and Intel Xeon power consumption, and ignore the fact that those require power-hungry FB-DIMMs and have chipsets that dissipate more power than the difference in CPU watts.

Your computer it NOT just a CPU -- it is an entire system that must be balanced. Go watch a Lotus Elise race some muscle-bound 7.7 litre Mustang and see which is a better balanced car. Clearly the Lotus is just as not competitive with that Mustang because it has a much smaller engine! Clearly that statement is just as true as AMD not being competitive with Intel.

Re:Right, AMD is not competitive.

[Hoplite3]

Yeah, but performance is not the top concern anymore. The three big questions are
(1) How much power does it draw?
(2) How much heat does it make?
(3) How loud is it?

The market is worried about how "livable" computers are. That's why laptop sales have grown so much.

AMD x2's are good chips (I have one and like it fine), but the market will turn on efficiency questions not performance.

Re:Right, AMD is not competitive.

[Anonymous Coward]

So you overclocked and they didn't? Somehow you think you can compare the systems?

Re:Right, AMD is not competitive.

[MojoStan]

A friend of mine and myself both upgraded our desktop PCs. They chose an Intel Core 2 Duo because "Intel wins in all the benchmarks." I bought AMD instead.

I knew right away, from your tone and your friend's quote, that you would buy for price/performance and your friend would buy for performance only. An unfair, biased comparison would follow. Did your friend know he or she was competing in a price/performance contest?

Their system is based around a E6600 ($270 at the time), mine is based around an X2 3600+ 65nm ($75 at the time).

"At the time" is not the "current lineup," which the GP was referring to. Way to go there, comparing a mid-range (at the time) Intel CPU to a low-end (at the time) AMD CPU. Don't mention that $270 currently gets you Intel's E6850 (3GHz, 1333MHz, 4MB) and almost gets you Intel's Quad Q6600 ($280). $75 currently gets you Intel's (Core 2 based) Pentium Dual-Core E2140.

Their system has 2gb of RAM, mine has 4gb of RAM.

RAM costs the same for both platforms.

My motherboard (with nVidia chipset) was $80 cheaper than their P5B Deluxe.

I'm sure "they" could have bought a significantly cheaper motherboard. Currently, an ASUS P5NSLI motherboard (with nForce 570 SLI Intel Edition chipset) is $45 cheaper (at Newegg) than your ASUS M2N-SLI Deluxe (with nForce 570 SLI AMD Edition chipset).

Overall my system was $400 cheaper -- with double the RAM.

You bought a low-end CPU and a mid-range motherboard. Your friend bought a mid-range CPU and a high-end motherboard. You also bought at a time when AMD drastically slashed prices in response to Intel kicking their arse in the mid-range and high-end. At the time, AMD was only competitive in the low-end (where Intel still only offered Netburst CPUs).

I go into my Asus M2N-SLI Deluxe BIOS and change the clock rate of my CPU from 1.9Ghz to 2.4Ghz with no ill effects and get the same # of 3D Marks as them because I have the same kind of video card (8600 GTS PCI-E).

Yeah, that's a fair comparison. Overclock your low-end AMD CPU and compare it to a mid-range Intel CPU at stock speeds.

They're happy because they bought "performance" (as sold to them via Intel marketing), and I'm happy because I bought the same performance (as proved by benchmarks) for a lot less.

If they're happy, then they probably didn't know they were competing in a price/performance contest with you.

For my workstation use in Linux compiling and rendering and working with large images, 4gb of RAM that run at the same speed as L2 cache (thanks to AMD's integrated memory controller) beats the piss out of that Intel setup (which has much lower memory bw and also half the RAM). For gaming use, I get the same # of 3D Marks and similar performance because an Intel 2.4Ghz CPU and an AMD 2.4Ghz CPU happen to be within a few % of each other on the same video card (which is the true bottleneck; don't lie to yourself and say it's that CPU that's 14-18x faster than RAM).

Today, a $280 Quad Q6600 on a $130 ASUS P5N-E (nForce 650i SLI) beats the piss out of an equivalently priced AMD workstation in compiling, rendering, and large images. If you're willing to risk stability and reliability by overclocking (like you did), then a $90 Pentium Dual-Core E2160 can be overclocked to 3.4GHz (according to X-bit Labs) and beat the piss out of any Athlon 64 X2 system with the same RAM, GPU, and class of motherboard.

I got the same performance for $400, but with more RAM. My CPU was $190 cheaper. My motherboard was also cheaper.

Your friend did not need to spend so much on his/her motherboard. Your friend did not overclock. Today, a cheaper system built around an overclocked Pentium Dual-Core E2160 and an nForce 570 SLI moth

Re:Right, AMD is not competitive.

[zippthorne]

Ugh, it bugs me that they always put those benchmarks on linear scales. Anything less than a factor of 2 (ok, maybe factor of square root of 2) difference is just noise. At the extreme end of it, people are stretching to come up with a reason they like their slower thing better, but speed improvements are like f/stops in lenses. A few percent is barely noticeable.

His comparison was flawed, but the difference in their unmodified machines would not be so much as to justify an additional $400.*

In fact, this is why I always step back my performance requirements to roughly the middle of the pack. The savings you get there can enable you to completely replace your entire machine twice as often as your state-of-the-art friends. If you time your purchases right, you can barely spend one year out of four with a machine inferior to your friends for the same money

*to a sane individual. Hardcore gamers excluded.

If you like spending $250+ on a CPU, sure.

[Inoshiro]

When I looked at upgrading my system, I had a choice. The Core 2 Duos that weren't crippled and had a proper amount of L2 cache started at $240. The AMD X2 systems with built-in memory controller and decent amounts of L2 cache started at $75.

Right now on any web site, you can order a X2 CPU with full dedicated L2 cache per core for around $70. The cheapest Core 2 Duo is the E4300 at $150. That has a bottlenecked 800Mhz FSB, not a fancy 2.0Ghz hypertransport bus like the X2. To get a 1066Mhz FSB C2D requires you go up to $190 or so.

Intel motherboards seem to require a premium as well. nVidia can make AM2 chipsets with firewire, dual ethernet, onboard 7.1 audio, multiple SATA and eSATA connectors, etc, for roughly $100 less than then equivalent Intel chipset board. Is that because Intel wants more $$ for its chipset licences?

So... when you do get this same base performance, it comes at a price. Honestly, you would be better served by getting an 8800 instead of an 8600 GeForce for the difference in CPU and motherboard costs. Plus, those SLI motherboards for AM2 are around $150 vs. the $220 + for Intel ones.

Re:If you like spending $250+ on a CPU, sure.

[Wavicle]

Right now on any web site, you can order a X2 CPU with full dedicated L2 cache per core for around $70. The cheapest Core 2 Duo is the E4300 at $150.

Newegg has the E4500 for $146. That comes with 2MB of shared L2 cache, which is twice the combined cache of the Brisbane. In a single-threaded game this means that most of the L2 is going to be used by one processor for the game giving the CPU access to almost 4x as much. Is the shared cache a problem?

That has a bottlenecked 800Mhz FSB, not a fancy 2.0Ghz hypertransport bus like the X2.

Let me start by saying that a dedicated cpu memory controller plus high-speed chip-to-chip interconnect is the way to go. Having said that, this comparison always annoys the hell out of me for a couple reasons:

1) FSB is wider than hypertransport.
2) Hypertransport data must be packetized

The fact that 800 < 2000 does not mean FSB < HT. There are other reasons, this one is naive and wrong.

Re:If you like spending $250+ on a CPU, sure.

[MojoStan]

The Core 2 Duos that weren't crippled and had a proper amount of L2 cache started at $240. The AMD X2 systems with built-in memory controller and decent amounts of L2 cache started at $75. If the "crippled" Core 2 Duos performa as well or better than "non-crippled" AMD X2 CPUs, then why would it matter if they had less L2 cache? Every performance review I've seen shows that Core 2 Duos with 2MB of shared L2 cache or even 1MB of L2 cache (Pentium Dual-Core E2xxx series) perform very well.

Right now on any web site, you can order a X2 CPU with full dedicated L2 cache per core for around $70. The cheapest Core 2 Duo is the E4300 at $150. That has a bottlenecked 800Mhz FSB, not a fancy 2.0Ghz hypertransport bus like the X2. To get a 1066Mhz FSB C2D requires you go up to $190 or so. For $75, you can buy a Pentium Dual-Core E2140 which performs very well against the Athlon X2 3800+ according to X-bit Labs (they go back-and-forth). Sure, today you can get an X2 4200+ for around $75, but AMD slashed prices in response to the Pentium Dual-Core E2xxx series.

Intel motherboards seem to require a premium as well. nVidia can make AM2 chipsets with firewire, dual ethernet, onboard 7.1 audio, multiple SATA and eSATA connectors, etc, for roughly $100 less than then equivalent Intel chipset board...

...those SLI motherboards for AM2 are around $150 vs. the $220 + for Intel ones.

The ASUS P5N32-SLI Premium/WiFi-AP uses the nForce 590 SLI Intel Edition chipset. It's $125 at Newegg.

Intel's Not Waiting for Anything

[Glasswire]

There seems to be a conspritorial thread running though alot of these comments which seems to assume that Intel already HAS 45nm processing up and running in volume and is deliberately holding it back just to make AMD looks bad. This is ridiculous for several reasons:

1) If Intel could produce volume 45nm right now it would - better chips, cheaper to make, higher performance, higher margin on the best ones - why would they hold back?
2) Even if Intel just cared about humiliating AMD, it would do it much more thoughly if Intel could bring out the 45nm stuff BEFORE Barcelona even ships. Believe me, if Intel could do that, it would.
3) Anyone who has any idea what's going on in the industry knows Intel is putting massive effort behind getting out the 45nm technology as soon as possible. There is NO financial upside in living with older process technology any longer than you have to. (Unless you're AMD and you don't have the latest process technology and have to bring out your flagship quad core on old 65nm process)

So, in summary, 45nm stuff may well give Barcelona a run for it's money, but there's no way Intel is holding it back for dramatic effect.

Re:Intel's Not Waiting for Anything

[smellsofbikes]

Maybe the CPU business is different than what I do (design analog power chips.) But I see clear reasons to believe, not in conspiracy, but in profit.

1. You spend, let's say, $10-100M to design a chip and its test systems, get it through quality and reliability testing, and into production. That's a one-time investment. From then on, every chip you make costs a few pennies of silicon, and a few dollars of testing, offsetting that enormous initial investment. You'd really like to, y'know, profit. The more chips you sell, the more you amortize that initial investment. Here's a strategy to put you out of business: make an incredibly fast new chip, costing $10M in R&D and fab, sell 10,000, then the next week make another chip, even faster, such that everyone buys it instead of the previous one. In order to make back what you've spent you have to charge a stupid amount of money. A successful strategy is to forecast how many you think you'll sell, design to see if you can meet that forecast, start selling them for a little more than you need to make your profit margin if you sell the amount you want to, then slowly cut prices down so that the product goes end-of-life at some point after it's paid itself back.

2. You're right. Intel isn't in the business of humiliating AMD, they're in the business of making a profit. There's some profit to be made in reputation, but given that computer buyers mostly fall into two camps, those who don't know what a CPU is and those who buy the highest-performing chip, regardless of who made it, I doubt humiliation is a good return on investment.

3. Companies spend a lot of money on research on where they're going, so they won't be surprised when they get there, but it's a much better idea to figure out the problems in small-scale, and then go to full production. The longer you can run on your old, already-paid-off fab equipment, the longer you can delay buying new equipment. There's a huge financial upside to not buying new equipment, and you know already that whatever you do, whatever equipment you use is going to have 70-80% utilization, so why not put all that hurt on old equipment, for as long as possible?

It's not being held back for dramatic effect. It's being held back because if you can sell the same thing tomorrow that you sold yesterday, you save money on developing the processes to bring the new stuff into full production.

AMD still has hypertransport and build in ram con.

[Joe The Dragon]

AMD still has hyper transport and build in ram controller and in mulit cpu setups it is better intel haveing 1FSB per cpu is better then the past for them but is still not as good.

Also AMD has more and better chipsets for there mulit cpu system with more pci-e lanes and DDR2 or DDR2 ECC ram.

And on the desktop side you can get a High end Nforce 590 board for the same price as a lower end intel board that does not even have TCP/IP Acceleration like the 590 and few other lower end nforce chips do have.

Score per buck

[Kokuyo]

I don't quite understand why people link AMD's demise with benchmarks all the time. That's just a small part of the whole truth.

Fact is, there are several types of clients in the computer market. Some are early adopters/hardcore users. They buy whatever earns them the highest benchmark scores. They are willing to pay 5 grand plus for a system just to have one sick fucker of a computer under their desk. I'd say they're a minority.
Then there's those who listen to the commercials. I don't know how it is in America but in Switzerland I have yet to see a tv commercial for AMD CPUs. So who's wondering why people still think there's only one CPU manufacturer?
A lot of people who know AMD isn't just a cheap chinese copy that will probably have trouble adding two and two in calc.exe will want to build a somewhat up to date system they can rely on to do its job for the next two or three years. They use some graphics tools, they run a few games, they browse the web, the skype from time to time and they watch their porn. Those people don't need the V12 1000 hp equivalents in the computer world. They need a midrange machine with reliable hardware. Overclocking? What for?
People like that, which includes me, buy what gives them a balance of most bang and reliability for the buck. I'll admit, I deviated from that path with my current system. I am running a Core2Duo. Why? Because AMD couldn't sell me a CPU when I needed one. And I am actually happy with my Intel. Do I see more power? Hell no. My stuff runs. Command and Conquer Tiberium Wars runs. World of Warcraft ran... until I got fed up with it.

I don't care for labels. I'll select the third or fourth newest chip unless it's only like ten bucks to the next faster one. I'll select like two gigs of memory upwards. I'll select a board that will work with my watercooling and be of agreeable quality. I'll select a somewhat actual but cheap video card. Somewhere in the middle of the range of available cards from my vendor of choice.

The point where about any PC made of parts from the last two years would run everything I need has been crossed years ago. Today reliability, noise, power consumption and such are factors... and the price. And I don't see Intel beating AMD in that regard anytime soon.

Re:Whatever...

[UncleTogie]

The K6 debuted in the days of the ORIGINAL Pentium, ergo comparing it to the P4 would be akin to comparing the K6 to the 8086.

At this link, you can check the debut date for the K6. Note at the top of the article that it quite clearly states that it was designed to compete with the original Pentiums.
At this link, you can check the debut date for the P4. Note that the P4 debuted *3* years *after* the K6 series...

If you believe that AMD had the foresight, manpower, and devel skills to beat a processor not even OUT for 3 years thereafter, it's time to adjust the tin hat, sir/ma'am.

AMD still much better for RAM-intensive jobs

[Phatmanotoo]

AMD is still way ahead of Intel when it comes to jobs which combine these two things:

• floating point (general purpose, not just video-encoding primitives)
• use of lots of RAM in "non-cache-friendly" ways

This is specially noticeable if you have a multi-core, multi-chip machine and start firing up several compute intensive jobs at the same time. If you don't believe me, try running several instances of the STREAM benchmak and see for yourself. Or go to www.spec.org and look at the SPEC CFP2006 Rates numbers (see this for instance, and compare).

Barcelona is supposed to increase scalability even a bit more from the current Opterons. Let's see if Intel comes up with something in this department. So far they have not, and that's why most scientific computing in the last 5 years is done on Opterons.

These guys are idiots

[adisakp]

From the article:

Until AMD launches the Barcelona, Intel have no reason to start selling 45 nm processors.

Umm, that's not true at all. Here are some reasons:

1) Lower cost - you get more 45nm CPU's per wafer than 65nm CPU's so they cost less if you have similar yield ratios.

2) Lower power systems are attractive now to large purchasers. On a system level, AMD is very competitive with Intel (and sometimes ahead of Intel) on performance per watt. This is very important to companies with huge server farms.

3) Higher single-threaded performance per core. The 45 nm shrink will allow them to run cooler and at higher clock speeds thus producing high-end high-margin CPU's that gamers and performance junkies crave.

4) The way to crush your opponents isn't to let them catch up before you move forward. Have you ever seen someone in a relay race wait for their opponent simply because they know the next runner on their own team is fast? You have to get ahead and stay ahead as far as possible. If you even let them have the appearance of catching up, you won't maintain your image of indomitable superiority.

Overclocking

[dreddnott]

Cyrix 200MX: overclocked to 83MHz bus, "never had a problem"
AMD K6-2 350MHz: overclocked to 400MHz, "never had a problem"
AMD K6-III 400MHz: overclocked to 450MHz, "never had a problem"
Didn't overclock the Athlon Slot A 700MHz, T-Bird 1.4GHz, either of my two Athlon XP 2500+ systems, or the Athlon 64 3400+, didn't get good steppings, so the payoff wouldn't have been much. I was also betrayed by a series of poor-quality VIA-chipset motherboards, I don't buy those anymore.
Intel Core 2 Duo E6600: overclocked to 3.0GHz, 1333MHz FSB, 667MHz RAM, stock voltage, no problems so far! This CPU is still faster than any retail Intel CPU available. I built a similarly overclocked system for a friend the next month, but with an 8800GTX instead of a 7950GX2. I'm quite jealous.

You have to try pretty hard nowadays to blow a system up. At the time I built the Intel system I calculated a significant savings over buying the 2.93GHz X6800 even after including the third-party cooler and high-end motherboard, and I ended up with a much, much faster FSB in the process. I may have even been able to get away with the stock heatsink/fan, but I'll never know now. The system itself is faster, cooler, and quieter than my non-overclocked Athlon 64, and I'm not about to mess with success.

People who have problems overclocking simply need to bone up on their processor-fu by checking out processor revisions and steppings, the previous successes and failures of others with identical CPU, and also by optimising case airflow (liquid cooling is a net loss in my opinion due to cost of entry, long-term maintenance, and the consequences of cooling system leakage or pump failure).