Now that Phenom II has limped in...

[MCMaChu]

Just wondering what AMD's next true architecture will do. What is it supposed to be and have they released any info on it. Or will there even be an AMD considering how crappy the market is. Thoughts?

 

[fazers_on_stun]

Well this thread already seems like a flame-thrower war, but anyway I believe AMD will have Bulldozer out in 2011 according to their roadmaps. There'll be a 6-core version of Shanghai out maybe next fall, but that probably doesn't meet your "next true architecture" test.

I've seen some old references to Bulldozer, from a couple years ago, where AMD promised it would have some incredible IPC gains over K8, but that was likely before AMD knew how difficult that would actually be. Then BD dropped off their roadmaps for about a year, and then resurfaced as part of "Fusion" I believe. In short, BD is sort of a moving target at this point. Could be an on-die GPU integrated with the CPU. Probably JDJ or somebody more familiar with BD can fill you in more.

If the Abu Dhabi deal does go through (i.e., no x86 license issues) then I would say AMD is OK for the next few years. If not, then it's hard to see them lasting by this time next year. They burned through $800M in cash and readily-available assets last quarter, and are down to $1.1B total. They lost more than they earned last quarter. And although the marketshare numbers for CPUs are not out yet, I've seen a report stating that AMD is purposely slowing their fab down in order to get rid of inventory.

What I think will actually happen is that AMD and Intel will reach some sort of understanding concerning the x86 license and the antitrust lawsuits, and hence AMD will continue on as usual.

Just my 3 cents worth...

 

[loneninja]

As far as I know AMD will be using K10 for some time yet, they'll just be adding more cores in 2010. I fear that their decision to do so will result in really high power consumption and little to no gains outside of highly multithreaded apps and multitasking.

 

[Dekasav]

Are they still going for a 2nd 45nm process with some sort of HKMG. If they do, maybe they can push clockspeeds into high 3Ghz range for stock. Not a new arch, but could be helpful.

 

[keithlm]

As far as I know AMD will be using K10 for some time yet, they'll just be adding more cores in 2010. I fear that their decision to do so will result in really high power consumption and little to no gains outside of highly multithreaded apps and multitasking.

I'll take the added real cores over the "fake" cores that some people actually somehow believe are important.

I remember working my butt off to talk my boss into getting me a hyperthreaded CPU. I got what I asked for. Shorty after ward I found out that the Java development applications we code worked much better if I disabled that stupid garbage (hyperthreading) that Intel pawned off onto us.

But now I'm supposed to trust Intel and believe that it now is a good idea.

That will happen. About the same time as a comet hits the Earth and we all die.

I will take MORE CORES before I will ever buy into the stupid hyperthreading idea.

People can show me stupid benchmarks which mean nothing in real use until they are blue in the face. I will not be stupid enough to buy into that idea again. And the "dynamic overclocking" they now propose as a good idea? I'll just put that into the same category as the failed hyperthreading.

Burn me once.... but it won't be happening again. Sadly they are upgrading my work PC soon and I will have to deal with the stupid "fake" technology they put into their new architecture because it is a corporate thing. Besides... we don't overclock the work machines. DUH. Give me REAL TECH not the fake crap that they want us to believe is real.

OH... and as to the "high power consumption". First of all I don't care. And second of all I don't care. Corporate PC's are never overclocked. The only thing important with them is that they do their job. Period. All this "enthusiast" stuff means nothing in that realm. (Oh and it is very likely I might be able to NOT get a new Intel machine... they might get me AMD. That makes me VERY happy because when I'm working I don't care about the tripe on these forums... I need performance. This weekend I have to go from a task that takes about 8 hours on an Opteron "test" machine to wasting my weekend getting things loaded on a Xeon machine. It will take about 3 times longer. And sadly the Xeon is about 600Mhz faster "theoretically".)

EDIT: Okay I lied. It won't be 3x longer. The Xeon machine will probably only take 12-14 hours instead of 8. But that is wasted time. Period.

Oh. And before some idiot jumps in and screams "But those are server applications": the tools we use run on Windows... for REAL applications we use Sun Starfire boxes. Duh. The stupid Business Objects tools run on a non-server box. (Of course some people consider these to be "Server" applications. But they aren't because they run on stupid x86 platforms. Stupidest thing that ever happened to the corporate world.)

HUMOR: About two years ago the (then) management decided to move to all Microsoft and all SQL Server. Recently that company was bought out and now the new parent company which is 4 times larger mandates moving to UNIX and ORACLE and not using any Microsoft products. Go figure. So I'll be back on a Starfire box soon. Praise whatever deity you believe in because none of this "Intel versus AMD" stuff will mean anything.

 

[enigma067]

Just wondering what AMD's next true architecture will do. What is it supposed to be and have they released any info on it. Or will there even be an AMD considering how crappy the market is. Thoughts?

"Limped"? AMD gives you more bang for your buck and your saying "limped"?

:lol:

http://www.youtube.com/watch?v=Z1EJKifq6JM

http://www.youtube.com/watch?v=dwcKyrHHQac

Presently at tigerdirect.com there is an $850 dollar difference between the two chips.

Flame on Fanboi !

 

[xx12amanxx]

For a low 235$ you get a quad that goes 3.7-3.9 on air with no problem's not really a limp more of a swift kick to the pricing nut's.

 

[jimmysmitty]

Are they still going for a 2nd 45nm process with some sort of HKMG. If they do, maybe they can push clockspeeds into high 3Ghz range for stock. Not a new arch, but could be helpful.

I have heard thats being held off (because of IBM mainly) for 32nm and there will not be a 45nm refresh. besides we don't know how HKMG will truly react to AMDs process itself.

Must wait and see what IBM says and gives the go ahead on that one.

"Limped"? AMD gives you more bang for your buck and your saying "limped"?

:lol:

http://www.youtube.com/watch?v=Z1EJKifq6JM

http://www.youtube.com/watch?v=dwcKyrHHQac

Presently at tigerdirect.com there is an $850 dollar difference between the two chips.

Flame on Fanboi !

Says the fanboy who uses the exact same videos each time he posts and never posts anything relevant or usefull.

For a low 235$ you get a quad that goes 3.7-3.9 on air with no problem's not really a limp more of a swift kick to the pricing nut's.

All depends on the person and what the build is. I say if you have a supporting AM2/AM2+ (preferably AM2+) mobo go for it. New build is iffy.

Keith, glad to see you in here with a pretty much useless "I hate Intel and AMD is the only one who made me feel good" rant but that has nothing to do at all with the OP. In fact you are soooo far off on the other side I can't even see you.

You didn't like HT. Have you actually tried a Core i7 with the newer enhanced SMT? Hell even Intel will admit that not for every application will it boost performance (mainly single threaded ones). But it does show that in most highly multithreaded apps it boosts performance. Or is that not enough?

Intel will probably hit 8 "real" cores before AMD does mainly considering that well hell 32nm is still on track for this year unless the world comes to an end.

And whats your problem with the dynamic OCing? For a server or people who do not OC its a great feature. It will boost it 2 steps up. Thats a free 266MHz to make apps go faster. I don't see a down side.

Either way to the OP, AMD has changed their roadmap a lot since their dump into the red. We don't have a definative answer from AMD on their next step. We do know that they and IBM along with 10 other companies are working on their own version of HKMG like Intel has but I have read a few reports that IBM is having a problem with it.

Could be the materials that they are using for their metal gate could be something else. Who knows. And I do believe that there will be a new socket replacing AM3 in what 2 years? And I think its called Socket G34 but haven't heard anything about it in a while.

 

[keithlm]

Keith, glad to see you in here with a pretty much useless "I hate Intel and AMD is the only one who made me feel good" rant but that has nothing to do at all with the OP. In fact you are soooo far off on the other side I can't even see you.

What can I say... I had too much beer last night.

But anyway: based on the OP's title he was trolling anyway so it doesn't really matter.

 

[jimmysmitty]

I would imagine if they were smart they would be working on 32nm right now and releasing 32nm with AM3, that would be the smart route.

Tell that to IBM. If they are having problems nothing AMD can do but wait until they get the good word from IBM.

And if AM3 is supposed to be out next month I highly doubt they would be able to even fathom switching processes over that fast. Thats IF AM3 is out next month. Plus, remember that AMDs FAbs are now not their own in a way meaning they have less power over them and have to work with them.

But we shall see. I predict they will have 32nm in 2010.

 

[fazers_on_stun]

^ According to the Foundry roadmap, 32nm should be out by Q3 of 2010. Looks like AMD moved it up half a year or so from 2011.

 

[MarkG]

Have you actually tried a Core i7 with the newer enhanced SMT? Hell even Intel will admit that not for every application will it boost performance (mainly single threaded ones). But it does show that in most highly multithreaded apps it boosts performance. Or is that not enough?

Seems to work well on the Atom too, though that's partly because it's such a 'low-tech' design which has plenty of stall cycles for the other thread to utilise.

The big difference is that the new Intel CPUs are designed for hyperthreading, whereas the P4 had it bolted on the side and hit various limitations as a result.

 

[yipsl]

As far as I know AMD will be using K10 for some time yet, they'll just be adding more cores in 2010. I fear that their decision to do so will result in really high power consumption and little to no gains outside of highly multithreaded apps and multitasking.

K10 Deneb based architecture won't necessarily be too high in power consumption if they do another die shrink. Since ATI's GPU's are going from 55nm to 40nm this year, we should expect 32nm CPU's by 2010. They may not be traditional SOI like today's Phenom II's but should involve HK/MG.

It will be an improvement, and Intel's 2010 CPU's should still have higher IPC's but Phenom II shows AMD's catching up. I don't expect the next AMD or Intel architecture before 2012, since the recession might last as long.

 

[MCMaChu]

What can I say... I had too much beer last night.

But anyway: based on the OP's title he was trolling anyway so it doesn't really matter.

Really?

Its more of an observation.
Ive owned both. AMD64, Q6600, and 9950. I am in no ways a fanboi.

From my perspective and other review sites, Phenom II did limp in. It fails to impress. Intel has had 45nm for how long? The Phenom II processor still fails, clock for clock to beat Intel's offerings. The only thing they got going is for previous AM2+ owners an upgrade path. AM3 unless by some miracle with the ddr3 memory allows it to blow past the intel offerings will have zero incentive for new people in the market to buy unless they are strapped for cash where it shines.



Also AMD made its rounds showing the Phenom II was also the be able to be clocked to 4.0Ghz on air. It doesn't.

Unless AMD made the Phenom II to a point where the price/perfomance justifies the purchase then the term limped in can be used.

 

[Dekasav]

Really?

Its more of an observation.
Ive owned both. AMD64, Q6600, and 9950. I am in no ways a fanboi.

From my perspective and other review sites, Phenom II did limp in. It fails to impress. Intel has had 45nm for how long? The Phenom II processor still fails, clock for clock to beat Intel's offerings. The only thing they got going is for previous AM2+ owners an upgrade path. AM3 unless by some miracle with the ddr3 memory allows it to blow past the intel offerings will have zero incentive for new people in the market to buy unless they are strapped for cash where it shines.



Also AMD made its rounds showing the Phenom II was also the be able to be clocked to 4.0Ghz on air. It doesn't.

Unless AMD made the Phenom II to a point where the price/perfomance justifies the purchase then the term limped in can be used.

Two things.
1. AM3 (or any CPU) doesn't have to "blow past the [competitor's] offerings" to have an incentive to buy. Price/performance, upgrade path, total system cost, and availability of good/better motherboards can give an average or below-average CPU (performance-wise) an incentive to "new people in the market."

2. Phenom II IS at a point where its price/performance justifies its purchase.

Phenom 1 limped in with a broken leg (TLB errata + performance/OC issues). Phenom II strolled nicely (but didn't do acrobatics like Core 2 did).

 

[MarkG]

Two things.
1. AM3 (or any CPU) doesn't have to "blow past the [competitor's] offerings" to have an incentive to buy.

Of course not. But Phenom II probably costs a similar amount to manufacture as i7 does, and AMD will have to sell it for a fraction of the price in order to remain competitive; that's good for end-users, but not so good for AMD's financial results.

 

[spathotan]

The smaller it gets the longer its going to take. The jump from 65nm to 45nm was here in a flash on the Intel side. However as you can obviously see 45nm to 32nm is taking 2 times longer, its also less of a shrink.

 

[MarkG]

However as you can obviously see 45nm to 32nm is taking 2 times longer, its also less of a shrink.

Both 65nm->45nm and 45nm->32nm are approximately the square root of two: i.e. you can fit twice as many transistors in the same area. That is the normal rate at which die shrinks occur.

Time-wise is a different matter of course, as is cost. And with AMD lagging behind, I doubt Intel have much reason to rush ahead and obsolete billions of dollars of investment in 45nm fabs.

 

[Raviolissimo]

Just wondering what AMD's next true architecture will do. What is it supposed to be and have they released any info on it. Or will there even be an AMD considering how crappy the market is. Thoughts?

i guess the answer depends partially on what happens to the world economy. above a certain volume, making CPU's is like a license to print money.

i wonder what the differences are between Phenom 2 and Core i7 in terms of putting 2 CPU's on a motherboard.

AMD could take back the performance crown by coming out with a dual or quad CPU version of the Phenom 2 before Intel Xeon'izes the Nehalem chips. i don't know how hard that is technically. they both have some practice with making processors for 2P and 4P workstation & server systems. i wonder if AMD's more extensive experience with the integrated memory controller will help them win this race to market.

AMD is really hurting financially. i would not be surprised to seem them go BK, bankrupt.

http://finance.yahoo.com/q?s=AMD

1.33 billion. heck, i could almost buy them.

i wish. i wouldn't be surprised to see Dell or HP buy them. their stock market value than it was after 9-11.

 

[yipsl]

Bankruptcy usually means reorganization. There's enough avoidance of monopoly incentive for Intel to not put the pressure on AMD as in the OEM rebate days, and the paper losses due to the ATI buyout and the spin off of the fabs will eventually disappear. In a sense, it's not real money losses as AMD is not selling products at a loss. Investor losses can be recouped as AMD retools and gains market share.

AMD is selling chipsets and GPU's at a profit and they aren't selling CPU's at a loss. At best, the margins for CPU's are rather thin, but I'd say that's true with i7 as well. Intel's Core 2 architecture, mature and affordable is the money maker there.

I'm tired of all the clock for clock mishegoss that some vent here. We know Intel has higher IPC right now. AMD had higher IPC back in the Netburst days. Both Intel and AMD kept their historical relationship because it's CPU's that compete, not instructions per clock.

A Q6600 is EOL but the fan of that CPU will probably argue that it's almost as good as a Phenom II because it's merely 2.4 gigahertz and Phenom II's are 2.8 or 3.0 (which makes up for the slightly lower IPC). The same held when the Athlon X2's at 2.0 gigahertz beat a Smithfield at 2.8. The more things change, the more they remain the same.

Both CPU's had markets and customers back then and both present day CPU's have markets and customer's now. At least the Q6600 isn't worthless like the Pentium D 805. It's still a viable CPU and I can't see anyone upgrading from it to either a Phenom II or an i7.

i wish. i wouldn't be surprised to see Dell or HP buy them. their stock market value than it was after 9-11.

Don't people understand the x86 license issue? Intel licenses their technology to AMD because AMD was once a producer of Intel designed CPU's when Intel didn't have the capacity to meet all customer's demands. AMD licensed x86-64 to Intel and though Intel's taking pot shots at AMD for allowing the new fab spinoff "access" to x86 technology AMD could fire back in court and it could get ugly.

It would get worse if a mere PC builder like Dell (which has it's own problems and might go the way of Compuadd in this decade) tried to buy into the x86 license. Only Samsung could possibly buy AMD, unless it really became "Arabian Micro Devices" as The Inquirer snarkily calls it.

Intel needs an x86 competitor and doesn't need hassles over x86-64. It is better for Intel that AMD survive bankruptcy reorganization and compete with as small a market share as viable otherwise Intel's in more problems with regulators than during the OEM rebate mess; plus in trouble if a player as big as itself gets the license. Intel needs AMD but needs AMD to be small.

I like AMD CPU's but I'm looking forward to the day when a consortium of companies go past x86 and x86-64 and we have a standard that doesn't lock Intel into major share of server and monopolistic share of desktop CPU license. AMD will probably be part of that consortium and Intel will follow, but it probably won't happen untill quantum computing, the way things are going. LOL

 

[MarkG]

AMD is selling chipsets and GPU's at a profit and they aren't selling CPU's at a loss.

They certainly aren't making any money.

At best, the margins for CPU's are rather thin, but I'd say that's true with i7 as well.

You think profit margins on $600 CPUs are 'rather thin'? I'm not sure what a 45mm die costs, but based on experience with earlier processes I'd suspect Intel are making at least 50% profit on those chips; you don't get anything like that margin selling big quad-core chips at $200 apiece.

 

[uguv]

based on the OP's title he was trolling anyway so it doesn't really matter.

Agreed. Phenom II isn't too bad. It's fairly competitive for the price.


Also, @Enigma067 - You're dumb. Are those AMD marketing videos supposed to actually convince anybody here that the pII 940 and the i7 965 are even in the same league? Compare the PII 940 to the Q9400, Q9550, or even i7 920. It actually stands up fairly well in most cases and makes a decent argument for going with PII. AMD marketing videos aren't going to influence many people on a site like this.

 

[fazers_on_stun]

Both 65nm->45nm and 45nm->32nm are approximately the square root of two: i.e. you can fit twice as many transistors in the same area. That is the normal rate at which die shrinks occur.

Time-wise is a different matter of course, as is cost. And with AMD lagging behind, I doubt Intel have much reason to rush ahead and obsolete billions of dollars of investment in 45nm fabs.

Recent news articles I've seen show Intel stating it is moving to 32nm on time, no plans to delay it despite the lousy economy. Westmere should be out in Q4 and available in quantity Q1 next year. Intel has already done most of the heavy lifting with 32nm and is using the D1D fab to scale it up for production quantities.

However, the bulk of CPUs next year will continue to be 45nm, and probably Penryns since they're cheaper to make I believe, plus no challenge from AMD at the high end, at least at the present.

 

[fazers_on_stun]

... and the paper losses due to the ATI buyout and the spin off of the fabs will eventually disappear. In a sense, it's not real money losses as AMD is not selling products at a loss. Investor losses can be recouped as AMD retools and gains market share.

Hmm, how do you explain AMD's cash & liquid assets decreasing to just $1.1B at the end of Q4, down by $800M for the year? I would call a 73% reduction in cash more than a "paper loss", esp. when AMD sold off some assets such as the old 200mm equipment. According to AMD's quarterly report, they lost $431M in "computing solutions" in Q4, and only a $10M loss in graphics for the same quarter. So, yes AMD is selling products at a loss. This current quarter is not shaping up any better, esp. with the P2 price cuts that went into effect a couple weeks ago.

AMD is selling chipsets and GPU's at a profit and they aren't selling CPU's at a loss. At best, the margins for CPU's are rather thin, but I'd say that's true with i7 as well. Intel's Core 2 architecture, mature and affordable is the money maker there.

From AMD's Priorities :

AMD's unit marketshare in each of these fields. We find that for Q3 2008, they were:

Notebooks: 11.3%

Desktops: 25.0%

Servers: 13.8%

We can see that AMD's unit marketshare differs considerably by sector, being strongest in desktops and weakest in notebooks. Unfortunately for AMD, the long-term trend in CPU growth is the opposite, with notebook CPU volume growing strongly while desktop CPU volume stagnates/slowly declines. I didn't list notebooks first just to be extra-nasty, Mercury's research also indicated that Q3 2008 was the first quarter ever where notebook CPU shipments exceeded those for desktops.

Next comes the revenue marketshare in the various segments. It's easy to see that AMD gets considerably less revenue per CPU, no matter what the segment than Intel.

How much less?

In servers, we find that in Q3 2008, AMD got 77% of the revenue for the average server chip that Intel did.

In notebooks, we find that in Q3 2008, AMD got 72% of the revenue for the average notebook chip that Intel did.

In desktops, we find that in Q3 2008, AMD got 63% of the revenue for the average desktop chip that Intel did.

So in the segment where AMD does best, they get (relatively) the least amount of money, but, no matter what the segment, AMD processors sell at a significant discount to Intel processors.

I suppose iSuppli will be releasing Q4 marketshare numbers in the next few weeks. And there was recent news about GPU sales being way off in Q4, although I can't find the link at the moment.

 

[fazers_on_stun]

Speaking of the devil, this just in:

For 2008 as a whole, Intel claimed an 80.4 percent share of the x86 processor market up from 77.1 percent in 2007. AMD’s share declined from 22.1 percent to 18.5, and Via’s rose from 0.8 to 1.1 percent. The total x86 processor market expanded 13.3 percent over 2007.

Desktop and notebook CPU shipments suffered an 18 percent drop in growth, even as server CPU shipments declined 25 percent. The report attributed the latter to businesses cutting back on infrastructure spending, and enterprises having finished their most recent server refresh cycle.