AMD has always been one generation behind Intel, at least as manufacturing process go. And that didn't stopped them to survive/be competitive/outperforming Intel/...
So, AMD doomsayers, quit your nonsense and be happy AMD is still around. We, consumers, are the only winners in the end.

Who said FBDIMM was screwup. Ever hear of Seaburg.
Duh!! While FBD would never break into desktop space due to its cost. In workstations and servers with seaburg chipset, well just look at the reviews.
Its a hit!

not sure who to believe, this article, or a quote from an AMD employee claiming 45nm is already coming off the line, and will be in market by mid 2008...article attached

http://www.tgdaily.com/content/view/34439/118/

and about rambus bandwidth, they were at fsb 800 the same like sdram pc100. PC 133 have 33% more bandwidth.Rambus works on 8 bit and sdram on 64 bit.

I read somewhere in Inq about 3d Chip design, that it is a much better/ efficient Process than current 2D lithography and AMD being after it. In 2009 things will be changed drastically, AMd will directly swith to the new process and catch up with a die size compareable to 32 nano. there are many side benefits, like being able to move the memory transistor away from die. Hard kick to intels Butt.

If this is true, then I really don't see how AMD could survive. Without a 45 nm process in 2008, Intel's Penryn will dominate Phenom's, and Nehalem will bury the whole K10 line --- servers and desktops.

AMD managed to survive in 2006 and 2007 despite being way behind Intel on desktops, but this would be the final blow.

BS, i wonder how much $$ in market development funds the ASLM bloke got to say that.

everyone keeps *assuming* a die shrink automatically leads to higher performance, such as the commenter above. It's like they have forgotten that Intel went to a die shrink with the Prescott and it didn't do squat but make it run hotter, while AMD was running behind in die shrinks then too but STILL cleaned Intel's clock with the Athlon64. So, stop this nonsense of assumptions. Remember, ignore paper releases and the rumor mill, especially from Intel, the king of paper releases and rumor mill fodder (i.e. remember they were saying the P4 will scale to 4Ghz or even 10 Ghz...RIIIIIIGGGGHHHHTTT).

I agree with the previous comment. AMD has only three advantages over Intel (1) the integrated memory controller (2) the hypertransport direct connect architecture (3) Intel’s incompetence (netburst, RDRAM and FBDimms). The memory controller has allowed AMD to get by with a smaller cache and hence smaller die size. This has negated one generation of Intel’s process advantage. Hypertransport has improved scaling in MP systems and allowed AMD many wins in this lucrative market. Once Intel has Nehalem with the integrated memory controller and a direct connect architecture, AMD only has Intel’s incompetence, which seems to have dried up as of late.

If AMD can achieve the die shrink to 45nm in mid 2008, they will be able to put two 45nm Barcelonas on one package, wire up a couple of Hypertransport channels between the chips and have an Octal processor from day one. At the moment, the 65nm Barcelona is too large to fit two chips in one package. An octal core would change the landscape: AMD’s octal core would be compared to Intel’s quad core and AMD’s quadcore would be compared to the dual core. Even with an IPC deficit of 20% per core, AMD would still win with double the cores.

Without an octal core to compete with Nehalem, AMD is lost.

I don´t think intel is the king of paper releases. The question is that a Shrink of an architeture will deliver more performance. That´s will happen to penryn... but the question is how far Nehalem will go... cause then we will have a big boost of performance over the phenom and over the penryn. AMD is really in trouble, but i hope they have something hidden, otherwise... processors market will be very boring, i think i will start to follow the GPU fight... until Intel do not starts delivering their own.

I really don't see this as an issue when you can get a Opteron 8218 HE @ 2.6 Ghz (~1.2 V) that's 90nm. AMD admitted some time ago that they were not able to give 65nm their full attention while giving birth to Barcelona. What AMD needs now is to evolve its 65nm tech to the level they have reached with 90nm. Lithography @ 45nm is somewhere around four times the price it is at 90nm and the first few runs will be duds anyway. So AMD needs to stick to it's guns, evolve 65nm, start weeding 90nm, and stay away from 45nm as long as it can otherwise we old faithful will be eating it on the hip.

with a die shrink, you get more CPUs per wafer, thus increasing profit (after the R&D is paid for), so even if a die shrink doesn't improve performance, it would add some much needed cash to AMD's pockets.

not produce. AMD already has 45nm SRAM wafers. I took that article to mean that they won't be moving to higher than 30% 45nm equipment in Fab 36 until 2009.

AMD was about a year behind Intel for the 65nm transition. It is not surprising that they will be also one year behind for the move to 45nm.

Intel has already started showing 32nm wafers. By the time AMD is producing 45nm chips, Intel will have moved to 32nm.

Despite the lagging manufacturing process, AMD can be an excellent price/performance choice in the low- and mid-range.

1. IMC and HT is pretty much packaged together so we’ll talk about them the same. The INQ has been saying that one CSI link is equivalent to about 25 GB/sec of transfer which outperforms HT v.3 I think.
2. Intel is not incompetent. You don’t get that big by sucking. I’m not going to praise NetBurst, but saying that the choice to go to rambus was a move of incompetence is an unfair statement. I like rambus and the bandwidth it allots. But you have to remember that it was price fixed. All those lawsuits? Yeah people were found guilty of price fixing.

And isn’t the MP market a pretty small market?

See the problem right now is that IFF Intel’s statements about Nehalem (posted at the INQ like two days ago) are correct, AMD is going to have a long uphill battle. They claimed that the performance increase from Core 2 to Nehalem will be similar to that of Presler to Core 2. That’s a huge increase and whether it involves clock speed and core count is irrelevant. The point is that Nehalem is a beast. Plus with 8 core (with HT) coming out next year, AMD better get creative really fast. 

My guess is that initial 8 core CPUs will be really expensive. the die of the Nehalems looked pretty big so my guess is that they'd reduce the cache to free up space for the other four cores and sell it for a lot of money.

C2D dominates X2 Athlon 64, Penryn will dominate the Phenom and even the Pentium dual core will clean up the cheap PC floors .... 
When Nehalem comes, I don't even want to think about. AMD can cut prices and make thier chips more attractive from performance/price perspective but we want AMD to survice, not bankrupt! AMD's glorious Athlon 64 time (from 03 to 05) has been gone... Hope AMD will do something quick... 
They are fighting 2 fronts, CPU and GPU. Both fronts have 2 issues: 1) behind competitor and 2) huge money losing business. 
When Intel chip sucked (03-05), it still made good profit... when AMD has trouble, AMD just lose big money. I guess size does matter

Don't tell me you didn't know this. AMD has a huge tendency of hyping up, and then under-delivering. Look at Barcelona--6 months late and 25% slower than initially promised.

The problem is, Barcelona die, even with its suckingly low 2MB L3 cache is about 270mm^2. Penryn (2 dies) is 280 mm^2. This implies, Nehalem will be approx 280*734 (million transistors)/800 (million transistors) = 257 mm^2. So, Nehalem will be smaller and faster. Don't expect any bargain prices from AMD.

The only way I see AMD making profit and gaining temporary lead is having an octal cpu out ASAP in desktop and server platforms ( as pointed out by anather reader) and they have to price it right. 

They have all the fundamentals correct for an octal core cpu as they have a mature HT tech inplace, only have to squeeze 2 quads in one package. Not easy but I think doable.

Another reason for AMD to go fabless....their own engineers are dead weight

Chris, unfair or not, netburst sucked baddd.
If it wasn't for the Israili division of Intel, they would be pushing netburst still. Intel was dual develping architectures, netburst and the core duo. the core duo has absolutely nothign to do with netburst. core 2 duo came from core duo, which in turn came from the pentuim M.... the pentuim M has its roots in PIII. no netburst here.

rambus sucked because of the price, the bandwidth was good BUT the error rate also sucked. So only half of intel was incompetent. It is unfair to say that bacause of one of their products they all suck I agree. Just dont try and feed the garbage that netburst was good. 
and yes I have both Intell and AMD cpus in my house, PIII, PIV, core2duo....sempron 64, and athlon x2.

AMD has the budget sector but they still perform quite well for the price. 
nm isnt everything.

HT3.0@4000MHz does 32 GB/s (per one 16-bit link).