AMD's 65nm chips

[dcv]

http://www.engadget.com/2006/12/05/amd-breaks-out-the-cigars-for-its-very-first-65nm-chips/
http://www.extremetech.com/article2/0,1697,2068252,00.asp

In news that is probably exponentially more exciting than AMD's lackluster 4x4, AMD has just released their 65nm Athlon 64 X2s, and they're priced as low as $169. AMD is claiming superior power management and energy efficienty to Intel's Core 2 Duo (at least when idle).

The real questions:

a) Will these chips be able to keep pace with the Core 2 Duo with its shorter paths, or is lower power consumption this new processor's only trick?
b) Will AMD be able to catch up to Intel's 45nm "Penryn" fast enough?

Either way - for $169, I still think the X2 4000+ is a great deal.

[bgwinkel]

Meh, this is like 90nm Prescott to 65nm Cedar Mill.
a) Will these chips be able to keep pace with the Core 2 Duo with its shorter paths, or is lower power consumption this new processor's only trick?

No and yes.
b) Will AMD be able to catch up to Intel's 45nm "Penryn" fast enough?
45nm for AMD comes out in 18 months...Intel next year, so no.

[amccabe]

I will be surprised if these do not overclock beyond 3.0 on air

and for the price, these should be a solid choice for people who don't need the best

[bgwinkel]

Doesn't the E6300 still beat the X2 4400+?

[dcv]

Doesn't the E6300 still beat the X2 4400+?

That's the 90nm X2 - the 65nm have yet to be benchmarked by any of the major sites. The E6300 will still probably come out the victor, but I'll be surprised if the 65nm architecture doesn't yield measurable improvement over its 90nm counterpart.

If AMD's energy numbers are half a good as they're claiming, then I think these chips may be of greater use to the general populous than one might give them credit for - these chips idle at 3.8W, vs. about 15W for a Core 2 Duo. I don't know about you, but my computer idles more than it burns, and I think that a few benchmark points for the times I do need to do heavy number-crunching may be worth the sacrifice if the processor is this energy efficient the rest of the time.

I think this may AMD's new marketing strategy, and I think it's a damn good one. If they can't beat Intel in benchmarks, beat them in other places where it matters - power consumption, and value for price.

[bgwinkel]

Isn't it a dumb shrink though? (dumb shrink is an industry term meaning a straight shrink without updating the design)

The new FX-74 processors are 65 nm parts, and Intel still holds the power/performance crown by almost a factor of 2.

http://techreport.com/reviews/2006q4/quad-fx/index.x?pg=13

and IIRC, all the theoretical consumption means is they are now on par with Intel, not better. Cool 'n' Quiet does a better job in theory, but in practice, it has to deal with the higher power consumption of the motherboard as well, which uses more power than the equivilent Intel board, and thus they are equal.

AMD and Intel use different power comsumption models for classifying their chips. Intel's give you averages and such, so their high is lower then the actual highest and their low is higher than their actual lowest. AMD gives actual max and mix values, so their high is higher and their low is much lower.

What we really have to do is hope they can actually get 65 nm parts in volume quickly (things aren't looking good as of right now) and then they need to get K8L out the door, then they'll be in competition with Intel...which at that point will drop to 45 nm and AMD will be screwed. Their current batch of chips can clock really, really well. It's like the days of Northwood where every chip could hit 3.0 GHz, they can't seem to make a Conroe chip that can't hit 2.8 Ghz (and that's the GHz of their current extreme edition)

[dcv]

Isn't it a dumb shrink though? (dumb shrink is an industry term meaning a straight shrink without updating the design)

The new FX-74 processors are 65 nm parts, and Intel still holds the power/performance crown by almost a factor of 2.


http://www.u.arizona.edu/~dcv/fx74.png

Actually, the FX-74 is still a 90nm part. These new chips are in fact AMD's first 65nm parts - so how beneficial the manufacturing process is (if at all) has yet to be tested - but compared to 90nm, 65nm is, from a purely physical point of view, a much shorter distance for electrons to travel. It is almost guaranteed to have some sort of performance boost based on that alone - even if it is a "dumb shrink."

With regards to power consumption, one also has to consider that the integrated memory controller on AMD's chips reduces power consumption on the northbridge. I've never heard that AMD motherboards consume more power than an equivalent Intel motherboard - in practice, I'm sure they're relatively on par, and in theory, AMD should be consuming less energy.

[dcv]

which at that point will drop to 45 nm and AMD will be screwed. Their current batch of chips can clock really, really well. It's like the days of Northwood where every chip could hit 3.0 GHz, they can't seem to make a Conroe chip that can't hit 2.8 Ghz (and that's the GHz of their current extreme edition)

Both AMD and Intel have their road maps laid out - it won't be a surprise when Intel drops 45 nm chips, and AMD drops theirs a few months later.

I would not go so far as to say that AMD are screwed - historically, they've often lagged slightly behind Intel. Suffice to say that they've been further behind than this before - and look where they ended up.

If anything is going to hurt AMD, it's probably going to be the fact that they're jumping sockets like hotcakes - my socket 754 was EOL'd pretty early, and you don't see too much going on with S939 (have those been EOL'd too?) - right now it's all about Socked AM2, and in a year, it'll be all about Socket 1207. Intel? Well most LGA775 motherboards, even older ones, support even the fastest Core 2's, often with as little as a BIOS flash.

As far as actual numbers go when it comes to both power consumption and benchmarks - I'll hold off final judgment until I see actual tested numbers from third parties.

Oh, and one more quick correction - the clock speed on the Extreme Edition Conroe is 2.93 GHz. I'm pretty sure Intel could hit that last 70 MHz if they felt like it (especially considering that the Woodcrest Xeon is built off of the same architecture and has hit 3.0 GHz).

[bgwinkel]

Both AMD and Intel have their road maps laid out - it won't be a surprise when Intel drops 45 nm chips, and AMD drops theirs a few months later.
Intel has 45nm roadmapped 9-12 months before AMD does.
would not go so far as to say that AMD are screwed - historically, they've often lagged slightly behind Intel.
And historically, there was talk of AMD declaring bankruptcy because they were losing so badly in 2002-2003, but then AMD64 saved them. At this rate, they need another AMD64-like release within a year or the 45-nm process shrink is going to kill them because it's so many billions of dollars.
If anything is going to hurt AMD, it's probably going to be the fact that they're jumping sockets like hotcakes
But that's directly linked to...
One also has to consider that the integrated memory controller on AMD's chips reduces power consumption on the northbridge. I've never heard that AMD motherboards consume more power than an equivalent Intel - in practice, I'm sure they're relatively on par, and in theory, AMD should be consuming less energy.
The reason they have so many sockets is that integrated memory controller. Every time they upgrade the memory controller, they have to revamp everything. They don't have to change sockets, but its certainly easier that way, and everyone has to buy a new motherboard anyway.

As for the 65nm thing, I read that off of some tech website in French, so I guess they were wrong (darn French!). However, they still eat a freaking ton of power. The higher consumption comes from their use of Nvidia chipsets. Nvidia's chipsets are really power hungry. The new 680 eats power like no other.

I'm pretty sure Intel could stabilize that last 70 MHz if they felt like it.
And the clock speed on my E6600 at stock is 2.39, and the clock speed on my 2.8 Northwood is 2.78. They could be saving it as a marketing gimmeck at 3.0, they've done it before. And the 2.8 comment was supposed to be 2.9, I'm not used to a laptop keyboard. So sorry about that.

[dcv]

Intel has 45nm roadmapped 9-12 months before AMD does.

Not disputing that. It hurts me, trust me - but I don't think the situation is as dire as you think it is.

At this rate, they need another AMD64-like release within a year

I think the deal they made with Dell is going to keep them afloat for a while yet (not to mention acquiring ATI, who, last I heard were still doing quite well). Besides, in the value realm, AMDs still sell like hotcakes. Just check out the sub $500 section of your local big box - having worked in one or two, I can say that AMD's Semprons and lower end A64s still power most of those HPs and eMachines.



The reason they have so many sockets is that integrated memory controller. Every time they upgrade the memory controller, they have to revamp everything. They don't have to change sockets, but its certainly easier that way, and everyone has to buy a new motherboard anyway.

I suppose the memory controller is a mixed blessing, then, but at least for now the market has stabilized on DDR2 - so the jump from AM2 to 1207 bugs me.


As for the 65nm thing, I read that off of some tech website in French, so I guess they were wrong (darn French!). However, they still eat a freaking ton of power. The higher consumption comes from their use of Nvidia chipsets. Nvidia's chipsets are really power hungry. The new 680 eats power like no other.

Even so, the chip you're talking about (the FX-74) is not the chip that I'm talking about (these new 65nm chips). I don't think it's fair to judge the power consumption of a new chip based on last month's model (man, has it really only been that long?).

And yeah, I bet the Nvidia 680 chipset eats power - it's DESIGNED to do that. It's an enthusiast's chipset.

Nvidia is not the only chipset manufacturer, though (and I'm willing to bet the lower end Nvidia chipsets aren't so power hungry, either) - VIA, SiS, ATI, etc. all make chipsets as well. I'd like to know how they stack up, power-consumption wise (though I'm too lazy right now to actually go and look - I hate when I do that!).

My personal theory is that, since now they own ATI (who can better tailor chipsets for them since they can work more closely together), AMD can do what intel does - develop not just a processor, but a platform that is AMD branded and engineererd from top to bottom for best performance and power consumption.


And the clock speed on my E6600 at stock is 2.39, and the clock speed on my 2.8 Northwood is 2.78. They could be saving it as a marketing gimmeck at 3.0, they've done it before. And the 2.8 comment was supposed to be 2.9, I'm not used to a laptop keyboard. So sorry about that.

Fair enough ;)

[rjhill1]

And historically, there was talk of AMD declaring bankruptcy because they were losing so badly in 2002-2003, but then AMD64 saved them.

Sauce, please. Considering this was near the peak of their dominance (yes, that's right, dominance), I *highly* doubt this would've been the time they would've had to file bankruptcy. Not only did they release the first consumer-level 64-bit proc for enthusiasts around that time, they had been beating Intel over the head for a while with the Athlon, since it was a cheaper proc and could outperform the p4. Not to forget the aforementioned fact AMD had been playing far behind Intel for most of their history, usually holding steady at around 5% market share.

Essentially we're seeing Intel's prediction come true of investing their time wisely in the Yonah series, and it's paying off in large dividends. Which is what they learned from AMD those years ago. It's probably to a resonant dominance now; one will trump the other by re-thinking the processor, then will themselves be trumped with another re-think.

[dparm]

Intel will just release something better the following month...this is how it works with competing products.

Intel releases A, AMD waits a month and releases B which is 10% better.
Intel waits a month and releases C, which is 10% better still.
AMD waits a month and releases D, which is 10% better still.

Companies intentionally hold-out on releasing products just to one-up the competition. Look what happened with the original XBOX and PS2...

[bgwinkel]

And yeah, I bet the Nvidia 680 chipset eats power - it's DESIGNED to do that. It's an enthusiast's chipset.
Yes, but it's a ridiculous amount of power, and their current chipsets (even the lower end ones) eat power as well. Intel chipsets are much more conservative (with the exception of some of the P965s for some odd reason)
Considering this was near the peak of their dominance (yes, that's right, dominance)
Er...That was the time of the Northwood core for Intel, which was price/performance better and overall faster then the Athlon XP line. It wasn't until the release of Prescott and AMDs newest interations of Athlon 64 that AMD finally took the crown, which didn't happen until early-mid 2004. Moreover, AMD never 'dominates' Intel in the market, they simply don't have the fab space, they'll dominate in performance, but not in the marketplace.
Essentially we're seeing Intel's prediction come true of investing their time wisely in the Yonah series, and it's paying off in large dividends. Which is what they learned from AMD those years ago. It's probably to a resonant dominance now; one will trump the other by re-thinking the processor, then will themselves be trumped with another re-think.
Except everything we have now has evolved directly from the P6...P7, Netburst, was the only exception, and we know how well that scaled...

Intel certainly didn't hold-out on the Conroe release...their shareholders needed that ASAP. Intel wasn't in huge trouble, simply because of the breadth of their portfolio and their influence in the corporate sector. AMD, on the other hand, really only had one major income source, and things like this happened:

SUNNYVALE -- April 17, 2002 --AMD today reported sales of $902,073,000 and a net loss of $9,163,000 for the quarter ended March 31, 2002.

SUNNYVALE, CA -- April 16, 2003 --AMD (NYSE:AMD) today reported sales of $715 million and a net loss of $146 million for the quarter ended March 30, 2003.

They were reporting net loss after net loss. No business model can survive that forever.

[rjhill1]

Well of course, but what I was getting at was the fact that that is actually the case now in processors. I mean previously, the market was EXTREMELY lopsided toward one or the other. First it was "Intel comes out w/ chip A, AMD makes a cheap knock-off of it. Intel comes out w/ chip B, AMD makes a cheap knock-off of it." Then it was "AMD comes out with several procs that are faster and cheaper than Intel's, Intel makes a cheap knock-off of their own previous proc to try and catch up." But now you have a situation where, as you said, they can one-up each other, instead of merely playing catch-up. I am very in favor of this situation, as when "Product F" comes out, I buy "Product B" at a quarter of its original price :P

[rjhill1]

Yes, but it's a ridiculous amount of power, and their current chipsets (even the lower end ones) eat power as well. Intel chipsets are much more conservative (with the exception of some of the P965s for some odd reason)

Er...That was the time of the Northwood core for Intel, which was price/performance better and overall faster then the Athlon XP line. It wasn't until the release of Prescott and AMDs newest interations of Athlon 64 that AMD finally took the crown, which didn't happen until early-mid 2004. Moreover, AMD never 'dominates' Intel in the market, they simply don't have the fab space, they'll dominate in performance, but not in the marketplace.

Except everything we have now has evolved directly from the P6...P7, Netburst, was the only exception, and we know how well that scaled...

Intel certainly didn't hold-out on the Conroe release...their shareholders needed that ASAP. Intel wasn't in huge trouble, simply because of the breadth of their portfolio and their influence in the corporate sector. AMD, on the other hand, really only had one major income source, and things like this happened:

SUNNYVALE -- April 17, 2002 --AMD today reported sales of $902,073,000 and a net loss of $9,163,000 for the quarter ended March 31, 2002.

SUNNYVALE, CA -- April 16, 2003 --AMD (NYSE:AMD) today reported sales of $715 million and a net loss of $146 million for the quarter ended March 30, 2003.

They were reporting net loss after net loss. No business model can survive that forever.

Well, I guess I had meant the speed wise, I coulda sworn I read for a short time they actually outsold Pentiums. But again, AMD vastly increased their market share from what they had before the Athlon, which, perhaps for those two years may have been costly (remember, there was a little event in 2001 known as 9/11 that nearly finished off what the burst of the .com bubble started, Intel's net income went down as well, leveraged, as you said, by the fact they're a more diverse company), but overall, they had the cash to take the hit (Remember, Microsoft announced they lost $4bil on the Xbox, and not only are they not going out of business, they decided to have another go @ it w/ the Xbox360). Thus, I'd still like to know the source of the bankruptcy rumors, since it would be an interesting read. Then again, everyone throws bankruptcy rumors around at the first hint of a problem. I'm sure someone yelled "OMG INTEL BANKRUPTCY" when the Athlon64 hit the market. Until I see papers filed in court, I don't give those much credence. Anyway, on the point of measuring who's winning and not based on sales/profits I found this (http://www.1up.com/do/feature?cId=3147953&did=1) to make the point. :)

[dcv]

AMD, on the other hand, really only had one major income source, and things like this happened:

SUNNYVALE -- April 17, 2002 --AMD today reported sales of $902,073,000 and a net loss of $9,163,000 for the quarter ended March 31, 2002.

SUNNYVALE, CA -- April 16, 2003 --AMD (NYSE:AMD) today reported sales of $715 million and a net loss of $146 million for the quarter ended March 30, 2003.

They were reporting net loss after net loss. No business model can survive that forever.

Lucky for AMD they're not in that financial situation anymore. While Intel was switching from Netburst to Core, AMD made more gains in marketshare than they ever had before, inked a deal with Dell, bought ATI, and (*very* important for a number of reasons - mostly production of chips and research) opened up another fab. Contrary to popular opinion, they are not in dire straights - their product lines and development cycles are still moving forward, and they have damn good engineering backing them up. The fact that their current chip architecture (which, at this point, is mature to the point of almost being old) runs as fast, cool, and power-efficient as it does - even in comparison to Intel's current Core 2 latest and greatest - is a real testament to their engineering prowess, and I very much look forward to their next chip.

Oh, and how could I forget - AMD also has a bit of a foothold in SFF and embedded devices - most prominently AMD Geode. While an interesting idea and great processor in and of itself, I am a big fan of the Geode mostly because it powers the OLPC. To get touchy-feely for a second - I find this act of philanthropy reason enough to prefer AMD as a company to Intel (well, that, and personal history with their products).

[bgwinkel]

While Intel was switching from Netburst to Core, AMD made more gains in marketshare than they ever had before
Realize that while they were taking losses like they, they were selling more chips than they ever had. (you can read their quarterly reports on their website) Their chips weren't as good as the Intel offerings at the time, and in order to get competitive they had to upgrade their fabs. As of right now Intel completely dominates in terms of advanced fabs capable of high yields. This is the main problem they face. Luckily for them, Intel isn't in the market to put them out of the market. AMD helps Intel and chip development. Intel will never make prices and components at a price where AMD could not be competitive...and believe me, as of right this second, thanks to a lot of back-stock, and great C2D yields on 300mm wafers they could.

At the same time, I detect a hint of AMD fanboy ;p

[rjhill1]

Luckily for them, Intel isn't in the market to put them out of the market. AMD helps Intel and chip development.
That has to be the weirder court results I've seen business-wise "Hey, we have rights to tech you develop, in return, you get rights to ours, even though we're direct competitors" Way cool for competition though, and that's all rock :)

[dcv]

At the same time, I detect a hint of AMD fanboy ;p

Who, me? Naw! ;)

I'm just saying that one should really watch themselves before making such a bold statement that Intel could just wipe AMD out in one swoop - this isn't 2002, and AMD isn't a company losing money anymore. AMD has business deals to ensure profitability for at least a little while. Their new fab will help their ability to produce higher yields of chips. Their engineers produced the Athlon 64. I'd not be so quick to discount AMD as just an annoyance to Intel.

[Unregistered]

I beleive that AMD still leads the sales in server models which are way better than the intel models, so i don't think they will die just yet- well i hope they don't die

[dparm]

I dunno, the new Xeons seem to be pretty damn popular (even in desktops, a la the Apple Mac Pro).

[picch]

I beleive that AMD still leads the sales in server models which are way better than the intel models, so i don't think they will die just yet- well i hope they don't die

Your statement was true about a year ago, however now that Intel Xeon processors use the Woodcrest core, the exact opposite is ture. The Xeon takes any Opteron processor out on the market right now in performance. I've personally used both processors in high demand servers and I can say that the Xeon processor by far out performs AMD's Opteron line. The Xeon also runs cooler (which is every server and network admin's dream especially when you have entire racks filled with the same servers), and if I'm not mistaking requires less electricity too.

Most companies that sell servers, bundle their machines with Xeon processors and maybe have 1 or 2 options for an Opteron box.