Tuesday, January 30, 2007

Q4 06 Intel and AMD: Who Won?

Sometimes it is easy to pick a clear winner and sometimes it is more difficult. The fourth quarter of 2006 for Intel and AMD is hard to call because of conflicting numbers for revenue share and volume share.

I am always sceptical of changes in share in a rising or falling market. For example, if AMD gains in a falling market, the numbers can be inflated and make AMD appear to be doing better than it really is. Likewise Intel can gain temporarily in a rising market (if AMD hits its capacity ceiling) and appear to be doing better than it really is. I figured I would have to wait until Q1 07 to see if Intel held onto its revenue share. However, now I'm faced with conflicting numbers so it looks like the actual scenario is different.

In order for Intel to have a temporary gain in Q4 06 they would have had to gain in both volume and revenue share, but this didn't happen. Intel gained revenue share but lost volume share. So, it does not appear that Intel had a temporary gain. We have to assume that the revenue share increase was genuine and not temporary. However, AMD gained overall in volume share. Since we know that Intel was not at a capacity ceiling this gain too must be genuine. This is confusing. However, it gets even more confusing because Intel and AMD appear to have swapped strategies. In the past, Intel has had a priority strategy of taking volume share, usually by lowering the price of its chips. In contrast, AMD has targeted the server market and has taken server share for the past three years. Yet, in Q4 06 AMD cut its prices (reducing its margins) and gained volume share in both desktop and mobile. Meanwhile, Intel gained back a small amount of share in servers. Clearly, AMD's actions reduced its margins while Intel's gains were most likely due to the improvements with Woodcrest and Tulsa over older Xeons.

Since both gained there is no obvious winner in Q4. So, we have to consider where AMD and Intel can move from here. In other words, if one company gained a temporary advantage that would be lost in Q1 07 while the other had a gain that could be built on in the following quarter, the company with the temporary gain would be the loser. Basically, half a step forward followed by another half step is better than one step forward followed by one step back. So, let's look at where each company is now after Q4 06 and where they can go in Q1 and Q2 07.

AMD's profits were only about 1/3rd of what they were in Q3 06 so the obvious need is either to raise prices or reduce costs. Raising the prices after lowering them would be out of the question so the only option would be reducing costs. FAB 36 is still ramping and as it ramps a greater volume of the production is on 300mm which does reduce cost. Likewise FAB 36 is converting to 65nm and this too reduces cost. So, by continuing with the same strategy AMD could expect to have better profits in Q1 07 although they may not quite match the Q3 06 numbers. Presumably, AMD could introduce Barcelona in Q2 with a higher price tag and with continually dropping costs on FAB 36 this should make the Q2 07 numbers better than Q1 07. I can't see any problem with this strategy so it looks like AMD could hold onto its current volume.

Intel lost volume share in desktop and mobile. However, we want to re-examine the issue of capacity. Although Intel was not limited overall, it was limited in terms of Conroe and Merom. Some could speculate that Intel gained where the market segment was covered by C2D (servers) but lost where capacity was constrained (desktop and mobile). This could then lead someone to assume that Intel would simply take back mobile and desktop share as production capacity of Merom and Conroe increased. I suppose this is possible but I don't think it is likely. For example, Intel's server gain was only 1.4%. If there were a strong preference for C2D then it seems that this gain should have been larger. I think a more likely scenario is that desktop is a function of performance and price, mobile is battery life and price, and servers are more complex with performance, power consumption, density, price, and support all being factors. Essentially, at the bottom range of the desktop segment, price would be dominant with performance secondary, but performance would become more important as we go up until it becomes the dominant factor at the high range with price secondary. Likewise, with mobile, cost is important at the bottom with battery life dominating at the top. With servers price and support probably dominate at the bottom with power consumption and density becoming stronger factors at the top. Performance also becomes a stronger factor as average workload increases. For example, a 10% increase in performance matters much more for a server that spends most of its time at 70% capacity than one that only gets above 20% a few times a day.

It is possible that AMD only gained temporarily due to capacity contraints of C2D for desktop and mobile. If this is the case, then Intel wouldn't have to do anything since their C2D capacity is steadily increasing. If the market prefers C2D to K8, Intel would simply gain the lost desktop and mobile share back as it is able to supply more of Conroe and Merom. So, if this is the case then Intel's strategy is solid and it will gain back its lost share. However, it is more likely that price is the biggest factor on the desktop so Intel would have to lower prices in addition to increasing volume of Conroe. Likewise, it appears that AMD will released a lower power chipset for mobile and this will take away Intel's battery life advantage. This would mean that Intel would have to change strategy to gain back desktop or mobile share. And, Intel's gains in server share could be in jeopardy once Barcelona is released in Q2 07. Basically, AMD will release a new architecture in 2007 just as Intel did in 2006. However, in spite of its new architecture Intel shows very little in terms of market gains. With increasing competition from AMD in 2007 both from a new architecture and new chipset offerings it really looks like Intel will have to be the one to change strategy. In this regard I would have to put AMD slightly ahead in Q4 06 but not by much. AMD will still have to deal with increasing volumes of C2D during the year so AMD is not guaranteed to be able to continue taking share. However, overall I think it is most likely that AMD will continue to nibble at Intel's share each quarter while Intel steadily grows to regain its former 2005 levels of revenue.

Overall, in 2006, there is no doubt which company is the winner. AMD added a second FAB and the ability to produce its own chipsets while gaining revenue and share while Intel lost both revenue and share in spite of releasing a new architecture. This puts AMD in a much stronger position in 2007 than it was in 2006 and it gains a new architecture on top of that. This makes for a very strong offering by AMD in 2007. However, if we take Intel's bag of tricks for 2007, turn it upside down, and shake it hard, not a lot falls out. Intel will increase the ratio of quad cores on its server chips and it will introduce a robust albeit large and expensive quad FSB chipset. It will also add a server chipset that uses DDR2 registered DIMMs instead of FBDIMM to reduce system cost. Apparently these are to try to avoid losing more server share in 2007. There really doesn't seem to be anything new for either desktop or mobile. AMD however will have new offerings for server, desktop, and mobile. For servers specifically, Barcelona should remove Intel's advantages with power consumption, SSE performance, and Integer performance. Barcelona should match Intel in terms of cache so this advantage (particularly in benchmarks) disappears as well. Intel has also enjoyed advantages in benchmarking because reviewers have been using operating systems that have poor NUMA support. This gives Intel an obvious advantage on multi-threaded code. However, having four cores on a single socket removes this advantage since a native AMD quad does not require NUMA. This should mean that Intel will lose almost all of its performance advantage when Kentsfield is benchmarked again versus a desktop version of Barcelona.

The only fair edge that Intel could gain would be by having a higher clock which is certainly possible. However, I'm guessing that Toms Hardware Guide will put its thumb on the scale several times to make sure that Intel has additional advantages. These cheats will probably include using the Intel Compiler instead of the PGI compiler. Whereas the Intel Compiler is only optimized for Intel processors and even turns off some hardware features for AMD processors, the PGI Compiler can generate code optimized for either Intel or AMD, or code that runs well on both. Since it supports both 32 and 64 bits, C++ and Fortran, Linux and Windows, and generates parallel code automatically while costing the same as the Intel Compiler, PGI will almost certainly become the standard compiler for software development. Using PGI instead of the Intel Compiler would go a long way to establishing credibility for review sites that claim to be impartial. I don't expect Toms Hardware Guide to do this and probably not Anandtech either. And, THG has other cheats that it can use. They will probably use high clocking memory with slow timings since this favors Intel. I've never yet seen THG bother to get lower clocked memory with faster timings that would work better with AMD even though it is usually readily available for the same price. They probably will avoid loading the extra cores since this allows one benchmark to use all of the cache. This situation would be very rare for real code but can give quite an artificial boost to benchmarks. They will probably also use their favorite cheat of overclocking the Intel chip while leaving the AMD chip stock. Overclocking belongs in a separate article much like using aftermarket parts to boost engine horsepower does not belong in a regular car review.

2007 shouldn't be bad for either company. I expect Intel to keep growing and gain the revenue that is lost from 2005. I expect AMD to keep slowly expanding and taking more share. I guess if both companies make more money then neither one really loses. The two companies' offerings should be very close in performance and features by Q3 or Q4 2007 unless Intel manages to maintain a substantial clock lead. I am really hoping for some genuine reviews because this should be the best horserace we've had since PIII versus K7. However, I'm not optimistic that either Toms Hardware Guide or Anandtech will see the handwriting on the wall by then. So, we may still have to rely on other sites to get any proper test results. Although I'm guessing that THG will still be chasing the Intel advertising dollars six months from now I'll mention again what proper testing would be: no overclocked chips, match the DIMM timings to the processor, do proper core loading for multi-core processors, publish the cpu activity graphs as Tech Report does, use the PGI compiler, and either compile each for its own best performance or compile for good performance for both. I'll admit that NUMA testing is more complicated because they would either have to publish both interleaved and non-interleaved scores to show whether the OS has good NUMA support or they would have to include some Linux comparison scores. Much like the Sea Biscuit story, this is a horserace that everyone has been waiting for. It's just a matter of finding a review site where Intel won't have a five length headstart.

Friday, January 26, 2007

AMD Q4 Earnings -- Good Or Bad?

AMD has delivered its Q4 2007 revenue numbers. The results can be pretty much as expected, disappointing, or evidence of a company going down in flames depending on how distorted the analyst's view is. Lately, the latter opinion seems to be the most prevalent with many articles mentioning the $574 Million loss. However, this view is, for the most part, inaccurate.

I think everyone with common sense expected that when AMD bought ATI that it would lose its Intel contracts. This is certainly what I expected. This loss of contracts created a big loss of revenue for ATI in the fourth quarter which again was expected. That it would take some money for this merger to be completed was again very much expected. I've seen a few people wail about the loss of the Intel contracts and make hasty projections that the ATI section will never again turn a profit and will only drag AMD down. Usually, this idea is lumped with the idea of AMD's several Billion debt and the projections then begin running to sale of ATI or even bankruptcy. These projections are something of a source of amusement but not much more. Let's see if we can look at this more realistically.

Intel has a big weakness in the area of motherboard integrated graphics where they are today almost unchallenged. ATI brings this capability to AMD in a way that is profoundly beneficial for both companies. It is this failure to recognize the benefits that is the most surprising to me. AMD benefits ATI tremendously because it has more money for R&D than ATI has. ATI also gets the benefit of early notification of AMD's technological directions. These two things are bound to give ATI the boost it needs to be more competitive with nVidia. ATI likewise gives AMD an all in one solution for OEM's who prefer this to putting together systems piecemeal. AMD will be able to deliver packages where the southbridge and graphics chips bear the AMD logo rather than just the cpu. This should give AMD a tremendous boost in corporate sales. This boost in turn provides a ready market for ATI's products which should easily make up for the lost Intel contracts.

Some people have suggested that AMD should have used its cash to build a third FAB rather than purchasing ATI. That does not seem reasonable to me. As a business gets larger some customers will expect more services to be provided and it is very likely that AMD had went as far as possible while providing only processors. This not only means that AMD can provide complete solutions but it also means that AMD can guarantee support when it is needed instead of when 3rd parties decide to provide it. This is something that would have done AMD some good as far back as the introduction of the K7 and would have done even more good when K8 was first introduced. Rather than trying to twist this around to some kind of problem for AMD, people should see it for what it is: the loss of yet another of Intel's previous advantages. AMD broke out of its single FAB limitations in 2006 and this lays the groundwork for broad platform support in 2007. I doubt there is anyone at Intel who while proclaiming the introduction of Core 2 Duo doesn't recognize the great strategic losses to AMD during 2006. Overall, AMD gained more with FAB 36 and ATI and than Intel gained with C2D.

For those who have missed this, it only takes a moment to look back at history. The progress that AMD made with K7 up to 2001 was largely pushed back in 2002. And, K7 was unable to make any real headway all through early 2003. With K8, however, AMD's volume share has grown 50% and Intel has shown no signs of being able to take this back. With a second FAB and ATI it is even less likely that Intel can take back any volume share in 2007. Let's look at the numbers:

For processor revenue from 2005 to 2006:
Intel shrunk 15.3%
AMD grew 34.6%

For processor revenue from 2004 to 2006:
Intel shrunk 2.5%
AMD grew 201.9%

I think anyone with common sense can see that Intel has slipped approximately 2 years in terms of processor revenue parity with AMD. I'm sure that Intel will have recovered to 2005 levels by the end of 2007 however I'm also certain that AMD will grow at least as much as Intel during the same time period. Of course, Intel enthusiasts will want to concentrate on just the Q4 numbers which seem to be the only bright spot in a very bad year for Intel.

If we ignore the acquisition costs then AMD's earnings were less than half of Q3's. Since AMD increased volume by 9% from Q3 this has to indicate a reduction in Average Selling Price (ASP). This would be the effect from Core 2 Duo. AMD's inventories are up 1% from Q3 which appears to be due to ramping of FAB 36. Intel's inventories are down very slightly. Q1 should provide a better inventory picture. Intel is slightly better off than AMD in terms of cash although AMD has about twice as much debt as cash while Intel has about four times more cash than debt so overall Intel is much better off. Compared to Q4 2005 Intel is down 11% while AMD is up 3%. Intel does appear to have gained back 0.8% of revenue share. However, with the boost in volume in Q4 this is probably only a temporary change.

Basically, Intel's numbers show that they are back on track and slowly growing again from the large losses earlier in the year. AMD's numbers show that they are trying to hold onto volume share and are currently willing to give up revenue to do that. This has been a drop for AMD of about 10% on ASP. This 10% accounts for the lower earnings compared to Q3 06. The outlook is also pretty boring. Obviously, Intel will continue to ramp C2D and not much more than that until late 2007. Then they will release a quad FSB northbridge allowing Woodcrest to move into the 4-way area and they begin slowly releasing 45nm. AMD will continue to ramp FAB 36 although this is partially offset by beginning to ramp down FAB 30 for the 300mm conversion. They will introduce K8L Barcelona in Q2 which should make 2007 K8L's year in the same way that 2006 was for C2D. AMD's margins should be up in Q1 and their channel distribution should be fine. The clearest signal from 2006 however is that the primacy of the desktop is fading fast. Prices are down for desktop compared to mobile and server and clearly these will be the areas that AMD will pursue. Likewise, AMD is growing faster in terms of commercial than consumer. Most of this stuff is pretty boring.

I see three battles taking place in 2007 that are not typically talked about. The first is with instruction extensions to x86. Up until AMD64, Intel was able to create new extensions which immediately became the standard. The best example of this is when AMD's 3D Now was brushed aside by SSE. However, it should be interesting to see whether or not Intel adopts AMD's planned instruction extensions in 2007. This is particularly noteworthy because adoption of these by Intel would tend to hurt Itanium. The second question is memory. It is no secret that Intel pushed JEDEC towards DDR2 at the expense of DDR. And it is no secret that this worked to Intel's advantage because the higher latency was offset by Intel's larger cache. However, it appears that AMD may be able to get JEDEC to extend DDR2 to faster speeds rather than just moving to DDR3. This would tend to be disadvantageous for Intel. Finally, it remains to be seen what the review sites will do during 2007. It is almost a certainty that by 2009 the typical pro-Intel stance will be broken however it is not clear when this might happen. It is somewhat possible that reviews could become more professional (such as not using the Intel Compiler) by the end of 2007 when head to head testing for quad cores should reach a peak. However, it is also possible that sites like Toms Hardware Guide and Anandtech could feel that Intel will retain its dominant position or make a strong comeback with 45nm. If this is the case they will continue to hold out in Intel's favor until sometime in 2009.

Friday, January 12, 2007

TechWeb -- Utter Confusion And Disappointment

Recently my blog was mentioned at TechWeb, Scientia's New AMD Blog. Although I do appreciate the mention I can't say that I appreciate the commentary by Max Fomitchev. After trying and failing over and over to post a corrective comment at TechWeb I decided to handle it here. I'll look at Fomitchev's comments and see how much his opinion is actually worth.

Scientia, the force behind AMDZone.com, has started a new blog devoted to all things AMD.

First of all, this is not an AMD blog. If I wanted to promote AMD, I assume I would do what other blogs do and simply cherry pick any news items that favored AMD or were bad for Intel and repost or link them here. I don't do that. Secondly, I'm not the driving force behind AMDZone. There are other knowledgeable people who post there. There are knowledgeable people on Toms Hardware Guide's ForumZ as well but too often the discussions there break down into pointless name calling. If the environment there were more civil it too would be a good forum.

Scienta clings to his beloved company

I'm not quite sure where Mr. Fomitchev got this idea. Let's compare me with Mr. Fomitchev.

As and AMD loyalist for 15 years I am utterly disappointed and confused: I cannot wait for K8L and need to upgrade now, and given the street prices for AMD and Intel CPUs I am left with no choice but to upgrade to Core 2 Duo

So, the supposedly objective Mr. Fomitchev has been an AMD loyalist for 15 years. I bought an Athlon 550 about 15 years ago however my next purchase was a Pentium 4 1.8Ghz. Since then I've bought an AMD notebook and a faster Pentium 4 system. So, I guess if I'm clinging to AMD I'm not doing a very good job of it with a 50/50 AMD/Intel purchase record. It looks like Mr. Fomitchev is trying to project his "utter confusion and disappointment" to me.

and refuses to acknowledge the benchmarks that favor Core 2 Duo over Athlon X2 . . . All the benchmarks cannot be wrong . . . No matter how much you analyze and criticize the benchmarks looking for AMD-favoring scenarios that have been downplayed the tests clearly indicate that Intel does have a significant lead in most applications and in power consumption . . . we must admit that Core 2 Duo on average is better than Athlon X2.

I'm wondering now if Mr. Fomitchev has ever actually read my blog. This characterization is nothing but fiction. Again, let's compare what I say with the supposedly objective Mr. Fomitchev.

Core 2 Duo getting better power consumption and 10-20% more performance

I generally give Intel a 15% integer lead so I'm surprised that Intel doesn't seem to show this kind of lead for a lot of the tests . . . it seems for many tasks, Intel only appears to be 10% faster.

So, let me see if I understand this. When Mr. Fomitchev says "10-20% faster" he fancies himself to be the essence of objectivity. Yet, when I say "10-15% faster" I'm somehow in denial and refusing to acknowledge C2D's superiority. This is obviously silly. And, perhaps he should actually read my blog instead of projecting.

It is true that my articles are often in defense of AMD, however, that is not because I lack objectivity. It is simply that when I find many articles on the web that are pro-Intel or incorrectly anti-AMD I don't feel any urgency to defend Intel. However, on AMDZone I have corrected people several times when they understated Intel's processor performance or features. In contrast, I have seen erroneous information about AMD processors go uncorrected on ForumZ several times.

The only other reason I can think of that Mr. Fomitchev would speak ill of my blog is that he seems to place far more importance on the testing done at Tom's Hardware Guide than I do. For example:

After all Tom's Hardware comparative CPU benchmarks clearly show that E6600 leads Athlon X2 5000+ by a good margin on all the text except for Sandra's Floating Point/Memory bench. So much to my disgust I had to order a Core 2 Duo system.

So, indeed, he likes the Tom's Hardware Guide testing. However, I've never felt any disgust at purchasing an Intel system so again I can only assume that he is projecting his feelings to me. In terms of the THG testing, however, my standards are simply higher than his. At TechReports they give a lot of information about their testing procedures whereas THG tends to be more secretive. TechReports shows core loading graphs during testing whereas THG does not. In fact, THG still favors single threaded benchmarks on multi-core processors. Although some have attempted to argue that single thread benchmarks represent what most people run and therefore are legitimate for testing this is an empty argument. If what most people actually run is single threaded code then the conclusion of any such review should be that there is no reason to purchase a multi-core processor; a single core would do just fine. However, there is no way that THG is going to reach such a conclusion because that would be saying that there is no reason to buy Intel's new Core 2 Duo and THG is not going to say that. Therefore, we end up with testing that does not support the conclusions and apparently people like Mr. Fomitchev fail to notice that.

If anyone can honestly say that there is a reason to buy a multi-core processor then they need to do proper testing where each core is loaded which THG does not do. THG has also shown a willingness to cheat in testing by using the Intel compiler when a neutral compiler is available (which will probably become the standard for software developers). THG also cheats by putting overclocked Intel processors up against stock AMD processors and by "tweaking" the performance of Intel processors by overclocking the FSB. I have nothing against overclocked testing but these should be separate articles and not in general reviews. Anandtech's recent blunder in measuring AMD's 65nm L2 cache speed along with the huge error in DailyTech on AMD's expected earnings also show that Anandtech is still untrustworthy.

TechReports makes the admirable statement that all of their testing should be repeatable. I don't know of any similar claim from Anandtech or Tom's Hardware Guide. I will readily admit that C2D could be faster than I have stated but without proper testing there is no way to know. Finally, if the web-press does turn against Intel and starts publishing a lot of erroneous information then I will have to start doing more aticles in Intel's defense. I don't see AMD as an underdog fighting for truth and freedom against the evil Intel. AMD is a pretty large electronics company that is working to gain enough share of the market to be able to share in standards creation. Intel's adoption of AMD64 and the recent speculation about DDR2-1066 could suggest that this is happening.