For AMD, more money means more problems

Success has its pitfalls, especially when you've awoken a sleeping giant.

For the past two years Advanced Micro Devices has made Intel, one of the world's most prominent companies, look bad. Better products and better timing have brought AMD significant market share and prominent new customers like Dell. But Intel is on the comeback with new processor designs better suited for the power-efficient multicore era, and it will beat AMD to the quad-core punch by using a design strategy that makes purists scoff but accountants happy.

The next six months will be tricky for AMD CEO Hector Ruiz. For starters, his company is taking a hit in the stock market after disclosing that its gross margin fell five points from the second quarter to the third, which ended last week. Part of that was due to the price war in the PC processor market, but AMD also is facing a challenge of overall demand rising before new manufacturing technologies are completely ready.

"It's a good problem to have," said Dean McCarron, an analyst with Mercury Research, since it means people want your products. "But AMD needs more factories." The company's design strategy for the quad-core era requires that it successfully navigate the transition to building smaller transistors at the same time it gets a new factory up and running.

AMD's current pickle is the result of its success, which makes it a little easier to swallow for company executives. Demand is high, but the company's dual-core processors still use its 90-nanometer manufacturing technology. Intel's chips, on the other hand, are built using the smaller transistors provided by its 65-nanometer manufacturing technology. Not only is AMD using larger transistors, but its dual-core Opteron and Athlon 64 processors contain two processing cores integrated onto a single piece of silicon, or a die. This design has given AMD great performance during the past few years, but resulted in processors that were almost twice the size of its single-core chips.

Individual chips are cut from round silicon wafers. Manufacturers obsess over reducing the number of defects on those wafers, but there's always going to be some number of chips on a wafer that simply don't work. The problem is that when each individual chip is relatively large, there's an increased risk that a portion of that chip might contain a defect. Since it costs the same to make a wafer whether a chipmaker gets six chips or 60 chips from that batch, maximizing yields--or the number of good die per wafer--is essential to this business.

AMD's 90-nanometer dual-core Opteron and Athlon 64 processors have a die size of 199 square millimeters. By chip design standards, that's considered a little large, McCarron said. When AMD starts making dual-core Opterons on its 65-nanometer manufacturing technology, that die size is expected to go down to something a little more comfortable that will allow AMD to produce more chips per wafer. An AMD representative declined to comment on the die size for its first 65-nanometer products.

On a conference call following AMD's earnings results last week, Chief Financial Officer Bob Rivet noted that the company would see a cost benefit from its move to 65-nanometer processors in the fourth quarter, since the cost of building the wafer can be spread over more chips. He also pointed out that AMD still hasn't made the full transition to 300-millimeter-wide wafers from 200-millimeter wafers. Obviously, the larger the wafer, the more chips that can be cut from that wafer, and--not counting the one-time expense of purchasing 300-millimeter equipment--the extra costs of the larger wafer are negligible.

[Lindy01]

AMD

is making alot of promises. I dont think they will make their .45nm promise. They still have not pushed out .65nm.<br /><br />It will be interesting to see how well AMD's performs in the next 24months. Intel has taken the performance, power, heat crown in a big way. Once their new 4/way CPU/chipset comes out based on the core technology they will really hand it to AMD. Have you seen the Intel "Sweet 16"...4 quad core CPU's?<br /><br />On top of loosing the performance crown, they are having shortages during FAB upgrades, and merging with ATI. Once Intel hits .45 in late 2007 it will get ugly if Intel starts a hard core price war....because then can way out produce AMD at cheaper prices.

[fred dunn]

Intel's Still depending on the FSB....

to communicate even between the dual core chips on the upcoming "quad core" and even barring that having more than one CPU depending on the FSB as does Intel's HUB architecture it becomes a bottleneck.<br />Why do you think Intel is "throwing" large amounts of L3 cache at their Multi-way processors, because it's the only way to somewaht compensate for the FSB bottleneck.<br /><br />Intel will not beat AMD's Opteron on multi processor servers until they implement their proposed CSI serial transport connection and that is promised in 2008-2009.<br /><br />AMD is already doubling the hypertransport speed and throwing in a lot of litle niceties in their upcoming "barcelona" processor.<br /><br />Don't count AMD out quite yet. Intel doesn't hold a torch to AMD in Multi processor systems and they have JUST caught up with them in the desktop.<br /><br />Yes, Intel can way outproduce AMD and saturate the Intel market but Intel is already hemmoraging and is having to cut back so I have a sneaking suspicion that if Paul Otellini doesn't want to be tarred and feathered by the shareholders that won't happen.

[visaris]

HT Correction and AMD prediction

"Each processor core will be connected by a fast Hypertransport link that allows signals to flow between the cores at the chip's clock speed without having to leave the die."<br /><br />Actually, this is not true. While HT (HyperTransport) is faster than Intel's outdated GTL+ FSB, AMD's quad-core chips are directly connected on die by the shared L3 cache. This is actually vastly faster than even HT would be. HT is only used to connect socket &lt;-&gt; socket and is never used to connect core &lt;-&gt; core within a socket. You should correct your article.<br /><br />----<br /><br />As far as my prediction, I think the near-term future (1-2 years) for AMD is looking very good. Let me explain why:<br /><br />If one looks at AMD's current situtation, it is easy to see they are trailing Intel a bit in (1) manufacturing capacity and feature size, and (2) in performance in the desktop market. Currently, these things are dragging on AMD a bit. I say that AMD looks good going forward because these two major problems are about to go away. <br /><br />(1) As AMD moves its processor lines onto larger 300mm wafers, AMD's capacity increases, and AMD's costs go down. Similaryly, AMD's move to 65nm will have the same effect on capacity and costs. AMD is ready to ship 65nm chips fabbed on 300mm wafers in 4Q06 and expects complete conversion by the end of 2Q2007. As we move a year or so into the future, we see AMD's manufacturing problems lessen dramatically.<br /><br />Also, it is interesting to note the amount of time it takes AMD and Intel to transition to a new node. Intel will spend around 5 to 6 quarters to transition totally to 65nm, while AMD will spend around 3 quarters. Intel may get to the starting line first, but Intel and AMD are crossing the finish line at pretty much the same time.<br /><br />(2) Intel does hold the performance lead on the desktop. There is no doubt about it. However, Intel is not planning on releasing a new core for another 2 years. This gives AMD two years to leapfrog Intel in performance, and AMD is set to do it in just one year. <br /><br />AMD's answer is not just a quad-core K8 chip. AMD's answer is a new core design. Many analysts and articles seem to forget this. Remember, Intel hopped AMD with their new Conroe core. Now, AMD is about to hop over Intel with their new Bercelona (incorrectly called K8L) core. The main advantage Core 2 (Conroe) has is its high IPC. Barcelona (K8L) contains a number of tweaks, updates, and enhancements that are all geared toward raising the IPC of the core, sometimes by up to 200%, as is the case with the FP pipeline.<br /><br />Will Barcelona (K8L) beat Intel's Core2 on the desktop? It is not totally clear. However, it is very clear, after one consideres the technical aspects of the core update, that Barcelona will put AMD and Intel on even footing in the desktop space. All that is left is cost, which has been adressed above. <br /><br />-- <br /><br />Bottom line:<br /><br />AMD's costs are set to constantly decrease (and decrease dramatically) over the next year or so, while Intel's costs are going to stay at a more or less constant (slightly decreasing) level for the next year.<br /><br />AMD is planning a new core for 2007 to raise performace. Intel, OTOH, has no new core planned until 2008, and has no way to raise performace levels except for rasing the CPU clock. We all know how pumping the clock worked for Intel with the P4 (which was designed specifically for high clocks), so I don't expect Intel to be able to pump the clocks much with Core2 (wich has a much shorter pipeline and was not specifically designed for high clocks). <br /><br />Expect AMD to make up ground in the next two years.

[LikeLinus]

?!?!

"Also, it is interesting to note the amount of time it takes AMD and Intel to transition to a new node. Intel will spend around 5 to 6 quarters to transition totally to 65nm, while AMD will spend around 3 quarters. Intel may get to the starting line first, but Intel and AMD are crossing the finish line at pretty much the same time."<br /><br />What does that mean? Intel has been selling and producing 65nm chips for several MONTHS now. AMD has not put out a single 65nm chip. Intel didn't "get to the starting line" they've already run the race and you can get any C2D on the market now in abundance. There is no shortage of these chips.<br /><br /><br />" Intel does hold the performance lead on the desktop. There is no doubt about it. However, Intel is not planning on releasing a new core for another 2 years. This gives AMD two years to leapfrog Intel in performance, and AMD is set to do it in just one year."<br /><br />This also makes NO sense. Intel is leading currently, right? Intel is planning quad-core desktop chips that will be released before AMD's. AMD is playing CATCH UP right now and will continue to for the next several years. AMD isn't set to beat Intel in a year. You think that Intel is sitting on it's hands and not coming up with new concepts and chips? AMD isn't planning two core revisions within a year. AMD's next core revision will be to EQUAL the C2D. So that leaves Intel and AMD equal, but it gives Intel the advantage of a YEAR head start in moving forward with 45nm products and moving to quad and more cores.<br /><br />Furthermore, "Barcelona (K8L) contains a number of tweaks, updates, and enhancements that are all geared toward raising the IPC of the core, sometimes by up to 200%"<br /><br />So they are raising their IPC 200% and you're not even sure that AMD will beat the C2D? But I thought you said AMD was getting ready to blow away Intel and they have 2 extra years to beat Intel.<br /><br /><br />"AMD's costs are set to constantly decrease (and decrease dramatically)"<br /><br />Wrong. AMD is going to have to move to 45nm and they have to bring up new facilities to continue their growth. You think that is all free?<br /><br />" AMD is planning a new core for 2007 to raise performace. Intel, OTOH, has no new core planned until 2008, and has no way to raise performace levels except for rasing the CPU clock."<br /><br />Again, absolutely NOT true. Intel is planning on adding quad cores to the desktop chips and moving towards a 45nm chip. AMD's new core = to compete with C2D, which has a YEAR head start!<br /><br />"Expect AMD to make up ground in the next two years." I thought AMD was going to beat Intel in 1 year? Why do they need 2 years to "make up ground".<br /><br /><br />Fanboyism aside....<br /><br />I think most people know that AMD will come out with a competitive product that will be good. But don't expect either company to sit and do nothing.

[Jon Skillings]

HT fix

Thanks for pointing out the HyperTransport error. The story has been corrected.

[dfmrrd]

Don't know if NUMA is good or bad for AMD

I mean, i guess local memory reference is tremendously fast, but remote is little slow. This puts the burden on OS to allocate memory for process locally and try to peg process to that CPU. But with threads, it complicates everything. these things move cpu to cpu and memory reference end up being remote.<br /><br />Where as with Xeon its all Uniform memory access.<br /><br /><br />ajay<br /><a class="jive-link-external" href="http://www.innerdep.com/" target="_newWindow">http://www.innerdep.com/</a>

[visaris]

NUMA

The question is this:<br /><br />Is it better to have (1) some fast (local) memory access and some slow (remote) memory access, or (2) all slow (remote) memory access?<br /><br />Many people have said things similar to what you are saying before. The bottom line is that AMD's slow (remote) access is faster than _all_ of Intel's FSB access. Therefore, the NUMA as a hinderance idea is a fallacy. <br /><br />AMD essentially has fast and normal, and Intel has only slow. Take your pick.